PILOTING ENVIRONMENTAL DNA FOR FISHERIES MONITORING IN THE UPPER
PEACE RIVER WATERSHED

by
Andrew McDermot-Fouts
B.Sc. University of Northern British Columbia, 2014

THESIS SUBMITTED IN PARTIAL FULFILLMENT OF
THE REQUIREMENTS FOR THE DEGREE OF
MASTER OF SCIENCE
IN
NATURAL RESOURCES AND ENVIRONMENTAL STUDIES

UNIVERSITY OF NORTHERN BRITISH COLUMBIA
March 2024
© Andrew McDermot-Fouts, 2024

ABSTRACT

Applications of environmental DNA (eDNA) represent one of the most significant recent
advances in aquatic species monitoring. eDNA has the potential to dramatically increase the
effectiveness and efficiency of fisheries monitoring across large geographic areas and particularly
for species that have traditionally been difficult to survey. The Williston Reservoir and its
surrounding watershed in northern British Columbia is a vast and heavily disturbed aquatic
ecosystem where fish stocks have been difficult to monitor with traditional survey techniques.
Although eDNA has many potential benefits, the validation process required in a novel
environment can be a barrier to adoption. Primer choice and validation, study design and
laboratory workflows are critical considerations for eDNA studies and must be evaluated when
developing a novel approach in a particular geographic area. My research advanced the validation
of species-specific eDNA assays for six prominent pelagic species in the Williston Reservoir and
a generic, metabarcoding primer set that was effective in detecting the diversity of fish in the
surrounding streams and rivers. The species-specific approach included the validation of four
published assays and the development of novel assays for detecting lake trout (Salvelinus
namaycush) and peamouth (Mylocheilus caurinus). Lake trout and bull trout (Salvelinus
confluentus) are very closely related species that are sympatric throughout much of northwestern
North America. The development of a novel lake trout assay that did not cross amplify bull trout
eDNA is an important achievement for species-specific eDNA monitoring. Reservoir samples
were collected from above and below the thermocline during the summer when the reservoir was
stratified to identify species-specific patterns of eDNA distribution throughout the water column.
Species detections in the reservoir were comparable for species-specific and metabarcoding assays
ii

when compared against gillnet catches in the reservoir. Abundant species that exhibit diel vertical
migrations were detected by eDNA and gillnets at all depths. eDNA copy numbers detected were
higher for samples with greater abundance and biomass of each species caught in the gillnets.
Samples collected from tributaries around the reservoir were tested with metabarcoding primers
and species detections were compared to results from a species-specific assay for Arctic grayling
and a snorkel survey. Metabarcoding results were comparable to those from the Arctic grayling
eDNA study and the snorkels survey, although the metabarcoding methods in this study were less
sensitive due to a reduced amount of sample replication. Piloting eDNA in the Williston Reservoir
and surrounding watershed has provided valuable insights into eDNA sampling design and the
overall workflow for future eDNA studies in the region.

TABLE OF CONTENTS
ABSTRACT...................................................................................................................................................... ii
LIST OF TABLES ............................................................................................................................................. vi
LIST OF FIGURES ......................................................................................................................................... viii
ACKNOWLEDGEMENS ................................................................................................................................. xii
PROLOGUE .................................................................................................................................................... 1
The Potential for eDNA ............................................................................................................................. 4
Research Objectives .................................................................................................................................. 6
CHAPTER 1: STUDY AREA .............................................................................................................................. 8
Williston Reservoir Limnology .................................................................................................................. 8
Tributaries of the Williston Reservoir ..................................................................................................... 10
Fish Species in the Williston Watershed ................................................................................................. 10
Pelagic Fish Species of the Williston Reservoir ....................................................................................... 11
Lake Whitefish..................................................................................................................................... 12
Kokanee............................................................................................................................................... 13
Peamouth ............................................................................................................................................ 14
Rainbow Trout..................................................................................................................................... 15
Bull Trout............................................................................................................................................. 16
Lake Trout ........................................................................................................................................... 16
Knowledge Gaps in the Williston Watershed ......................................................................................... 17
Conventional Survey Approaches for Pelagic Fish in the Williston Reservoir ........................................ 19
Conventional Survey Approaches in Streams and Rivers ....................................................................... 20
CHAPTER 2: DEVELOPMENT AND VALIDATION OF SPECIES-SPECIFIC eDNA ASSAYS FOR PELAGIC FISH IN
THE WILLISTON RESERVOIR ........................................................................................................................ 21
INTRODUCTION ....................................................................................................................................... 21
METHODS ................................................................................................................................................ 25
Creating a Mitochondrial DNA Sequence Database ........................................................................... 25
In Silico Screening of Published Assays ............................................................................................... 25
Novel Assay Design ............................................................................................................................. 26
In Vitro Assay Validation ..................................................................................................................... 30
In Situ Assay Testing in the Williston Reservoir .................................................................................. 33
RESULTS .................................................................................................................................................. 40
iv

In Silico Screening and Novel Assay Design ........................................................................................ 40
In Vitro Validation ............................................................................................................................... 42
In Situ Assay Testing in the Williston Reservoir .......................................................................................... 44
DISCUSSION............................................................................................................................................. 50
Assay Design and In Vitro Validation .................................................................................................. 50
In Situ Assay Validation in the Williston Reservoir ............................................................................. 55
CHAPTER 3: VALIDATION OF eDNA METABARCODING ASSAYS FOR FISH SPECIES IN THE WILLISTON
WATERSHED ................................................................................................................................................ 63
INTRODUCTION ....................................................................................................................................... 63
METHODS ................................................................................................................................................ 65
Creating a Mitochondrial DNA Sequence Database ........................................................................... 65
In Silico Screening................................................................................................................................ 66
In Vitro Validation ............................................................................................................................... 68
In Situ Assay Testing in the Williston Watershed................................................................................ 68
RESULTS .................................................................................................................................................. 74
In Silico Validation ............................................................................................................................... 74
ASV Assignment and Species Resolution ............................................................................................ 79
Non-fish Sequences ............................................................................................................................ 80
Mock Community ................................................................................................................................ 81
In Situ Assay Testing on Reservoir eDNA Samples .............................................................................. 82
In Situ Assay Testing on Williston Tributary eDNA Samples ............................................................... 84
DISCUSSION............................................................................................................................................. 87
eDNA Metabarcoding in the Williston Reservoir ................................................................................ 87
eDNA Metabarcoding in Williston Tributaries .................................................................................... 90
Evaluating eDNA Metabarcoding for use in the Williston Watershed ............................................... 94
EPILOGUE .................................................................................................................................................... 98
BIBLIOGRAPHY .......................................................................................................................................... 104
APPENDIX A: IN VITRO qPCR ASSAY VALIDATION .................................................................................... 116
Synthetic DNA Templates ..................................................................................................................... 116
Temperature Gradient Tests ................................................................................................................. 118
Specificity Tests ..................................................................................................................................... 120
Sensitivity Tests..................................................................................................................................... 124
APPENDIX B: CANDIDATE qPCR ASSAYS................................................................................................... 130

v

Candidate Assay Design and Optimization ....................................................................................... 131
Candidate Assay Specificity ............................................................................................................... 132
APPENDIX C: PELAGIC SPECIES QPCR ASSAY VALIDATIONS FOR THE WILLISTON RESERVOIR ................ 136
APPENDIX D: ENVIRONMENTAL GENOMICS FACILITY (GREAT LAKES INSTITUTE OF ENVIRONMENTAL
RESEARCH) IN-HOUSE PCR CLEAN-UP PROCEDURE.................................................................................. 139

LIST OF TABLES
Table 1. 1. The 19 fish species known to occur in the Williston Watershed and their general habitat
preferences. ......................................................................................................................................... 11
Table 2. 1. Published qPCR assays for bull trout, kokanee and rainbow trout selected for validation and
novel assays for lake trout and peamouth that were developed in this study for use in the Williston
Watershed. .......................................................................................................................................... 26
Table 2. 2. Primer and probe mismatches of the LT_CYTB assay with closely related, non-target species
from the Williston Watershed. Mismatches between the primer and probe and non-target species
sequences are underlined and coloured. Critical mismatches (So et al., 2020) are coloured red and
other mismatches are coloured blue. ................................................................................................... 28
Table 2. 3. Primer and probe mismatches of the PCC_CYTB assays with closely related, non-target
species from the Williston Watershed. Mismatches between the primer and probe and non-target
species sequences are underlined and coloured. Critical mismatches (So et al., 2020) are coloured red
and other mismatches are coloured blue. Mismatches with polymorphic sites (for longnose dace)
were also noted, although they may not always be a mismatch.......................................................... 29
Table 2. 4. Characteristics and secondary structure analysis of the LT_CYTB and PCC_CYTB assays as
calculated with IDT’s OligoAnalyzer Tool......................................................................................... 30
Table 2. 5. List of species whose extracted DNA was included in the specificity testing of each qPCR
assay. Numbers represent the number of samples per species that each assay was tested against. .... 32
Table 2. 6. Optimal annealing temperatures determined from temperature gradient tests with each assay
and a synthetic DNA template (gBlock). ............................................................................................ 42
Table 2. 7. Limits of detection (LOD) for the six assays tested in this study. Concentrations of the
synthetic DNA standards were adjusted after comparison with ddPCR results from the dilution of
highest concentration in the series. Standard LOD is the lowest dilution level with >95% detection
amongst replicates. Modeled LOD was assessed with a statistical model for digital PCR developed
by Hunter et al. (2017). Modeled LOD was not applicable (N/A) when the relationship between
expected and measured DNA concentrations remained linear and a “concentration plateau” was not
observed at low concentrations. For both Standard and Modeled LOD, the DNA concentration of the
sample and the concentration in the 20 µL PCR reaction mixture are shown. ................................... 44
Table 3. 1. Primer sequences of the metabarcoding assays selected for testing on the fish species in the
Williston Watershed. The underlined “C” base in the MiFish-U_F primer was changed from the “T”
in the original sequence to perfectly match all 19 fish species considered in this study. UniA
sequences were added to the 5’ ends of forward primers and UniB sequences were added to 5’ ends
of reverse primers for barcoding and sequencing on the Ion Torrent platform. ................................. 67
vi

Table 3. 2. Location, year and collection information for the eDNA samples collected from around the
Williston watershed and analyzed with the metabarcoding primers in this study (Stamford et al.,
2020, 2022). ........................................................................................................................................ 69
Table 3. 3. FishCYTB primer mismatches with the CYTB sequences of 19 species from the Williston
Watershed. .......................................................................................................................................... 76
Table 3. 4. Read counts for each of the 12 species included in the mock community samples.................. 82
Table 3. 5. Species detections above (<10 m depth) and below (> 10 m depth) the thermocline in the
Williston Reservoir at the Teare Creek and Factor Ross stations with eDNA and gillnets. eDNA
metabarcoding species detections with a single assay are noted as M for MiFish or F for FishCYTB.
Detections with both metabarcoding assays are noted as B. Detection confidence with eDNA
metabarcoding is indicated as high (green), moderate (yellow) or low (red). Species-specific
detections with qPCR assays and species caught by gill net are noted as X.Species not tested for with
qPCR are noted with a dash (-). .......................................................................................................... 83
Table 3. 6. Species detections with the MiFish (M), FishCYTB (F), or both (B) metabarcoding assays in
tributary rivers and creeks to the Williston Reservoir. Detection confidence is indicated as high
(green), moderate (yellow) or low (red).............................................................................................. 84
Table 3. 7. Species detections with the MiFish (M), FishCYTB (F) or both (B) metabarcoding assays
from the middle, upper and headwaters reaches of the Ingenika River. Snorkel counts for the middle
Ingenika River were 279 mountain whitefish, 18 bull trout, 19 rainbow trout, 27 Arctic grayling.
Snorkel counts for the upper Ingenika River were 259 mountain whitefish, 37 bull trout, 19 rainbow
trout, 23 Arctic grayling. ..................................................................................................................... 86
Table A 1. Limits of detection for the six assays tested in this study. Concentrations of the synthetic DNA
standards were adjusted after comparison with ddPCR. Standard LOD is the lowest dilution level
with >95% detection amongst replicates. Modeled LOD was assessed with a statistical model for
digital PCR developed by Hunter et al. (2017). Modeled LOD was not applicable (N/A) when the
relationship between expected and measured DNA concentrations remained linear and a
“concentration plateau” was not detected at low concentrations. ..................................................... 125
Table A 2. Characteristics and secondary structure analysis of the unsuccessful candidate assays for bull
trout and lake trout as calculated with IDT’s OligoAnalyzer Tool. .................................................. 130
Table A 3. The five validation levels described by Thalinger et al. (2021) and the validation level
achieved by each of the six assays tested in this study. .................................................................... 138
Table A 4. Reagents and amounts used in bead solution recipe for PCR clean-up. ................................. 139

vii

LIST OF FIGURES
Figure 1. 1. The Williston Reservoir watershed in northern British Columbia. The Finlay Reach is the
flooded valley of the south flowing Finlay River, the Parsnip Reach is the flooded valley of the north
flowing Parsnip River, and the Peace Reach encompasses the east flowing Peace River valley
between the confluence of the Parsnip and Finlay tributaries and the W.A.C. Bennet dam near
Hudson’s Hope, BC. Arrows indicate the Factor Ross and Teare Creek sample locations in the
Finlay Reach where eDNA samples were collected. Gillnet and trawl surveys were conducted at the
Teare Creek and Finlay Forks locations................................................................................................ 9
Figure 2. 1. Dissolved oxygen (DO) and temperature profile measured at the Teare Creek sample location
on 16 August 2021. eDNA sample depths are indicated by arrows.................................................... 36
Figure 2. 2. Tests for sample inhibition and integrity with the lambda (top) and ePlant (bottom) assays.
Droplet charts are from tests of first elution samples from Teare Creek (TC) and Factor Ross (FR) at
sample depths that are indicated by numbers in the sample codes. The master mix for all samples
was spiked with lambda DNA at a concentration of 2 copies per µL. The low amplitude of positive
droplets for lambda and the absence of positive droplets for ePlant at FR20 is indicative of
inhibition. Droplets that were positive for PCR amplification are coloured and negative droplets are
grey. .................................................................................................................................................... 45
Figure 2. 3. Total number of eDNA copies from each species detected in 10 µL of the first elution of all
samples at Factor Ross (left) and Teare Creek (right). Counts of peamouth eDNA copies are omitted
because tests of the PCC_CYTB assay were not comparable. ............................................................ 46
Figure 2. 4. Species detections using eDNA by depth at Factor Ross (left) and Teare Creek (right) sample
locations. A total of three 1-litre eDNA samples were collected from each depth and their extractions
were tested with six species-specific qPCR assays. ............................................................................ 47
Figure 2. 5. Gillnet catches by species at each set depth between the surface and 20 m at Teare Creek
(left) and Finlay Forks (right). Each net panel spanned two vertical metres in the water column. .... 49
Figure 2. 6. Total biomass (left) and total count (right) of each species caught in gillnets at Teare Creek.
............................................................................................................................................................ 49
Figure 3. 1. Neighbour joining tree for the MiFish amplicons (214-218 bp) from 18 species in the
Williston Reservoir created using the Tamura-Nei distance model. Percent consensus based on 500
bootstrap iterations is shown at the nodes. The scale bar indicates the relative genetic distance
between amplicon sequences for each species. ................................................................................... 77
Figure 3. 2. Neighbour joining tree for the FishCYTB amplicons (209 bp) from 19 species in the
Williston Reservoir created using the Tamura-Nei distance model. Percent consensus based on 500
bootstrap iterations is shown at the nodes. The scale bar indicates the relative genetic distance
between amplicon sequences for each species. ................................................................................... 78
Figure A 1. Synthetic DNA template sequence used to test the Sock_COI and Rain_COI assays. Primers
and probes are annotated for the Sock_COI and Rain_COI assays as H-Sock-COI and H-Rain-COI,
respectively. The template also included sequences for other salmon species not tested for in this
study. ................................................................................................................................................. 116
Figure A 2. Synthetic DNA template sequence used to test the COCL_CYTB and BUT1_COCL assays.
Forward primers are coloured dark green, reverse primers are coloured light green, and probes are
coloured red. ..................................................................................................................................... 117

viii

Figure A 3. Synthetic DNA template sequence used to test all the candidate assays for lake trout. Forward
primers are coloured dark green, reverse primers are coloured light green, and probes are coloured
red. The primers and probe labeled LT4 were the most successful candidates and became the novel
LT_CYTB assay featured in this study. ............................................................................................. 117
Figure A 4. Synthetic DNA template sequence used to test the novel peamouth assay, PCC_CYTB.
Forward primers are coloured dark green, reverse primers are coloured light green, and probes are
coloured red. ..................................................................................................................................... 118
Figure A 5. Temperature gradient results for the kokanee assay, Sock_COI............................................ 118
Figure A 6. Temperature gradient results for the rainbow trout assay, Rain_COI. .................................. 119
Figure A 7. Temperature gradient results for the lake whitefish assay, COCL_CYTB. ............................ 119
Figure A 8. Temperature gradient results for the bull trout assay, BUT1_CYTB. .................................... 119
Figure A 9. Temperature gradient results for the lake trout assay, LT_CYTB. ......................................... 120
Figure A 10. Temperature gradient results for the peamouth assay, PCC_CYTB. ................................... 120
Figure A 11. Fluorescence amplitude of droplets plotted against DNA samples of target and non-target
salmonid species for the kokanee assay, Sock_COI. Droplets that were positive for PCR
amplification are coloured and negative droplets are coloured grey. Trace amounts of lake whitefish
contamination were observed with lake trout samples collected from gillnets in the Williston
Reservoir. .......................................................................................................................................... 121
Figure A 12. Fluorescence amplitude of droplets plotted against DNA samples of target and non-target
salmonid species for the rainbow trout assay, Rain_COI. Droplets that were positive for PCR
amplification are coloured and negative droplets are coloured grey. ............................................... 122
Figure A 13. Fluorescence amplitude of droplets plotted against DNA samples of target and non-target
salmonid species for the lake whitefish assay, COCL_CYTB. Droplets that were positive for PCR
amplification are coloured and negative droplets are coloured grey. ............................................... 122
Figure A 14. Fluorescence amplitude of droplets plotted against DNA samples of target and non-target
salmonid species for the bull trout assay, BUT1_CYTB. Droplets that were positive for PCR
amplification are coloured and negative droplets are coloured grey. ............................................... 123
Figure A 15. Fluorescence amplitude of droplets plotted against DNA samples of target and non-target
salmonid species for the lake trout assay, LT_CYTB. Droplets that were positive for PCR
amplification are coloured and negative droplets are coloured grey. ............................................... 123
Figure A 16. Fluorescence amplitude of droplets plotted against DNA samples of target and non-target
Leuciscidae species for the peamouth assay, PCC_CYTB. Droplets that were positive for PCR
amplification are coloured and negative droplets are coloured grey. ............................................... 124
Figure A 17. ddPCR-based LOD assessment of the kokanee assay, Sock_COI, with a dilution series of
synthetic kokanee DNA. Synthetic DNA concentrations along the x-axis include a 1:5 standard
dilution with eight points and 6 replicates each. The log of measured concentrations is along the yaxis. Replicate averages (including zeros) are denoted by a red X. The Standard LOD is the lowest
dilution level with 100% detection across all replicates. The Modeled LOD could not be calculated
for the Sock_COI assay because the dilution series did not exhibit a concentration plateau. ........... 126
Figure A 18. ddPCR-based LOD assessment of the rainbow trout assay, Rain_COI, with a dilution series
of synthetic rainbow trout DNA. Synthetic DNA concentrations along the x-axis include a 1:5
standard dilution with eight points and 6 replicates each. The log of measured concentrations is along
the y-axis. Replicate averages (including zeros) are denoted by a red X. The Standard LOD is the
lowest dilution level with 100% detection across all replicates. The Modeled LOD is defined at the
point of intersection between linearly related standard concentrations (solid line) and instrumental
responses that do not vary with concentration (concentration plateau; dashed line). Upper and lower

ix

95% confidence limits are shown (dotted lines). The upper 95% confidence limit was used as a
conservative bound for the Modeled LOD (Hunter et al. 2017). ...................................................... 126
Figure A 19. ddPCR-based LOD assessment of the lake whitefish assay, COCL_CYTB, with a dilution
series of synthetic lake whitefish DNA. Synthetic DNA concentrations along the x-axis include a 1:5
standard dilution with eight points and 6 replicates each. The log of measured concentrations is along
the y-axis. Replicate averages (including zeros) are denoted by a red X. The Standard LOD is the
lowest dilution level with 100% detection across all replicates. The Modeled LOD is defined at the
point of intersection between linearly related standard concentrations (solid line) and instrumental
responses that do not vary with concentration (concentration plateau; dashed line). Upper and lower
95% confidence limits are shown (dotted lines). The upper 95% confidence limit was used as a
conservative bound for the Modeled LOD (Hunter et al. 2017). ...................................................... 127
Figure A 20. ddPCR-based limit of detection (LOD) assessment of the bull trout assay, BUT1_CYTB,
with a dilution series of synthetic bull trout DNA. Synthetic DNA concentrations along the x-axis
include a 1:5 standard dilution with eight points and 6 replicates each. The log of measured
concentrations is along the y-axis. Replicate averages (including zeros) are denoted by a red X. The
Standard LOD is the lowest dilution level with 100% detection across all replicates. The Modeled
LOD could not be calculated for the Sock_COI assay because the dilution series did not exhibit a
concentration plateau. ....................................................................................................................... 127
Figure A 21. ddPCR-based LOD assessment of the lake trout assay, LT_CYTB, with a dilution series of
synthetic lake trout DNA. Synthetic DNA concentrations along the x-axis include a 1:5 standard
dilution with eight points and 6 replicates each. The log of measured concentrations is along the yaxis. Replicate averages (including zeros) are denoted by a red X. The Standard LOD is the lowest
dilution level with 100% detection across all replicates. The Modeled LOD is defined at the point of
intersection between linearly related standard concentrations (solid line) and instrumental responses
that do not vary with concentration (concentration plateau; dashed line). Upper and lower 95%
confidence limits are shown (dotted lines). The upper 95% confidence limit was used as a
conservative bound for the Modeled LOD (Hunter et al. 2017). ...................................................... 128
Figure A 22. ddPCR-based limit of detection (LOD) assessment of the peamouth assay, PCC_CYTB, with
a dilution series of synthetic peamouth DNA. Synthetic DNA concentrations along the x-axis include
a 1:5 standard dilution with eight points and 6 replicates each. The log of measured concentrations is
along the y-axis. Replicate averages (including zeros) are denoted by a red X. The Standard LOD is
the lowest dilution level with 100% detection across all replicates. The Modeled LOD is defined at
the point of intersection between linearly related standard concentrations (solid line) and
instrumental responses that do not vary with concentration (concentration plateau; dashed line).
Upper and lower 95% confidence limits are shown (dotted lines). The upper 95% confidence limit
was used as a conservative bound for the Modeled LOD (Hunter et al. 2017). ............................... 129
Figure A 23. Fluorescence amplitude of ddPCR droplets plotted against DNA samples of target and nontarget species from the Williston Reservoir for the BT2_CYTB candidate bull trout assay (top) and
the BUT1_CYTB assay (bottom) from Wilcox et al. (2013). Some of the lake whitefish tissue
samples that DNA was extracted from were contaminated with bull trout DNA. This is evident from
the small number of positive droplets in some but not all lake whitefish samples. .......................... 133
Figure A 24. Fluorescence amplitude of ddPCR droplets plotted against DNA samples of target and nontarget species from the Williston Reservoir for the BT3_CYTB candidate bull trout assay. Some of
the lake whitefish tissue samples that DNA was extracted from were contaminated with bull trout
DNA. This is evident from the small number of positive droplets in some but not all lake whitefish
samples.............................................................................................................................................. 134

x

Figure A 25. Fluorescence amplitude of ddPCR droplets plotted against DNA samples of target and nontarget species from the Williston Reservoir for the LT1_CYTB candidate lake trout assay. Some of
the lake whitefish and bull trout tissue samples that DNA was extracted from were contaminated
with lake trout DNA. This is evident from the small number of positive droplets in some but not all
lake whitefish samples. ..................................................................................................................... 135

xi

ACKNOWLEDGEMENS
I would like to thank my supervisor Dr. J. Mark Shrimpton for all of his help and guidance
throughout my graduate research. I would also like to thank the other members of my research
committee, Dr. Margaret Docker from the University of Manitoba, and Dr. Brent Murray from
the University of Northern British Columbia for their help with my project and their careful
review of my thesis. I would like to thank Dr. Heath from the University of Windsor for serving
as my external examiner and for the thoughtful comments he made. I am grateful for the help
that Caitlin Pitt provided in the lab at UNBC for my ddPCR analysis. I owe a special thank-you
to Shelby Toews, Jonathon Leblanc and James Watkins from the University of Windsor for their
help with my metabarcoding sample processing and with the bioinformatics analysis. I would
like to thank Bryce O’Connor and staff from Chu Cho Environmental for their help with data
collection in the Williston Reservoir. I would alsso like to thank Mike Stamford for sharing his
eDNA samples from around the Williston Watershed and John Haggen for helping with the
snorkel survey data. I would never have been able to complete a graduate degree without the
personal funding provided by Veterans Affairs Canada, the Natural Sciences and Engineering
Research Council of Canada, and UNBC. This project was made possible with funding from the
Fish and Wildlife Compensation Program, the Genomic Network for Fish Identification, Stress
and Health, and Genome Canada.

xii

PROLOGUE

Effective management of a fish stock requires an understanding of fish abundance, species
composition, size and age composition, spatial distribution patterns, and an assessment of the
accuracy and precision of the information gathered (Kubečka et al., 2009). Large lakes and
reservoirs are difficult to study because sampling methods can be limited in their effectiveness and
even cost-prohibitive on a large scale. The approaches used to monitor species within the aquatic
environment often require intensive effort to determine species presence and considerably more
effort to accurately assess population size. Multiple sampling techniques are often used to
compensate for gear biases, and the approach may also differ depending on the type or “size” of
habitat (Kubečka et al., 2009; Rudstam et al., 2012; Winfield et al., 2009). Fisheries scientists are
often forced to settle for an estimate of the “true picture” of a stock because the resources needed
for an accurate assessment are seldom available (Kubečka et al., 2009). Stock assessments provide
a snapshot that is only valid for a particular moment in time and multiple snapshots are needed to
truly understand fish population dynamics. Knowledge of temporal variation in numbers of fish
and use of habitat is essential when habitat is altered through anthropogenic disturbance.
The creation of large reservoirs by the impoundment of river systems causes profound
environmental changes which can drastically affect fish populations (Blackman, 1992; Ross,
1997). Many kilometres of lotic habitat are flooded and converted to hectares of lentic habitat.
Reservoirs usually experience a short period of high productivity shortly after they are created,
followed by a gradual loss of nutrients and productivity over time (Coxson et al., 2018; Schindler
et al., 2009). Species composition in a reservoir can change dramatically as the original riverine
species decline and lacustrine species become dominant in the reservoir and as reservoir
1

productivity fluctuates (Blackman, 1992). The effects on riverine species can extend far beyond
the reservoir if much of the remaining fluvial habitat is restricted to higher gradient headwaters
and if the reservoir becomes a migration barrier, isolating critical habitats and populations (Clarke
et al., 2007; Stamford & O’Connor, 2022). Reservoirs often present many unique challenges for
fisheries sampling that are not encountered in natural lakes. For example, the flooded forests that
underly many temperate reservoirs can snag gears with nets or anchor lines and can hide fish that
might otherwise be detected by hydroacoustic methods (Pillipow & Langston, 2002; Sebastian et
al., 2003, 2009). Monitoring fish stocks in the hundreds of tributary streams that drain into a
reservoir can be an even more daunting task. Sampling approaches in streams can be even more
varied and resource intensive than those designed for a reservoir due to factors such as the variety
of habitat types, access issues, and the amount of effort required. Conventional approaches to
fisheries inventories in streams and reservoirs are resource intensive, costly and have many
limitations, often leaving fisheries managers to make difficult decisions based on information that
may be outdated, inaccurate, or simply unavailable.
Combining extensive hydroacoustic surveys with limited gillnetting has become an
internationally adopted method for fisheries assessments in large reservoirs and lakes (Kubečka et
al., 2009; Winfield et al., 2009). Acoustic methods measure the distribution and abundance of fish
by transmitting sound pulses (pings) through the water column and analyzing the return echoes.
The time delay between the transmission of a ping and the reception of a return echo is analyzed
to determine fish distribution. Simultaneously, the intensity of the returning echo is analyzed to
determine fish size (Rudstam et al., 2012). Because hydroacoustic methods only determine the size
of individual fish, they cannot easily differentiate between species without the biological data that
is provided by traditional fishing gears. Gillnetting surveys are most commonly used in

2

combination with hydroacoustic surveys to determine the biological characteristics of each species
in the surveyed environment (Kubečka et al., 2009; Rudstam et al., 2012). Every sampling
technique has limitations and combining more than one technique is often required to assess fish
stocks in lakes and reservoirs (Kubečka et al., 2009).
Combining hydroacoustic methods with traditional gears can improve the accuracy and
detail of fisheries surveys but there are strengths and weaknesses to any combination of techniques.
For example, the first few metres of the water column is a blind spot for downward looking
hydroacoustic transducers and can be problematic if fish are concentrated near the surface
(Rudstam et al., 2012; Sebastian et al., 2003). Gillnetting results have been relied upon to
characterize the distribution and abundance of fish near the surface that cannot be effectively
targeted by hydroacoustic surveys, but gillnetting can be highly selective (Olin et al., 2009; Zale
et al., 2012). Any biases associated with traditional gears are incorporated into the acoustic
estimates (Rudstam et al., 2012). The probability that fish will encounter a gillnet and be retained
depends a variety of factors including their behaviour and activity, swimming speed, size and
morphology (Olin et al., 2009). Gillnets do not efficiently capture rare species (Sard et al., 2019;
Shelton et al., 2019; Zale et al., 2012), fish that are too small or too large for the mesh sizes in the
net, fish that swim too slowly to become entangled and fish who are not active at the times,
locations and depths where the nets are set (Olin et al., 2009; Vašek et al., 2009). Furthermore,
gillnet catch rates decline over time as they accumulate fish (Olin et al., 2009). Incorporating an
active sampling gear, such as trawling, can overcome many of the pitfalls of gillnetting and
hydroacoustic methods (Kubečka et al., 2009) but may still fall short of providing a true picture of
the stock being assessed. Fast swimming fish, such as large piscivores, tend to avoid trawls (Bethke
et al., 1999) and the depth of the trawl can select for certain species over others (Olin et al., 2009).

3

A considerable drawback of hydroacoustic and trawling surveys are their associated cost and
complexity (Kubečka et al., 2009). Hydroacoustic surveys require specialized equipment and a
sampling crew with expert knowledge and skills. Trawling is expensive and requires a specialized
boat and a trained crew. Qualified crews and appropriate equipment for hydroacoustic and trawling
surveys are often a limited resource (see Sebastian et al., 2003) and may contribute to the omission
of trawls for long term surveys of lentic environments.
The Potential for eDNA
Genetic analysis of fish DNA obtained directly from water samples (environmental DNA)
has become a powerful tool for biodiversity monitoring (Jerde, 2019; Schenekar, 2023). Two
approaches to detecting species with environmental DNA (eDNA) are methods designed to detect
a single species or methods that detect multiple species (Thomsen & Willerslev, 2015). Single
species detection uses the polymerase chain reaction (PCR) or quantitative PCR (qPCR) to identify
one or a few known species in a sample with species-specific primers and probes. Multi-species
detection (eDNA metabarcoding) uses next generation sequencing (NGS) in combination with
generic primers to identify a group of species in an eDNA sample by analyzing genetic barcodes.
Single species methods have greater specificity, sensitivity, and quantification ability but can only
detect one target species at a time (Bylemans et al., 2019; Thomsen & Willerslev, 2015).
Metabarcoding is more expensive and time consuming overall compared to species-specific
methods but is much more efficient and cost effective for diverse systems with many known
species or in systems where information about species composition is lacking. Metabarcoding is
also useful for comparing habitat use, species interactions and in identifying potentially critical
habitats (biodiversity hotspots) (Balasingham et al., 2018).

4

eDNA has the potential to supplement or even replace traditional methods of fisheries
inventory and improve the quality of data gathered for a relatively low cost. eDNA can usually
detect rare species with greater efficiency and sensitivity compared to conventional sampling gears
such as gill netting (Evans, Shirey, et al., 2017; Jerde et al., 2011; Thomsen et al., 2012) and has
proven to be particularly useful for monitoring lake trout (Salvelinus namaycush) (Lacoursière‐
Roussel, Rosabal, et al., 2016; Littlefair et al., 2020). Fish eDNA tends to have a smoother
distribution in space and time than the fish themselves, and integrates information about fish
presence and abundance over a larger area than most traditional methods of fish capture (Pont et
al., 2018; Shelton et al., 2019). eDNA metabarcoding can provide a more accurate and efficient
measure of community diversity, detecting more species with fewer samples and less effort (Sard
et al., 2019).
The distribution of eDNA in a waterbody can provide valuable information about the
distribution of species. Seasonal thermal stratification and the thermal preferences of fish species
influences the distribution of eDNA in lakes (Klobucar et al., 2017; Littlefair et al., 2020). Thermal
stratification in lakes separates the epilimnion and hypolimnion in summer (Wetzel, 2001). Fish
distribute throughout the water column in accordance with their thermal niche, and in response to
predators and prey. Littlefair et al. (2020) found that eDNA distributions also became stratified in
summer. The eDNA of species such as lake trout that remain below the thermocline were only
detected in water samples taken from the hypolimnion and species with a preference for warmer,
shallower water were only detected in samples taken from the epilimnion. However, collecting
eDNA samples at various depths throughout the water column presents several challenges with
respect to sample contamination and potential depth limits of sampling equipment. At least two

5

studies have employed long hoses and electric pumps to sample at specific depths in the water
column (Klobucar et al., 2017; Littlefair et al., 2020).
Many studies have demonstrated that eDNA particle concentrations are positively
correlated with species abundance (Klobucar et al., 2017; Lacoursière‐Roussel, Rosabal, et al.,
2016; Sard et al., 2019; Shelton et al., 2019; Stoeckle et al., 2017; Takahara et al., 2012; Yates et
al., 2020) and could provide a rapid, cost-effective indicator of abundance or biomass for fisheries
stock assessment (Rourke et al., 2021). However, the eDNA particle concentration at a particular
sample location depends on a variety of biotic and abiotic factors that influence the amount of
genetic material being shed into the environment, how and where it is transported, and how long
it persists before it is degraded (Hansen et al., 2018; Klymus et al., 2015; Rourke et al., 2021;
Takahara et al., 2012). The stochasticity of biotic and abiotic processes in natural environments
complicates the relationship between eDNA copy number and fish abundance, but recent
approaches have overcome enough of this uncertainty to provide a reasonably accurate measure
of fish abundance with eDNA by standardizing against estimates with traditional gears (Chambert
et al., 2018; Lacoursière‐Roussel, Rosabal, et al., 2016; Shelton et al., 2019).
Research Objectives
My graduate research has piloted and validated the use of eDNA survey methods for
monitoring fish species. I conducted my research in the Williston Watershed in north-central
British Columbia. I outline the study area and knowledge gaps for the watershed in Chapter 1. In
Chapter 2, I identified existing qPCR assays that were developed to detect kokanee (Oncorhynchus
nerka), rainbow trout (Oncorhynchus mykiss), lake whitefish (Coregonus clupeiformis), and bull
trout in other geographic regions and validated them for their use in the Williston Reservoir. I
developed and validated a novel qPCR assay for lake trout because previously published qPCR
6

assays were not specific enough to discriminate between sympatric lake trout and bull trout. There
were no previously developed qPCR assays for peamouth, so I developed and validated a novel
peamouth assay for use in the Williston Reservoir. By modeling my sampling design and eDNA
analysis on those published in Littlefair et al. (2020), I was able to analyze the vertical distribution
of eDNA in the Williston Reservoir during summer stratification. My eDNA sampling in the
Williston Reservoir was conducted concurrently with the 2021 pelagic fisheries survey led by the
BC Ministry of Forest Lands and Natural Resource Operations and Rural Development
(FLNRORD). In Chapter 3, I compared eDNA metabarcoding and species-specific qPCR methods
for detecting fish in the Williston Reservoir and in a variety of the surrounding tributaries. Both
eDNA methods were compared against conventional sampling gears in the reservoir and in
streams, and I investigated the relationship between eDNA concentration and fish abundance in
the Williston Reservoir. I re-analyzed many eDNA samples collected by Stamford et al. (2020,
2022) from streams in the Williston Watershed with eDNA metabarcoding and compared those
results to the original eDNA detections for Arctic grayling using qPCR.

7

CHAPTER 1: STUDY AREA

Williston Reservoir Limnology
The Williston Reservoir is British Columbia’s largest body of freshwater (surface area of
1779 km2, Stockner et al., 2005). The reservoir was created in 1968 by the construction of the
W.A.C. Bennett Dam on the upper Peace River and is comprised of three reaches (Figure 1. 1): the
Finlay Reach which is the flooded valley of the south flowing Finlay River, the Parsnip Reach
which is the flooded valley of the north flowing Parsnip River and the Peace Reach which
encompasses the east flowing Peace River valley between the confluence of the Parsnip and Finlay
tributaries and the dam. The Finlay and Parsnip reaches lie within a wide, flat bottomed rocky
mountain trench bounded by the Omineca Mountains to the west and the Rocky Mountains to the
east. The Peace Reach is a V-shaped basin and is much narrower and deeper than the Finlay and
Parsnip reaches. The Williston Reservoir is a dimictic system with two periods of deep mixing
(May and November) and two periods of stratification (June to October and February to April).
During summer stratification, the epilimnion depth varies between reaches and depending on the
duration and strength of wind events. Average summer epilimnion depths measured in 2000 ranged
from 20 m to 30 m in the Finlay reach, 15 m to 20 m in the Parsnip Reach, and 30 m to 40 m in
the Peace Reach (Stockner et al., 2005). The Williston Reservoir experienced an initial increase in
productivity during the first decade after impoundment, but reservoir wide productivity has since
declined dramatically to an ultra-oligotrophic state (Harris et al., 2006; Stockner et al., 2005). The
Finlay Reach had the highest measures of productivity among the three reaches in 2000, followed
by the Parsnip Reach and Peace Reach (Stockner et al., 2005).

8

Figure 1. 1. The Williston Reservoir watershed in northern British Columbia. The Finlay Reach is the
flooded valley of the south flowing Finlay River, the Parsnip Reach is the flooded valley of the north
flowing Parsnip River, and the Peace Reach encompasses the east flowing Peace River valley between the
confluence of the Parsnip and Finlay tributaries and the W.A.C. Bennet dam near Hudson’s Hope, BC.
Arrows indicate the Factor Ross and Teare Creek sample locations in the Finlay Reach where eDNA
samples were collected. Gillnet and trawl surveys were conducted at the Teare Creek and Finlay Forks
locations.

9

Tributaries of the Williston Reservoir
The Williston Watershed encompasses all the tributaries that drain into the Williston
Reservoir upstream of the W.A.C. Bennett Dam and includes 10 major rivers (drainage >1,500
km2), seven large rivers (400-1,500 km2), 49 medium rivers (50-400 km2), and 262 small rivers
(<50 km2) (Langston & Blackman, 1993). The creation of the Williston Reservoir drastically
altered the lotic systems within the basin by inundating a large portion of many streams and rivers,
limiting the availability and connectivity of stream habitats for fish (Fish and Wildlife
Compensation Program, 2020).
Fish Species in the Williston Watershed
A diverse assemblage of 19 fishes belonging to four orders and five families have been
identified in the streams, rivers, lakes and reservoir that are encompassed within the Williston
Watershed (Table 1. 1)(Blackman, 1992; Cowie & Blackman, 2003, 2004; Fish and Wildlife
Compensation Program, 2020; Langston & Blackman, 1993; Langston & Murphy, 2008; Province
of British Columbia, 2020; Sebastian et al., 2003). Lentic habitats in the watershed are more
diverse, and often more productive than those in the reservoir. Similarly, the diversity of species
in lentic habitats is greater than in the reservoir. Diversity in local species assemblages often only
includes a small subset of the species in the entire watershed and are frequently dominated by a
few, very abundant species. Many of the species present in the watershed are not widely distributed
and are considered rare.

10

Table 1. 1. The 19 fish species known to occur in the Williston Watershed and their general habitat
preferences.
Common Name
Prickly sculpin
Slimy sculpin
Lake trout
Bull trout
Rainbow trout
Kokanee
Lake whitefish
Mountain whitefish
Pygmy whitefish
Arctic grayling
Longnose sucker
White sucker
Largescale sucker
Peamouth
Northern pikeminnow
Redside shiner
Lake chub
Longnose dace
Burbot

Family
Cottidae
Cottidae
Salmonidae
Salmonidae
Salmonidae
Salmonidae
Salmonidae
Salmonidae
Salmonidae
Salmonidae
Catostomidae
Catostomidae
Catostomidae
Leuciscidae
Leuciscidae
Leuciscidae
Leuciscidae
Leuciscidae
Gadidae

Scientific Name
Cottus asper
Cottus cognatus
Salvelinus namaycush
Salvelinus confluentus
Oncorhynchus mykiss
Oncorhynchus nerka
Coregonus clupeiformis
Prosopium williamsoni
Prosopium coulterii
Thymallus arcticus
Catostomus catostomus
Catostomus commersonii
Catostomus macrocheilus
Mylocheilus caurinus
Ptychocheilus oregonensis
Richardsonius balteatus
Couesius plumbeus
Rhinichthys cataractae
Lota lota

Habitat preference
lotic/lentic (littoral, benthic)
lotic/lentic (littoral, benthic)
lentic (pelagic)
lotic/lentic
lotic
lentic
lentic
lotic
lentic/lotic
lotic
lentic/lotic
lentic/lotic
lotic
lentic
lentic (littoral)/lotic (large, slow)
lotic/lentic
lentic
lotic
lentic/lotic

Pelagic Fish Species of the Williston Reservoir
The fish community assemblage in the Williston Reservoir has changed dramatically over
the last 50 years (Barrett & Halsey, 1975; Blackman, 1992; Pillipow & Langston, 2002; Sebastian
et al., 2003; Stockner et al., 2005), and recent surveys indicate it is still quite dynamic (DWB
Consulting Services Ltd., 2019, 2020, 2021; Langston, 2012; Plate et al., 2012; Sebastian et al.,
2009). The Williston Reservoir converted more than 600 km of mainstem and large river tributary
habitat to reservoir environment and changed the fish community composition in favour of species
adapted to lacustrine habitats (Blackman, 1992). The littoral zone in the reservoir is essentially
abiotic because extensive annual drawdowns for hydroelectric power limit the growth of aquatic
plants and associated aquatic insect communities. As a result, plankton is the primary food source
in this ultra-oligotrophic system (Blackman, 1992; Stockner et al., 2005). The Williston
Reservoir’s fish community has historically been dominated by planktivores, but species

11

composition and relative abundances have fluctuated in the decades since impoundment (Barrett
& Halsey, 1975; Blackman, 1992; Pillipow & Langston, 2002; Plate et al., 2012; Sebastian et al.,
2003, 2009). Recent surveys indicate that lake whitefish, kokanee, and peamouth are the most
abundant pelagic fish in the Williston Reservoir (O’Connor, 2022; Pillipow & Langston, 2002;
Plate et al., 2012; Sebastian et al., 2003, 2009). Piscivorous species in the reservoir are much less
abundant in the pelagic zone and include rainbow trout, bull trout, and lake trout (O’Connor, 2022;
Pillipow & Langston, 2002; Sebastian et al., 2003, 2009). Bull trout and lake trout are apex
predators in the reservoir and exist at very low densities. Lake trout and bull trout populations have
fluctuated since the reservoir was created (Barrett & Halsey, 1975; Blackman, 1992; Pillipow &
Langston, 2002; Sebastian et al., 2003), and lake trout are a potential threat to the long-term
persistence of bull trout in pelagic habitats (Hagen & Weber, 2019). Recent pelagic species
inventories of the Williston Reservoir have been completed approximately every 8-10 years
(O’Connor, 2022; Pillipow & Langston, 2002; Sebastian et al., 2009). The status of the pelagic
fish community is largely unknown in the intervals between inventory years, but other studies have
suggested that significant changes in population dynamics can occur over shorter time spans. For
example, aerial enumerations of kokanee spawning in the Williston Watershed suggested that
kokanee abundance declined sharply in the years before the 2021 pelagic survey (DWB Consulting
Services Ltd., 2019, 2020, 2021), in contrast to findings from the pelagic survey (Weir et al., 2023).
The most abundant pelagic species documented in the Williston Reservoir and the population
changes observed in past surveys are summarized in the following sections.
Lake Whitefish
Lake whitefish were the dominant species in fisheries inventories of the Williston
Reservoir in 1974, 1988, and 2000 (Blackman, 1992; Pillipow & Langston, 2002; Sebastian et al.,
12

2003), but were overtaken by kokanee in 2008 and 2021 (O’Connor, 2022; Sebastian et al., 2009).
Lake whitefish have an optimal temperature range of 8 to 14°C (McPhail, 2007), and pelagic
adults are known to actively select depths where water temperatures are below 16°C (Gorsky et
al., 2012). Juvenile lake whitefish are more tolerant of warmer temperatures than adults and may
occupy water as warm as 19.5°C (McPhail, 2007). Divergence into limnetic and benthic forms has
been described in British Columbia and the Yukon (McPhail, 2007; Roberge & Slaney, 2001).
Limnetic lake whitefish feed on zooplankton near the surface and are known to exhibit diel vertical
migrations throughout the spring and summer (Gorsky et al., 2012). Lake whitefish in the Williston
Reservoir appear to prefer near-surface habitats in summer (< 7.5 m) and have been caught in
similar numbers in surface nets set in both near and off-shore areas (Sebastian et al., 2003).
Surveys in the Peace Reach have detected lake whitefish at both shallow (< 15 m) and mid-water
depths (15-30 m) at night (O’Connor, 2022; Sebastian et al., 2003, 2009).
Kokanee
Kokanee are native to Thutade Lake at the headwaters of the Finlay River, and to Arctic
and Tacheeda Lakes in the headwaters of the Parsnip River. A small population of native kokanee
established themselves in the Williston Reservoir shortly after impoundment (Langston, 2012) and
recent genetic research has identified the Thutade Lake population as their source (Wilson &
Shrimpton, 2020). Kokanee originating from Hill Creek (Arrow Reservoir) and Meadow Creek
(Kootenay Lake) of the Columbia River were introduced into the Williston Reservoir between
1990 and 1998 (Langston & Murphy, 2008; Langston, 2012). Kokanee populations increased
dramatically following the introduction of non-native stocks to the reservoir. Pelagic fish surveys
in 2008 and 2012 identified kokanee as the new dominant planktivorous species in the reservoir,
surpassing lake whitefish (Plate et al., 2012; Sebastian et al., 2009). Aerial enumerations identified
13

a sharp increase in the number and distribution of spawning kokanee between 2002 to 2006
(Langston, 2012) to an apparent peak abundance in 2010 when kokanee were observed spawning
in the majority of tributaries to the Williston Reservoir (DWB Consulting Services Ltd., 2019).
Recent spawner enumerations suggested that kokanee populations have declined since 2010 (DWB
Consulting Services Ltd., 2019, 2020, 2021). In contrast, the pelagic survey conducted in 2021
suggests kokanee abundance in the reservoir may be double that of the 2008 survey (Weir et al.,
2023).
Kokanee behaviour and distribution in the water column is characterized by diel vertical
migrations that balance feeding opportunities and predation risks. In many lakes and reservoirs,
kokanee ascend to the food rich middle and upper strata of the water column to feed on
zooplankton during low light periods at dusk and dawn, remaining in shallow waters overnight. At
dawn, kokanee descend to depths of 60 metre or more where they remain during the day to avoid
predators (Clark & Levy, 1988; Hardiman et al., 2004; McPhail, 2007; Scheuerell & Schindler,
2003). All age groups of rearing kokanee (ages 0-4) are found almost exclusively in the pelagic
zones of the Williston Reservoir (Sebastian et al., 2009) and have an optimal temperature range of
6 to 13°C (McPhail, 2007). Despite these temperature preferences, kokanee in the Williston
Reservoir have been particularly vulnerable to capture in overnight gillnets set at the surface (02.4 m). Catch rates have declined steeply in nets set below 5 m, although deep gillnetting sets have
rarely been used in pelagic surveys of the Williston Reservoir (O’Connor, 2022; Sebastian et al.,
2003, 2009).
Peamouth
Peamouth travel in schools and juveniles are often abundant in shallow weedy zones of
lakes (McPhail, 2007). Little is known about peamouth temperature preferences but they appear
14

to be more tolerant of warmer temperatures than other pelagic species (McPhail, 2007; Roberge &
Slaney, 2001). Adults are known to utilize the pelagic habitats of the Williston Reservoir and have
rivaled lake whitefish and kokanee for abundance in fisheries surveys between 1988 and 2012
(Blackman, 1992; Pillipow & Langston, 2002; Plate et al., 2012; Sebastian et al., 2009). In the
summer months, peamouth exhibit a diel migration pattern between surface and inshore waters in
the evenings to deeper, offshore habitats in the morning (McPhail, 2007). Peamouth in the
Williston Reservoir are abundant near the bottom in near-shore habitats (Barrett & Halsey, 1975;
Blackman, 1992). Pelagic surveys have caught high numbers of peamouth near the surface in
gillnets (Pillipow & Langston, 2002), but trawling and gillnetting results suggest that they are not
abundant in the pelagic zone at depths greater than 5 m (Sebastian et al., 2003). Peamouth are a
member of the minnow family which has recently undergone a taxonomic reclassification from
Cyprinidae to Leuciscidae (Mandrak et al., 2023; Schönhuth et al., 2018; Tan & Armbruster,
2018).
Rainbow Trout
Rainbow trout populations in the Williston Reservoir have declined since its creation
(Blackman, 1992; Plate et al., 2012) and now only constitute a small portion of the pelagic fish
community (Sebastian et al., 2009). In large lakes and reservoirs, rainbow trout usually remain
within 50 m of shore and below the 18°C isotherm (McPhail, 2007). Sub-adult and adult rainbow
trout have been caught in the pelagic zone of the Williston Reservoir, but most frequently near the
surface (< 7.5 m) in transition zones between near-shore and off-shore habitat (Sebastian et al.,
2003, 2009).

15

Bull Trout
Bull trout in large, oligotrophic lakes may forage in the littoral zone during spring and fall,
but avoid water temperatures above 15 °C in summer by moving below the thermocline and into
the pelagic zone (McPhail & Baxter, 1996). Large bull trout in the Williston Reservoir prey heavily
on kokanee (Hagen & Weber, 2019) but constitute a small proportion (< 5 % in 2008) of the pelagic
fish community (Sebastian et al., 2009). Bull trout in the Williston Reservoir are part of the
Western Arctic population, which is listed as a species of Special Concern under the federal
Species at Risk Act (Species At Risk Public Registry, 2019). Bull trout are a provincially blue
listed species in B.C. (B.C. Conservation Data Center, 2021) due to their low abundance and high
vulnerability to a variety of threats. Although bull trout abundance in the Parsnip Reach appears
to have declined recently, populations in the Peace Reach appear to be stable and populations in
the Finlay Reach appear to have increased in abundance (Hagen & Weber, 2019). Bull trout have
been caught in gillnets set at all depths from the surface to 30 m, but have been caught most
abundantly in surface nets (< 7.5 m) (Sebastian et al., 2003).
Lake Trout
Lake trout are native to the Upper Peace Watershed and colonized the Williston Reservoir
naturally by dispersing from populations upstream (Hagen & Weber, 2019). Lake trout prefer cool
water and will thermoregulate by selecting depths with temperatures below 10 °C (Hardiman et
al., 2004; McPhail, 2007). Recent acoustic tagging studies have determined that lake trout make
extensive seasonal migrations throughout the Williston Reservoir ranging between 50 to 115 km
in length and remain at depths below the thermocline (between 20-30 m) in summer (Culling &
Euchner, 2018). Lake trout are top predators that prey heavily upon kokanee and lake whitefish
but have not been considered a major contributor to the reservoir’s pelagic fish community (< 1 %
16

in 2008) in the past (Blackman, 1992; Sebastian et al., 2003, 2009). Lake trout data have been
nearly or completely absent from most of the past pelagic surveys in the Williston Reservoir
because gillnets have rarely been set at depths below 20 m (Sebastian et al., 2003) and large
piscivores are often able to evade trawl nets (Bethke et al., 1999; Olin et al., 2009). Some
researchers have speculated that lake trout populations may have experienced a recent increase but
better monitoring is needed to understand how their populations are changing (Culling & Euchner,
2019; Hagen & Weber, 2019).
Knowledge Gaps in the Williston Watershed
The Williston Reservoir is an extraordinarily large and heavily perturbed lacustrine
ecosystem that has experienced dramatic physical and ecological changes since its creation.
Several studies have provided critical information on the ecology of the fish community in the
Williston Reservoir but substantial knowledge gaps still exist (Fish and Wildlife Compensation
Program, 2020). Monitoring of introduced kokanee revealed a rapid increase in their abundance
and spawning range throughout the reservoir to its apparent peak in 2010 (DWB Consulting
Services Ltd., 2019, 2020, 2021; Langston, 2012). The influence that fluctuations in kokanee
abundance have had on the fish community in the Williston Reservoir and their causes are not well
understood. Kokanee, lake whitefish and peamouth occupy a similar ecological niche in the
reservoir and changes in the distribution and abundance of one species will undoubtedly affect the
others. Kokanee are also an important forage species for lake trout and bull trout, and may have
competitive interactions with stream dwelling species.
Lake trout and bull trout are top predators in the reservoir with a high degree of niche
overlap, including their preference for other salmonids as a prey source (Culling & Euchner, 2019;
Donald & Alger, 1993; Hagen & Weber, 2019; Johnson & Martinez, 2000; McPhail & Baxter,
17

1996; Sebastian et al., 2009). Lake trout are the more dominant of the two species and can displace
native bull trout in lakes and reservoirs after a relatively short period of co-existence (Donald &
Alger, 1993; Fredenberg, 2002). Hagen and Weber (2019) completed a comprehensive
information synthesis on the conservation status of bull trout in the Williston Reservoir and
identified a potentially increasing lake trout population as a threat. A better understanding of lake
trout abundance and habitat use is a major information gap for this system. The structure and
complexity of the aquatic food web is an important factor influencing the outcomes of lake trout
and bull trout competition (Fredenberg, 2002).
Overall, there is limited information available on historical fish abundance and distribution
within the tributary streams of the Williston Reservoir (Fish and Wildlife Compensation Program,
2020). The inventories and life history studies conducted in the Williston Watershed have focused
largely on Arctic grayling (e.g., Blackman, 2002; Cowie & Blackman, 2003, 2004; Stamford et
al., 2020, 2022; Strohm et al., 2020), bull trout (e.g., Hagen et al., 2020; Hagen & Spendlow, 2019;
Hagen & Weber, 2019), and kokanee (e.g., Coxson et al., 2018; DWB Consulting Services Ltd.,
2019, 2020, 2021; Wilson & Shrimpton, 2020). However, large-scale inventories of entire fish
communities have rarely been completed, and the status and trends of many fish species are
unknown.
The Fish and Wildlife Compensation Program (FWCP) is a partnership between BC Hydro,
the province of British Columbia, Fisheries and Oceans Canada, Indigenous Nations and public
stakeholders with a mandate to conserve and enhance fish and wildlife in the upper Peace River
Basin Ecosystem (Fish and Wildlife Compensation Program, 2020). A few of the research
priorities for the FWCPs Peace Region Rivers, Lakes and Reservoirs Action Plan include:

18

·

to better understand the relationships between reservoir productivity, kokanee, lake
trout and other fish populations;

·

to develop a cost-effective method for monitoring lake trout populations to
understand the potential competitive effects on bull trout;

·

to conduct research on priority, but lesser-known, fish populations related to
conservation status, critical habitats, and key limiting factors; and

·

work with Indigenous Nations and stakeholder communities to characterize
culturally important fish species in priority freshwater habitat locations.

Conventional Survey Approaches for Pelagic Fish in the Williston Reservoir
Fish inventories of the Williston Reservoir since 2000 have assessed the distribution and
abundance of pelagic species with a combination of hydroacoustic surveys and gillnetting
(O’Connor, 2022; Pillipow & Langston, 2002; Plate et al., 2012; Sebastian et al., 2003, 2009).
Mid-water (5-35 m) trawl surveys were included in the 2000 and 2021 surveys, but not in any
other survey before or after. An accurate and complete picture of the fish community in the
Williston Reservoir has been difficult to produce due to the limitations of conventional survey
methods and an apparent focus on the most abundant species and the shallow water habitats they
frequent. Hydroacoustic surveys have had difficulty distinguishing between similarly sized species
that occur together in the water column (Sebastian et al., 2009). Lake whitefish, kokanee, and
peamouth in the Williston Reservoir have overlapping size ranges and are abundantly distributed
in surface and mid-water depths to 30 m during night surveys. Bull trout and lake trout are also
difficult to differentiate because of similarities in body sizes and distributions. Gillnets have
primarily been set at shallow depths to target abundant species such as lake whitefish, kokanee
and peamouth. Bull trout have been caught in surface gillnets but lake trout were not caught in a
19

gillnet in a pelagic survey between 1990 and 2012 (Blackman, 1992; Pillipow & Langston, 2002;
Plate et al., 2012; Sebastian et al., 2003, 2009). Mid-water (5-35 m) trawling in 2000 did not
capture any bull trout and only one lake trout (Sebastian et al., 2003). Bull trout and lake trout are
both assumed to be relatively rare in the Williston Reservoir, but pelagic surveys have poorly
defined their relative abundances despite the potential value of this information for managing these
high priority species. Fisheries inventories of the Williston Reservoir would benefit greatly from
survey methods with improved species resolution and detection sensitivity at a reasonable cost.
Conventional Survey Approaches in Streams and Rivers
A variety of survey techniques have been applied in the tributaries around the Williston
Reservoir including electrofishing (Coxson et al., 2018; Stamford & O’Connor, 2022), snorkel
surveys (Cowie & Blackman, 2003, 2004; Strohm et al., 2020), beach seines (Blackman, 2002;
Stamford & O’Connor, 2022), redd counts (Hagen & Spendlow, 2019), angling (Stamford &
O’Connor, 2022), and aerial enumeration by helicopters (DWB Consulting Services Ltd., 2019,
2020, 2021; Langston, 2012). Stream surveys have often targeted a few focal species for a specific
purpose, but inventories that include all species across a wide range of habitats have rarely been
completed and would be very difficult and costly to do with conventional survey methods.

20

CHAPTER 2:
DEVELOPMENT AND VALIDATION OF SPECIES-SPECIFIC eDNA ASSAYS FOR
PELAGIC FISH IN THE WILLISTON RESERVOIR

INTRODUCTION
The Williston Reservoir is an extraordinarily large and heavily perturbed lacustrine
ecosystem that has experienced dramatic physical and ecological changes since its creation in
1968. Recent surveys have found that kokanee, lake whitefish and peamouth, all planktivores, are
the most abundant fish species in the pelagic zone of the Williston Reservoir. Piscivorous species
are much less abundant in the pelagic zone, and include rainbow trout, bull trout and lake trout
(O’Connor, 2022; Plate et al., 2012; Sebastian et al., 2009). An accurate and complete picture of
the fish community in the Williston Reservoir has been difficult to produce due to the challenges
associated with studying such a large inland waterbody, the limitations of conventional survey
methods, and an apparent focus on the most abundant species. Fish inventories of the Williston
Reservoir since 2000 have assessed the distribution and abundance of pelagic species with a
combination of hydroacoustic surveys, gillnetting and trawls (Pillipow & Langston, 2002; Plate et
al., 2012; Sebastian et al., 2003, 2009). These types of conventional pelagic surveys are costly,
have occurred relatively infrequently, and poorly characterize the status of rare, difficult to capture
species such as bull trout and lake trout. The shortcomings of past approaches to fisheries
inventories have been recognized and addressed in recently updated research priorities for the
Williston Reservoir (Fish and Wildlife Compensation Program, 2020). Fisheries priorities include
a better understanding of relationships between kokanee and lake trout, maximizing the population
viability of bull trout through monitoring and developing a cost-effective method for monitoring
lake trout populations.

21

Organisms continuously shed DNA into their surrounding environments in the form of fur,
scales, skin, mucus, waste, and other tissues. The genetic material that an organism leaves behind
in its environment is referred to as environmental DNA or eDNA. The eDNA of aquatic animals
can be collected in a water sample, extracted, and analyzed to identify the species that occur in
those habitats without the need to observe or capture them (Thomsen et al., 2012; Thomsen &
Willerslev, 2015). A useful feature of eDNA in aquatic environments is its smoother and more
continuous distribution in space and time than the organisms themselves (Shelton et al., 2019). At
the same time, eDNA either settles out of the water column (Canals et al., 2021) or degrades over
a period of days to weeks (Rourke et al., 2021; Takahara et al., 2012). The rapid degradation of
eDNA in fresh water means that detections indicate the recent presence of an animal, and
potentially misleading signals from the past are generally avoided (Thomsen & Willerslev, 2015).
One of the first successful applications of eDNA was in the early detection of invasive species
(Jerde et al., 2011). Many studies have demonstrated that eDNA can detect rare or cryptic aquatic
species with higher sensitivity than conventional sampling gears, and often at a lower cost (Deiner
et al., 2017; Evans, Shirey, et al., 2017; Jerde et al., 2011; Jerde, 2019; Thomsen et al., 2012).
Early eDNA research focused on detecting species presence, and determining the distributions and
habitat preferences of a species (Thomsen & Willerslev, 2015). Over a decade ago, researchers
began to explore the correlation between aquatic species abundance and/or biomass and eDNA
concentrations (Takahara et al., 2012). Since then, a great deal of research has investigated the
correlation between fish abundance and/or biomass and eDNA concentrations to improve the
accuracy and precision of eDNA methods that may soon be a cost-effective alternative, or at least
a valuable supplement, to traditional inventory methods (Rourke et al., 2021).

22

There is tremendous potential for eDNA in the upper Peace River Watershed and molecular
methods for species detection have already been used successfully to monitor the distribution of
Arctic grayling in tributaries of the Williston Reservoir (Stamford et al., 2020, 2022; Strohm et al.,
2019). Detecting fish species using eDNA methods, rather than directly sampling them, can reduce
impacts on sensitive species and increases the power of field surveys for rare and elusive species
(Goldberg et al., 2016). eDNA methods involve the careful design and validation of molecular
markers, thoughtful sampling design, care in avoiding contamination, complex laboratory
procedures, and an interpretation of data that requires careful consideration of multiple factors.
There are several methods used to detect eDNA but the most accessible and simplest to use is a
method known as quantitative polymerase chain reaction (abbreviated as qPCR). Developing a
robust, species-specific, qPCR based assay for eDNA detection takes a great deal of time and
expertise to do properly (Langlois et al., 2021), and this may discourage the use of eDNA in
situations where it would otherwise be appealing. Fortunately, eDNA research has developed an
ever-expanding catalogue of qPCR assays, and groups like the Genomic Network for Fish
Identification, Stress and Health (GEN-FISH) are working towards the goal of developing and
validating eDNA assays for the detection all freshwater fish species in Canada (GEN-FISH, 2020).
However, the performance of published qPCR assays is dependent upon a variety of factors and
assays must be tested and optimized when being used by a different laboratory, with different
reaction conditions (Bustin & Huggett, 2017; Klymus et al., 2020), and when used in a different
geographic region than for which they were originally designed (Goldberg et al., 2016; Thalinger
et al., 2021). The results of an eDNA study can be interpreted with confidence after the entire
workflow used to detect a species’ DNA from an environmental sample has been validated
(Thalinger et al., 2021).

23

In this study, I sought to pilot targeted eDNA methods for surveying the pelagic fish
community in the Williston Reservoir. This included the validation of previously published qPCR
assays for kokanee, rainbow trout, lake whitefish and bull trout as well as the design and validation
of novel qPCR assays for lake trout and peamouth. Assays were validated in the lab and on
environmental samples that were collected from the Finlay Reach of the Williston Reservoir.
eDNA sampling in the Finlay Reach was conducted concurrently with a pelagic fisheries inventory
of the entire Williston Reservoir in the summer of 2021. The eDNA results were compared to the
gillnet catch results in the Finlay Reach.
The final validation state of each qPCR assay was evaluated based on the criteria and
principles described by Thalinger et al. (2021) to determine if eDNA assays are ready for use in
routine species monitoring. Five levels of the validation scale incorporate the entire workflow used
to detect a species’ DNA in an environmental sample, from sample collection through to the
interpretation of PCR results. Levels 1 and 2 include assay design, basic in silico analysis, in vitro
optimizations, testing on target species tissue and specificity testing on closely related, non-target
species. eDNA results from assays that have only been validated to level 2 cannot be interpreted
with confidence until further levels of validation have been achieved. Level 3 and 4 validation
criteria includes testing on environmental samples (in situ validations). More extensive specificity
testing that includes co-occurring non-target species and sensitivity testing that establishes
detection limits are considered important criteria for assay validation at level 4. Level 5 validation
includes statistical modelling of detection probability and investigations of ecological and physical
factors that may influence eDNA in the environment. Level 3 validation is considered essential
and the results from assays validated to level 4 and 5 can be interpreted with confidence (Thalinger
et al., 2021).

24

METHODS
Creating a Mitochondrial DNA Sequence Database
A diverse assemblage of fishes belonging to four orders and five families inhabit the
Williston Watershed. I created a reference sequence database of mitochondrial genes for 19 species
(Table 1. 1) that have been identified in the Williston Reservoir and its surrounding watershed
(Province of British Columbia, 2020; Sebastian et al., 2009). Available sequences from the
cytochrome oxidase subunit 1 (COI) gene, NADH dehydrogenase subunit 1 (NADH) gene, and
cytochrome b (CYTB) gene were downloaded from the National Center for Biotechnology
Information’s (NCBI) Nucleotide Database (https://www.ncbi.nlm.nih.gov/nucleotide/) and
imported to Geneious Prime version 2021.2.2 (www.geneious.com). Sequences were organized in
Geneious Prime by gene and by species and aligned using the default parameters in the MUSCLE
alignment algorithm. Intraspecific genetic variation was incorporated into a consensus sequence
for each species by coding polymorphic sites with International Union of Pure and Applied
Chemistry (IUPAC) codes. Published assays for lake trout, bull trout, rainbow trout, kokanee, lake
whitefish and peamouth were evaluated based on the alignments of consensus sequences.
In Silico Screening of Published Assays
Published qPCR assays were available for lake trout, bull trout, kokanee, rainbow trout,
and lake whitefish but not for peamouth. Published lake trout assays were only available for the
COI gene (Hernandez et al., 2020; Lacoursière‐Roussel, Rosabal, et al., 2016) and were a nearly
complete match with bull trout DNA. Candidate assays were screened for potential crossamplification of non-target DNA from species in the Williston Watershed. Primer and probe
sequences for each assay were compared against the consensus sequence alignments of the 19 fish
species from Table 1. 1 using Geneious Prime. The minimum criteria for in silico specificity were

25

two mismatches with all non-target species in at least two of the three component sequences (the
two primers, or a primer and a probe) and at least one mismatch in the other (Currier et al., 2018).
Published assays for bull trout, kokanee, rainbow trout and lake whitefish were selected for in vitro
validation based on these criteria (Table 2. 1).
Table 2. 1. Published qPCR assays for bull trout, kokanee and rainbow trout selected for validation and
novel assays for lake trout and peamouth that were developed in this study for use in the Williston
Watershed.
Species and
Assay Name
Kokanee

Target
Gene
COI

Primers/
Probe
Forward

Sequence 5’ to 3’
GGAAACCTTGCCCACGCG

COI

Reverse
Probe
Forward

AAAAGTGGGGTCTGGTACTGAG
CTCTGTTGACTTAACCATC
ACCATTATTAACATAAAACCTCCAG

121

Rasmussen
Hellberg et al.
(2010)

CYTB

Reverse
Probe
Forward

GTAATGCCTGCTGCCAGGA
CGTTTGAGCCGTGCTA
CAAACCTCCTTTCTGCCGTG

198

Hernandez et al.
(2020)

CYTB

Reverse
Probe
Forward

AGTTGATCCCTGCTGGGTTG
TTGTGCAGTGAATCTGA
AGTACTTCACCTTCTGTTTCTGCATG

134

Wilcox et al.
(2013)

Reverse
Probe
Forward
Reverse
Probe
Forward
Reverse
Probe

CAATATAGCTACGAAACCGAGGAGG
CCGACAAAATCTCA
GCCTTCCACTTCCTATTCCCA
GCTAGGGTTGTTAGGCCAAGTAG
ACGCCGATAAAATCTCGTTCCACCC
GCACTCACATCACTGACATTG
AGGACCCCTCCTAATTTGTTC
TACTCCCCCACACATCCAACCAGAG

191

This study

165

This study

Sock_COI
Rainbow trout

Rain_COI
Lake whitefish
COCL_CYTB
Bull trout
BUT1_CYTB
Lake trout
LT_PCC

CYTB

Peamouth
PCC_CYTB

CYTB

Amplicon
length (bp)
152

Citation
Rasmussen
Hellberg et al.
(2010)

Novel Assay Design
I designed novel qPCR assays to detect lake trout and peamouth (Table 2. 1). Both assays
were developed for the CYTB gene due to its divergence among species in the Williston

26

Watershed. The CYTB gene was the only mitochondrial gene examined with enough sequence
divergence for a targeted assay that could discriminate between lake trout and bull trout. Integrated
DNA Technologies’ (IDT) PrimerQuest Tool (https://www.idtdna.com/) was used as a starting
point for designing qPCR primers and probes. The primer and probe sets generated by the
PrimerQuest Tool were transferred to Geneious Prime for assessment against target and non-target
species sequences. The IDT generated assays identified promising regions for primers and probes
and were modified to optimize specificity.
Assay specificity was achieved primarily by maximizing the number and influence of basepair mismatches between target and non-target species (Currier et al., 2018; Klymus et al., 2020;
So et al., 2020). The number of mismatches between primer and probe sequences is generally
considered proportional to an assay’s specificity, but the effect of a mismatch can vary depending
on its type and position (Klymus et al., 2020; So et al., 2020; Wilcox et al., 2013). Mismatches
were prioritized in the primer sequences because they influence assay specificity more than
mismatches in the probe (Wilcox et al., 2013). Mismatches were positioned as close as possible to
the 3’ end of primers and as close as possible to the 5’ end of probes to maximize their effect
(Klymus et al., 2020; Langlois et al., 2021; So et al., 2020; Wilcox et al., 2013). Purine-purine
mismatch pairings (i.e., A-A, A-G, or G-A) and C-C mismatch pairings, referred to as “critical”
mismatches by So et al. (2020), were incorporated into primers and probes where possible because
they reduce amplification efficiency of non-target DNA more than other mismatch types.
Optimizing the influence of mismatches was particularly important for the lake trout assay because
it was not possible to incorporate more than two base pair mismatches with bull trout into any of
the primers or probes (Table 2. 2).

27

Multiple candidate primer and probe sequences for each species were assessed individually
and in combination for characteristics and interactions that could affect assay sensitivity, PCR
conditions and efficiency, and that may have had some additional impact on specificity (Table 2.
4 and Table A3). Primer and probe melting temperatures (Tm) were analyzed using IDT’s

OligoAnalyzer Tool (https://www.idtdna.com/) with the default parameter settings for qPCR
(Klymus et al., 2020; Prediger, 2013). Forward and reverse primers were designed to have a Tm
between 60-64 °C and to be within 2 °C of one another. Probe melting temperatures were designed
to have a Tm 4-8 °C higher than the primers. The GC content of primers and probes were kept
between 35-65%, and regions with 4 or more consecutive Gs were avoided. The number of Gs and
Cs in the last 5 bases of the 3’ end of primers, known as a GC clamp (Thornton & Basu, 2011),
were noted for their potential influence on primer binding and assay specificity, but was considered
less important than incorporating mismatches. G bases were avoided at the 5’ end of probes
because of their potential to dampen the signal from green and yellow dyes (Klymus et al., 2020;
Prediger, 2013). Probes were designed to be compatible with IDT’s double quenched probes
(3IABkFQ and ZEN) and FAM or HEX fluorophores (https://www.idtdna.com/). Amplicon
lengths were kept between 100 and 250 base pairs to maintain high PCR efficiency while also
allowing for verification by Sanger sequencing (Hernandez et al., 2020; Klymus et al., 2020).
Table 2. 2. Primer and probe mismatches of the LT_CYTB assay with closely related, non-target species
from the Williston Watershed. Mismatches between the primer and probe and non-target species
sequences are underlined and coloured. Critical mismatches (So et al., 2020) are coloured red and other
mismatches are coloured blue.
Assay/Species
LT_CYTB
Bull trout
Lake whitefish
Kokanee

F-primer Mismatches (5’-3’)
GCCTTCCACTTCCTATTCCCA
GCCTTTCACTTCCTATTCCCT
GCCTTTCACTTCTTATTCCCC
GCCTTTCACTTCCTGTTCCCT

R-primer Mismatches (5’-3’)
GCTAGGGTTGTTAGGCCAAGTAG
GCTAGGGTTGTTAGGCCAAGCAA
GCCAAGGATGTTAGTCCTAGTAG
GCTAAGGATGTTAGACCAAGAAG

28

Probe Mismatches (5’-3’)
ACGCCGATAAAATCTCGTTCCACCC
ACGCCGACAAAATCTCATTCCACCC
ATGCCGATAAAATCTCATTCCACCC
ATGCCGATAAAATCTCGTTCCACCC

Table 2. 3. Primer and probe mismatches of the PCC_CYTB assays with closely related, non-target
species from the Williston Watershed. Mismatches between the primer and probe and non-target species
sequences are underlined and coloured. Critical mismatches (So et al., 2020) are coloured red and other
mismatches are coloured blue. Mismatches with polymorphic sites (for longnose dace) were also noted,
although they may not always be a mismatch.
Assay/Species
PCC_CYTB
Redside shiner
N. pikeminnow
Lake chub
Longnose dace

F-primer Mismatches (5’-3’)
GCACTCACATCACTGACATTG
GCTCTTACATCACTAACTCTA
GCCCTTACATCATTAGCACTA
GCTCTCACATCATTAGCCCTA
RCTCTCRCATCCCTYACRYTA

R-primer Mismatches (5’-3’)
AGGACCCCTCCTAATTTGTTC
AGGACTCCCCCCAGTTTGTTT
AGGACCCCTCCTAGTTTATTT
AGAACCCCTCCCAGCTTGTTA
AGKACYCCTCCYARYTTGTTW

Probe Mismatches (5’-3’)
TACTCCCCCACACATCCAACCAGAG
TACCCCACCACATATCCAACCAGAG
CACTCCCCCACATATCCAGCCTGAA
CACCCCGCCACACATCCAACCGGAG
TACCCCRCCRCACATCCARCCRGAR

The potential for secondary structure formation between primers and probe was assessed
with the OligoAnalyzer Tool (Integrated DNA Technologies, 2021). Hairpin structures were
ignored if a predicted Tm was lower than the anticipated PCR annealing temperatures. Primers and
probes that produced hairpin structures with a Tm at or above the anticipated annealing temperature
were rejected or redesigned. All possible combinations of dimer structures between primers and
probes were analyzed and evaluated with respect to the difference in free energy, or ΔG. Dimer
structures with a predicted ΔG of -9 kcal/mol or greater were considered problematic. Primers and
probes that had the potential to create strong dimer structures were redesigned if possible. In
situations where primer or probe design was constrained by other criteria, the potential for
secondary structures was weighted against other priorities, such as assay specificity. Multiple
candidate assays were designed and tested for each species, but the body of this study only
describes the successful assays that were validated on environmental DNA. The unsuccessful assay
candidates are described in Appendix B.

29

Table 2. 4. Characteristics and secondary structure analysis of the LT_CYTB and PCC_CYTB assays as
calculated with IDT’s OligoAnalyzer Tool.

Assay
LT_
CYTB
PCC_
CYTB

Primer/
Probe
Forward
Reverse
Probe
Forward
Reverse
Probe

Sequence (5’-3’)
GCCTTCCACTTCCTATTCCCA
GCTAGGGTTGTTAGGCCAAGTAG
ACGCCGATAAAATCTCGTTCCACCC
GCACTCACATCACTGACATTG
AGGACCCCTCCTAATTTGTTC
TACTCCCCCACACATCCAACCAGAG

Tm
(°C)
63.7
64.2
68.2
61.4
61.5
68.7

Hairpin
Tm (°C)
None
52.8
36.4
33.8
50.6
25.2

Dimer Structures ΔG (kcal/mol)
Self
F/R
F/
R/
Primers
Probe
Probe
-3.14 -7.81
-3.42
-7.48
-9.28
-5.19
-3.53 -3.42
-3.9
-4.64
-6.24
-3.17

Final screening of assays included testing through NCBI’s Primer-Blast program (Ye et
al., 2012) to identify potential non-targets in the NCBI nt/nr database that might amplify with the
assay. Any relevant non-target species that were identified as being a close match were included
in in vitro specificity testing.
In Vitro Assay Validation
Assay Optimization
The published Sock_COI, Rain_COI, COCL_CYTB, and BUT1_CYTB primer/probe sets
were purchased from Applied Biosystems (https://www.thermofisher.com) in the TaqMan format
with Minor Groove Binding (MGB) probe chemistry and VIC fluorophores. The novel primers
and probes for lake trout (LT_CYTB) and peamouth (PCC_CYTB) were purchased from IDT in the
PrimeTime® qPCR Assay format with double quenched probes (3IABkFQ and ZEN) and FAM
fluorophores (Integrated DNA Technologies Inc., Coralville, IA). Assays were ordered with
default primer and probe concentrations and were added to all PCR reactions at a final 1x
concentration. The 1x concentrations for TaqMan assays were 900 nM of primers and 250 nM of
probe. The 1x concentrations for IDT assays were 500 nM primers and 250 nM probe.

30

Optimal PCR annealing temperatures were determined for each qPCR assay by running a
temperature gradient test on the Bio-Rad QX200 Droplet Digital PCR system (Life Science,
Mississauga, ON) at eight annealing temperatures between 50 and 60 °C. Assays were tested
against synthetic DNA templates (Integrated DNA Technologies Inc.)(Appendix A).
Assay Specificity
All six of the assays in Table 2. 1 were tested for amplification of their target species and
for cross-amplification of closely related and representative species from the Williston Watershed
(Table 2. 5). Genomic DNA used for specificity testing was extracted from the fin tissues of 15
fish species with a DNeasy Blood and Tissue kit (Qiagen, Mississauga, ON) following the
manufacturer’s protocol. Assays were tested against extracted DNA from at least two individuals
from non-target species in the same family. Some assays were tested against three to six samples
from non-target species with a higher risk of cross-amplification due to factors such as a low
number of primer mismatches (e.g., LT_CYTB assay and bull trout) (Table 2. 5). Assays were
tested on the QX200 Droplet Digital PCR system in 20 µL final reaction volumes that included 10
µL of 2x ddPCR Supermix for Probes (no dUTPs) (Life Science, Mississauga, ON), 1 µL of 20x
concentrated assay, 5 µL of extracted DNA (2-6 ng) and 4 µL of nuclease free water (IDT). The
thermocycling protocols for all assays were: initial denaturation at 95 °C for 10 min, 45 cycles of
denaturation at 95 °C for 30 seconds and annealing at 58 °C for 60 seconds, followed by a final
incubation at 98 °C for 10 min.

31

Table 2. 5. List of species whose extracted DNA was included in the specificity testing of each qPCR
assay. Numbers represent the number of samples per species that each assay was tested against.
Species
Slimy sculpin
Lake trout
Bull trout
Rainbow trout
Kokanee
Lake whitefish
Mountain whitefish
Arctic grayling
White sucker
Peamouth chub
Northern pikeminnow
Redside shiner
Lake chub
Longnose dace
Burbot

Sock_COI

Rain_COI

2
2
7
6
4
4
2

2
2
7
6
4
4
2

1
1

1
1

Assay Name
COCL_CYTB BUT1_CYTB
1
1
4
6
3
6
2
4
2
4
6
5
2
3
1
2
1
2
1
2
1
2

1

1

LT_CYTB
1
6
6
4
4
5
3
2
2
2
2

1

PCC_CYTB
1
1
1
1
1
1
1
1
1
3
3
3
3
3
1

Assay Sensitivity
Sensitivity was measured on the Bio-Rad QX200 Droplet Digital PCR system (Life
Science, Mississauga, ON) with dilution series experiments. A synthetic DNA template (gBlock,
Integrated DNA Technologies Inc.) of each target amplicon was used to create 1:5 standard serial
dilutions at eight dilution points with six molecular replicates of each standard. Synthetic templates
were based on consensus sequences generated in Geneious Prime from the NCBI sequence
database. Use of synthetic templates enabled an accurate estimation of template concentrations.
The gBlock sequences can also incorporate multiple species’ templates into the same synthetic
DNA molecule. A gBlock with multiple templates can be used in duplex reactions with two assays
and the equimolar ratio of the templates is useful for evaluating the performance of both assays
relative to one another. Targeted concentrations for each standard dilution point were between
10,000 and 0.13 copies/20 µL ddPCR reaction. However, the end-point ddPCR measurements of
synthetic template concentrations varied for each batch of gBlock used. The standard

32

concentrations for each dilution series (Table A1) were calculated from the mean concentration of
gBlock detected by ddPCR at the most concentrated dilution point and dividing by five for each
subsequent standard. The limit of detection (LOD) was defined and assessed in two ways: standard
and modeled LOD. Standard LOD was defined as the lowest concentration of target analyte that
was detected with 95% confidence (Klymus et al., 2019). Detections were required in all six
replicates of a dilution level with ddPCR to achieve 95% confidence level or better for this study.
Modeled LOD was assessed with a statistical model for digital PCR developed by Hunter et al.
(2017). The modeled LOD was defined as the lowest amount of analyte that could be both detected
and distinguished from a source of overestimation bias in qPCR known as the concentration
plateau; as DNA concentrations approach zero, measured concentrations may be higher than
expected and result in a nonlinear standard dilution series (Hunter et al., 2017). The modeled LOD
was determined by analyzing the dilution series data with R code provided by Hunter et al. (2017).
In Situ Assay Testing in the Williston Reservoir
eDNA Field Sampling
I sampled eDNA in the Finlay Reach of the Williston Reservoir near the Teare Creek and
Factor Ross sampling stations established by previous pelagic fish surveys (Blackman, 1992;
Pillipow & Langston, 2002; Sebastian et al., 2003, 2009). Teare Creek samples were collected at
Latitude: 56.40, Longitude: -124.47 and Factor Ross samples were collected at Latitude: 56.59,
Longitude: -124.65 (Figure 1. 1). The Teare Creek station was sampled on August 16, 2021 and
the Factor Ross station was sampled on August 17, 2021. Sampling dates were chosen to coincide
with summer stratification of the water column and to avoid periods when mature rainbow trout
or bull trout and kokanee migrate out of the pelagic zone of the reservoir to spawn. Temperature
and dissolved oxygen profiles were measured at Teare Creek from the surface to a depth of 50 m

33

to characterize and confirm lake stratification in the Finlay Reach at the time of sampling. The
thermocline extended from 10 m depth down to 25 m (Figure 2. 1). Temperatures above 10 m
ranged from 17.1 °C to 17.9 °C. Temperatures between 25 and 50 m depths ranged from 6.8 to 5.3
°C.
Water samples were taken at five depths dispersed between the surface and a maximum
depth of 30 m at the Teare Creek and Factor Ross locations. Three 1-L replicate water samples
were taken at each depth. Water was sampled at the surface by submerging sterilized plastic sample
bottles until filled. Water was sampled at 5, 10, 20 and 30 m with a Van Dorn water sampler
(Wildco, Yulee, FL) and then transferred into sterilized sample bottles. Each of the three 1-L
sample replicates from a given depth were collected from separate deployments of the Van Dorn
to that depth. Samples collected at the surface (0 m) and at depths of 5 and 10 m were taken from
above the thermocline. Samples collected at 20 m were near the bottom of the thermocline where
temperatures were approximately 10 °C. Samples collected from 30 m were below the thermocline
where temperatures were between 5 and 6 °C (Figure 2. 1).
To prevent contamination, the Van Dorn sampler was cleaned before use at each site by
submerging it in a bucket of 30% bleach, followed by submersion in two separate buckets of sterile
fresh water to rinse away the bleach. Two 1-L bottles of distilled water were taken to and from the
field as non-template control samples in the same cooler where water samples were stored. Water
samples were stored in a cooler on the boat or in a fridge at 4 °C until they were filtered.
Water samples were filtered onto 47 mm, 1.5 µm pore glass microfiber filters (Cytiva,
Marlborough, MA). Filters were placed in a Nalgene™ Single Use Analytical Filter Funnel
(ThermoFisher Scientific) that was attached to a sidearm Erlenmeyer flask. Negative pressure to
draw water through the filter quickly was created with a peristaltic pump (Geotech Environmental
34

Equipment Inc., Denver, CO) attached to the side arm of the flask. Teare Creek samples were
filtered 24 hours after collection and Factor Ross samples were filtered within 5 hours of collection.
Teare Creek sample filtration was delayed due to a windstorm that forced the field crew to seek
shelter on the reservoir. Samples were kept in a cooler during the day, transferred to a fridge
overnight and filtered the next morning. Filters were stored in individual coin envelopes and all
three replicates from each depth were preserved together in plastic specimen bags filled with 90 g
of self-indicating silica beads (Fisher Scientific, Ottawa, ON). All negative controls of distilled
water were filtered, stored, and preserved in the same way as field samples from each location.
Preserved filters were stored at room temperature for 72 hours and then transferred to a -80 °C
freezer until DNA was extracted.

35

Figure 2. 1. Dissolved oxygen (DO) and temperature profile measured at the Teare Creek sample location
on 16 August 2021. eDNA sample depths are indicated by arrows.

eDNA Extraction
DNA was extracted in a room dedicated to low quality DNA sources and a negative control
was included with each batch of extractions to monitor for contamination. DNA was extracted
from filters using a DNeasy Blood and Tissue kit. Filters were rolled and cut into 1 mm slices with
sterile tweezers and scissors and then placed in a sterile 2 mL centrifuge tube with two 5/32-inch
steel grinding beads. The tubes were shaken in a Genogrinder 2000 (SPEX SamplePrep, Metuchen,

36

NJ) at 1500 oscillations/minutes for 90 s. Tubes were centrifuged at 6000 x g for 1 minute before
adding 870 µL of Buffer ATL and 30 µL of Proteinase K. Tubes were incubated at 56 °C with 150
rpm shaking for 30 min, followed by a second shake in the Genogrinder at 1000 oscillations/minute
for 60 s. Tubes were incubated at 56 °C and 150 rpm shaking for another 90 min. After incubation,
tubes were vortexed for 15 seconds and centrifuged for 3 minutes at 10 000 x g. Supernatant was
transferred to a sterile 1.5 mL centrifuge tube. Buffer AL and 95% ethanol were added to the 1.5mL centrifuge tube in volumes equal to the volume of the supernatant. The remainder of the
extraction procedure followed the DNeasy Blood and Tissue Kit’s spin column procedure with the
modification of using 2 x 50 µL AE buffer for the final elution step. Eluted DNA samples were
stored at -20 °C until used for polymerase chain reaction (PCR) analysis.
ddPCR Analysis of eDNA
Extracted eDNA samples were analyzed on a Bio-Rad QX200 Droplet Digital PCR system
(Life Science, Mississauga, ON) according to the manufacturer’s instructions (Bio-Rad
Labratories, Inc., 2022). The 20 µL ddPCR reactions contained 1x ddPCR Supermix for Probes
(Life Science, Mississauga, ON), 1 µL of each assay (primers and probe) at 20x concentration, and
5 µL of eDNA sample. Each reaction was partitioned into nanoliter sized water-in-oil microdroplets by a Droplet Generator (Bio-Rad) and the prepared droplets were transferred into a 96well PCR plate. The PCR plate was subsequently heat-sealed with pierceable foil using a PX1™
PCR plate sealer (Bio-Rad) and then amplified in a C1000 Touch™ deep-well thermal cycler (BioRad). The thermocycling protocol was: initial denaturation at 95 °C for 10 minutes, then 45 cycles
of denaturation at 95 °C for 30 s, annealing at 58 °C for 60 s, followed by a final incubation at 98
°C for 10 min. After cycling, the 96-well plate was placed into a Droplet Reader (Bio-Rad).
Droplets of each sample were analyzed sequentially, and fluorescent signals of each droplet were

37

measured individually by a detector. The first elutions (i-elutions) of all eDNA samples were
amplified in two PCR replicates with the Sock_COI, Rain_COI, COCL_CYTB, BUT1_CYTB and
LT_CYTB assays. The PCC_CYTB assay only assessed a single PCR replicate of the second elution
(ii-elution) of eDNA samples, due to the limited volume of sample available. Filtration and
extraction negative controls were amplified with all assays and negative PCR controls were
included on each plate by substituting nuclease free water (IDT) for DNA.
The potential for samples to produce false negative results was evaluated by testing for
DNA integrity and inhibition. DNA integrity was evaluated by testing for plant DNA with the
ePlant assay (i.e., IntegritE-DNA™; Veldhoen et al., 2016). Inhibition was evaluated by detecting
a known quantity of Lambda DNA that was added to the ddPCR mastermix as a positive control
with a Lambda assay (Murray, 2021). One replicate from each sample depth at the Teare Creek
and Factor Ross locations were tested for integrity and inhibition in a single 20 µL final reaction
volume. Each reaction included 10 µL of 2x ddPCR Supermix for Probes (no dUTPs) (Life
Science, ON, Canada), 0.5 µL of 40x concentrated lambda assay, 0.5 µL of 40x concentrated
ePlant assay (Veldhoen et al., 2016), 5 µL of the first elution of each eDNA sample, and 4 µL of
nuclease free water spiked with lambda DNA at a concentration of 2 copies per µL. Integrity and
inhibition tests used a similar thermocycling protocol as for other environmental samples but with
an annealing temperature of 56 °C.
The minimum criterion for a positive species detection in eDNA samples was set at 3
positive droplets between all PCR replicates of a sample, following the recommendations from
Bio Rad’s Droplet Digital PCR Applications Guide (Bio Rad). Positive droplets were pooled from
the two PCR replicates of each sample tested with every assay except PCC_CYTB. Species

38

detections were evaluated independently for each of the three eDNA sample replicates collected
from each depth in the water column.
Gillnet Sampling
Pelagic gillnets were used to validate the eDNA results. Floating and sinking monofilament
“experimental” gillnets were set at the Teare Creek Station between August 14 and 15, 2021
(O’Connor, 2022). Gillnetting was also planned for the Factor Ross Station, but many days of high
winds prevented the survey team from getting out on the reservoir. Gillnets were set at other
locations around the reservoir and the one set at Finlay Forks on August 4, 2021 was the closest to
Teare Creek and Factor Ross (Figure 1. 1). Each net was comprised of six panels, 2.4 m deep and
15.3 m long connected end-to-end for a total length of 94.1 m. Stretch mesh sizes of the six panels
were 25 mm, 76 mm, 51 mm, 89 mm, 38 mm, and 64 mm. At Teare Creek, one floating net was
set at the surface and five sinking nets were set at 5-metre depth increments from the surface to a
maximum depth of 25 m. At Finlay Forks, nets were only set at 5-metre increments between 5 and
20 m depth. Nets were deployed parallel to the shoreline and in locations where the depth of the
reservoir was at least 70 m to avoid littoral species (O’Connor, 2022). High wind and waves on
August 15 prevented the recovery of the Teare Creek gillnets until that evening, resulting in a 24hour soak time. The high wind and waves during the day also released the anchor buoy that held
the deep end of the chain of nets. This caused the net set at 25 m to sink and become entangled in
sunken debris, ultimately preventing its recovery. All fish captured in the nets were identified to
species, weighed, and measured for fork length.

39

RESULTS
In Silico Screening and Novel Assay Design
The published assays for kokanee, rainbow trout, lake whitefish and bull trout in Table 2.
1 were all successfully validated in silico for their application in the Williston Reservoir. The

corresponding qPCR assays for kokanee and rainbow trout were designed to be specific enough to
differentiate between one another as well as other sympatric salmonids. The Sock_COI and
Rain_COI assays both had a total of seven mismatches with their congener and at least eight with
all the other salmonids in the Williston Watershed. Primer mismatches were often closer to the 3’
end and the Sock_COI assay and had a single critical mismatch with rainbow trout and lake
whitefish near the 5’ end of the forward primer. The COCL_CYTB lake whitefish assay had the
fewest total mismatches with lake trout (seven) and bull trout (eight). Both COCL_CYTB primers
had mismatches at the 3’ ends with lake trout and bull trout, but no critical mismatches. The
BUT1_CYTB bull trout assay had the fewest mismatches with lake trout, the other species within
the genus Salvelinus: one mismatch at the fifth position from the 3’ end of the forward primer,
three mismatches near the middle and 5’ end of the reverse primer, and two mismatches in the
probe. Interestingly, the forward primer of the BUT1_CYTB assay included two sites of
intraspecific polymorphism for bull trout. Individual sequences downloaded from NCBI for bull
trout indicated that the polymorphic sites were both either G or A and both haplotypes appeared
to be common. The BUT1_CYTB assay was designed with G bases at both polymorphic sites in
the forward primer, one of which was the only mismatch with lake trout. The BUT1_CYTB assay
narrowly met the minimum in silico mismatch criteria for inclusion in this study and had two
mismatches with some bull trout haplotypes.

40

Multiple primer and probe combinations were assessed and tested for lake trout and
peamouth (Table 2. 4 and Table A2), and the most successful assay candidates for both species
were chosen for further validation (Table 2. 4). Specificity was the most important design aspect
for the lake trout assay, and mismatches between lake trout and bull trout were prioritized. No
more than two mismatches with bull trout could be incorporated into each primer and probe
sequence (Table 2. 2). Mismatch influence was optimized by including a critical A-G mismatch in
the forward primer and by positioning all mismatches as close to the 3’ end of primers as possible.
The risk of cross-amplification with the most abundant pelagic salmonids in the Williston
Reservoir was also low due to the number of total mismatches with the LT_CYTB primers and their
relative influence due to mismatch position and type (Table 2. 2). Prioritizing assay specificity
constrained the options for primer locations and resulted in trade-offs with the potential for
secondary structures in the reverse primer (Table 2. 4). The hairpin Tm of the LT_CYTB reverse
primer was relatively high, and dimer structures that involved the reverse primer had a moderately
high ΔG values. Secondary structures were formed primarily with the GGCC sequence located in
the middle of the reverse primer. Options were limited for redesigning the reverse primer while
still incorporating key mismatches, and often resulted in higher ΔG values for potential secondary
structures, or substantially reduced assay specificity. The LT_CYTB primer Tm was designed to be
high enough to avoid optimal PCR annealing temperatures that would overlap with the high hairpin
Tm of the reverse primer.
Peamouth and other leuciscid species in the Williston Watershed have highly divergent
CYTB sequences which made designing the PCC_CYTB assay relatively uncomplicated. Hairpin
Tms and ΔG values for dimer structures were low (Table 2. 4). Primer and probe mismatches with
the three other leuciscid species in the Williston Watershed were numerous, and their positions

41

were biased towards the 3’ end of the primers (Table 2. 3). As many as two A-G critical mismatches
were incorporated into the forward primer, and near the 3’ end with two of the three other leuciscid
species.
In Vitro Validation
Annealing Temperature Optimization
Temperature gradient tests revealed the optimal annealing temperatures for each assay
(Table 2. 6), although most assays performed well at all temperatures above 52°C (Appendix A).
The BUT1_CYTB bull trout assay was particularly robust, performing similarly at all temperatures.
The kokanee assay performed best at 54 °C, but droplet amplitude was only slightly reduced at 56
°C and 58 °C. The lake trout assay performed well at or above 58 °C. Below 58 °C, the average
amplitude of positive droplets was reduced, and the range of amplitudes for positive droplets
increased (Figure A 9). An annealing temperature of 58 °C was suitable for all the assays tested
and considered to be appropriate for duplexing assays in a single reaction, or when using more
than one assay on the same PCR plate.
Table 2. 6. Optimal annealing temperatures determined from temperature gradient tests with each assay
and a synthetic DNA template (gBlock).
Assay
Sock_COI
Rain_COI
COCL_CYTB
BUT1_CYTB
LT_CYTB
PCC_CYTB

Target Species
kokanee
rainbow trout
lake whitefish
bull trout
lake trout
peamouth

Optimal Annealing
Temperature (°C)
54
56-58
56-58
50-60
60
54-56

Assay Specificity
All six qPCR assays in this study demonstrated species specificity when tested against
DNA extracted from the tissues of non-target species. The baseline amplitude for lake trout

42

template was slightly higher than for other species when tested with the COCL_CYTB and
BUT1_CYTB assays (Figure A 13 and A 14). However, the difference in droplet amplitude was
negligible and droplets were unambiguously identified as negative. The COCL_CYTB and
BUT1_CYTB assays had relatively few primer and probe mismatches with lake trout DNA, but
both assays were specific enough to discriminate between lake trout and their target species. The
novel lake trout assay did not show any evidence of cross-reaction with bull trout DNA despite
only having two mismatches in each primer and the probe (Figure A 15).
Assay Sensitivity
The six assays tested had Standard LODs ranging from concentrations of 0.86 to 9.12
copies/µL in the 5 µL of sample that was added to the ddPCR reaction, translating to a range of
4.32 to 45.6 total copies in a 20 µL ddPCR reaction (Table 2. 7). Modeled LODs did not differ
from Standard LODs for the Rain_COI, COCL_CYTB, and LT_CYTB assays. Modeled LODs for
the Sock_COI and BUT1_CYTB assays could not be calculated because their dilution series
remained linear to the lowest dilution level. The Standard and Modeled LODs for the PCC_CYTB
assay differed substantially. Standard LOD for the PCC_CYTB assay was relatively high because
the assay only detected template in 5 out of 6 of the replicates at the next dilution point that
represented 1.82 copies/µL in the original sample, or 9.12 copies/20 µL ddPCR reaction.

43

Table 2. 7. Limits of detection (LOD) for the six assays tested in this study. Concentrations of the
synthetic DNA standards were adjusted after comparison with ddPCR results from the dilution of highest
concentration in the series. Standard LOD is the lowest dilution level with >95% detection amongst
replicates. Modeled LOD was assessed with a statistical model for digital PCR developed by Hunter et al.
(2017). Modeled LOD was not applicable (N/A) when the relationship between expected and measured
DNA concentrations remained linear and a “concentration plateau” was not observed at low
concentrations. For both Standard and Modeled LOD, the DNA concentration of the sample and the
concentration in the 20 µL PCR reaction mixture are shown.
Assay
Sock_COI
Rain_COI
COCL_CYTB
BUT1_CYTB
LT_CYTB
PCC_CYTB

Standard LOD
(copies/µL of sample)
1.22
1.28
0.86
0.99
0.99
9.12

Standard LOD
(copies/20 µL rxn)
6.08
6.40
4.32
4.96
4.96
45.6

Modeled LOD
(copies/µL of sample)
N/A
1.28
0.86
N/A
0.99
0.36

Modeled LOD
(copies/20 µL rxn)
N/A
6.40
4.32
N/A
4.96
1.82

In Situ Assay Testing in the Williston Reservoir
ddPCR Analysis of eDNA
All the samples collected from the Williston Reservoir had low concentrations of eDNA
and the detected concentrations varied between species. Most species detections came from eDNA
samples collected above the thermocline. The number of samples with positive detections, eDNA
concentrations, and the diversity of detected species all declined with increasing depth. However,
these observations were more pronounced at Factor Ross than at Teare Creek. Tests with lambda
and ePlant assays indicated that Factor Ross samples from 20 m may have been inhibited (Figure
2. 2). The lambda assay detected lambda DNA that was included in each reaction as a positive
control at a concentration of 2 copies per µL. The ePlant assay detected chloroplast DNA present
in water. Inhibition was indicated in the Factor Ross samples from 20 m by the low amplitude of
positive droplets for lambda and the absence of positive droplets for ePlant (Figure 2. 2). Despite
this evidence for inhibition, kokanee were detected in two out of three replicate samples from
Factor Ross at 20 m and at concentrations as high as 15 copies per reaction.

44

Figure 2. 2. Tests for sample inhibition and integrity with the lambda (top) and ePlant (bottom) assays.
Droplet charts are from tests of first elution samples from Teare Creek (TC) and Factor Ross (FR) at
sample depths that are indicated by numbers in the sample codes. The master mix for all samples was
spiked with lambda DNA at a concentration of 2 copies per µL. The low amplitude of positive droplets
for lambda and the absence of positive droplets for ePlant at FR20 is indicative of inhibition. Droplets
that were positive for PCR amplification are coloured and negative droplets are grey.

The total number of eDNA copies detected from all samples at Factor Ross and at Teare
Creek demonstrated similar trends of eDNA abundance for most species. Kokanee eDNA was by
far the most abundant at Factor Ross, followed by eDNA from lake whitefish, bull trout and
rainbow trout (Figure 2. 3). Compared to Factor Ross, kokanee eDNA was much less abundant at
Teare Creek and was nearly identical to lake whitefish in total eDNA copies detected at that
location. The total abundance of lake whitefish, bull trout and rainbow trout eDNA was
comparable between sites. The amount of bull trout and rainbow trout eDNA was always much
less, relative to the amount of kokanee and lake whitefish eDNA.

45

60

180
160

50

Total eDNA Copies Detected

Total eDNA Copies Detected

140
120
100
80
60
40
20
0

40
30
20
10
0

kokanee

lake
bull trout
whitefish

rainbow
trout

kokanee

lake
bull trout
whitefish

rainbow
trout

Figure 2. 3. Total number of eDNA copies from each species detected in 10 µL of the first elution of all
samples at Factor Ross (left) and Teare Creek (right). Counts of peamouth eDNA copies are omitted
because tests of the PCC_CYTB assay were not comparable.

Kokanee were detected in the greatest number of samples at Teare Creek and Factor Ross
and had the highest concentrations of eDNA in individual samples. Kokanee were detected at every
sample depth other than 10 m at Teare Creek (Figure 2. 4). The number of samples that kokanee
were detected in declined with depth, and so did the sample eDNA concentrations. Kokanee eDNA
concentrations were as high as 21 copies per reaction (16 positive droplets) in samples above the
thermocline at Factor Ross but were lower than 10 copies per reaction (8 positive droplets)
elsewhere. Lake whitefish were the second most detected species and were also detected more
frequently and with higher eDNA concentrations at sample depths above the thermocline. Lake
whitefish eDNA concentrations were as high as 15 copies per reaction (10 positive droplets) in
one surface sample from Factor Ross, but positive samples from other depths had lower
concentrations. Lake whitefish eDNA concentrations from Teare Creek did not vary greatly in any
of the positive samples.

46

0

5

5

Sample Depth (m)

Sample Depth (m)

0

10

rainbow trout
10

peamouth

20

20

30

30

0

1

2

bull trout

lake whitefish
kokanee

0

3

1

2

Positive Samples

Positive Samples

Figure 2. 4. Species detections using eDNA by depth at Factor Ross (left) and Teare Creek (right) sample
locations. A total of three 1-litre eDNA samples were collected from each depth and their extractions
were tested with six species-specific qPCR assays.

Peamouth, bull trout, rainbow trout and lake trout were detected in few or no samples, and
eDNA concentrations were low in positive samples for those species. The peamouth assay was
only tested on a single PCR replicate of the second elution of eDNA extractions, but peamouth
were still the third most frequently detected species at both sample locations. Peamouth were only
detected from Teare Creek at 30 m, and only above the thermocline at Factor Ross. Bull trout and
rainbow trout were detected at the surface, but the positive bull trout sample from Teare Creek and
both positive rainbow trout samples only contained the minimum concentration of eDNA required
for a positive detection (3 positive droplets between both PCR replicates of a sample). Lake trout
were not detected in any samples.

47

Gillnet Survey
Gillnets at Teare Creek were set above and throughout the thermocline but not below it.
Gillnets set at the surface and at 5 metres were above the thermocline where temperatures were
warmest. Depths between 10 and 23 m were fished with three gillnets where water temperatures
transitioned from approximately 16 °C down to 7 °C (Figure 2. 1). A small number of peamouth
and kokanee were caught in gillnets set at the surface and all other fish were caught in the nets at
15 m. Kokanee were the smallest but most numerous fish caught in the nets, averaging 65 g each.
Kokanee made up 67% of the individual count, but only 41% of the total biomass (Figure 2. 6).
Lake whitefish were the second most abundant species in the gillnets at 21% of the total count.
Lake whitefish had an average mass of 186 g and made up 37% of the total biomass. Peamouth
were the third most numerous species caught, but they only made up 5.5% of the total biomass,
averaging 76 g each. A single bull trout that weighed 880 g was caught in the gillnet at 15 m and
a single rainbow trout weighing 96 g was caught at 5 metres. Lake trout were not caught in gillnets
at Teare Creek.
The gillnets at Finlay Forks were only set at four depths between 5 and 20 m deep (Figure
2. 5). Kokanee were the most numerous species in the nets at Finlay Forks. Although most were
caught at 10 and 15 m, there was not much of a difference in catch abundance between all four
depths. Lake trout were caught at all four depths at Finlay Forks and were the second most
abundant species in the nets. Few lake whitefish were caught and only at 5 and 10 m depth.

48

0

5

Net Depth

Net Depth

5

10

10

15

15

20

20

0

10

20

0

30

5

10

Fish Count

Fish Count
longnose sucker

rainbow trout

lake trout

bull trout

peamouth

lake whitefish

lake whitefish

kokanee

kokanee

Figure 2. 5. Gillnet catches by species at each set depth between the surface and 20 m at Teare Creek
(left) and Finlay Forks (right). Each net panel spanned two vertical metres in the water column.
3000

2000

Fish Count

Biomass (g)

2500

1500
1000
500
0

45
40
35
30
25
20
15
10
5
0

Figure 2. 6. Total biomass (left) and total count (right) of each species caught in gillnets at Teare Creek.

49

DISCUSSION
Assay Design and In Vitro Validation
I successfully developed two novel qPCR assays and validated them alongside four
published assays to pilot species-specific eDNA methods for monitoring the pelagic fish
community in the Williston Reservoir. Published assays for lake whitefish, kokanee, rainbow trout
and bull trout were developed by others for use outside of northern British Columbia on real-time
qPCR platforms. This study validated the published assays so that they can be used with confidence
in the Williston Reservoir and with ddPCR. The development of novel lake trout and peamouth
assays were necessary for the implementation of species-specific eDNA surveys in the Williston
Reservoir. Lake trout monitoring is a priority for fisheries managers (Fish and Wildlife
Compensation Program, 2020), and developing a specific assay that could differentiate lake trout
from bull trout was essential. The LT_CYTB assay performed well in this study, and additional
field testing will increase confidence in detection results. Peamouth are not a high priority species
for monitoring in the Williston Reservoir, but eDNA results would not be as comparable to
traditional surveys without the ability to detect all six of the dominant fish species in the pelagic
community. The peamouth assay performed well but also requires additional field testing to
increase detection confidence.
Species-specific eDNA monitoring relies upon qPCR assays that amplify only the DNA of
their target species. The qPCR assays tested in this study demonstrated that they could reliably
distinguish between target and non-target species from the Williston Watershed. Assay specificity
is situation specific and heavily influenced by the design of primers and probes (Klymus et al.,
2020; Thalinger et al., 2021). The Sock_COI, Rain_COI and COCL_CYTB assays were originally
designed and validated by others against most of the same species that co-occur in the Williston

50

Watershed (Hernandez et al., 2020; Rasmussen Hellberg et al., 2010) and included species-specific
point mutations close to the 3’ ends of primers. This study validated these three assays for use in
the Williston Reservoir and with the ddPCR platform. The BUT1_CYTB assay was designed
primarily to discriminate between bull trout and invasive brook trout (S. fontinalis) (Wilcox et al.,
2013). Although there were five mismatches between the BUT1_CYTB primers and lake trout
DNA, they were not located near the 3’ end. Wilcox et al. (2013) observed some minor crossreactivity between the BUT1_CYTB assay and lake trout DNA with real-time PCR, but my results
with ddPCR were unambiguous. The ddPCR platform is not dependent upon reaction efficiency
and the digital format of presence and absence enhances the delineation of a true positive from
false positives (Burns et al., 2010; dMIQE Group, 2020; Doi et al., 2015). The slight increase in
negative droplet amplitude I observed when testing the BUT1_CYTB assay against lake trout DNA
(Figure A 14) is likely due to similar low-level cross-reactions observed by Wilcox et al. (2013).
Despite this small but noticeable difference, there was still a clear separation between the droplet
amplitudes achieved when the BUT1_CYTB assay reacted with bull trout DNA versus lake trout
DNA (Figure A 14). The LT_CYTB and PCC_CYTB assays were developed with primers that
maximized the influence of the available mismatches with closely related species from the
Williston Reservoir. This was especially important for the LT_CYTB assay as there were very few
mitochondrial DNA sequence differences between lake trout and bull trout. In vitro testing of the
LT_CYTB and PCC_CYTB assays demonstrated that they were highly specific (Appendix A). The
in vitro success of these novel assays is a testament to the importance and value of carefully
designing and thoroughly analyzing the performance of novel assays in silico.
Sensitivity is an essential aspect of assay performance and validation and is particularly
important for detecting some of the rare species in the Williston Reservoir. The standard LOD

51

definition from Klymus et al. (2019) used in this study is the most widely accepted definition for
qPCR assays used in eDNA studies (Thalinger et al., 2021). However, Hunter et al. (2017) argued
that the standard definitions and strategies for assessing the LOD of an eDNA assay can be either
too stringent or subjective and may result in biased estimates of species presence. Klymus et al.
(2019) acknowledged that detections below the LOD are expected in eDNA studies and may still
represent true positives. Detections below the LOD are important and should not be disregarded,
but the level of confidence in detections below the LOD is less than desired (Klymus et al., 2019).
PCR assays may also overestimate low DNA concentrations due to fixed instrumental responses
that cannot be monitored or corrected with experimental controls (Hunter et al., 2017). The
“concentration plateau” describes the nonlinear relationship between expected and measured DNA
concentrations observed in dilution series experiments. The statistically modeled LOD described
by Hunter et al. (2017) accounts for nonlinear responses in a serial dilution experiment and was
used as a double check for validating assay sensitivity.
Five of the six assays tested in this study demonstrated high sensitivity. LOD values for
the kokanee, rainbow trout, lake whitefish, bull trout and lake trout assays ranged from 4.32 to
6.40 copies per 20 µL reaction (Table 2. 7). These values are very close to the most sensitive LOD
theoretically possible for real-time qPCR at 3 copies per reaction, which assumes a Poisson
distribution, a 95% chance of including at least one copy in the PCR and single copy detection
(Bustin et al. 2009). Digital PCR provides greater precision than real-time qPCR (dMIQE Group,
2020) and the theoretical LOD on the Bio-Rad QX200 Droplet Digital PCR system is one copy
per reaction (Bio-Rad Labratories, Inc., 2022). The Modeled LOD is based on a mathematically
determined value that accounts for instrumental responses that may otherwise overestimate the
sensitivity of an assay. Modeled LOD values could not be determined for the Sock_COI and

52

BUT1_CYTB assays because a concentration plateau was not observed. However, the dilution
series results for both assays remained linear throughout the range tested, including the dilution
level that defined Standard LOD.
The peamouth assay did not demonstrate the same level of sensitivity as the other assays.
Five out of six PCR replicates were successful at the dilution level corresponding to 9.12 copies
per reaction, and single replicate failed to detect any DNA. As a result, the lowest concentration
that we tested that could meet the Standard LOD criterion of 95% confidence was five times higher
(Table 2. 7). However, the accuracy of LOD values depends on the serial dilution ratio, the number
of replicates tested at each dilution level, and can be affected by inconsistent lab procedures and
simple human errors. A smaller dilution ratio, like 1:2 instead of 1:5, would have allowed for more
accurate LOD determinations by reducing the incremental changes in copy numbers between
concentration levels. Studies that have tested assays with real-time qPCR platforms often include
many more than six replicates (Klymus et al., 2019). A downside of ddPCR is the increased cost
compared to real-time qPCR: a factor that limited the number of replicates that were used in
dilution series experiments for this study. Factors such as the stochastic distribution of DNA
molecules in very low concentration samples, pipetting errors, inhibition of the PCR or even
degraded target DNA are all possible causes for a single PCR replicate to fail to detect low
concentrations of DNA (Hunter et al., 2017). The Standard and Modeled LOD values differed by
two dilution levels for the peamouth assay. The Modeled LOD is an unbiased estimate of
sensitivity that prevents an overestimation of the LOD due to instrumental factors or artifacts in
ddPCR, but provides a less stringent definition than the Standard LOD described by Klymus et al.
(2019). It seems likely that the true LOD of the PCC_CYTB assay lies between the values

53

determined by the Standard and Modeled LODs reported by this study, and repeating the dilution
series experiment would provide clarity.
The performance of the eDNA assays tested in this study were affected by the entire
workflow, including the PCR platform and reaction conditions for which the assays were
optimized. Droplet digital qPCR has many advantages over real-time qPCR and is particularly
well suited to eDNA studies because it can detect trace levels of target DNA with high confidence
(dMIQE Group, 2020). The increased sensitivity of ddPCR has been demonstrated in other studies
(Burns et al., 2010; Doi et al., 2015; Sahu et al., 2021). The LOD for the COCL_CYTB assay
published by Hernandez et al. (2020) was 8 copies per reaction on a real-time qPCR platform, and
4.32 copies per reaction in this study (Table 2. 7). ddPCR is an end-point measurement that is
independent of reaction efficiency and is less influenced by matrix materials such as other DNA
and inhibitors (Bio-Rad Labratories, Inc., 2022; Burns et al., 2010; dMIQE Group, 2020; Doi et
al., 2015). Reactions that produce droplets with intermediate fluorescence that do not separate
nicely into positive or negative partitions (a phenomenon known as “rain”) may not affect the
ultimate quantification of DNA with ddPCR, but it is a sign of poor reaction efficiency, suboptimal annealing temperatures or mismatches between the assay and the target sequence (dMIQE
Group, 2020). The LT_CYTB assay had a relatively high potential for forming dimer structures
with the reverse primer (Table 2. 4) but performed well on the ddPCR platform. The temperature
gradient tests clearly indicated that the LT_CYTB assay performs best at temperatures at or above
58 °C, but the potential for poor reaction efficiency likely contributed to the increased range in
droplet amplitudes observed at lower annealing temperatures (Figure A 9). Wilcox et al. (2013)
found that PCR cycle threshold (Ct) values (the cycle number that indicates the detection of target
DNA) for the BUT1_CYTB assay were strongly skewed downwards when bull trout DNA was

54

diluted into lake trout DNA at ratios of 1:10 and 1:100. However, because DNA molecules are
partitioned in ddPCR, the likelihood of lake trout DNA influencing the quantification of bull trout
DNA would have been minimal. The ddPCR platform and the reaction conditions used in this
study have optimized the six assays tested, and the workflow I followed has validated their future
use in eDNA studies of the Williston Reservoir.
In Situ Assay Validation in the Williston Reservoir
The eDNA and gillnet results from the Finlay Reach demonstrate that molecular methods
are at least as sensitive as traditional survey gears, especially for rare species. Aside from a single
longnose sucker which was not targeted with an eDNA assay, all species caught in the gillnets at
Teare Creek were also detected by eDNA. Abundant species like kokanee, lake whitefish and
peamouth should not have been difficult to detect with either method, and agreement between
gillnets and eDNA on those species was expected. Lake trout, bull trout and rainbow trout are rare
species in the Williston Reservoir, and detecting them with pelagic gillnets has required a
significant amount of sampling effort in the past (Pillipow & Langston, 2002; Sebastian et al.,
2003, 2009). The gillnets set at Teare Creek for this study were also part of a larger pelagic survey
of the entire reservoir in 2021, but Teare Creek was the only location where bull trout were caught
in a gillnet (O’Connor, 2022). eDNA has proven to be a very sensitive and efficient method for
detecting rare species compared to conventional sampling gears (Evans, Shirey, et al., 2017; Jerde
et al., 2011; Thomsen et al., 2012). Bull trout and rainbow trout were detected with eDNA at both
Factor Ross and Teare Creek. Their eDNA was detected in very few samples and at very low
concentrations per sample, but this is not unexpected for rare species. The LT_CYTB assay was
tested on environmental samples for the first time in this study, and the results must be interpreted
cautiously. However, it is reassuring that lake trout were undetected by both eDNA and gillnets in

55

the Finlay Reach, and the results provide partial in situ validation for the novel lake trout assay
(Thalinger et al., 2021). Of all the gillnets set around the Williston Reservoir in 2021, only one
location in the Peace Reach and the nets set at Finlay Forks caught lake trout (O’Connor, 2022).
A total of 12 lake trout were caught at Finlay Forks (Figure 2. 5), which suggests that lake trout
are easily detected with gillnets when they are present. The gillnet results also suggest that lake
trout were not abundant outside of the Finlay Forks and Peace Reaches of the reservoir in early
August of 2021. The agreement between eDNA and gillnet results in the Finlay Reach gives
confidence in eDNA as a sensitive survey tool for presence and absence of species and advances
the validation of the individual assays used in this study.
The vertical distribution of species detections with eDNA were consistent with the gillnet
results, but eDNA integrates information about fish presence over larger areas and timescales than
traditional sampling methods (Shelton et al., 2019). Kokanee had the broadest ranging vertical
distribution of any species in the Finlay Reach as detected with both eDNA and gillnets. Although
the gillnets only sampled to a depth of 20 m, a nighttime trawl survey conducted at Teare Creek
on August 10, 2021 only caught kokanee and only at depths between 23 and 30 m (Weir et al.,
2023). The gillnet at Teare Creek was left in place for 24 hours due to poor weather, but the catch
is likely more representative of the nighttime distribution of fish. Gillnet catch rates are typically
highest at dusk and dawn (Vašek et al., 2009) and are reduced as fish accumulate in the nets (Olin
et al., 2009). The gillnet and trawl catches demonstrate that kokanee were distributed both above
and below the thermocline from the surface to a depth of 30 m. These results were biased towards
nighttime fish distributions and suggest that kokanee preferred greater depths at night. eDNA has
a smoother distribution in space and time than the fish themselves and should be detectable at all
of the depths that a fish has frequented in the hours to days prior to sample collection (Pont et al.,

56

2018; Shelton et al., 2019). This was exemplified in the Williston Reservoir by the mostly
continuous detection of kokanee and lake whitefish eDNA throughout the water column. The
eDNA results suggest that kokanee, lake whitefish, and perhaps even peamouth utilize depths
above and below the thermocline during late summer and are not confined to any zone in the water
column. The gillnet results offer an incomplete characterization of vertical fish distributions, but
one that overlaps with the distribution interpreted from eDNA. Characterizing the vertical
distributions of rare species like bull trout and rainbow trout is difficult given the limited dataset
of a few chance detections with eDNA and gillnets.
The distribution and abundance of fish eDNA in the Williston Reservoir is influenced by
the ecology and behaviour of each species as well as the ecology of eDNA itself. Kokanee, lake
whitefish and peamouth were all detected in at least one sample from depths below the
thermocline, but most species detections were in samples collected from the epilimnion. Kokanee,
lake whitefish and peamouth are planktivorous species that are known to exhibit diel vertical
migrations in summer (Clark & Levy, 1988; Gorsky et al., 2012; Hardiman et al., 2004; McPhail,
2007; Scheuerell & Schindler, 2003). Although these three species spend time above and below
the thermocline, eDNA production is positively correlated with metabolic rates and feeding
(Klymus et al., 2019; Rourke et al., 2021), which are both increased for planktivores when they
are in the warmer waters of the epilimnion. Fish metabolism is sensitive to temperature (Chabot et
al., 2016) and would be elevated in the warmer waters of the epilimnion compared to the cooler
hypolimnion water. Kokanee, lake whitefish and peamouth migrate to shallower depths above the
thermocline to feed on plankton, and actively feeding fish can shed as much as 10 times more
DNA than non-feeding fish (Klymus et al., 2015).

57

Rainbow trout do not stray far from shoreline features in large reservoirs (McPhail, 2007)
and have been caught most frequently in gillnets near the surface and near to shore in the Williston
Reservoir (Pillipow & Langston, 2002; Sebastian et al., 2009). The detection of rainbow trout
eDNA at the surface is consistent with historical gillnet results and likely an indication of feeding
in this location. The eDNA sample locations were further from shore than is typical for rainbow
trout, but the Williston Reservoir is subject to heavy wind which may transport eDNA near the
surface of the water long distances from where it originated. At the same time, eDNA at the surface
degrades faster than eDNA at depth because of increased exposure to UV radiation from sunlight,
elevated temperatures, and an increase in both the abundance and metabolism of microbes that
break down DNA (Takahara et al., 2012; Thomsen et al., 2012). eDNA production is elevated in
the warm and productive waters of the epilimnion where many species of fish feed actively. The
distribution of eDNA above the thermocline is likely broadened by the mixing action of heavy
winds, increasing detection sensitivity of eDNA but potentially reducing the spatial accuracy of
detections.
Factors affecting the distribution and abundance of eDNA in the hypolimnion can make
detecting species below more difficult than above the thermocline. Bull trout and lake trout are
cold water stenotherms and avoid the warm surface waters in summer, although they may exhibit
vertical migrations into the epilimnion to pursue prey (McPhail & Baxter, 1996; Riley et al., 2021).
Bull trout and lake trout metabolic rates would be slowed by the cooler temperatures in deeper
waters, and their larger body size would result in a lower metabolic rate relative to mass compared
to smaller species. Fish metabolic rates scale nonlinearly with body size and larger fish release less
eDNA into the environment per unit of mass compared to smaller bodied fish (Chabot et al., 2016;
Yates et al., 2020). If bull trout move into the epilimnion to feed, they could release more eDNA

58

at the surface than they do at depth, even if they spent most of their time in deeper water. This may
explain why gillnets caught bull trout at 15 m, but eDNA only detected them at the surface. eDNA
produced below the thermocline may have a patchier distribution than eDNA in the wind mixed
epilimnion because the water column below the thermocline is subject to much less movement and
internal mixing (Wetzel, 2001). In addition, eDNA will sink in the water column like any other
particulate organic matter, and especially if it is attached to or part of larger particulates such as
feces, carcasses, or scales (Canals et al., 2021). Consequently, the probability of detection may be
greater for eDNA in the epilimnion where it is suspended longer due to wind mixing, while eDNA
in the hypolimnion may sink relatively quickly and only be detectable in the water column for a
short time before it settles to the bottom. Acoustic monitoring studies in the Williston Reservoir
have observed that lake trout tend to congregate around discrete habitat features (Culling &
Euchner, 2019) which would further limit the distribution of their eDNA. The 2021 gillnet results
from the entire reservoir (O’Connor, 2022) also provide evidence for a patchy lake trout
distribution. Bull trout and lake trout are rare species in the Williston Reservoir, and their eDNA
is expected to be sparsely distributed. If either species venture into the surface waters to feed, they
may release a disproportionate amount of eDNA there compared to below the thermocline. Wind
mixing would likely disperse any eDNA left at the surface and reduce its concentration, but
potentially increase the likelihood of detection. Bull trout and lake trout eDNA below the
thermocline is also likely to be found in low concentration, but likely has a patchier distribution.
Klobucar et al. (2017) found that reliably detecting species that remain below the thermocline in
summer requires a substantial amount of sampling effort, including a large number and spatial
distribution of samples as well as a large volume of water filtered per sample. The distribution of

59

eDNA in the hypolimnion is likely more contracted than in the epilimnion but should more
accurately represent the recent presence of fish in the area sampled.
A recent review supported the findings for a positive correlation between fish abundance
and the concentration of eDNA (Rourke et al., 2021). The correlation between eDNA
concentration and species abundance is complicated by the same factors that affect the detection
and distribution of eDNA (Yates et al., 2019), although the relationship can be improved with
careful study design and data modeling (Chambert et al., 2018; Fukaya et al., 2018; Shelton et al.,
2019; Yates et al., 2020). It is reassuring to see a similarity between the relative abundance and
biomass of the species caught in gillnets and eDNA concentrations detected (Figure 2. 3 and Figure
2. 6). Kokanee were the dominant species caught in the gillnets as measured by the number of
individuals and by total biomass. Similarly, kokanee eDNA was more concentrated in the water
column than eDNA from other species at Factor Ross. eDNA from lake whitefish, bull trout and
rainbow trout were detected at concentrations that were comparable between Factor Ross and
Teare Creek, and each species ranked similarly in relative eDNA concentration compared to the
gillnet catch abundance and biomass. However, the relative concentration of kokanee eDNA at
Teare Creek differed from Factor Ross and was more like the concentration of lake whitefish
eDNA at Teare Creek. This appears to be a slight departure from the relationship between eDNA
concentration and gillnet catch abundance observed for other species and at Factor Ross. It is
difficult to explain what may have caused this difference, but the relationship between kokanee
eDNA concentration and relative abundance appears to be reduced at Teare Creek.
The eDNA results from Teare Creek are discontinuous and less comparable to the gillnet
catch, but these differences may have been caused by sample degradation and the timing of the
sampling event. Water samples were collected from Teare Creek in the morning on August 16 but

60

were not filtered and preserved until the following day. This delay in sample filtration was due to
a windstorm that forced the field crew to seek shelter on the reservoir and delayed the return to
camp until late in the evening. Every effort was made to keep the samples cool during the day but
conditions were not optimal on the boat. The water samples were kept in a fridge overnight and
filtered the next morning. The eDNA samples were also collected at Teare Creek the day after the
gill net was set in the same location. The sequencing of eDNA sample collection after the
gillnetting was not ideal and may have increased the amount of eDNA in the water column at Teare
Creek. Water samples from Factor Ross were collected early in the morning and filtered hours
later that same day. Although the Factor Ross samples were collected more than 20 km from Teare
Creek, the eDNA results from Factor Ross compare better to the Teare Creek gillnet catch, and
eDNA detections were more continuous throughout the water column. eDNA degradation in the
water samples from Teare Creek may explain the lack of any detections at 10 m, and the reduction
in kokanee detections relative to those at Factor Ross.
The results of this study have demonstrated that detecting rare species in an enormous,
unproductive waterbody like the Williston Reservoir can be challenging and provides insights into
how eDNA methods for large, unproductive lakes could be improved. Increasing the volume of
water filtered per sample, as well as the number and distribution of samples would increase
detection probabilities and confidence. eDNA studies of lake trout in smaller, presumably more
productive lakes in Ontario and Quebec were able to detect fish from 500-1000 mL volumes per
replicate collected below the thermocline in late summer (Lacoursière‐Roussel, Côté, et al., 2016;
Littlefair et al., 2020). However, Klobucar et al. (2017) filtered 5 L per replicate to detect Arctic
char (Salvelinus alpinus) that were in low abundance in relatively unproductive Alaskan lakes.
Stamford et al. (2020) filtered volumes of 3-5 L per replicate to detect Arctic grayling from

61

tributaries to the Williston Reservoir. Sampling larger volumes of water is highly recommended
for future eDNA studies in the Williston Reservoir. The 1-L replicate samples collected for this
study confidently detected relatively abundant species like kokanee and lake whitefish, but eDNA
concentrations were low, very low, or non-existent for rare species like bull trout and lake trout.
The clear, ultra-oligotrophic water sampled from the Williston Reservoir did not clog any of the
filters or reduce filtering efficiency and larger sample volumes than the 1-L used in this study
could easily have been processed through a single filter. Schabacker et al. (2020) recently
developed an eDNA collection method for lentic systems that has improved detection sensitivity
and requires fewer samples than the traditional filter methods used in this study. The method uses
a 64-µm mesh tow net to filter more than 3000 L of water per sample. Although the present study
was only meant to pilot eDNA methods, increasing the number and distribution of sample locations
would also have increased the probability of detecting rare species. When the Williston Reservoir
stratifies in late summer, the thermocline sets up very deep in the water column, making sampling
efforts below it more difficult. Furthermore, the eDNA of rare species like bull trout and lake trout
may be difficult to detect below the thermocline. Sampling eDNA in summer when a lake stratifies
can provide insight into fish distribution and habitat usage, but sampling during spring or fall
turnover in a lake may simplify sampling efforts and improve detection probabilities for rare
species. Studies that collected eDNA samples in early spring (Lacoursière‐Roussel, Côté, et al.,
2016) and fall (Klobucar et al., 2017; Littlefair et al., 2020) when lakes were isothermal found that
the eDNA of coldwater stenotherms like lake trout was distributed evenly throughout the water
column. Samples collected under isothermal conditions also had higher concentrations of rare
species eDNA compared to samples collected during stratification in summer (Klobucar et al.,
2017; Littlefair et al., 2020).

62

CHAPTER 3:
VALIDATION OF eDNA METABARCODING ASSAYS FOR FISH SPECIES IN THE
WILLISTON WATERSHED

INTRODUCTION
Rapid biodiversity monitoring is critical to conservation management in ecosystems that
are affected by climate change and other anthropogenic pressures. Freshwater habitats have been
classified as one of the most vulnerable ecosystems worldwide (WWF, 2020). Traditional
sampling methods for monitoring fish have various forms of bias (Kubečka et al., 2009; Zale et
al., 2012), are not easily scalable across vast geographic ranges and habitat types, and often fail to
detect rare or cryptic species (Evans, Shirey, et al., 2017; Jerde et al., 2011; Thomsen et al., 2012).
Biodiversity monitoring with molecular techniques that can detect the DNA released by organisms
into their environment, referred to as environmental DNA (eDNA), is a recent development that is
revolutionizing the way biodiversity is monitored, particularly in aquatic habitats (Beng & Corlett,
2020; Cristescu & Hebert, 2018; Deiner et al., 2017; Jerde, 2019; Ruppert et al., 2019). Although
eDNA-based methods have been developed relatively recently, they are now frequently applied in
biodiversity monitoring, especially for fish in aquatic environments (Schenekar, 2023).
Two approaches are commonly used to detect species with eDNA: single species detection
and multi-species detection, also known as eDNA metabarcoding. Single species detection
involves amplifying eDNA in a polymerase chain reaction (PCR) or quantitative PCR (qPCR)
using species-specific primers. Species are detected if their template DNA is present and is
amplified in the PCR reaction. Sampled water should contain DNA from all the species present in
the environment, but characterizing community composition by detecting one species at a time is
cost-inefficient and limited by the amount of sample available (Thomsen & Willerslev, 2015).

63

Metabarcoding assays employ generic primers that target highly conserved gene regions that flank
polymorphic regions (Balasingham et al., 2018). In a metabarcoding study, community DNA is
PCR-amplified and then sequenced on a Next Generation Sequencing (NGS) platform. Species are
detected and differentiated based on the identification of unique amplicon sequences, or species
barcodes.
eDNA metabarcoding is powerful and cost-efficient but the method has some limitations
and a complex process for analyzing results that must be considered before it is implemented in a
particular situation. Although metabarcoding primers are designed to amplify a wide range of taxa,
they may not detect all species present in the sampling environment due to biases in primer affinity
for certain species (Deagle et al., 2014; Morey et al., 2020; Olds et al., 2016). A multiple assay
approach is a popular solution to the shortfalls of individual markers and can maximize species
detections and confidence (Evans, Li, et al., 2017; Hänfling et al., 2016; Morey et al., 2020; Olds
et al., 2016; Snyder et al., 2020). However, analyzing environmental samples with multiple
metabarcoding primer sets increases the costs and time associated with analysis and may not be
necessary or appropriate if a single primer set is sufficient to identify all the target species in a
particular environment (Morey et al., 2020). Next Generation Sequencing produces a massive
number of sequences that must be filtered through a bioinformatics pipeline before the results can
be interpreted. Bioinformatic filtering removes sequencing errors and attempts to assign the
diversity of sequence reads to individual species. Matching sequences to species is complicated if
there is a limited amount of genetic variation in the amplicon for closely related species.
Metabarcoding primers are designed for highly conserved genes so they limit the number of primer
mismatches across taxa, but they sacrifice amplicon sequence diversity. Primers can be designed

64

for less conserved genes with more amplicon diversity and better species resolution, but this may
affect primer bias between species.
Aquatic species conservation in the rivers, lakes and reservoirs of the Upper Peace River
Basin is constrained by limits on the time and resources available to monitor such a vast and often
remote area (Fish and Wildlife Compensation Program, 2020). Species-specific eDNA methods
have been used to monitor Arctic grayling (Thymallus arcticus) in rivers and streams that drain
into the Williston Reservoir since 2018 (Stamford et al., 2020, 2022; Strohm et al., 2019). eDNA
has a great deal of potential for monitoring aquatic species in the Williston Watershed, and
metabarcoding methods could provide a rapid and cost-effective means of investigating fish
distributions and community compositions. In this study, I evaluated two metabarcoding primer
sets for use on the fish species in the Williston Reservoir and its surrounding watersheds. eDNA
metabarcoding results were compared against gillnet sampling in the Williston Reservoir and a
snorkel survey that was conducted in the Ingenika River. eDNA samples from other tributaries
around the reservoir were also tested with metabarcoding primers. All reservoir samples were
tested with species-specific primers for six dominant pelagic species, and tributary samples were
tested with species-specific primers for Arctic grayling. Detection results from the metabarcoding
analysis were compared against the species-specific eDNA analysis.
METHODS
Creating a Mitochondrial DNA Sequence Database
A reference sequence database of mitochondrial genes was created for 19 fish species
(Table 1. 1) that have been identified in the Williston Reservoir and its surrounding watershed
(Province of British Columbia, 2020; Sebastian et al., 2009). Available mitochondrial DNA
(mtDNA) sequences from the 12S gene, cytochrome oxidase subunit 1 gene (COI), and
65

cytochrome b (CYTB) gene were downloaded from the National Center for Biotechnology
Information’s (NCBI) Nucleotide Database (https://www.ncbi.nlm.nih.gov/nucleotide/) and
imported to Geneious Prime version 2021.2.2 (www.geneious.com). Sequences were organized in
Geneious Prime by gene and by species and aligned using the default parameters in the MUSCLE
alignment algorithm. Intraspecific genetic variation was incorporated into a consensus sequence
for each species by coding polymorphic sites with IUPAC codes.
In Silico Screening
A diverse assemblage of fishes belonging to four orders and seven families inhabit the
Williston Watershed. Primer sequences from published metabarcoding assays that were validated
in other freshwater fish studies were uploaded into Geneious Prime and screened for use with the
fish species in the Williston Watershed. Primer sequences were compared against the consensus
sequences in the Geneious database, and base pair mismatches with forward and reverse primers
were tallied. Assays with the fewest number of mismatches with the species in the Williston
Watershed were favoured during the selection process. Genetic distances between the amplicons
for each species were calculated in Geneious Prime with the Tamura-Nei distance model and
visually assessed on a neighbour-joining tree. Consensus trees were created based on the entire
amplicon, including the primer sequences. Genetic distances between amplicon sequences were
visualized in the consensus trees to assess an assay’s ability to resolve each of the 19 species in
the Williston Watershed. No single metabarcoding assay was able to resolve all species in the
Williston Watershed, but many eDNA metabarcoding studies have overcome this problem by
testing their samples with multiple assays (Evans, Li, et al., 2017; Hänfling et al., 2016; Morey et
al., 2020; Olds et al., 2016). The MiFish-U (MiFish; M) and FishCYTB-L (FishCYTB; F) assays
were selected for testing in the Williston Watershed (Table 3. 1) because they had the lowest

66

number of primer mismatches among all of the assays considered and when used in tandem, the
genetic distances between species amplicons was assessed to be sufficient for resolving at least
most of the species in the Williston Watershed. The MiFish primer set amplifies a portion of the
12S gene. The highly conserved nature of the 12S gene minimized the number of mismatches for
all species. The MiFish forward primer had a single mismatch at the second base position with
sequences from every species in the Williston Reservoir. This was corrected by changing that base
from a thymine (T) to a cytosine (C) so that the primer set perfectly matched every species. The
downside to a highly conserved target gene like 12S is that species resolution can be more difficult.
The FishCYTB assay targeted a segment of the Cytochrome B gene (CYTB), which has greater
interspecific sequence divergence than the 12S gene. However, the primer regions for the CYTB
gene were more likely to include mismatches. The MiFish and FishCYTB were chosen because
the in silico analysis predicted that they would be able to amplify and correctly identify the DNA
for all of the species in the Williston Watershed and together.
Table 3. 1. Primer sequences of the metabarcoding assays selected for testing on the fish species in the
Williston Watershed. The underlined “C” base in the MiFish-U_F primer was changed from the “T” in
the original sequence to perfectly match all 19 fish species considered in this study. UniA sequences were
added to the 5’ ends of forward primers and UniB sequences were added to 5’ ends of reverse primers for
barcoding and sequencing on the Ion Torrent platform.
Primer Set
MiFish-U

FishCYTB-L

Ion Torrent

Primer
Name
MiFish-U_F
(modified)
MiFish-U_R
FishCYTBL_F
FishCYTBL_R
UniA
UniB

Primer Sequence (5’-3’)
GCCGGTAAAACTCGTGCCAGC

mtDNA
Marker
12S

Amplicon
size (bp)
238-242

CATAGTGGGGTATCTAATCCCAGTTTG
GCCTACGCYATYCTHCGMTCHATYCC

CYTB

233

GGGTGTTCNACNGGYATNCCNCCAATTCA
ACCTGCCTGCCG
ACGCCACCGAGC

67

Citation
Miya et al.
(2015)
Snyder et al.
(2020)

In Vitro Validation
Metabarcoding assay performance was evaluated with a DNA mock community. The mock
community contained DNA from 12 fish species from the Williston Watershed, including six
salmonids (lake trout, bull trout, rainbow trout, kokanee, lake whitefish, mountain whitefish,
Arctic grayling), two minnows (northern pikeminnow, peamouth), and a representative species
from 3 other families (white sucker, slimy sculpin, and burbot). DNA used in the mock community
was extracted from fin tissues using a Qiagen DNeasy Blood and Tissue kit (Qiagen, ON, Canada)
following the manufacturer’s instructions and quantified with a Qubit Fluorometer (ThermoFisher
Scientific). The DNA from each species was diluted to between 0.05 and 0.11 ng/µL and 10 µL of
each was added to the mock community sample. The final concentration of each species’ DNA in
the mock community was between 4.88 and 11.8 pg/µL.
In Situ Assay Testing in the Williston Watershed
Pelagic eDNA Sampling from the Williston Reservoir
Pelagic eDNA sampling was conducted as described in Chapter 2.
eDNA Sampling from Tributaries of the Williston Reservoir
The tributary eDNA samples tested in this study were a subsample of those collected from
streams in the Finlay, Nation and Ospika regions in 2019, and from the Parsnip and Peace regions
in 2020 and 2021 (Stamford et al., 2020, 2022). Stamford et al. (2020; 2022) analyzed the eDNA
samples for Arctic grayling using a species-specific qPCR assay. Collection methods for eDNA
samples from the tributaries are presented in Stamford et al. (2020) and Stamford et al. (2022). In
this study, I analyzed a total of 44 replicates from 29 sample locations in representative regions of

68

the Williston watershed (Table 3. 2). The negative control samples that were included with the
original field samples were incorporated into the metabarcoding analysis.
Table 3. 2. Location, year and collection information for the eDNA samples collected from around the
Williston watershed and analyzed with the metabarcoding primers in this study (Stamford et al., 2020,
2022).
Location

Sample Region

Sample
Year

No. of Sites
Sampled

Total No. of
Replicates

Akie Creek

Finlay

2019

1

2

Pesika Creek

Finlay

2019

1

2

Ingenika River – middle

Finlay

2019

3

6 (2 for each site)

Ingenika River – upper

Finlay

2019

4

8 (2 for each site)

Ingenika River – headwaters

Finlay

2019

2

4 (2 for each site)

Ospika River

Ospika

2019

7

10

Manson River

Nation

2019

2

2

Carbon Creek

Peace

2020

2

2

Clearwater River

Peace

2020

2

2

Nabesche Creek

Peace

2020

2

2

Colbourne Creek

Parsnip

2021

1

2

Reynolds Creek

Parsnip

2021

2

2

29

44

TOTAL

eDNA Extraction
DNA from both reservoir and tributary eDNA samples was extracted at UNBC using the
same modified protocol described in Chapter 2.
PCR-based Library Preparation and Next-generation Sequencing
A preliminary polymerase chain reaction (PCR1) was performed on all eDNA samples at
UNBC to amplify target DNA. PCR1 samples were all amplified in 25 µL reaction volumes using
a Qiagen Multiplex PCR kit (Cat. No. 206143). Reaction mixtures included 12.5 µL of Qiagen
Multiplex PCR Master Mix (1x final concentration) and 2.5 µL of primers (0.3 µM for MiFish or
0.6 µM for FishCYTB). The remaining 10 µL volume varied among samples. The mock
community and contrasting dilution series samples contained 2 µL of extracted DNA mixture and
69

8 µL of nuclease free water (IDT). Tributary eDNA samples contained 3 µL of sample (i-elutions)
and 7 µL of nuclease free water. Reservoir eDNA samples (from Teare Creek and Factor Ross
locations) contained 10 µL of the ii-elution. Negative PCR controls contained 10 µL of nuclease
free water (IDT) only. Thermocycler protocols for the reaction mixtures with MiFish primers was:
initial denaturation at 95 °C for 15 minutes, then 35 cycles of denaturation at 94 °C for 30 seconds,
annealing at 63 °C for 90 seconds and elongation at 72 °C for 30 seconds followed by a final
incubation at 72 °C for 10 minutes. The thermocycler protocol for reaction mixtures with
FishCYTB primers was: initial denaturation at 95 °C for 15 minutes, then 35 cycles of denaturation
at 94 °C for 30 seconds, annealing at 57 °C for 90 seconds and elongation at 72 °C for 30 seconds
followed by a final incubation at 72 °C for 10 min. All reservoir samples were run for an additional
10 PCR cycles, for a total of 45 cycles. PCR1 products were visualized on 2% agarose gels (E-Gel
precast agarose gels, ThermoFisher Scientific) to verify amplification success. PCR1 samples were
submitted to the Environmental Genomics Facility at the Great Lakes Institute for Environmental
Research Laboratory at the University of Windsor (Windsor, Ontario, Canada).
PCR1 samples were further processed and sequenced unidirectionally on the Ion
GeneStudio S5 sequencer (Applied Biosystems, Burlington, Ontario) at the University of Windsor.
A PCR clean up was performed on PCR1 samples using Sera-Mag beads (Cytiva, Marlborough,
MA) following an in-house protocol (Appendix D). A second polymerase chain reaction (PCR2)
incorporated the IonA adaptor (required for the Ion Torrent sequencing reaction), a unique 10-12
base-pair sequence tag (barcode) to distinguish each individual sample, and the P1 adaptor
required for attaching the prepared amplicons to the Ion Spheres. PCR2 samples were amplified
in 25 µL reaction volumes that contained 8.4 µL of ultrapure water, 2.5 µL of both 10X Taq buffer
and MgSO4, 1 µM of UniA barcode, 1 µM of UniB adaptor, 1 µM of dNTPs, 0.1 µL of Taq DNA

70

Polymerase High Purity (Bio Basics), and 10 µL of the bead-cleaned product from PCR1. The
PCR2 reactions were amplified with an initial denaturation at 95 °C for 1 minute, followed by 8
cycles of: 95 °C for 15 seconds, 55 °C for 15 seconds, and 72 °C for 30 seconds, and a final
elongation of 72 °C for 7 minutes.
Samples were normalised by pooling varying amounts based on band intensities after
running a 2% agarose gel; 2 µL were added for samples with strong bands, 5 µL were added for
samples with faint bands, and 10 µL was added for samples that did not appear to have bands. The
final two normalised amplicon libraries, 12S (MiFish) and CYTB (FishCYTB), were then run on
2% TAE agarose gels stained with gel red at approximately 96 V for 4 hours. Run time was
increased while using a low voltage at this step to allow for sufficient separation of the 233 basepair bands for the MiFish amplicons and 238-242 base-pair FishCYTB (amplicons plus all required
adaptor sequences) from those containing amplicons with only one adaptor sequence attached. The
233 base-pair MiFish amplicons and 238-242 base-pair FishCYTB amplicons were extracted from
the gel by visualising the bands on the GelDoc Imaging System (BioRad) using a clean, sterile
scalpel, and purified using the GenCatch Gel Extraction Kit (Epoch Life Sciences), following the
manufacturer’s protocol. The quality and concentration of the 12S and CYTB amplicon libraries
were analysed using a High Sensitivity DNA Kit on the Agilent 2100 Bioanalyzer. All
concentrations were normalized to 60 pmol/L and the libraries were loaded on to the Ion Torrent
Chef for templating onto the 530 chip (Life Technologies). The following day, the loaded 530 chip
was put onto the Ion GeneStudio S5 sequencer along with the Ion 510 & 520 & 530 Chef Kit
chemistries (Thermo Fisher) for sequencing.

71

Bioinformatic Analysis of Next-generation Sequencing Data
Bioinformatics analysis were performed with Mothur version 1.48.0 (Schloss et al., 2009)
and QIIME2 version 2023.9 (Bolyen et al., 2019). Raw sequences were demultiplexed in Mothur,
and a manifest file was imported into QIIME2. Primers and adapters were trimmed from sequences
with the Cutadapt plugin (Martin, 2011) followed by denoising with DADA 2 (Callahan et al.,
2016). Filtering parameters included a minimum sequence length of 100 bp, and a quality score
cutoff of 2. All amplicon sequence variants (ASVs) were compared against the entire NCBI
nucleotide database with the Basic Local Alignment Search Tool (BLASTn). BLAST formatters
included were a minimum percent identity match of 97%, a maximum number of alignments per
ASV set to 100, and a single High-scoring Segment Pair (HSP).
The resulting ASVs were manually filtered by reviewing the BLAST results. Any ASVs
that were matched to non-fish species (e.g., birds, mammals, invertebrates, and bacteria) were
discarded. ASVs with unidentified species hits or no hits were also discarded. The BLAST results
were analyzed for all ASVs that were matched to fish species to determine the most appropriate
species assignment. Species assignments that were unambiguously identified as one of the 19
species known to occur in the Williston Watershed (100% identity, or only one species result)
were retained. The top hits for some ASVs were for fish species that were not native to the study
area. In many cases, BLAST results identified multiple species with the same or very similar values
for identity, E-value, and bit scores. In these cases, the ASV was assigned to the highest scoring
species from the Williston Watershed, if one was identified. If multiple species from the same
genus that occur or had potential to occur in the Williston Watershed were identified with the same
BLAST scores, the ASV was only classified to the genus level. Several ASVs produced by both
metabarcoding assays were only resolved to the genus level for Salvelinus. Confidence in species

72

assignments for an ASV were improved by comparison against the results from the mock
community samples and by comparing the species results from both primers. Any ASVs that were
only matched to fish species that do not occur in the Williston Watershed were discarded.
Species detections were evaluated based on three levels of bioinformatic stringency and
their related detection certainty in a similar manner to Evans, Li, et al. (2017). Bioinformatic
stringency is related to the level of certainty that a detection with eDNA represents the true
presence of a species in the environment. Under the low-stringency and low confidence scenario,
a species was considered detected if it was found in at least one sample using at least one of the
two metabarcoding assays. For the moderate-stringency, moderate confidence scenario, species
presence was inferred by the detection of its DNA in at least two samples, or if the DNA was
detected by both assays in a single sample. The high-stringency and high confidence scenario
required the detection of a species in a minimum of two samples by both assays.
qPCR Data Comparison
All the eDNA samples assessed with metabarcoding primers in this study were previously
analysed with species-specific qPCR assays. The reservoir samples from Teare Creek and Factor
Ross were analysed with species-specific assays for kokanee, rainbow trout, lake whitefish, bull
trout, lake trout and peamouth (Chapter 2). PCR protocols for the reservoir samples differed
slightly between qPCR and metabarcoding analysis. Most notably, qPCR analysis was performed
on two 5 µL replicates of the i-elution for each sample, while PCR1 for metabarcoding samples
was performed on a single replicate of the ii-elution. The eDNA samples from tributaries of the
Williston Reservoir were previously tested for Arctic grayling by Stamford et al. (2020) and
Stamford et al. (2022). Summaries of the qPCR data are provided in this study for comparison
between the two eDNA methods. The PCR protocols for the qPCR and metabarcoding approach

73

also differed for samples collected from tributary streams to the Williston Reservoir. The qPCR
protocols for Arctic grayling utilized two 5 µL replicates of the i-elution for each sample, while
PCR1 for metabarcoding samples was performed on a single 3 µL replicate of the same sample
for each metabarcoding assay.
Gillnet Sampling in the Williston Reservoir
The eDNA results from the Williston Reservoir were compared with catch results from
pelagic gill nets set at Teare Creek. Gillnet sampling methods are described in Chapter 2.
Snorkel Surveys of the Ingenika River
In 2019, a snorkel survey of the Ingenika River was paired with eDNA sampling for Arctic
Grayling. A detailed description of the Ingenika snorkel survey is provided in Strohm et al. (2020),
and the eDNA sampling procedure that accompanied it is described in Stamford et al. (2020). Poor
visibility in 2019 reduced the number of sites that could be snorkeled compared to those where
eDNA was collected. eDNA samples selected for metabarcoding analysis in this study were from
the middle and upper river sections that were snorkeled, as well as four replicates from two
headwater sites that were not paired with snorkeling.
RESULTS
In Silico Validation
The MiFish and FishCYTB assays were selected for use in tandem because their strengths
are complementary, their weaknesses in species resolution have minimal overlap, and they are
both well suited to the species assemblage in the Williston Watershed. The modified MiFish
primers did not have any mismatches with the species in the Williston Watershed, but the
amplicons for closely related species had low genetic distances between them. Largescale sucker

74

were the only species without a sequence in the NCBI nucleotide database that covered the MiFish
region of the 12S gene and were excluded from the in silico analysis for that assay. MiFish
amplicon divergence was low between closely related species and only differed by a single base
pair for kokanee and rainbow trout, as well as between prickly and slimy sculpins (Figure 3. 1).
The FishCYTB primers include six degenerate bases in the forward primer and five in the reverse
primer (Table 3. 1). There were few mismatches with the FishCYTB primers across all species
(Table 3. 3). Most species mismatches occur in the middle or at the 5’ end of the primers and none
are within five positions from the 3’ end. Mismatches are most significant for kokanee and rainbow
trout. Both species have a single mismatch at the sixth base position from the 3’ end of the reverse
primer, and the total number of mismatches are slightly higher than for other species (Table 3. 3).
FishCYTB amplicons differed greatly between most species, including those belonging to the
same genus (Figure 3. 2). The one exception was the low genetic distance between bull trout and
lake trout amplicons.

75

Table 3. 3. FishCYTB primer mismatches with the CYTB sequences of 19 species from the Williston
Watershed.
Species
Prickly sculpin
Slimy sculpin
Lake trout
Bull trout
Rainbow trout
Kokanee
Lake whitefish
Mountain whitefish
Pygmy whitefish
Arctic grayling
Longnose sucker
White sucker
Largescale sucker
Peamouth chub
Northern pikeminnow
Redside shiner
Lake chub
Longnose dace
Burbot

FishCYTB Primer Mismatches
Forward Reverse Total
0
0
0
0
0
0
1
0
1
2
0
2
2
1
3
2
2
4
1
0
1
0
0
0
2
0
2
1
0
1
1
1
2
0
1
1
0
1
1
1
1
2
2
1
3
1
1
2
0
1
1
0
0
0
2
0
2

76

Figure 3. 1. Neighbour joining tree for the MiFish amplicons (214-218 bp) from 18 species in the
Williston Reservoir created using the Tamura-Nei distance model. Percent consensus based on 500
bootstrap iterations is shown at the nodes. The scale bar indicates the relative genetic distance between
amplicon sequences for each species.

77

Figure 3. 2. Neighbour joining tree for the FishCYTB amplicons (209 bp) from 19 species in the
Williston Reservoir created using the Tamura-Nei distance model. Percent consensus based on 500
bootstrap iterations is shown at the nodes. The scale bar indicates the relative genetic distance between
amplicon sequences for each species.

78

ASV Assignment and Species Resolution
A total of 54,845 reads were generated from all samples amplified with the MiFish assay.
Only 8,338 reads from 28 ASVs were retained for final analysis after discarding ASVs that did not
have any matches with fish species from the Williston Watershed. The FishCYTB assay generated
1,118,323 total reads from the same samples. Of those, 80,706 FishCYTB amplicon reads were
retained from 51 ASVs that matched fish species in the Williston Watershed.
The MiFish (12S) assay was unable to discriminate between kokanee and rainbow trout,
however both species were unambiguously matched to ASVs produced with the FishCYTB assay.
Five MiFish ASVs had identical BLAST scores for both kokanee and rainbow trout. Neither
species was detected with the MiFish assay in any of the tributary samples despite multiple
detections of rainbow trout with the FishCYTB assay, and by snorkeling observations that
confirmed the presence of rainbow trout in the Ingenika River. Detections for the MiFish rainbow
trout/kokanee ASVs in the reservoir samples were assumed to be kokanee because they were much
more abundant in the gillnet catch and because the FishCYTB assay only detected kokanee in the
same samples.
Many ASVs from both metabarcoding assays identified species from the genus Salvelinus,
but identification to species was not always clear. Three MiFish ASVs and four FishCYTB assays
were unambiguously identified as bull trout. Two MiFish ASVs and one FishCYTB ASV were
matched with identical BLAST scores to lake trout, Dolly Varden (Salvelinus malma), Arctic char
(Salvelinus alpinus), as well as other species from the same genus that are not native to North
America. Bull trout were either not identified in the top 100 hits for these ASVs or they received
lower BLAST scores than the other species. These Salvelinus ASVs were detected in the mock
community samples, reservoir samples from Teare Creek and Factor Ross, as well as in some of

79

the Ingenika River headwaters samples. Three MiFish ASVs and six FishCYTB ASVs had equal
BLAST scores for Dolly Varden, Arctic char, brook trout (Salvelinus fontinalis) and bull trout, but
not for lake trout. All the ASVs that could not be unambiguously identified to a single species
were classified to the genus level only.
Sculpin were not as easily identified to species with the MiFish assay as they were with
FishCYTB. Spoonhead sculpin (Cottus ricei) received the highest BLAST scores for all sculpin
ASVs produced by the MiFish assay. Spoonhead sculpin are not known to occur in the Williston
Watershed upstream of the W.A.C. Bennet Dam on the Peace River, but they have been observed
downstream of it (McPhail, 2007). Slimy or prickly sculpin were the next best match to the MiFish
ASVs and only one of those two species was identified for each OUT. The BLAST scores for the
slimy or prickly sculpins were only slightly lower than the scores for spoonhead sculpin. The
FishCYTB assay unambiguously matched ASVs to either slimy or prickly sculpins but not to
spoonhead sculpin. Cross referencing the MiFish results with the FishCYTB results suggested that
slimy or prickly sculpin was a more appropriate species assignment for the MiFish ASVs.
Non-fish Sequences
BLAST results for the MiFish and FishCYTB sequences included matches to a wide range
of non-fish species, many of which are associated with aquatic environments. The number and
diversity of non-fish species identified was much higher for the FishCYTB assay than for MiFish.
A large portion of non-fish sequences from both assays were attributed to bacteria, protists, and
aquatic invertebrates. Mammal species included beavers (Castor canadensis – identified by both
assays) and a variety of other rodent species, porcupine (Erethizon dorsatum) least weasel
(Mustela nivalis), little brown myotis (Myotis lucifugus), moose (Alces alces), caribou (Rangifer
tarandus), elk (Cervus canadensis nannodes), snowshoe hare (Lepus americanus), and wolf

80

(Canis lupis). Bird species included the hermit thrush (Catharus guttatus) and wren (Tryoglodytes
sp.).
Mock Community
At least nine of the 12 species included in the mock community samples were detected by
at least one of the metabarcoding assays (Table 3. 4). Although two replicates of the mock
community were tested with each assay, only one of the MiFish replicates was sequenced
successfully. Slimy sculpin and rainbow trout were not detected by either of the assays. Bull trout
and kokanee were only detected by FishCYTB. The read count for many species was low for all
samples, but the MiFish amplified mock community sample had a much lower read count
compared to the FishCYTB amplified mock community samples. Lake trout and bull trout DNA
were both included in the mock community samples. BLAST results for one MiFish ASV and one
FishCYTB ASV matched many species from the genus Salvelinus including lake trout and bull
trout. Because the mock community included only lake trout and bull trout DNA and bull trout
were more confidently identified with a FishCYTB ASV, the ambiguous Salvelinus FishCYTB
ASV was suspected to represent potential detections of lake trout DNA. However, only the
ambiguous, genus level Salvelinus ASV was detected in the MiFish mock community sample and
may have been an amplification of DNA from bull trout or lake trout.

81

Table 3. 4. Read counts for each of the 12 species included in the mock community samples.
Species
Slimy sculpin
Salvelinus sp.
Bull trout
Rainbow trout
Kokanee
Lake whitefish
Mountain whitefish
Arctic grayling
White sucker
Peamouth
Northern pikeminnow
Burbot

MiFish

FishCYTB-1

FishCYTB-2

41

186
23

212
47

47
97
449
21
26
269
92
31

35
141
468
87
25
280
80
48

14
39
4
21
20
10

In Situ Assay Testing on Reservoir eDNA Samples
Kokanee and lake whitefish were detected throughout the water column by both the MiFish
and FishCYTB assays at Teare Creek and Factor Ross with high confidence (Table 3. 5). Peamouth
were detected in fewer samples, but detection confidence was still high at Teare Creek and Factor
Ross. Kokanee, lake whitefish and peamouth were the most abundantly caught species in gillnets
at Teare Creek. The gillnet catch also included one rainbow trout, one bull trout and one longnose
sucker (Table 3. 5). Bull trout and longnose sucker were detected with moderate confidence by the
metabarcoding primers, but rainbow trout was not detected at either location. Species detected by
eDNA metabarcoding but not by gillnets at Teare Creek included mountain whitefish and slimy
sculpin (both moderate confidence). Four species (white sucker, prickly sculpin, pygmy whitefish
and redside shiner) were detected at Factor Ross in addition to those detected at Teare Creek, but
with low confidence. Multiple ASVs that could be confidently identified only to genus level for
Salvelinus were detected at both locations. All reservoir samples were tested for the presence of
lake trout and Arctic grayling with species-specific qPCR assays but neither species was detected.

82

Table 3. 5. Species detections above (<10 m depth) and below (> 10 m depth) the thermocline in the
Williston Reservoir at the Teare Creek and Factor Ross stations with eDNA and gillnets. eDNA
metabarcoding species detections with a single assay are noted as M for MiFish or F for FishCYTB.
Detections with both metabarcoding assays are noted as B. Detection confidence with eDNA
metabarcoding is indicated as high (green), moderate (yellow) or low (red). Species-specific detections
with qPCR assays and species caught by gill net are noted as X.Species not tested for with qPCR are
noted with a dash (-).

Species
Prickly sculpin
Slimy sculpin
Bull trout
Lake trout
Salvelinus sp.
Rainbow trout
Kokanee
Lake whitefish
Mountain
whitefish
Pygmy
whitefish
Arctic grayling
Longnose
sucker
White sucker
Peamouth
Redside shiner

FACTOR ROSS eDNA
Metabarcoding
qPCR
<10 m >10 m <10 m >10 m
M
M
B
B
X
B

B

B
B

B
B

X
X
X

M

F

TEARE CREEK eDNA
Metabarcoding
qPCR
<10 m
>10m
<10 m >10 m
M
B
F
X

-

M

F
B
B

X
X
X

X

B
B

X
X

-

-

M

F

-

-

M

-

-

-

-

M

M

-

-

-

-

M
B
F

X
-

-

F

-

X
-

F
F

-

B
-

TEARE CREEK
Gillnet
<10 m
>10m

X

X
X
X

X
X

X
X

Species detections for eDNA differed somewhat between qPCR and metabarcoding
methods. Kokanee, rainbow trout, bull trout, lake trout, lake whitefish, Arctic grayling and
peamouth were the only species tested for with qPCR. Results were similar between the two eDNA
methods for kokanee, lake whitefish, peamouth and bull trout (Table 3. 5). Multiple ASVs that
could only be confidently attributed to genus level for Salvelinus were detected above and below
the thermocline at Teare Creek and Factor Ross. Lake trout were not detected in the gillnets or
with the novel qPCR assay that was being evaluated for use in the Williston Reservoir (Chapter
2). Rainbow trout were not detected with either metabarcoding assay. Rainbow trout were detected
by qPCR, but the number of eDNA copies detected was very low (See Chapter 2).

83

In Situ Assay Testing on Williston Tributary eDNA Samples
A total of 15 species and ASVs that matched the Salvelinus genus were detected in samples
from tributary streams around the Williston Reservoir. Slimy sculpin, mountain whitefish and bull
trout were detected with at least moderate confidence in every stream that was tested (Table 3. 6).
Longnose sucker and rainbow trout were detected in most streams, and all other species were
detected to a lesser extent.

1
2

F
F
F

F

F

F

F
F
F
F

4:6:8
3:4:6
2:5:5
3:5:6
2:5:6

B2
F
F

F

Total species detected
by confidence level
(High:Mod:Low)

B
B
B
B
B
F
F

B
F
F

F
F2
F2

Burbot

B
B
B
B
B
B
B

B
B
B
B
B
B
B

Redside shiner

F

Northern pikeminnow

B

F

F

Longnose dace

F

B

B

Arctic grayling

B

F

B
B
B
B

Kokanee

M

F
F
F

B
F

Bull trout
B
B
F
B

B
B
B
B

F
B

Pygmy whitefish

Mountain whitefish

Largescale sucker

White sucker

Longnose sucker

Rainbow trout

F

Salvelinus sp.

Sample Location
Akie Creek
Pesika Creek
Middle Ingenika
Upper Ingenika
Ingenika
Headwaters
Ospika River
Manson River
Carbon Creek
Clearwater Creek
Nabesche Creek
Colbourne Creek
Reynolds Creek1

Slimy sculpin

Prickly sculpin

Table 3. 6. Species detections with the MiFish (M), FishCYTB (F), or both (B) metabarcoding assays in
tributary rivers and creeks to the Williston Reservoir. Detection confidence is indicated as high (green),
moderate (yellow) or low (red).

F
B

F

F

2

F2

F

F

B
B

4:5:8
1:5:11
2:2:5
3:4:5
3:4:5
2:5:5
0:8:9

One of the MiFish amplified samples from Reynolds Creek was not successfully sequenced.
Streams where Arctic grayling was detected with qPCR.

The FishCYTB assay detected more species and had a greater number of detections per species
than the MiFish assay did. The most abundant and ubiquitously distributed species (slimy sculpin,
mountain whitefish and bull trout) were almost always detected with both metabarcoding assays.
Most of the other detections for a species were with the FishCYTB assay. Rainbow trout were
84

frequently detected, but only by the FishCYTB assay. The only instance where the MiFish assay
had a detection and the FishCYTB did not, was for one of the Salvelinus ASVs in the Upper
Ingenika River (Table 3. 6 and Table 3. 7).
eDNA Metabarcoding Comparison with Snorkel Surveys
The large number of samples collected from the Ingenika River allowed me to assess
distribution of species along the length of a river and make direct comparisons to the snorkel
surveys that were performed in the same reaches. Snorkel surveys in the middle and upper reaches
of the Ingenika River observed mountain whitefish, bull trout, Arctic grayling and rainbow trout,
but did not record observations of other species (Strohm et al., 2020). Counts of bull trout, rainbow
trout and Arctic grayling ranged from 18 to 37 individuals per reach. Mountain whitefish counts
were considerably higher with 279 and 259 individuals in the middle and upper reaches,
respectively. Both eDNA metabarcoding assays detected slimy sculpin and mountain whitefish in
every sample from the Ingenika River (Table 3. 6). Rainbow trout and either bull trout or Salvelinus
sp. were detected in nearly every sample. Bull trout were detected in all three reaches of the
Ingenika, while Salvelinus sp. ASVs were detected almost exclusively in the headwaters reach
(Table 3. 7). Arctic grayling were detected in two out of six samples from the middle Ingenika
reach. Longnose suckers were only detected with FishCYTB and only in the upper river and
headwaters samples. Burbot were detected with both assays in the middle reach and longnose dace
were detected in a single sample from the upper reach.
eDNA Metabarcoding Comparison with qPCR
All the tributary samples I analyzed with eDNA metabarcoding were originally tested with
a qPCR assay for Arctic grayling by Stamford et al. (2020; 2022). Arctic grayling were detected
by metabarcoding and qPCR in both samples from Pesika Creek and Reynolds Creek (Table 3. 6).
85

Metabarcoding detected Arctic grayling in two out of six samples from the middle Ingenika River
where qPCR detected them (Table 3. 7), and in two out of four samples where qPCR detected them
from the Ospika River. Metabarcoding detected Arctic grayling in one sample from Akie Creek
when qPCR did not, and qPCR detected Arctic grayling in one sample from Reynolds Creek when
metabarcoding did not (Table 3. 6). The MiFish assay only detected Arctic grayling in a single
sample from the Ospika River and all other metabarcoding detections were with the FishCYTB
assay.

F
F
F
F
F
F

F
F
F
F
F
F
F
F
F
F

F
F
F
B
B
F

*
*
*
F*
F*
*

F
F

B

F

B
F
B
B
B

*Samples where Arctic grayling was detected with species-specific qPCR analysis.

86

M

M

B
B
F
F

Burbot

F
F

Longnose
dace

F
F
F
F
F
F

Arctic
grayling

B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B

Salvelinus
sp.

Bull trout

F

Rainbow
trout

B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B

Mountain
whitefish

Sample
Mid 1A
Mid 1B
Mid 2A
Mid 2B
Mid 3A
Mid 3B
Upper 1A
Upper 1B
Upper 2A
Upper 2B
Upper 4A
Upper 4B
Upper 5A
Upper 5B
Headwater 1A
Headwater 1B
Headwater 1C
Headwater 2A
Headwater 2B

Longnose
sucker

Slimy
sculpin

Table 3. 7. Species detections with the MiFish (M), FishCYTB (F) or both (B) metabarcoding assays
from the middle, upper and headwaters reaches of the Ingenika River. Snorkel counts for the middle
Ingenika River were 279 mountain whitefish, 18 bull trout, 19 rainbow trout, 27 Arctic grayling. Snorkel
counts for the upper Ingenika River were 259 mountain whitefish, 37 bull trout, 19 rainbow trout, 23
Arctic grayling.

DISCUSSION
eDNA Metabarcoding in the Williston Reservoir
eDNA metabarcoding was effective in detecting most fish species in pelagic habitats of the
Williston Reservoir when compared to qPCR and gillnetting surveys. Rainbow trout were the only
species detected by qPCR and gillnetting but not by either metabarcoding assay. Although the
MiFish assay was unable to differentiate between rainbow trout and kokanee, signal detection with
this assay in the reservoir was likely kokanee as rainbow trout were detected by the FishCYTB in
many of the tributary samples. Metabarcoding detection probability would have been reduced
compared to qPCR due to the types of samples both methods analyzed. The qPCR analyses were
conducted on i-elutions of the reservoir samples and metabarcoding was conducted on the iielutions of the same samples. Subsequent elutions of the same sample typically yield less DNA
than the first (Qiagen, 2020) and since rainbow trout were barely detected by qPCR in the first
elution, it is not surprising that they were not detected by metabarcoding on the ii-elutions. Like
the qPCR analysis, species detection confidence with metabarcoding was proportional to the catch
abundance in the gillnets. Kokanee and lake whitefish were the most abundant species caught in
the gillnets and were detected with high confidence by eDNA metabarcoding. The Williston
Reservoir is both enormous and unproductive; consequently, detection confidence with eDNA
could be improved by increasing the sampling effort. This could be done by filtering larger
volumes per sample, collecting more samples, and by using innovative sample techniques that
capture higher concentrations of eDNA, like the mesh tow net method described by Schabacker et
al. (2020). Detection probabilities are also affected by the level of replication during other stages
of the workflow such as PCR and sequencing (Ficetola et al., 2015). Increasing the number of

87

PCR1 replicates in this study would likely have increased our detection probabilities for rare
reservoir species like rainbow trout.
The metabarcoding analysis detected multiple species that were absent from any of the
gillnet surveys conducted throughout the Williston Reservoir (O’Connor, 2022) including slimy
sculpin, mountain whitefish, pygmy whitefish and redside shiner (Table 3. 5). Slimy sculpin and
redside shiner are less susceptible to being captured in gillnets because of their small body size
and behaviour (Kubečka et al., 2009; Olin et al., 2009). Rare species such as pygmy whitefish
would have a very low probability of capture in traditional gears but may be readily detected by
eDNA due to its increased sensitivity (Evans, Shirey, et al., 2017; Jerde et al., 2011; Thomsen et
al., 2012). Pygmy whitefish and redside shiner were detected with low confidence using eDNA
metabarcoding, but slimy sculpin and mountain whitefish were detected with moderate confidence.
Mountain whitefish would have been susceptible to capture in gillnets but were absent from any
of the nets set in the reservoir in 2021 (O’Connor, 2022). Therefore, the detection of mountain
whitefish eDNA in the pelagic zone does not appear to be the result of their presence in the same
area. It is possible that some of the unexpected species detections were the result of eDNA that
was transported into the pelagic zone of the reservoir from elsewhere. Mountain whitefish and
slimy sculpin were detected in every tributary of the reservoir that I tested (Table 3. 6), and it is
possible that their DNA was transported into the reservoir by streamflow (Evans, Li, et al., 2017)
and could have been transported further by wind and water currents in the reservoir itself.
However, eDNA will only remain detectable in the environment for a relatively short period that
ranges from a few days to a couple of weeks before decaying (Rourke et al., 2021; Takahara et al.,
2012) or settling out of the water column (Canals et al., 2021; Jane et al., 2015; Stamford et al.,
2020). The distance that eDNA can be transported from any tributary into the pelagic zone of the

88

Williston Reservoir is unknown, but other studies in lakes have demonstrated that eDNA dispersal
is limited by summer stratification (Littlefair et al., 2020). Mountain whitefish and other species
may utilize near-shore areas in the reservoir that were not targeted with gillnets but that were close
enough to the eDNA sample sites to facilitate detection of transported eDNA. The distance that
eDNA could be dispersed in the epilimnion of the Williston Reservoir may be quite large when
strong winds persist for days, as they did immediately prior to the collection of samples for this
study.
The difficulty that the MiFish and FishCYTB assays had in resolving species from the
Salvelinus genus is problematic for the use of eDNA metabarcoding in the Williston Reservoir.
Some ASVs were unambiguously matched to bull trout, but many could not be confidently
identified beyond the genus level. Lake trout and bull trout are high priority species for fisheries
management in the Williston Reservoir (Fish and Wildlife Compensation Program, 2020) and
effective eDNA monitoring should be able to differentiate the two species. Lake trout DNA was
included in the mock community but did not produce any ASVs that were unambiguously matched
to lake trout in the BLAST results. At least one ASV from both metabarcoding assays that was
detected in the mock community was suspected to have been the result of lake trout DNA. These
same ASVs were detected in some of the reservoir samples and in the headwater samples from the
Ingenika River. Lake trout were not identified in any of the reservoir samples with a novel qPCR
assay (although the assay was being validated for that purpose at the time), and they were not
caught in any of the gillnets set at Teare Creek. It is possible that the Salvelinus ASVs were
detecting lake trout in the reservoir, but there is no compelling evidence to support this idea. It
appears equally likely that those same Salvelinus ASVs could have been detecting bull trout
eDNA. The utility of eDNA metabarcoding in the Williston Reservoir could be improved if an

89

alternate combination of primers were able to resolve all the pelagic species that are priorities for
fisheries management.
eDNA Metabarcoding in Williston Tributaries
The probability of detection with eDNA metabarcoding was proportional to the relative
abundance of each species in the sampling environment. Mountain whitefish were much more
abundant than any of the other species observed during the 2019 snorkel surveys in the Ingenika
River (Strohm et al., 2020) and accounted for 80% of the species captured in a 2021 survey of the
Ospika River (Stamford & O’Connor, 2022). Mountain whitefish were detected in every eDNA
sample I tested with high confidence (Table 3. 7). Detections of less abundant species were often
reduced with metabarcoding compared to qPCR and snorkel observations. Bull trout and rainbow
trout were detected in most samples, although to a lesser extent than mountain whitefish. This
result was consistent with the relative abundance of each species observed in the Ingenika River
snorkel surveys and in surveys of the Ospika River (Stamford & O’Connor, 2022). Arctic grayling
were observed more frequently than rainbow trout and bull trout in the middle reach of the Ingenika
River during the snorkel survey, but Arctic grayling were detected in fewer samples than the other
two species with eDNA metabarcoding. The reduced detection confidence with eDNA
metabarcoding can be partially explained by the poor performance of the MiFish assay. Across all
the tributary samples tested, the MiFish assay never detected rainbow trout while bull trout and
Arctic grayling were detected much less frequently with MiFish than they were with the
FishCYTB assay. Arctic grayling were also detected less often with metabarcoding in the Ingenika
River and Ospika River compared to the qPCR analysis of the same samples. Although Arctic
grayling were detected in more samples with qPCR than with metabarcoding, the difference in
sensitivity between the two methods was not substantial. Differences in lab procedures likely

90

contributed to the discrepancy between metabarcoding and qPCR detection results. The qPCR
workflow that was used for detecting Arctic grayling in the tributary samples included amplifying
eDNA on a droplet digital PCR (ddPCR) platform. ddPCR has many advantages over traditional
PCR including increased sensitivity, and it is less influenced by matrix materials such as non-target
DNA and inhibitors (Bio-Rad Labratories, Inc., 2022; Burns et al., 2010; dMIQE Group, 2020;
Doi et al., 2015). The small amount of sample that was tested in the metabarcoding analysis relative
to qPCR would have reduced the probability of detection for metabarcoding. The PCR1 reactions
for the metabarcoding analysis were performed on a single replicate for the MiFish and FishCYTB
assays that tested a total of 6 µL of sample. The qPCR reactions were amplified in two PCR
replicates that included a total of 10 µL of the sample. Increasing the number of PCR replicates
has demonstrated a reduction in the rate of false negatives and to increase the detection probability
for rare species in metabarcoding studies (Ficetola et al., 2015). Increasing the number of PCR1
replicates could have increased the detection probability for eDNA metabarcoding in this study to
a level that would have been comparable to qPCR. This was supported in the analysis of reservoir
samples from this study where the total sample volumes tested by qPCR and metabarcoding were
the same (albeit from different elutions) and the detection results from both methods were similar.
The eDNA sample distribution and number of samples that were analyzed in this study was
not sufficient to describe the full diversity of species accurately and confidently in any single
watercourse and may have contributed to the discrepancy between species detections with
metabarcoding and snorkeling in the Ingenika River. The eDNA of very abundant and evenly
distributed species like mountain whitefish and slimy sculpin should likewise be abundant and
evenly distributed in the environment. This is supported by the detection of mountain whitefish
and slimy sculpin in every sample analyzed by eDNA metabarcoding. However, relatively rare

91

species like Arctic grayling often have a patchier distribution resulting in eDNA concentrations
that vary throughout the environment (Eichmiller et al., 2014; Rourke et al., 2021). eDNA samples
collected by Stamford et al. (2020) in the Ingenika River were spaced 1.5 km apart based on the
assumption that Arctic grayling eDNA would remain suspended in the water column over that
distance. That assumption proved to be reasonable for Arctic grayling, although false negatives in
some sample replicates suggested that eDNA may not always travel that far, and concentrations
are reduced with distance. The 1.5 km distance between samples may not be an appropriate spacing
for all species or at all times of the day or year. Species-specific behaviour can affect the rates of
eDNA shedding (Klymus et al., 2015) and habitat preferences or environmental factors may affect
eDNA shedding rates and downstream transport of eDNA (Barnes et al., 2014; Barnes & Turner,
2016; Jane et al., 2015). Sample size is an important determinant of the accuracy and precision
with which eDNA metabarcoding can estimate species richness in any system (Evans, Li, et al.,
2017). The distribution and size of samples selected for this exploratory study of eDNA
metabarcoding in the Williston Watershed are insufficient to confidently detect all the species
present. When an appropriate sample design is implemented, eDNA metabarcoding has proven
effective in determining species richness in environments that can be difficult to sample with
traditional fish-capture methods (Balasingham et al., 2018; Evans, Shirey, et al., 2017; Hänfling
et al., 2016).
The issue of resolving closely related species with metabarcoding was not as problematic
in the tributaries as it was in the reservoir due to the difference in species assemblages between
the two environments. The inability of the MiFish assay to differentiate between kokanee and
rainbow trout was overshadowed by the fact that the MiFish rainbow trout/kokanee ASV was
never detected in any of the tributaries, even when rainbow trout were observed near sample

92

locations. Kokanee and rainbow trout eDNA would be present together in many of the same stream
habitats around the Williston Reservoir from the time kokanee adults begin migrating to spawn in
the fall until fry migrate out of the same streams in the spring. Although differentiating between
species would not be possible with the MiFish assay during that period, it would not be a problem
with the FishCYTB assay. Both assays had some difficulty resolving species of the genus
Salvelinus, but bull trout are likely the only species of that genus present in streams and rivers of
the Williston Watershed. Lake trout complete their entire life cycle in lacustrine environments
(McPhail, 2007; Riley et al., 2021) and have never been observed in streams or river habitats
around the Williston Reservoir (Culling & Euchner, 2019; Zemlack & Langston, 1994). Therefore,
it would be safe to assume that any Salvelinus ASVs detected in tributary samples were for bull
trout. In the headwaters reach of the Ingenika River, genus level Salvelinus ASVs were detected
with high confidence while ASVs that unambiguously matched to bull trout were detected less
(Table 3. 7). This result contrasted with other reaches of the Ingenika River where bull trout ASVs
were detected frequently and there was only a single Salvelinus detection. This contrasting pattern
of detections is an interesting observation which may indicate some genetic differences between
bull trout that occupy the headwaters of the Ingenika River and those that are present elsewhere.
Bull trout in the Ingenika headwaters may be a stream-resident population that has become at least
partially isolated from bull trout with fluvial or adfluvial life histories. Little is known about the
distribution of Dolly Varden in the Williston Watershed but they have been identified in tributaries
of Thutade Lake at the headwaters of the Finlay River (Hagen, 2000) and more recently in the
headwaters of the Ingenika River (J. Hagen, personal communication, 15 April 2024). Bull trout
and Dolly Varden are sympatric in the Thutade watershed and although hybridization has been
confirmed there, it appears to occur infrequently (Hagen, 2000; Redenbach & Taylor, 2003).

93

Differentiating between bull trout and Dolly Varden with metabarcoding or species-specific eDNA
methods would be challenging due to the genetic similarity between the two species and their
tendency to hybridize (Redenbach & Taylor, 2003; Taylor et al., 2001).
Evaluating eDNA Metabarcoding for use in the Williston Watershed
MiFish and FishCYTB Assay Performance
I chose to evaluate the MiFish and FishCYTB metabarcoding assays for their potential
applications in the Williston Reservoir and the surrounding watershed. In this study, the FishCYTB
assay outperformed MiFish assay and was effective enough to be used on its own in the tributaries.
The FishCYTB assay generated approximately 20 times more raw reads, 10 times more reads that
were matched to fish species in the Williston Watershed, and nearly twice as many final ASVs
compared to MiFish. Both assays failed to detect slimy sculpin and rainbow trout in the mock
community, but the results from the in situ samples suggest that this may not have been an issue
with the assays. Slimy sculpin were detected by both assays, and rainbow trout were detected by
FishCYTB in the Williston samples. The FishCYTB assay did not consistently fail to detect any
species like the MiFish assay did with rainbow trout. The FishCYTB assay detected more species
in more samples compared to the MiFish assay, and the MiFish results added little to the
FishCYTB results. Littlefair et al. (2020) used the MiFish assay successfully on its own in the
Experimental Lakes Area of Ontario with many of the same species. The species assemblage there
may not have been comprised of as many closely related species, and the absence of bull trout
would have made lake trout detections less ambiguous. Littlefair et al. (2020) also analyzed more
samples and more technical PCR replicates than I did and used a different bioinformatics pipeline.
Metabarcoding studies have often used multiple genetic markers to increase overall detection
probability. The 12S, 16S and CYTB genes are commonly used in combination, and CYTB

94

markers have often detected more species than those designed for 12S (Evans et al., 2016; Evans,
Li, et al., 2017; Hänfling et al., 2016; Morey et al., 2020; Olds et al., 2016). The performance of
the 12S and CYTB markers relative to one another varied in those studies due to a variety of factors
that included the relative length of the amplicons for each marker (Hänfling et al., 2016), the
diversity of the species assemblage being tested (Morey et al., 2020) and the sampling effort
applied (Evans, Li, et al., 2017). Amplicon lengths for the MiFish and FishCYTB amplicons were
nearly identical (Table 3. 1), the diversity of the species assemblage in the Williston Watershed is
relatively low, and the sampling effort applied in this study was small relative to the size of the
reservoir. The results of this study are exploratory and not comparable to those from larger applied
studies with more comprehensive sample designs and more sophisticated bioinformatics analysis.
However, the same methods were applied with the MiFish and FishCYTB assay in this study and
their performance can be evaluated relative to one another.
A significant shortfall of both the MiFish and FishCYTB assays was the inability to resolve
bull trout and lake trout. Substituting MiFish for an assay that could discriminate between these
two species would improve the overall performance of eDNA metabarcoding for the Williston
Watershed, but this level of species resolution may not be possible with any metabarcoding
marker. Metabarcoding primers are usually designed for highly conserved genes to minimize
primer mismatches and maximize species coverage, often at the expense of specificity. It is
important to note, however, that an eDNA metabarcoding assay includes the entire workflow from
field sampling to bioinformatics analysis and the sensitivity and specificity of an assay is
determined by more than just the primers used (Deiner et al., 2017; Ficetola et al., 2015; Goldberg
et al., 2016). The CYTB gene has a relatively high degree of interspecific species diversity and
was a good candidate for differentiating between lake trout and bull trout. The CYTB gene was

95

chosen as the best candidate for that purpose when I designed a species-specific lake trout assay
for the Williston Reservoir (Chapter 2), although it was still quite challenging. The FishCYTB
assay has proven effective enough to use on its own to detect at least 15 of the 17 species that exist
in the tributaries of the Williston Reservoir, and species-specific qPCR methods appear to be better
suited to differentiating between such closely related species as bull trout and lake trout.
Metabarcoding vs. qPCR
eDNA metabarcoding is particularly well suited to large-scale inventories of streams in the
Williston Watershed and could efficiently characterize changes in community structure across
space and time. Compared to traditional sampling methods, eDNA metabarcoding can be cost
effectively scaled up to inventory vast geographic ranges with minimal effort, has increased
sensitivity for detecting rare and cryptic species, and is easily adapted to all habitat types (Jerde et
al., 2013; Olds et al., 2016; Sard et al., 2019). Although I only tested a small number of samples
from 13 streams and rivers around the reservoir, the detection results have revealed interesting
patterns in species distribution and co-occurrence (Table 3. 6). Species occurrence data for entire
fish communities provided by eDNA metabarcoding can be used to screen for critical habitats,
biodiversity hotspots, and species interactions that may provide more insights to fisheries
managers than species-specific approaches (Balasingham et al., 2018). Monitoring with eDNA
metabarcoding is also well suited to the early detection of invasive species (Jerde et al., 2013). The
Williston Watershed is a massive geographic region that has changed dramatically since the
creation of the reservoir and will continue to change in the future because of climate change,
industrial development, forestry, and angling pressure, among other factors. This study has
provided several insights into the potential for eDNA metabarcoding as a method for large-scale

96

fisheries monitoring in the Williston Watershed as well as many lessons learned that will inform
any future applications of the methodology.
Metabarcoding has advantages over a species-specific approach when the monitoring goal
is to characterise diverse and dynamic fish communities, but species-specific methods appear to
be more appropriate for monitoring focal species. The pelagic fish community in the reservoir is
dominated by only six species, three of which are relatively abundant (kokanee, lake whitefish and
peamouth), and three that are relatively rare (bull trout, lake trout and rainbow trout) (O’Connor,
2022; Plate et al., 2012; Sebastian et al., 2009). Species-specific qPCR assays are preferred for
detecting a small number of rare species because the analysis is faster, not complicated by
bioinformatics pipelines, highly specific, and relatively inexpensive (Bylemans et al., 2019; Morey
et al., 2020). My results from the Williston Reservoir do not show a substantial difference in
sensitivity between methods for the species that were detected, but differentiating between bull
trout and lake trout detections is an important requirement for pelagic species monitoring that was
not addressed well with the metabarcoding assays. The metabarcoding assays tested in this study
appeared to have a lower probability of detecting Arctic grayling in tributary samples, although
sensitivity could be increased by testing more PCR1 replicates (Ficetola et al., 2015). Speciesspecific assays appear to have an advantage over metabarcoding in the relationship they produce
between eDNA concentrations and species abundance or biomass (Rourke et al., 2021), which
would be a useful application for eDNA in the reservoir. Metabarcoding has many potential
applications in the Williston Watershed, but a species-specific approach is better suited to studying
the pelagic species in the Williston Reservoir.

97

EPILOGUE
Assay validation is critical for eDNA studies and even previously developed assays must
be validated for their application in a particular geographic area and with a particular workflow
(Goldberg et al., 2016; Thalinger et al., 2021). My research advanced the validation of speciesspecific qPCR assays for the six prominent pelagic species in the Williston Reservoir and a
metabarcoding assay that was effective in detecting fish in the surrounding streams and rivers. The
assays for kokanee, rainbow trout and bull trout have been used previously in Northern British
Columbia (Dr. Brent Murray, pers. comm.), but the lake whitefish assay from Hernandez et al.
(2020) was only developed recently, and this study may be the first to validate it in western North
American. Novel qPCR assays were developed for lake trout and peamouth to cover all six of the
most abundant pelagic species in the Williston Reservoir but have also addressed challenges for
eDNA that are not unique to this situation. Bull trout and lake trout are high priority species for
fisheries management in the Williston Reservoir (Hagen & Weber, 2019) and throughout western
North America (Donald & Alger, 1993; Fredenberg, 2002; Martinez et al., 2009; Riley et al., 2021;
Species At Risk Public Registry, 2019). The bull trout assay used in this study was previously
validated for specificity with other closely related species, including lake trout (Wilcox et al.,
2013). However, in silico validations of published lake trout assays developed in eastern Canada
(Hernandez et al., 2020; Lacoursière‐Roussel, Côté, et al., 2016) indicated that they would crossamplify bull trout DNA. The development and validation of a lake trout assay that can discriminate
between bull trout is an important achievement for eDNA monitoring in lakes and reservoirs across
western North America where these two species are sympatric. This study found that the
FishCYTB metabarcoding assay could detect at least 15 of the 19 species in the Upper Peace River
Watershed reliably enough that it could be used on its own. Having validated eDNA assays for
species in the Williston Reservoir and the surrounding watershed is a prerequisite for any future
98

eDNA studies and will reduce the barriers to the application of eDNA methods in fisheries
management in the region.
Piloting eDNA in the Williston Reservoir and surrounding watershed has provided
valuable insights into eDNA sampling design and the overall workflow for future eDNA studies.
Detecting and characterizing the distribution of pelagic fish within the Williston Reservoir is
challenging with any means of sampling because of the enormous size of the reservoir and the
relatively low abundance of pelagic fish. Although my research affirmed the sensitivity of eDNA,
my results demonstrated that pelagic eDNA concentrations are very low, especially for rare species
like bull trout and lake trout. Filtering larger volumes of water is recommended to increase
detection confidence and minimize false negative results, but the methods used in my research
have practical limits. Collecting and storing large volumes of water for filtration on shore can be
logistically challenging and filter efficiency can be reduced quickly as particles accumulate and
reduce filter porosity. The high-volume eDNA water sampling method developed by Schabacker
et al. (2020) is an alternative to traditional eDNA sampling techniques that could address the
challenges with sampling logistics and eDNA sensitivity in the Williston Reservoir, and has the
potential to increase pelagic eDNA sampling efficiency.
By modeling my sampling design and eDNA analysis on those published by Littlefair et
al. (2020), I was able to analyze the vertical distribution of eDNA in the Williston Reservoir during
summer stratification. Seasonal thermal stratification and the thermal preferences of fish species
influences the distribution of eDNA in lakes (Klobucar et al., 2017; Littlefair et al., 2020).
Although a stratified water column can limit the transportation and distribution of eDNA, many of
the pelagic species in the Williston Reservoir exhibit diel vertical migrations. This resulted in
eDNA detections above and below the thermocline for species who used habitats throughout the
99

water column. eDNA appeared to be distributed more homogeneously and abundantly above the
thermocline, particularly for the relatively abundant planktivorous species. Piscivorous species
were relatively rare and were not detected frequently enough to make any assumptions about their
vertical distribution. I suggested that the distribution of eDNA below the thermocline is likely to
be patchier but may provide a more precise representation of fish distributions due to the reduced
potential for transportation relative to the wind mixed layers above the thermocline. Most of the
pelagic species in the reservoir appear to be more detectable above the thermocline during summer
stratification, while sampling below the thermocline is likely to provide more useful information
about the distribution of lake trout. The distribution of eDNA for all species is likely to be more
homogenous and easily detectable from a single depth when the reservoir water column is
isothermal in spring and fall (Klobucar et al., 2017; Lacoursière‐Roussel, Côté, et al., 2016;
Littlefair et al., 2020). Sampling during those periods would require less effort to detect all species,
but it may not be possible to draw as many conclusions about distribution.
eDNA has the potential to supplement traditional methods of fisheries inventory in the
Williston Reservoir and may be a cost-effective and efficient solution for monitoring rare species
across extremely large geographic areas. eDNA can detect rare species with greater efficiency and
sensitivity compared to conventional sampling gears such as gill netting (Evans, Shirey, et al.,
2017; Jerde et al., 2011; Thomsen et al., 2012) and has proven to be particularly useful for
monitoring lake trout (Lacoursière‐Roussel, Rosabal, et al., 2016; Littlefair et al., 2020).
Developing a cost-effective method for monitoring of lake trout is a priority action for the
Williston Reservoir (Fish and Wildlife Compensation Program, 2020) and traditional methods for
monitoring lake trout are unrealistic for this waterbody. Two common survey methods for
monitoring lake trout include Spring Littoral Index Netting (SLIN) and Summer Profundal Index

100

Netting (SPIN) (Jessup & Millar, 2011). Both methods would require hundreds of gillnet sets on
a waterbody the size of the Williston Reservoir. Gillnetting in the Williston Reservoir is
challenging due to several factors that include the flooded forests on the bottom of the reservoir
that are difficult to anchor in and can entangle nets, and extreme weather that can be difficult to
impossible to work in safely. Gillnetting at such high intensity across the reservoir would also
result in large numbers of fish mortalities. A great deal of recent eDNA research has focused on
understanding the relationship between the concentration of eDNA and the abundance of fish to
develop eDNA as a method of stock assessment (Rourke et al., 2021). Several studies have
demonstrated that eDNA particle concentrations are positively correlated with species abundance
(Klobucar et al., 2017; Lacoursière‐Roussel, Rosabal, et al., 2016; Sard et al., 2019; Shelton et al.,
2019; Stoeckle et al., 2017; Takahara et al., 2012; Yates et al., 2020). eDNA research is actively
pursuing the development of protocols that will provide reliable estimations of abundance (Rourke
et al., 2021; Schenekar, 2023). Once achieved, eDNA will provide a rapid, cost-effective estimate
of abundance for fisheries stock assessment in large waterbodies like the Williston Reservoir,
where traditional means of stock assessment are challenging, if not impossible.
Although methods are still being developed, standardized, and refined, eDNA is a valuable
tool with current utility and incredible potential for fisheries monitoring in the Williston
Watershed. eDNA has been used successfully for monitoring Arctic grayling in the tributaries of
the Williston Reservoir and has helped to better understand the extent of their current distribution
(Stamford et al., 2020, 2022; Stamford & O’Connor, 2022). eDNA could be used in a similar way
to effectively monitor other rare or important species but would not necessarily require the same
sampling effort. eDNA analyses consume small amounts of the total sample collected and any
unused volumes can be stored for long periods and analyzed for other purposes later. My

101

metabarcoding analysis reused several of the eDNA samples collected for Arctic grayling research,
highlighting the potential cost savings and efficiency of eDNA for addressing multiple research
objectives. eDNA has been collected from many of the streams from around the Williston
Reservoir and reanalyzing those samples with a metabarcoding assay, like in this study, would be
an effective means of conducting a large-scale inventory of tributary streams without any
additional sampling costs. Expanding further, standardizing eDNA sampling throughout the Upper
Peace River Watershed would be beneficial as a means of addressing multiple monitoring
objectives. eDNA is revolutionizing biodiversity monitoring in aquatic environments and future
applications of this technology are likely to become standard practices (Schenekar, 2023). This
study has laid the groundwork for the future of eDNA in the Upper Peace River Watershed and
will help to facilitate its adoption as eDNA-based methods are added to the standard toolbox for
fisheries management.
As the number and magnitude of threats to aquatic environments increases, it is imperative
that technologies for monitoring aquatic species evolve to identify problems and solutions more
efficiently. Applications of eDNA represent one of the most significant recent advances in aquatic
species monitoring that has many advantages over traditional survey methods (Jerde, 2019;
Schenekar, 2023). eDNA is a rapidly growing field of research and is increasingly becoming a
standard tool for biodiversity monitoring around the world (Schenekar, 2023). In Canada, research
groups such as the Genomic Network for Fish Identification, Stress and Health (GEN-FISH, 2020)
are developing and validating eDNA “toolkits” that will make eDNA methods quick, easy,
reliable, inexpensive and broadly applicable across Canada for all freshwater species. Although
eDNA methods are unlikely to completely replace traditional sampling methods, they are currently
able to provide a substantial amount of the information required for a “complete picture” of a fish

102

population with much fewer resources. eDNA can identify species composition and spatial
distribution, but traditional methods that capture individual fish will likely always be required to
gather information about fish size and age. A great deal of research has focused on understanding
how eDNA can provide a reliable estimate of fish abundance and biomass (Rourke et al., 2021),
and it may only be a matter of time before this is achieved. eDNA and similar genetic methods for
species identification have been broadly integrated into biodiversity monitoring and conservation
programs across Canada. For example, Parks Canada uses eDNA to monitor aquatic species across
the country (Parks Canada Agency, 2023) and programs like the Canadian Aquatic Biomonitoring
Network are implementing a technique similar to metabarcoding to replace traditional taxonomic
analysis (Canada, 2014). Like any developing technology, eDNA will continue to improve and
new applications will be developed, but it is clearly a technology that can solve many of the
challenges that fisheries managers are facing today.

103

BIBLIOGRAPHY
Balasingham, K. D., Walter, R. P., Mandrak, N. E., & Heath, D. D. (2018). Environmental DNA detection of
rare and invasive fish species in two Great Lakes tributaries. Molecular Ecology, 27(1), 112–127.
https://doi.org/10.1111/mec.14395
Barnes, M. A., & Turner, C. R. (2016). The ecology of environmental DNA and implications for
conservation genetics. Conservation Genetics, 17(1), 1–17. https://doi.org/10.1007/s10592-0150775-4
Barnes, M. A., Turner, C. R., Jerde, C. L., Renshaw, M. A., Chadderton, W. L., & Lodge, D. M. (2014).
Environmental conditions influence eDNA persistence in aquatic systems. Environmental Science
& Technology, 48(3), 1819–1827. https://doi.org/10.1021/es404734p
Barrett, D. T., & Halsey, T. G. (1975). Fishery resources of Williston Reservoir: A preliminary report
(54411).
https://a100.gov.bc.ca/pub/acat/documents/r54411/Fishery_Williston_1531173190351_117089
8330.pdf
B.C. Conservation Data Center. (2021). Species Summary: Salvelinus confluentus pop. 10, bull trout—
Western Arctic populations. B.C. Ministry of Environment.
https://a100.gov.bc.ca/pub/eswp/speciesSummary.do?id=312190
Beng, K. C., & Corlett, R. T. (2020). Applications of environmental DNA (eDNA) in ecology and
conservation: Opportunities, challenges and prospects. Biodiversity and Conservation, 29(7),
2089–2121. https://doi.org/10.1007/s10531-020-01980-0
Bethke, E., Arrhenius, F., Cardinale, M., & Håkansson, N. (1999). Comparison of the selectivity of three
pelagic sampling trawls in a hydroacoustic survey. Fisheries Research, 44(1), 15–23.
https://doi.org/10.1016/S0165-7836(99)00054-5
Bio-Rad Labratories, Inc. (2022). Droplet digitial PCR applications guide. https://www.biorad.com/webroot/web/pdf/lsr/literature/Bulletin_6407.pdf
Blackman, B. G. (1992). Fisheries resources of Williston Reservoir twenty years after impoundment (239;
p. 35). Peace/Williston Fish and Wildlife Compensation Program.
https://a100.gov.bc.ca/pub/acat/documents/r20547/Blackman1992pwfwcp_report_no_239_130
6777274296_dcd031b80f9bf542c185a57ea06fab0ea4cc48cc03ec8e9efcc883468de66e88.pdf
Blackman, B. G. (2002). The distribution and relative abundance of Arctic grayling (Thymallus arcticus) in
the Parsnip, Anzac and Rable Rivers (PWFWCP Report No. 254).
https://a100.gov.bc.ca/pub/acat/documents/r35844/pwfwcp_report_no_254_1363355935405_e
2062bc3ef9224c80e01f7c590775e3a71bd5a459de5bd248f57e7106b055d6b.pdf
Bolyen, E., Rideout, J. R., Dillon, M. R., Bokulich, N. A., Abnet, C. C., Al-Ghalith, G. A., Alexander, H., Alm,
E. J., Arumugam, M., Asnicar, F., Bai, Y., Bisanz, J. E., Bittinger, K., Brejnrod, A., Brislawn, C. J.,
Brown, C. T., Callahan, B. J., Caraballo-Rodríguez, A. M., Chase, J., … Caporaso, J. G. (2019).
Reproducible, interactive, scalable and extensible microbiome data science using QIIME 2. Nature
Biotechnology, 37(8), Article 8. https://doi.org/10.1038/s41587-019-0209-9

104

Burns, M. J., Burrell, A. M., & Foy, C. A. (2010). The applicability of digital PCR for the assessment of
detection limits in GMO analysis. European Food Research and Technology, 231(3), 353–362.
https://doi.org/10.1007/s00217-010-1277-8
Bustin, S., & Huggett, J. (2017). qPCR primer design revisited. Biomolecular Detection and Quantification,
14, 19–28. https://doi.org/10.1016/j.bdq.2017.11.001
Bylemans, J., Gleeson, D. M., Duncan, R. P., Hardy, C. M., & Furlan, E. M. (2019). A performance
evaluation of targeted eDNA and eDNA metabarcoding analyses for freshwater fishes.
Environmental DNA, 1(4), 402–414. https://doi.org/10.1002/edn3.41
Callahan, B. J., McMurdie, P. J., Rosen, M. J., Han, A. W., Johnson, A. J. A., & Holmes, S. P. (2016).
DADA2: High-resolution sample inference from Illumina amplicon data. Nature Methods, 13(7),
Article 7. https://doi.org/10.1038/nmeth.3869
Canada, E. and C. C. (2014, August 12). CABIN news: Assessing freshwater health through DNA and
community based moinitoring [Guidance]. https://www.canada.ca/en/environment-climatechange/services/canadian-aquatic-biomonitoring-network/news-events.html
Canals, O., Mendibil, I., Santos, M., Irigoien, X., & Rodríguez-Ezpeleta, N. (2021). Vertical stratification of
environmental DNA in the open ocean captures ecological patterns and behavior of deep-sea
fishes. Limnology and Oceanography Letters, n/a(n/a). https://doi.org/10.1002/lol2.10213
Chabot, D., Steffensen, J. F., & Farrell, A. P. (2016). The determination of standard metabolic rate in
fishes. Journal of Fish Biology, 88(1), 81–121. https://doi.org/10.1111/jfb.12845
Chambert, T., Pilliod, D. S., Goldberg, C. S., Doi, H., & Takahara, T. (2018). An analytical framework for
estimating aquatic species density from environmental DNA. Ecology and Evolution, 8(6), 3468–
3477. https://doi.org/10.1002/ece3.3764
Clark, C. W., & Levy, D. A. (1988). Diel vertical migrations by juvenile sockeye salmon and the
antipredation window. The American Naturalist, 131(2), 271–290.
Clarke, A. D., Telmer, K. H., & Shrimpton, J. M. (2007). Habitat use and movement patterns for a fluvial
species, the Arctic grayling, in a watershed impacted by a large reservoir: Evidence from otolith
microchemistry. Journal of Applied Ecology, 44(6), 1156–1165. https://doi.org/10.1111/j.13652664.2007.01350.x
Cowie, D. M., & Blackman, B. G. (2003). 2001-2002 Arctic grayling (Thymallus arcticus) fry surveys in the
Omineca and Osilinka Rivers (PWFWCP Report No. 276).
Cowie, D. M., & Blackman, B. G. (2004). 2003 Actic grayling (Thymallus arcticus) fry surveys in the
Ingenika River (PWFWCP Report No. 291; p. 19).
Coxson, D. S., Huber, D. P. W., Shrimpton, J. M., Theilman, A. C., & Wiensczyk, A. (2018). Ecosystem
impact of nutrient enrichment by Kokanee in the Williston Reservoir Watershed (Pea-F18-F-2296).
Natural Resources and Environmental Studies Institute.
http://a100.gov.bc.ca/appsdata/acat/documents/r55493/PEA_F18_F_2296_1547227412356_722
4430300.pdf

105

Cristescu, M., & Hebert, P. (2018). Uses and misuses of environmental DNA in biodiversity science and
conservation. Annual Review of Ecology, Evolution, and Systematics, 49.
https://doi.org/10.1146/annurev-ecolsys-110617-062306
Culling, B., & Euchner, T. (2018). Peace Reach lake trout movements—Year 2 (PEA-F18-F-2313; p. 30).
Fish and Wildlife Compensation Program - Peace.
Culling, B., & Euchner, T. (2019). Peace Reach lake trout movements—Year 3 (Peace Project No. PEAF19-F-2622; p. 27). Fish and Wildlife Compensation Program - Peace.
Currier, C. A., Morris, T. J., Wilson, C. C., & Freeland, J. R. (2018). Validation of environmental DNA
(eDNA) as a detection tool for at-risk freshwater pearly mussel species (Bivalvia: Unionidae).
Aquatic Conservation: Marine and Freshwater Ecosystems, 28(3), 545–558.
https://doi.org/10.1002/aqc.2869
Deagle, B. E., Jarman, S. N., Coissac, E., Pompanon, F., & Taberlet, P. (2014). DNA metabarcoding and the
cytochrome c oxidase subunit I marker: Not a perfect match. Biology Letters, 10(9), 20140562.
https://doi.org/10.1098/rsbl.2014.0562
Deiner, K., Bik, H. M., Mächler, E., Seymour, M., Lacoursière-Roussel, A., Altermatt, F., Creer, S., Bista, I.,
Lodge, D. M., Vere, N. de, Pfrender, M. E., & Bernatchez, L. (2017). Environmental DNA
metabarcoding: Transforming how we survey animal and plant communities. Molecular Ecology,
26(21), 5872–5895. https://doi.org/10.1111/mec.14350
dMIQE Group. (2020). The digital MIQE guidelines update: Minimum information for publication of
quantitative digital PCR experiments for 2020. Clinical Chemistry, 66(8), 1012–1029.
https://doi.org/10.1093/clinchem/hvaa125
Doi, H., Uchii, K., Takahara, T., Matsuhashi, S., Yamanaka, H., & Minamoto, T. (2015). Use of droplet
digital PCR for estimation of fish abundance and biomass in environmental DNA surveys. PLOS
ONE, 10(3), e0122763. https://doi.org/10.1371/journal.pone.0122763
Donald, D. B., & Alger, D. J. (1993). Geographic distribution, species displacement, and niche overlap for
lake trout and bull trout in mountain lakes. Canadian Journal of Zoology, 71(2), 238–247.
https://doi.org/10.1139/z93-034
DWB Consulting Services Ltd. (2019). Williston Watershed kokanee spawner distribution and aerial
enumeration surveys (2018) (PEA-F19-F-2895-DC-103364). Fish and Wildlife Compensation
Program - Peace Region.
DWB Consulting Services Ltd. (2020). Williston Watershed kokanee spawner distribution and aerial
enumeration survey (2019) (PEA-F20-F-3359-DC-106374; p. 51).
DWB Consulting Services Ltd. (2021). Williston Watershed kokanee spawner distribution and aerial
enumeration surveys (2020) (PEA-F21-F-3363-DC109837; p. 57). Fish and Wildlife Compensation
Program - Peace.
Eichmiller, J. J., Bajer, P. G., & Sorensen, P. W. (2014). The relationship between the distribution of
common carp and their environmental DNA in a small lake. PLOS ONE, 9(11), e112611.
https://doi.org/10.1371/journal.pone.0112611

106

Evans, N. T., Li, Y., Renshaw, M. A., Olds, B. P., Deiner, K., Turner, C. R., Jerde, C. L., Lodge, D. M.,
Lamberti, G. A., & Pfrender, M. E. (2017). Fish community assessment with eDNA metabarcoding:
Effects of sampling design and bioinformatic filtering. Canadian Journal of Fisheries & Aquatic
Sciences, 74(9), 1362–1374. https://doi.org/10.1139/cjfas-2016-0306
Evans, N. T., Olds, B. P., Renshaw, M. A., Turner, C. R., Li, Y., Jerde, C. L., Mahon, A. R., Pfrender, M. E.,
Lamberti, G. A., & Lodge, D. M. (2016). Quantification of mesocosm fish and amphibian species
diversity via environmental DNA metabarcoding. Molecular Ecology Resources, 16(1), 29–41.
https://doi.org/10.1111/1755-0998.12433
Evans, N. T., Shirey, P. D., Wieringa, J. G., Mahon, A. R., & Lamberti, G. A. (2017). Comparative cost and
effort of fish distribution detection via environmental DNA analysis and electrofishing. Fisheries,
42(2), 90–99. https://doi.org/10.1080/03632415.2017.1276329
Ficetola, G. F., Pansu, J., Bonin, A., Coissac, E., Giguet-Covex, C., De Barba, M., Gielly, L., Lopes, C. M.,
Boyer, F., Pompanon, F., Rayé, G., & Taberlet, P. (2015). Replication levels, false presences and the
estimation of the presence/absence from eDNA metabarcoding data. Molecular Ecology
Resources, 15(3), 543–556. https://doi.org/10.1111/1755-0998.12338
Fish and Wildlife Compensation Program. (2020). Peace Region rivers, lakes and reservoirs action plan.
https://fwcp.ca/app/uploads/2020/08/Action-Plan-Peace-Region-Rivers-Lakes-Reservoirs-Aug-112020.pdf
Fredenberg, W. (2002). Further evidence that lake trout displace bull trout in mountain lakes.
Intermountain Journal of Science, 8(3), 10.
Fukaya, K., Murakami, H., Yoon, S., Minami, K., Osada, Y., Yamamoto, S., Masuda, R., Kasai, A.,
Miyashita, K., Minamoto, T., & Kondoh, M. (2018). Estimating fish population abundance by
integrating quantitative data on environmental DNA and hydrodynamic modelling. Molecular
Ecology, n/a(n/a). https://doi.org/10.1111/mec.15530
GEN-FISH. (2020). About GEN-FISH. GEN-FISH. https://gen-fish.ca/about/
Goldberg, C. S., Turner, C. R., Deiner, K., Klymus, K. E., Thomsen, P. F., Murphy, M. A., Spear, S. F.,
McKee, A., Oyler-McCance, S. J., Cornman, R. S., Laramie, M. B., Mahon, A. R., Lance, R. F., Pilliod,
D. S., Strickler, K. M., Waits, L. P., Fremier, A. K., Takahara, T., Herder, J. E., & Taberlet, P. (2016).
Critical considerations for the application of environmental DNA methods to detect aquatic
species. Methods in Ecology and Evolution, 7(11), 1299–1307. https://doi.org/10.1111/2041210X.12595
Gorsky, D., Zydlewski, J., & Basley, D. (2012). Characterizing seasonal habitat use and diel vertical activity
of lake whitefish in Clear Lake, Maine, as determined with acoustic telemetry. Transactions of the
American Fisheries Society, 141(3), 761–771. https://doi.org/10.1080/00028487.2012.675905
Hagen, J. (2000). Reproductive isolation between Dolly Varden (salvelinus malma) and bull trout (s.
confluentus) in sympatry: The role of ecological factors [University of British Columbia].
https://doi.org/10.14288/1.0099497

107

Hagen, J., & Spendlow, I. (2019). Bull trout spawner abundance within the Williston Reservoir Watershed
2018, and critical spawning habitats of the Omineca River Watershed (FWCP Project No. PEA-F19F-2626).
Hagen, J., Spendlow, I., & Pillipow, R. (2020). Critical spawning habitats and abundance of bull trout in
the Williston Reservoir watershed, 2019 (PEA-F20-F-2956; p. 57). Fish and Wildlife Compensation
Program - Peace.
Hagen, J., & Weber, S. (2019). Limiting factors, enhancement potential, critical habitats, and
conservation status for bull trout of the Williston Reservoir Watershed: Information synthesis and
recommended monitoring framework (p. 153). FWCP.
Hänfling, B., Lawson Handley, L., Read, D. S., Hahn, C., Li, J., Nichols, P., Blackman, R. C., Oliver, A., &
Winfield, I. J. (2016). Environmental DNA metabarcoding of lake fish communities reflects longterm data from established survey methods. Molecular Ecology, 25(13), 3101–3119.
https://doi.org/10.1111/mec.13660
Hansen, B. K., Bekkevold, D., Clausen, L. W., & Nielsen, E. E. (2018). The sceptical optimist: Challenges
and perspectives for the application of environmental DNA in marine fisheries. Fish and Fisheries,
19(5), 751–768. https://doi.org/10.1111/faf.12286
Hardiman, J. M., Johnson, B. M., & Martinez, P. J. (2004). Do predators influence the distribution of age0 kokanee in a Colorado reservoir? Transactions of the American Fisheries Society, 133(6), 1366–
1378. https://doi.org/10.1577/T03-1234.1
Harris, S. L., Langston, A. R., Stockner, J. G., & Vidmanic, L. (2006). A limnological assessment of four
Williston reservoir embayments, 2004. (p. 62).
Hernandez, C., Bougas, B., Perreault-Payette, A., Simard, A., Côté, G., & Bernatchez, L. (2020). 60 specific
eDNA qPCR assays to detect invasive, threatened, and exploited freshwater vertebrates and
invertebrates in Eastern Canada. Environmental DNA, 2(3), 373–386.
https://doi.org/10.1002/edn3.89
Hunter, M. E., Dorazio, R. M., Butterfield, J. S. S., Meigs-Friend, G., Nico, L. G., & Ferrante, J. A. (2017).
Detection limits of quantitative and digital PCR assays and their influence in presence–absence
surveys of environmental DNA. Molecular Ecology Resources, 17(2), 221–229.
https://doi.org/10.1111/1755-0998.12619
Integrated DNA Technologies. (2021). Using OligoAnalyzer Tool to analyze hairpins and dimers | IDT.
Integrated DNA Technologies. https://www.idtdna.com/pages/support/faqs/how-do-i-use-theoligoanalyzer-tool-to-analyze-possible-hairpins-and-dimers-formed-by-my-oligo
Jane, S. F., Wilcox, T. M., McKelvey, K. S., Young, M. K., Schwartz, M. K., Lowe, W. H., Letcher, B. H., &
Whiteley, A. R. (2015). Distance, flow and PCR inhibition: eDNA dynamics in two headwater
streams. Molecular Ecology Resources, 15(1), 216–227. https://doi.org/10.1111/1755-0998.12285
Jerde, C. L. (2019). Can we manage fisheries with the inherent uncertainty from eDNA? Journal of Fish
Biology, 1–13. https://doi.org/10.1111/jfb.14218

108

Jerde, C. L., Chadderton, W. L., Mahon, A. R., Renshaw, M. A., Corush, J., Budny, M. L., Mysorekar, S., &
David M., L. (2013). Detection of Asian carp DNA as part of a Great Lakes basin-wide surveillance
program. Canadian Journal of Fisheries and Aquatic Sciences. https://doi.org/10.1139/cjfas-20120478
Jerde, C. L., Mahon, A. R., Chadderton, W. L., & Lodge, D. M. (2011). “Sight-unseen” detection of rare
aquatic species using environmental DNA. Conservation Letters, 4(2), 150–157.
https://doi.org/10.1111/j.1755-263X.2010.00158.x
Jessup, L., & Millar, N. (2011). Application of a new method for monitoring lake trout abundance in
Yukon: Summer profundal index netting (SPIN) (Yukon Fish and Wildlife Branch Report TR-11-11).
https://open.yukon.ca/sites/default/files/new_method_to_monitor_lake_trout_SPIN_1.pdf
Johnson, B. M., & Martinez, P. J. (2000). Trophic economics of lake trout management in Reservoirs of
differing productivity. North American Journal of Fisheries Management, 20(1), 127–143.
https://doi.org/10.1577/1548-8675(2000)020<0127:TEOLTM>2.0.CO;2
Klobucar, S. L., Rodgers, T. W., & Budy, P. (2017). At the forefront: Evidence of the applicability of using
environmental DNA to quantify the abundance of fish populations in natural lentic waters with
additional sampling considerations. Canadian Journal of Fisheries and Aquatic Sciences.
https://doi.org/10.1139/cjfas-2017-0114
Klymus, K. E., Merkes, C. M., Allison, M. J., Goldberg, C. S., Helbing, C. C., Hunter, M. E., Jackson, C. A.,
Lance, R. F., Mangan, A. M., Monroe, E. M., Piaggio, A. J., Stokdyk, J. P., Wilson, C. C., & Richter, C.
A. (2019). Reporting the limits of detection and quantification for environmental DNA assays.
Environmental DNA, 2(3), 271–282. https://doi.org/10.1002/edn3.29
Klymus, K. E., Richter, C. A., Chapman, D. C., & Paukert, C. (2015). Quantification of eDNA shedding rates
from invasive bighead carp Hypophthalmichthys nobilis and silver carp Hypophthalmichthys
molitrix. Biological Conservation, 183, 77–84. https://doi.org/10.1016/j.biocon.2014.11.020
Klymus, K. E., Ruiz Ramos, D. V., Thompson, N. L., & Richter, C. A. (2020). Development and testing of
species-specific quantitative PCR assays for environmental DNA applications. Journal of Visualized
Experiments, 165, 61825. https://doi.org/10.3791/61825
Kubečka, J., Hohausová, E., Matěna, J., Peterka, J., Amarasinghe, U. S., Bonar, S. A., Hateley, J., Hickley,
P., Suuronen, P., Tereschenko, V., Welcomme, R., & Winfield, I. J. (2009). The true picture of a lake
or reservoir fish stock: A review of needs and progress. Fisheries Research, 96(1), 1–5.
https://doi.org/10.1016/j.fishres.2008.09.021
Kutyavin, I. V., Afonina, I. A., Mills, A., Gorn, V. V., Lukhtanov, E. A., Belousov, E. S., Singer, M. J.,
Walburger, D. K., Lokhov, S. G., Gall, A. A., Dempcy, R., Reed, M. W., Meyer, R. B., & Hedgpeth, J.
(2000). 3′-Minor groove binder-DNA probes increase sequence specificity at PCR extension
temperatures. Nucleic Acids Research, 28(2), 655–661.
Lacoursière-Roussel, A., Côté, G., Leclerc, V., & Bernatchez, L. (2016). Quantifying relative fish
abundance with eDNA: A promising tool for fisheries management. Journal of Applied Ecology,
53(4), 1148–1157. https://doi.org/10.1111/1365-2664.12598

109

Lacoursière-Roussel, A., Rosabal, M., & Bernatchez, L. (2016). Estimating fish abundance and biomass
from eDNA concentrations: Variability among capture methods and environmental conditions.
Molecular Ecology Resources, 16(6), 1401–1414. https://doi.org/10.1111/1755-0998.12522
Langlois, V. S., Allison, M. J., Bergman, L. C., To, T. A., & Helbing, C. C. (2021). The need for robust qPCRbased eDNA detection assays in environmental monitoring and species inventories. Environmental
DNA, 3(3), 519–527. https://doi.org/10.1002/edn3.164
Langston, A. R. (2012). Williston Watershed kokanee spawner distribution and enumeration surveys
(2002 – 2006) (Peace Region Report No. 357; p. 22). FWCP.
Langston, A. R., & Blackman, B. G. (1993). Fisheries resources and enhancement potentials of selected
tributaries of the Williston Reservoir Volume II (Report No. 70; p. 185). Peace/Williston Fish and
Wildlife Compensation Program.
https://a100.gov.bc.ca/pub/acat/documents/r35366/pwfwcp_report_no_070_1360354625130_a
b5d1d2216f01b7cbdb42136e10249e8fdb1d3d173d713a69571c05c9f2d0dea.pdf
Langston, A. R., & Murphy, E. B. (2008). The history of fish introductions (to 2005) in the Peace/Williston
Fish and Wildlife Compensation Program area. (PWFWCP Report No. 325; p. 62). FWCP.
Littlefair, J. E., Hrenchuk, L. E., Blanchfield, P. J., Rennie, M. D., & Cristescu, M. E. (2020). Thermal
stratification and fish thermal preference explain vertical eDNA distributions in lakes. bioRxiv,
2020.04.21.042820. https://doi.org/10.1101/2020.04.21.042820
Mandrak, N. E., Curry, A. R., Dumont, P., Reist, J. D., Taylor, E. B., & Watkinson, D. A. (Eds.). (2023).
Freshwater fisheries in Canada: Historical and contemporary perspectives on the resources and
their management, Chapter 1. In Freshwater Fisheries in Canada: Historical and Contemporary
Perspectives on the Resources and Their Management. American Fisheries Society.
https://doi.org/10.47886/9781934874707.ch1
Martin, M. (2011). Cutadapt removes adapter sequences from high-throughput sequencing reads.
EMBnet.Journal, 17(1), Article 1. https://doi.org/10.14806/ej.17.1.200
Martinez, P. J., Bigelow, P. E., Deleray, M. A., Fredenberg, W. A., Hansen, B. S., Horner, N. J., Lehr, S. K.,
Schneidervin, R. W., Tolentino, S. A., & Viola, A. E. (2009). Western lake trout woes. Fisheries,
34(9), 424–442. https://doi.org/10.1577/1548-8446-34.9.424
McPhail, J. D. (2007). The freshwater fishes of British Columbia. University of Alberta Press.
McPhail, J. D., & Baxter, J. S. (1996). A review of bull trout (Salvelinus confluentus) life-history and habitat
use in relation to compensation and improvement opportunities (Fisheries Management Report
No. 104, 35 p.; p. 39). Department of Zoology, U.B.C.
Miya, M., Sato, Y., Fukunaga, T., Sado, T., Poulsen, J. Y., Sato, K., Minamoto, T., Yamamoto, S.,
Yamanaka, H., Araki, H., Kondoh, M., & Iwasaki, W. (2015). MiFish, a set of universal PCR primers
for metabarcoding environmental DNA from fishes: Detection of more than 230 subtropical
marine species. Royal Society Open Science, 2(7), 150088. https://doi.org/10.1098/rsos.150088

110

Morey, K. C., Bartley, T. J., & Hanner, R. H. (2020). Validating environmental DNA metabarcoding for
marine fishes in diverse ecosystems using a public aquarium. Environmental DNA, 2(3), 330–342.
https://doi.org/10.1002/edn3.76
Murray, B. W. (2021). FWCP Report on developing an aquatic survey eDNA degradation rate assay (PEAF21-F-3217).
O’Connor, B. (2022). Williston pelagic fish survey gillnetting summary 2021 (PEA-F22-F-3587-DCA;
Prepared for the Peace Region Fish and Wildlife Compensation Program, p. 13).
Olds, B. P., Jerde, C. L., Renshaw, M. A., Li, Y., Evans, N. T., Turner, C. R., Deiner, K., Mahon, A. R.,
Brueseke, M. A., Shirey, P. D., Pfrender, M. E., Lodge, D. M., & Lamberti, G. A. (2016). Estimating
species richness using environmental DNA. Ecology and Evolution, 6(12), 4214–4226.
https://doi.org/10.1002/ece3.2186
Olin, M., Malinen, T., & Ruuhijärvi, J. (2009). Gillnet catch in estimating the density and structure of fish
community—Comparison of gillnet and trawl samples in a eutrophic lake. Fisheries Research,
96(1), 88–94. https://doi.org/10.1016/j.fishres.2008.09.007
Parks Canada Agency, G. of C. (2023, July 24). Environmental DNA at Parks Canada—Nature and science.
https://parks.canada.ca/nature/science/especes-species/adne-edna
Pillipow, R. A., & Langston, A. R. (2002). Williston Reservoir fish assessments 2000, pelagic netting
summary (PWFWCP Report No. 261; p. 47).
Plate, E. M., Bocking, R. C., & Degan, D. J. (2012). Williston fish index in the vicinity of W.A.C. Bennett
Dam. Peace Water Use Plan: Monitoring programs for the Peace spill protocol (GMSMON-4 WAC
Bennett Dam Entrainment Study) (p. 61). BC Hydro.
Pont, D., Rocle, M., Valentini, A., Civade, R., Jean, P., Maire, A., Roset, N., Schabuss, M., Zornig, H., &
Dejean, T. (2018). Environmental DNA reveals quantitative patterns of fish biodiversity in large
rivers despite its downstream transportation. Scientific Reports, 8(1), Article 1.
https://doi.org/10.1038/s41598-018-28424-8
Prediger, E. (2013). How to design primers and probes for PCR and qPCR | IDT. Integrated DNA
Technologies. https://www.idtdna.com/pages/education/decoded/article/designing-pcr-primersand-probes
Province of British Columbia. (2020). Habitat wizard lakes report: Williston Lake.
Qiagen. (2020). DNeasy blood & tissue handbook—QIAGEN.
https://www.qiagen.com/us/resources/resourcedetail?id=68f29296-5a9f-40fa-8b3d1c148d0b3030&lang=en
Rasmussen Hellberg, R. S., Morrissey, M. T., & Hanner, R. H. (2010). A multiplex PCR method for the
identification of commercially important salmon and trout species (Oncorhynchus and Salmo) in
North America. Journal of Food Science, 75(7), C595–C606. https://doi.org/10.1111/j.17503841.2010.01752.x

111

Redenbach, Z., & Taylor, E. B. (2003). Evidence for bimodal hybrid zones between two species of char
(Pisces: Salvelinus) in northwestern North America. Journal of Evolutionary Biology, 16(6), 1135–
1148. https://doi.org/10.1046/j.1420-9101.2003.00619.x
Riley, S., Hansen, M., Krueger, C., Noakes, D., & Muir, A. (2021). The lake charr: Biology, ecology,
distribution, and management (pp. 1–12). https://doi.org/10.1007/978-3-030-62259-6_1
Roberge, M., & Slaney, T. (2001). Life history characteristics of freshwater fishes occuring in British
Columbia, with major emphasis on Lake habitat requirements. Fisheries and Aquatic Sciencs,
2574, 189 pp.
Ross, M. R. (1997). Fisheries conservation and management. Prentice Hall.
Rourke, M. L., Fowler, A. M., Hughes, J. M., Broadhurst, M. K., DiBattista, J. D., Fielder, S., Walburn, J. W.,
& Furlan, E. M. (2021). Environmental DNA (eDNA) as a tool for assessing fish biomass: A review of
approaches and future considerations for resource surveys. Environmental DNA, 00, 1–25.
https://doi.org/10.1002/edn3.185
Rudstam, L. G., Jech, J. M., Parker-Stetter, S. L., Horne, J. K., Sullivan, P. J., & Mason, D. M. (2012).
Chapter 13: Fisheries acoustics. In Fisheries Techniques (3rd ed., pp. 597–636). American Fisheries
Society.
Ruppert, K. M., Kline, R. J., & Rahman, M. S. (2019). Past, present, and future perspectives of
environmental DNA (eDNA) metabarcoding: A systematic review in methods, monitoring, and
applications of global eDNA. Global Ecology and Conservation, 17, e00547.
https://doi.org/10.1016/j.gecco.2019.e00547
Sahu, R., Vishnuraj, M. R., Srinivas, C., Dadimi, B., Megha, G. K., Pollumahanti, N., Malik, S. S.,
Vaithiyanathan, S., Rawool, D. B., & Barbuddhe, S. B. (2021). Development and comparative
evaluation of droplet digital PCR and quantitative PCR for the detection and quantification of
Chlamydia psittaci. Journal of Microbiological Methods, 190, 106318.
https://doi.org/10.1016/j.mimet.2021.106318
Sard, N. M., Herbst, S. J., Nathan, L., Uhrig, G., Kanefsky, J., Robinson, J. D., & Scribner, K. T. (2019).
Comparison of fish detections, community diversity, and relative abundance using environmental
DNA metabarcoding and traditional gears. Environmental DNA, 1(4), 368–384.
https://doi.org/10.1002/edn3.38
Schabacker, J. C., Amish, S. J., Ellis, B. K., Gardner, B., Miller, D. L., Rutledge, E. A., Sepulveda, A. J., &
Luikart, G. (2020). Increased eDNA detection sensitivity using a novel high-volume water sampling
method. Environmental DNA, 2(2), 244–251. https://doi.org/10.1002/edn3.63
Schenekar, T. (2023). The current state of eDNA research in freshwater ecosystems: Are we shifting from
the developmental phase to standard application in biomonitoring? Hydrobiologia, 850(6), 1263–
1282. https://doi.org/10.1007/s10750-022-04891-z
Scheuerell, M. D., & Schindler, D. E. (2003). Diel vertical migration by juvenile sockeye salmon: Empirical
evidence for the antipredation window. Ecology, 84(7), 1713–1720.
https://doi.org/10.1890/0012-9658(2003)084[1713:DVMBJS]2.0.CO;2

112

Schindler, E. U., Sebastian, D., Vidmanic, L., Andrusak, H., Stockner, J., Bassett, M., & Ashley, K. I. (2009).
Arrow Lakes Reservoir fertilization experiment, Year 8 (2006) report (Fisheries Project Report No.
RD 125). Fish and Wildlife Science and Allocation, Ministry of Environment, Province of British
Columbia.
Schloss, P. D., Westcott, S. L., Ryabin, T., Hall, J. R., Hartmann, M., Hollister, E. B., Lesniewski, R. A.,
Oakley, B. B., Parks, D. H., Robinson, C. J., Sahl, J. W., Stres, B., Thallinger, G. G., Van Horn, D. J., &
Weber, C. F. (2009). Introducing mothur: Open-source, platform-independent, communitysupported software for describing and comparing microbial communities. Applied and
Environmental Microbiology, 75(23), 7537–7541. https://doi.org/10.1128/AEM.01541-09
Schönhuth, S., Vukić, J., Šanda, R., Yang, L., & Mayden, R. L. (2018). Phylogenetic relationships and
classification of the Holarctic family Leuciscidae (Cypriniformes: Cyprinoidei). Molecular
Phylogenetics and Evolution, 127, 781–799. https://doi.org/10.1016/j.ympev.2018.06.026
Sebastian, D. C., Andrusak, G., Scholten, G., & Langston, A. (2009). An index of fish distribution and
abundance in Peace Arm of Williston Reservoir based on hydroacoustic and gillnet surveys
(GMSMON #13; p. 98).
Sebastian, D. C., Scholten, G. H., & Woodruff, P. E. (2003). Williston Reservoir fish assessment: Results of
hydroacoustic, trawl and gill net surveys in August, 2000 (p. 72) [Stock Management Report No.
16]. Province of BC.
Shelton, A. O., Kelly, R. P., O’Donnell, J. L., Park, L., Schwenke, P., Greene, C., Henderson, R. A., &
Beamer, E. M. (2019). Environmental DNA provides quantitative estimates of a threatened salmon
species. Biological Conservation, 237, 383–391. https://doi.org/10.1016/j.biocon.2019.07.003
Snyder, M. R., Stepien, C. A., Marshall, N. T., Scheppler, H. B., Black, C. L., & Czajkowski, K. P. (2020).
Detecting aquatic invasive species in bait and pond stores with targeted environmental (e)DNA
high-throughput sequencing metabarcode assays: Angler, retailer, and manager implications.
Biological Conservation, 245, 108430. https://doi.org/10.1016/j.biocon.2020.108430
So, K. Y. K., Fong, J. J., Lam, I. P. Y., & Dudgeon, D. (2020). Pitfalls during in silico prediction of primer
specificity for eDNA surveillance. Ecosphere, 11(7). https://doi.org/10.1002/ecs2.3193
Species At Risk Public Registry. (2019). Species profile: Bull trout western Arctic populations. Government
of Canada. https://wildlife-species.canada.ca/species-riskregistry/species/speciesDetails_e.cfm?sid=1202
Stamford, M., & O’Connor, B. (2022). Fish sampling at eDNA sites for Finlay Reach Arctic grayling (FWCP
Report No. PEA-F22-F-3408).
Stamford, M., Strohm, J., Euchner, T., Pitt, C., & Murray, B. (2022). Williston grayling distribution:
Parsnip, Peace, Dinosaur (FWCP Project Number PEA-F21-F-3198).
Stamford, M., Strohm, J., Pitt, C., & Murray, B. (2020). 2019 Williston Arctic grayling distribution: eDNA
monitoring (PEA-F20-F-2965).
https://a100.gov.bc.ca/pub/acat/documents/r59037/PEA_F20_F_2965_1611779660377_D0D794
3A93.pdf

113

Stockner, J., Langston, A., Sebastian, D., & Wilson, G. (2005). The limnology of Williston Reservoir: British
Columbia’s largest lacustrine ecosystem. Water Quality Research Journal, 40(1), 28–50.
https://doi.org/10.2166/wqrj.2005.003
Stoeckle, M. Y., Soboleva, L., & Charlop-Powers, Z. (2017). Aquatic environmental DNA detects seasonal
fish abundance and habitat preference in an urban estuary. PLOS ONE, 12(4), e0175186.
https://doi.org/10.1371/journal.pone.0175186
Strohm, J., Bonderud, E., & Stamford, M. (2019). 2018 Williston Arctic grayling distribution: eDNA
monitoring (PEA-F19-F-2659; p. 42).
Strohm, J., Hagen, J., & Stamford, M. (2020). Ingenika River Arctic grayling 46 years post-flooding: 2019
snorkeling counts in index sites (FWCP Project No. PEA-F20-F-2963; p. 42). Chu Cho Environmental
LLP.
Takahara, T., Minamoto, T., Yamanaka, H., Doi, H., & Kawabata, Z. (2012). Estimation of fish biomass
using environmental DNA. PLOS ONE, 7(4), e35868.
https://doi.org/10.1371/journal.pone.0035868
Tan, M., & Armbruster, J. W. (2018). Phylogenetic classification of extant genera of fishes of the order
Cypriniformes (Teleostei: Ostariophysi). Zootaxa, 4476(1), Article 1.
https://doi.org/10.11646/zootaxa.4476.1.4
Taylor, E. B., Redenbach, Z., Costello, A. B., Pollard, S. M., & Pacas, C. J. (2001). Nested analysis of genetic
diversity in northwestern North American char, Dolly Varden (Salvelinus malma) and bull trout
(Salvelinus confluentus). Canadian Journal of Fisheries & Aquatic Sciences, 58(2), 406.
https://doi.org/10.1139/f00-262
Thalinger, B., Deiner, K., Harper, L. R., Rees, H. C., Blackman, R. C., Sint, D., Traugott, M., Goldberg, C. S.,
& Bruce, K. (2021). A validation scale to determine the readiness of environmental DNA assays for
routine species monitoring. bioRxiv, 2020.04.27.063990.
https://doi.org/10.1101/2020.04.27.063990
Thomsen, P. F., Kielgast, J., Iversen, L. L., Wiuf, C., Rasmussen, M., Gilbert, M. T. P., Orlando, L., &
Willerslev, E. (2012). Monitoring endangered freshwater biodiversity using environmental DNA.
Molecular Ecology, 21(11), 2565–2573. https://doi.org/10.1111/j.1365-294X.2011.05418.x
Thomsen, P. F., & Willerslev, E. (2015). Environmental DNA – An emerging tool in conservation for
monitoring past and present biodiversity. Biological Conservation, 183, 4–18.
https://doi.org/10.1016/j.biocon.2014.11.019
Thornton, B., & Basu, C. (2011). Real-time PCR (qPCR) primer design using free online software.
Biochemistry and Molecular Biology Education, 39(2), 145–154.
https://doi.org/10.1002/bmb.20461
Vašek, M., Kubečka, J., Čech, M., Draštík, V., Matěna, J., Mrkvička, T., Peterka, J., & Prchalová, M. (2009).
Diel variation in gillnet catches and vertical distribution of pelagic fishes in a stratified European
reservoir. Fisheries Research, 96(1), 64–69. https://doi.org/10.1016/j.fishres.2008.09.010

114

Veldhoen, N., Hobbs, J., Ikonomou, G., Hii, M., Lesperance, M., & Helbing, C. C. (2016). Implementation
of novel design features for qPCR-Based eDNA assessment. PLOS ONE, 11(11), e0164907.
https://doi.org/10.1371/journal.pone.0164907
Weir, T., Peck, K., & Johner, D. (2023). Williston Reservoir 2021 hydroacoustic, trawl and gillnet survey—
Synthesis report. Province of British Columbia, Ministry of Forests, FIsh & Wildlife Branch.
Wetzel, R. G. (2001). Limnology: Lake and river ecosystems (3rd ed). Academic Press.
Wilcox, T. M., McKelvey, K. S., Young, M. K., Jane, S. F., Lowe, W. H., Whiteley, A. R., & Schwartz, M. K.
(2013). Robust detection of rare species using environmental DNA: The importance of primer
specificity. PLOS ONE, 8(3), e59520. https://doi.org/10.1371/journal.pone.0059520
Wilson, P. N., & Shrimpton, J. M. (2020). Genetic population structure and demographics of kokanee
introduced to the Williston Reservoir (PEA-F19-F-2870-DCA; p. 41). Fish and Wildlife Compensation
Program - Peace.
https://a100.gov.bc.ca/pub/acat/documents/r58750/PEA_F9_F_2870_DCA_1594143450750_413
9699754.pdf
Winfield, I. J., Fletcher, J. M., James, J. B., & Bean, C. W. (2009). Assessment of fish populations in still
waters using hydroacoustics and survey gill netting: Experiences with Arctic charr (Salvelinus
alpinus) in the UK. Fisheries Research, 96(1), 30–38. https://doi.org/10.1016/j.fishres.2008.09.013
WWF. (2020). Living planet report 2020—Bending the curve of biodiversity loss.
Yates, M. C., Fraser, D. J., & Derry, A. M. (2019). Meta-analysis supports further refinement of eDNA for
monitoring aquatic species-specific abundance in nature. Environmental DNA, 1(1), 5–13.
https://doi.org/10.1002/edn3.7
Yates, M. C., Glaser, D. M., Post, J. R., Cristescu, M. E., Fraser, D. J., & Derry, A. M. (2020). The
relationship between eDNA particle concentration and organism abundance in nature is
strengthened by allometric scaling. Molecular Ecology, n/a(n/a).
https://doi.org/10.1111/mec.15543
Ye, J., Coulouris, G., Zaretskaya, I., Cutcutache, I., Rozen, S., & Madden, T. L. (2012). Primer-BLAST: A tool
to design target-specific primers for polymerase chain reaction. BMC Bioinformatics, 13, 134.
https://doi.org/10.1186/1471-2105-13-134
Zale, A. V., Parrish, D. L., Sutton, T. M., & American Fisheries Society (Eds.). (2012). Fisheries techniques
(3rd ed). American Fisheries Society.
Zemlack, R. J., & Langston, A. R. (1994). Finlay River lake trout spawning investigation (PWFWCP Report
No. 80; p. 6pp plus appendices). Peace/Williston Fish and Wildlife Compensation Program.
http://a100.gov.bc.ca/appsdata/acat/documents/r35380/pwfwcp_report_no_080_13603613386
72_fced2e50d9abf765520f00550f7bc885edc696f901cbea99139799182d018168.pdf

115

APPENDIX A:
IN VITRO qPCR ASSAY VALIDATION
Synthetic DNA Templates
Synthetic DNA templates were designed for each qPCR assay based on consensus
sequences for each species. Synthetic DNA templates were purchased from Integrated DNA
Technologies Inc. as gBlocks® Gene Fragments. Synthetic DNA templates were used in the
temperature gradient and sensitivity testing of qPCR assays. The synthetic template sequences
that all the qPCR assays in this study were tested on are shown in the following four figures
(Figures A1 to A4).

Figure A 1. Synthetic DNA template sequence used to test the Sock_COI and Rain_COI assays. Primers
and probes are annotated for the Sock_COI and Rain_COI assays as H-Sock-COI and H-Rain-COI,
respectively. The template also included sequences for other salmon species not tested for in this study.

116

Figure A 2. Synthetic DNA template sequence used to test the COCL_CYTB and BUT1_COCL assays.
Forward primers are coloured dark green, reverse primers are coloured light green, and probes are
coloured red.

Figure A 3. Synthetic DNA template sequence used to test all the candidate assays for lake trout. Forward
primers are coloured dark green, reverse primers are coloured light green, and probes are coloured red.
The primers and probe labeled LT4 were the most successful candidates and became the novel LT_CYTB
assay featured in this study.

117

Figure A 4. Synthetic DNA template sequence used to test the novel peamouth assay, PCC_CYTB.
Forward primers are coloured dark green, reverse primers are coloured light green, and probes are
coloured red.

Temperature Gradient Tests
Optimal PCR annealing temperatures were determined for each qPCR assay by running a
temperature gradient test on the Bio-Rad QX200 Droplet Digital PCR system (Life Science,
Mississauga, ON) at eight annealing temperatures between 50 and 60 °C. Assays were tested
against synthetic DNA templates (Integrated DNA Technologies Inc.).

Figure A 5. Temperature gradient results for the kokanee assay, Sock_COI.

118

Figure A 6. Temperature gradient results for the rainbow trout assay, Rain_COI.

Figure A 7. Temperature gradient results for the lake whitefish assay, COCL_CYTB.

Figure A 8. Temperature gradient results for the bull trout assay, BUT1_CYTB.

119

Figure A 9. Temperature gradient results for the lake trout assay, LT_CYTB.

Figure A 10. Temperature gradient results for the peamouth assay, PCC_CYTB.

Specificity Tests
qPCR assays were tested for amplification of their target species and for crossamplification of closely related and representative species from the Williston Watershed. Genomic
DNA used for specificity testing was extracted from the fin tissues of 15 fish species using the
DNeasy Blood and Tissue kit (Qiagen, Mississauga, ON) following the manufacturer’s protocol.
Assays were tested against extracted DNA from at least two separate individuals from all nontarget species in the same family. Assays were tested on the QX200 Droplet Digital PCR system
in 20 µL final reaction volumes that included 10 µL of 2x ddPCR Supermix for Probes (no dUTPs)

120

(Life Science, Mississauga, ON), 1 µL of 20x concentrated assay, 5 µL of extracted DNA (2-6 ng)
and 4 µL of nuclease free water (IDT). The thermocycling protocols for all assays were: initial
denaturation at 95 °C for 10 min, 45 cycles of denaturation at 95 °C for 30 seconds and annealing
at 58 °C for 60 s, followed by a final incubation at 98 °C for 10 minutes. The following figures are
representative results from species in the same family found in the Williston Reservoir.

Figure A 11. Fluorescence amplitude of droplets plotted against DNA samples of target and non-target
salmonid species for the kokanee assay, Sock_COI. Droplets that were positive for PCR amplification are
coloured and negative droplets are coloured grey. Trace amounts of lake whitefish contamination were
observed with lake trout samples collected from gillnets in the Williston Reservoir.

121

Figure A 12. Fluorescence amplitude of droplets plotted against DNA samples of target and non-target
salmonid species for the rainbow trout assay, Rain_COI. Droplets that were positive for PCR
amplification are coloured and negative droplets are coloured grey.

Figure A 13. Fluorescence amplitude of droplets plotted against DNA samples of target and non-target
salmonid species for the lake whitefish assay, COCL_CYTB. Droplets that were positive for PCR
amplification are coloured and negative droplets are coloured grey.

122

Figure A 14. Fluorescence amplitude of droplets plotted against DNA samples of target and non-target
salmonid species for the bull trout assay, BUT1_CYTB. Droplets that were positive for PCR amplification
are coloured and negative droplets are coloured grey.

Figure A 15. Fluorescence amplitude of droplets plotted against DNA samples of target and non-target
salmonid species for the lake trout assay, LT_CYTB. Droplets that were positive for PCR amplification
are coloured and negative droplets are coloured grey.

123

Figure A 16. Fluorescence amplitude of droplets plotted against DNA samples of target and non-target
Leuciscidae species for the peamouth assay, PCC_CYTB. Droplets that were positive for PCR
amplification are coloured and negative droplets are coloured grey.

Sensitivity Tests
Sensitivity was measured for on the Bio-Rad QX200 Droplet Digital PCR system (Life
Science, Mississauga, ON) with dilution series experiments. A synthetic DNA template (gBlock,
Integrated DNA Technologies Inc.) of each target amplicon was used to create 1:5 standard serial
dilutions at eight dilution points with six molecular replicates of each standard. Synthetic templates
were based on consensus sequences generated in Geneious Prime from the NCBI sequence
database. Targeted concentrations for each standard dilution point were between 10 000 and 0.13
copies/20 µL ddPCR reaction. However, the end-point ddPCR measurements of synthetic template
concentrations varied for each batch of gBlock used. The standard concentrations for each dilution
series (Table A1) were calculated from the mean concentration of gBlock detected by ddPCR at
the most concentrated dilution point and dividing by five for each subsequent standard. The limit
of detection (LOD) was defined and assessed in two ways. Standard LOD was defined as the
lowest concentration of target analyte that was detected with 95% confidence (Klymus et al.,
2019). Detections were required in all six replicates of a dilution level to achieve 95% confidence
level for this study. Modeled LOD was assessed with a statistical model for digital PCR developed
124

by Hunter et al. (2017). The modeled LOD was defined as the lowest amount of analyte that could
be both detected and distinguished from a source of overestimation bias in qPCR known as the
concentration plateau; as DNA concentrations approach zero, measured concentrations may be
higher than expected and result in a nonlinear standard dilution series (Hunter et al., 2017). The
modeled LOD was determined by analyzing the dilution series data with R code provided by
Hunter et al. (2017). Modeled LOD could not be calculated for the Sock_COI and BUT1_CYTB
assays because their dilution series data remained linear and detections were completely absent at
the lowest dilution levels.
Table A 1. Limits of detection for the six assays tested in this study. Concentrations of the synthetic DNA
standards were adjusted after comparison with ddPCR. Standard LOD is the lowest dilution level with
>95% detection amongst replicates. Modeled LOD was assessed with a statistical model for digital PCR
developed by Hunter et al. (2017). Modeled LOD was not applicable (N/A) when the relationship
between expected and measured DNA concentrations remained linear and a “concentration plateau” was
not detected at low concentrations.
Assay
Sock_COI
Rain_COI
COCL_CYTB
BUT1_CYTB
LT_CYTB
PCC_CYTB

Dilution Series Standard Concentrations
(copies/rxn)
3800, 760, 152, 30.4, 6.08, 1.22, 0.24, 0.05
4000, 800, 160, 32, 6.40, 1.28, 0.26, 0.05
2700, 540, 108, 21.6, 4.32, 0.86, 0.17, 0.03
3100, 620, 124, 24.8, 4.96, 0.99, 0.20, 0.04
15500, 3100, 620, 124, 24.8, 4.96, 0.99, 0.20
5700, 1140, 228, 45.6, 9.12, 1.82, 0.37, 0.07

125

Standard LOD
(copies/rxn)
6.08
6.40
4.32
4.96
4.96
45.6

Modeled LOD
(copies/rxn)
N/A
6.40
4.32
N/A
4.96
1.82

Figure A 17. ddPCR-based LOD assessment of the kokanee assay, Sock_COI, with a dilution series of synthetic
kokanee DNA. Synthetic DNA concentrations along the x-axis include a 1:5 standard dilution with eight points and
6 replicates each. The log of measured concentrations is along the y-axis. Replicate averages (including zeros) are
denoted by a red X. The Standard LOD is the lowest dilution level with 100% detection across all replicates. The
Modeled LOD could not be calculated for the Sock_COI assay because the dilution series did not exhibit a
concentration plateau.

Figure A 18. ddPCR-based LOD assessment of the rainbow trout assay, Rain_COI, with a dilution series of
synthetic rainbow trout DNA. Synthetic DNA concentrations along the x-axis include a 1:5 standard dilution with
eight points and 6 replicates each. The log of measured concentrations is along the y-axis. Replicate averages
(including zeros) are denoted by a red X. The Standard LOD is the lowest dilution level with 100% detection across
all replicates. The Modeled LOD is defined at the point of intersection between linearly related standard
concentrations (solid line) and instrumental responses that do not vary with concentration (concentration plateau;
dashed line). Upper and lower 95% confidence limits are shown (dotted lines). The upper 95% confidence limit was
used as a conservative bound for the Modeled LOD (Hunter et al. 2017).

126

Figure A 19. ddPCR-based LOD assessment of the lake whitefish assay, COCL_CYTB, with a dilution series of
synthetic lake whitefish DNA. Synthetic DNA concentrations along the x-axis include a 1:5 standard dilution with
eight points and 6 replicates each. The log of measured concentrations is along the y-axis. Replicate averages
(including zeros) are denoted by a red X. The Standard LOD is the lowest dilution level with 100% detection across
all replicates. The Modeled LOD is defined at the point of intersection between linearly related standard
concentrations (solid line) and instrumental responses that do not vary with concentration (concentration plateau;
dashed line). Upper and lower 95% confidence limits are shown (dotted lines). The upper 95% confidence limit was
used as a conservative bound for the Modeled LOD (Hunter et al. 2017).

Figure A 20. ddPCR-based limit of detection (LOD) assessment of the bull trout assay, BUT1_CYTB, with a dilution
series of synthetic bull trout DNA. Synthetic DNA concentrations along the x-axis include a 1:5 standard dilution
with eight points and 6 replicates each. The log of measured concentrations is along the y-axis. Replicate averages
(including zeros) are denoted by a red X. The Standard LOD is the lowest dilution level with 100% detection across
all replicates. The Modeled LOD could not be calculated for the Sock_COI assay because the dilution series did not
exhibit a concentration plateau.

127

Figure A 21. ddPCR-based LOD assessment of the lake trout assay, LT_CYTB, with a dilution series of synthetic
lake trout DNA. Synthetic DNA concentrations along the x-axis include a 1:5 standard dilution with eight points and
6 replicates each. The log of measured concentrations is along the y-axis. Replicate averages (including zeros) are
denoted by a red X. The Standard LOD is the lowest dilution level with 100% detection across all replicates. The
Modeled LOD is defined at the point of intersection between linearly related standard concentrations (solid line) and
instrumental responses that do not vary with concentration (concentration plateau; dashed line). Upper and lower
95% confidence limits are shown (dotted lines). The upper 95% confidence limit was used as a conservative bound
for the Modeled LOD (Hunter et al. 2017).

128

Figure A 22. ddPCR-based limit of detection (LOD) assessment of the peamouth assay, PCC_CYTB, with a dilution
series of synthetic peamouth DNA. Synthetic DNA concentrations along the x-axis include a 1:5 standard dilution
with eight points and 6 replicates each. The log of measured concentrations is along the y-axis. Replicate averages
(including zeros) are denoted by a red X. The Standard LOD is the lowest dilution level with 100% detection across
all replicates. The Modeled LOD is defined at the point of intersection between linearly related standard
concentrations (solid line) and instrumental responses that do not vary with concentration (concentration plateau;
dashed line). Upper and lower 95% confidence limits are shown (dotted lines). The upper 95% confidence limit was
used as a conservative bound for the Modeled LOD (Hunter et al. 2017).

129

APPENDIX B:
CANDIDATE qPCR ASSAYS
I developed and tested a single candidate qPCR assay for peamouth (Table 2. 1), two
candidate assays for bull trout (Table A2) and three for lake trout (Table 2. 1 and Table A2).
Table A2 describes the characteristics of candidate assays that did not perform well during
preliminary screening and that were rejected. Assay performance during in vitro specificity
testing was the primary consideration for determining assay success. Assays were rejected if they
amplified the extracted DNA from non-target species. The performance of the two candidate bull
trout assays were compared against the published BUT1_CYTB assay from Wilcox et al. (2013).
The LT2_CYTB lake trout assay shared the same forward primer and probe sequence as the
successful lake trout assay that was eventually named LT_CYTB (Table 2. 1) but had a different
reverse primer sequence. The LT2_CYTB assay was developed as an alternate to the LT_CYTB
assay if it did not perform well during in vitro testing. Due to the success of the LT_CYTB, the
LT2_CYTB assay was never tested in vitro.
Table A 2. Characteristics and secondary structure analysis of the unsuccessful candidate assays for bull
trout and lake trout as calculated with IDT’s OligoAnalyzer Tool.
Species/
Assay
Name
Bull trout
BT2_CYTB

Target
Gene
CYTB

Bull trout
BT3_CYTB

CYTB

Lake
trout
LT1_CYTB
Lake
trout
LT2_CYTB

CYTB

CYTB

Primers/
Probe
Forward
Reverse
Probe
Forward
Reverse
Probe
Forward
Reverse
Probe
Forward
Reverse
Probe

Sequence 5’ to 3’
CCTTCTGAGGAGCCACTGTAATC
AATAACGAAGGGGAATAGGAAGTGA
TGTCCCTTACGTAGGAGGTGCCCTT
CGCCTTTCACTTCCTATTCCCC
GCCAAGCAATATAGCTACGAAACC
ACGCCGACAAAATCTCATTCCACCC
CTTCTGAGGAGCCACTGTGATT
GATAACGAATGGGAATAGGAAGTGG
CGTCCCTTATGTGGGAGGTGCCCT
GCCTTCCACTTCCTATTCCCA
AGGCCAAGTAGTATAGCTACGAAG
ACGCCGATAAAATCTCGTTCCACCC

130

(bp)
23
25
25
22
24
25
22
25
24
21
24
25

Tm
(°C)
63.9
63.7
70.0
64.6
63.7
68.7
63.9
62.9
70.9
63.7
63.2
68.2

Hairpin
Tm (°C)
47.7
1.6
50.9
None
53.8
23.8
14.6
25.9
57.3
None
44.5
34.4

Dimer Structures ΔG (kcal/mol)
F/R
F/
R/
Self
Primers Probe Probe
-4.67
-8.19
-4.67
-9.68
-3.61
-8.22
-3.61
-5.46
-3.89
-3.61
-8.26
-3.61
-3.3
-7.96
-6.37
-7.71
-3.61
-7.71
-3.14
-7.81
-3.42
-6.53
-9.28
-5.19

Candidate Assay Design and Optimization
eDNA assays frequently incorporate conjugated minor groove binder (MGB) groups at the
3’ end of probes because an MGB probe can be shorter and more specific than an unmodified
probe (Kutyavin et al., 2000). Addition of the MGB stabilizes the DNA duplex between probe and
template, especially at A+T rich sites, reducing the length of probe needed to achieve the high Tm
required for PCR. MGB probes increase specificity because mismatches within the MGB region
are more destabilizing than mismatches in non-MGB probes, resulting in a greater reduction in Tm.
All four of the published qPCR assays tested in this study (Table 2. 1) were designed with MGB
probes. The MGB probes likely contributed in varying degrees to the specificity of those assays
but the BUT1_CYTB assay may not have been as specific without it, due to the reduced influence
of primer mismatches with lake trout DNA. The LT_CYTB and PCC_CYTB assays were designed
without MGB probes because the benefits of the MGB chemistry for those assays were not deemed
necessary. Primer mismatches have a much larger influence on assay specificity than mismatches
in the probe (Wilcox et al., 2013). The LT_CYTB and PCC_CYTB assays were designed with a
focus on primer specificity and in vitro testing demonstrated that the assays were specific enough
without an MGB probe. The novel assays also had the advantage of being designed specifically
for the species assemblage in the Williston Watershed. Probe length was not a problem for the
LT_CYTB and PCC_CYTB assays because the probe regions were relatively G+C rich and
appropriate Tms were achieved with 25mer non-MGB probes. A shorter MGB probe also reduces
the distance between the fluorophore and quencher, reducing background fluorescence and
potentially providing increased assay sensitivity (Kutyavin et al., 2000). The LT_CYTB and
PCC_CYTB assays were both designed with double quenched probes, which is an alternative
method for reducing background fluorescence in a longer probe (Prediger, 2013). A considerable
downside of MGB assays is the increased cost compared to non-MGB assays. The TaqMan MGB
131

assays purchased for this study were sold in 360-reaction quantities from ThermoFisher Scientific
for $434 CAD. The IDT PrimeTime® Std qPCR Assays (non-MGB) with double quenched probes
purchased for this study were sold in 500-reaction quantities for $170 CAD. All the assays tested
in this study demonstrated that they were species-specific, regardless of their probe chemistry. The
LT_CYTB and PCC_CYTB assays could easily be modified to incorporate MGB probes, but the
benefits to assay specificity do not appear to justify the added costs for eDNA monitoring in the
Williston Reservoir.
Optimal PCR annealing temperatures were determined for each candidate qPCR assay by
running a temperature gradient test on the Bio-Rad QX200 Droplet Digital PCR system (Life
Science, Mississauga, ON) at eight annealing temperatures between 50-60 °C. Assays were tested
against synthetic DNA templates (Integrated DNA Technologies Inc.). An annealing temperature
of 58 °C was suitable for all the assays tested and was the annealing temperature used in all the
subsequent in vitro testing for specificity.
Candidate Assay Specificity
The BT2_CYTB bull trout assay that I developed demonstrated similar specificity to the
BUT1_CYTB assay developed by Wilcox et al. (2013) (Figure A 23). However, the variance in
droplet amplitude for the BUT1_CYTB assay was less than that for the BT2_CYTB assay. The
large variance in droplet fluorescence observed in the BT2_CYTB assay may have made it more
difficult to clearly identify a positive detection compared to the BUT1_CYTB assay. The
BT3_CYTB assay was also reasonably specific, but there was a visible amount of cross
amplification with lake trout DNA (Figure A 25). Lake trout DNA was not amplified as
efficiently as bull trout DNA and the difference in fluorescence amplitude was enough to
differentiate between the extracted DNA of both species. However, this difference may not have

132

been as great in an environmental sample that contained the DNA of both species. The
BUT1_CYTB assay outperformed both candidate bull trout assays that I developed and was the
only bull trout assay chosen for further validation in this study.

Figure A 23. Fluorescence amplitude of ddPCR droplets plotted against DNA samples of target and nontarget species from the Williston Reservoir for the BT2_CYTB candidate bull trout assay (top) and the
BUT1_CYTB assay (bottom) from Wilcox et al. (2013). Some of the lake whitefish tissue samples that
DNA was extracted from were contaminated with bull trout DNA. This is evident from the small number
of positive droplets in some but not all lake whitefish samples.

133

Figure A 24. Fluorescence amplitude of ddPCR droplets plotted against DNA samples of target and nontarget species from the Williston Reservoir for the BT3_CYTB candidate bull trout assay. Some of the
lake whitefish tissue samples that DNA was extracted from were contaminated with bull trout DNA. This
is evident from the small number of positive droplets in some but not all lake whitefish samples.

The LT1_CYTB candidate lake trout assay performed similarly in the specificity tests to
the BT3_CYTB assay. The LT1_CYTB assay demonstrated low amplitude cross amplification
with extracted bull trout DNA and although the difference in fluorescence amplitude was enough
to differentiate between the extracted DNA of both species, it was not ideal (Figure A 25). The
LT1_CYTB assay was therefore not selected for further validation in this study.

134

Figure A 25. Fluorescence amplitude of ddPCR droplets plotted against DNA samples of target and nontarget species from the Williston Reservoir for the LT1_CYTB candidate lake trout assay. Some of the
lake whitefish and bull trout tissue samples that DNA was extracted from were contaminated with lake
trout DNA. This is evident from the small number of positive droplets in some but not all lake whitefish
samples.

135

APPENDIX C:
PELAGIC SPECIES QPCR ASSAY VALIDATIONS FOR THE WILLISTON
RESERVOIR
A validation scale devised by Thalinger et al. (2021) provides criteria for determining if
eDNA assays are ready for use in routine species monitoring. The five levels of the validation
scale incorporate the entire workflow used to detect a species’ DNA in an environmental sample,
from sample collection through to the interpretation of PCR results (Table A3). Levels 1 and 2
include assay design, basic in silico analysis, in vitro optimizations, testing on target species tissue
and specificity testing on closely related, non-target species. eDNA results from assays that have
only been validated to level 2 cannot be interpreted with any confidence until further levels of
validation have been achieved.
More extensive specificity testing that includes co-occurring non-target species and
sensitivity testing that establishes an LOD are considered important criteria for assay validation at
level 4. The in silico assay design and screening and in vitro testing completed for the six assays
in this study meet the criteria for validation at level 1, 2 and much of level 4. Level three and four
validation criteria includes testing on environmental samples (in situ validations). Confidence in
eDNA results build with the level of validation. Level three validation is considered essential and
the results from assays validated to level four and five can be interpreted with confidence
(Thalinger et al., 2021).
The novel lake trout assay, LT_CYTB, could not be validated beyond level two because it
did not successfully detect lake trout eDNA in samples collected from the Williston Reservoir.
However, the eDNA and gillnet results from Teare Creek and the eDNA results from Factor Ross
all agreed with one another, but a verified negative result is not the same as a verified positive
result. Collecting and analyzing samples from other locations in the Williston Reservoir where
136

lake trout are known to be abundant, such as Finlay Forks, would likely result in a positive
detection and help to validate the LT_CYTB to at least level three.
The novel peamouth assay, PCC_CYTB, achieved a few positive detections from the eDNA
samples taken in the Finlay Reach, but the quality and quantity of those eDNA samples was poor.
The presence of peamouth in the area was confirmed by the gillnet results and that adds confidence
to the positive eDNA detections. However, more extensive field testing of the PCC_CYTB assay
is needed to be confident in the results.
Although in vitro specificity testing for this study was robust, I did not determine detection
probabilities of any of the assays with statistical models. Further, this was only a pilot study and
did not produce enough data to rigorously investigate the ecological and physical factors
influencing eDNA in the Williston Reservoir. These are key criteria that must be met to validate
an eDNA assay to level five and should be the focus of future eDNA studies in the Upper Peace
River watershed.

137

Table A 3. The five validation levels described by Thalinger et al. (2021) and the validation level
achieved by each of the six assays tested in this study.

Level
1
2

Description
Incomplete
Partial

3

Essential

4

Substantial

5

Operational

Criteria
Basic in silico analysis.
In vitro specificity testing.
Target species successfully detected from
an eDNA sample. Specifics of eDNA
extraction and concentration reported.
LOD determined. Extensive field testing
and in vitro testing on co-occurring, nontarget species.
Comprehensive specificity testing,
detection probability from statistical
modelling and investigations of ecological
and physical factors potentially
influencing eDNA in the environment.

138

Assays
COCL_ BUT1_
CYTB
CYTB
X
X
X
X

Sock
_COI
X
X

Rain
_COI
X
X

X

X

X

X

X

X

LT_
CYTB
X
X

PCC_
CYTB
X
X

X

Partial

Partial

X

Partial

Partial

APPENDIX D:
ENVIRONMENTAL GENOMICS FACILITY (GREAT LAKES INSTITUTE OF
ENVIRONMENTAL RESEARCH) IN-HOUSE PCR CLEAN-UP PROCEDURE
SPRI (Sera-Mag) bead solution recipe for PCR clean-up
Table A 4. Reagents and amounts used in bead solution recipe for PCR clean-up.
Reagent
PEG-8000 powder
5M NaCl
Washed Sera-Mag Beads
ddH2O

Mass
10g
-

Volume
25 mL
1 mL
24 mL

1. Add 10g of PEG-8000 to a 50mL Falcon tube.
2. To the powder, add NaCl and fill with ddH2O up to the 50 mL mark.
3. Shake/vortex Falcon tube until all PEG has dissolved. Note: Shaking for five minutes and
then letting it sit for a few minutes usually works the best.
4. Resuspend stock solution of Sera-Mag beads by shaking until there is no longer
“sediment.”
5. Wash the beads: Transfer 1mL bead suspension to a 1.7mL tube and pellet the beads on a
magnetic rack (~3 min). Keeping the tube on the magnetic rack, remove the storage
buffer and add 1mL of TE. Remove the tube from the magnetic rack, vortex to mix well
and place back on magnetic rack. Remove TE buffer and repeat the TE wash one more
time. Fully resuspend the beads in 1mL TE and mix well.
6. Add bead suspension to the Falcon tube and mix immediately. Wrap Falcon tube in
aluminum foil and store at 4°C.
Bead-Cleaning Steps
1. Mix 15uL of the bead solution with 10uL of PCR product thoroughly by pipetting 10
times.
· This step binds PCR products 100bp and larger to the magnetic beads.
· Let the mixed samples incubate for 3-5 minutes at room temperature for
maximum recovery.
2. Place the reaction plate onto a magnetic plate for 5-10 minutes to separate beads from the
solution.
· The separation time is dependent on the size of the reaction.
· Wait for the solution to clear before proceeding to the next step.
3. Aspirate the cleared solution form the reaction plate and discard.
· This step must be performed while the reaction plate is situated on the magnetic
plate.
139

· Do not disturb the collection of magnetic beads.
4. Dispense 200uL of 70% ethanol to each well of the reaction plate and incubate for 30
seconds at room temperature.
· Aspirate out the ethanol and discard.
· Repeat for a total of two washes. It is important to perform these steps with the
reaction plate situated on the magnetic plate.
· Do not disturb the magnetic beads.
· Be sure to remove all of the ethanol from the bottom of the well as it may contain
residual contaminants.
5. Place the reaction plate on bench top to air-dry.
· Be sure to allow the plate to dry completely.
· The plate should be left to air-dry for 10-20 minutes on a bench top to allow
complete evaporation of residual ethanol.
6. Add 40uL of elution buffer (ddH2O or TE) to each well of the reaction plate and mix the
solution with beads by pipetting 10 times.
· Put the reaction on the magnetic plate for 5 minutes and wait for the solution to
become clear.
· Remove all solution to a new plate and keep it at -20°C.

140