Does size matter? An experimental evaluation of ... - ACS Publications

CSIRO Land and Water, GPO Box 1700, Canberra, ACT 2601, Australia. 9. 10. Page 1 of 31. ACS Paragon Plus Environment. Environmental Science & ...
0 downloads 0 Views 713KB Size
Subscriber access provided by University of Winnipeg Library

Environmental Measurements Methods

Does size matter? An experimental evaluation of the relative abundance and decay rates of aquatic eDNA. Jonas Bylemans, Elise M Furlan, Dianne M Gleeson, Christopher M Hardy, and Richard P Duncan Environ. Sci. Technol., Just Accepted Manuscript • DOI: 10.1021/acs.est.8b01071 • Publication Date (Web): 14 May 2018 Downloaded from http://pubs.acs.org on May 14, 2018

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 31

Environmental Science & Technology

1

Does size matter? An experimental evaluation of the

2

relative abundance and decay rates of aquatic eDNA.

3 4

Jonas Bylemans,*, †, ‡ Elise M. Furlan,†, ‡ Dianne M. Gleeson,†, ‡ Christopher M. Hardy,§, ‡

5

Richard P. Duncan†

6 7



Institute for Applied Ecology, University of Canberra, Canberra, ACT 2617, Australia

8



Invasive Animals CRC, University of Canberra, Canberra, ACT 2617, Australia

9

§

CSIRO Land and Water, GPO Box 1700, Canberra, ACT 2601, Australia

10

ACS Paragon Plus Environment

1

Environmental Science & Technology

Page 2 of 31

11

KEYWORDS

12

Environmental DNA, fish, eDNA target length, quantitative real-time PCR

13

ABSTRACT

14

Environmental DNA (eDNA) is increasingly used to monitor aquatic macro-fauna. Typically,

15

short mitochondrial DNA fragments are targeted because these should be relatively more abundant

16

in the environment as longer fragments will break into smaller fragments over time. However,

17

longer fragments may permit more flexible primer design and increase taxonomic resolution for

18

eDNA metabarcoding analyses, and recent studies have shown that long mitochondrial eDNA

19

fragments can be extracted from environmental water samples. Nuclear eDNA fragments have

20

also been proposed as targets but little is known about their persistence in the aquatic environment.

21

Here we measure the abundance of mitochondrial eDNA fragments of different length, and short

22

nuclear eDNA fragments, originating from captive fish in experimental tanks, and test whether

23

longer mitochondrial and short nuclear fragments decay faster than short mitochondrial fragments

24

following fish removal. We show that, when fish are present, shorter mitochondrial fragments are

25

more abundant in water samples than both longer mitochondrial fragments and the short nuclear

26

eDNA fragment. However, the rate of decay following fish removal was similar for all fragment

27

types, suggesting that the differences in abundance resulted from differences in the rates at which

28

different fragment types were produced rather than differences in their decay rates.

29

ACS Paragon Plus Environment

2

Page 3 of 31

30

Environmental Science & Technology

ABSTRACT ART

31 32

INTRODUCTION

33

Environmental DNA (eDNA) extracted from environmental samples can be used to monitor the

34

presence/absence of particular species and obtain biodiversity estimates.1–3 A key issue with using

35

eDNA for monitoring purposes is that DNA in the environment can decay rapidly. Over time,

36

longer DNA fragments break into smaller fragments causing the latter to be relatively more

37

common in environmental samples.4–8 Mitochondrial DNA fragments are also likely to be more

38

abundant in environmental samples than nuclear DNA fragments because mitochondrial DNA is

39

present in higher copy numbers per cell and is believed to be less susceptible to environmental

40

decay. Several studies have shown that recovery rates from environmental samples are lower for

41

long relative to short fragments, and for nuclear relative to mitochondrial DNA fragments.4,6,9–11

42

Consequently, most studies have used short mitochondrial DNA barcodes (i.e. typically between

43

50 and 200 base pairs (bp)), particularly for ancient and highly degraded samples (e.g. ice cores,

44

permafrost, fossil bones, faeces, etc.).4,9,12–16

ACS Paragon Plus Environment

3

Environmental Science & Technology

Page 4 of 31

45

Environmental DNA from water samples is increasingly being used as a monitoring tool.17–19

46

While the optimization of sampling, capturing and extraction protocols has received considerable

47

attention, the relative abundance and decay of DNA fragments of different size and origin (i.e.

48

mitochondrial versus nuclear) has not been formally evaluated.20–23 Within the water column, the

49

relative abundance of different DNA fragments depends on both their rate of production and the

50

rate at which fragments are removed from the water column (i.e. through decay, settling, advection

51

and dispersion).24 While the source of eDNA is often unclear (e.g. from faeces, urine, mucous,

52

etc.), the abundance of eDNA in the environment is known to be correlated with species’ biomass,

53

ontogeny and metabolism, which are likely associated with differences in production rates.24–29 On

54

the other hand, the decay of eDNA is known to be influenced by temperature, pH, UV-radiation

55

and microbial activity.27,30–33 While DNA fragments can decay rapidly in some sample types, such

56

as sediment and faecal samples,4,8,34 recent evidence suggests that the aquatic matrix preserves

57

eDNA relatively well. For example, DNA fragments believed to be more susceptible to decay (i.e.

58

long mitochondrial fragments and nuclear fragments) have been successfully extracted and

59

amplified from water samples.35–38 One reason for these findings may be the fact that aquatic

60

eDNA occurs mainly within whole organelles and/or cells, meaning it may be partly protected

61

from environmental factors that could promote decay.39,40 However, a thorough evaluation of the

62

relative abundance and decay rates of different eDNA fragments is needed as the ability to use

63

non-standard fragments (i.e. longer mitochondrial fragments and nuclear fragments) could

64

advance eDNA-based monitoring of aquatic biodiversity through more flexible primer design and

65

increased taxonomic resolution of barcodes for whole community analyses (i.e. eDNA

66

metabarcoding).

ACS Paragon Plus Environment

4

Page 5 of 31

Environmental Science & Technology

67

Here we use replicate tanks stocked with three different densities of common goldfish (Carassius

68

auratus) to quantify the relative abundance and decay rates of mitochondrial eDNA fragments of

69

different size and a short nuclear eDNA fragment. We use these results to evaluate whether the

70

differences in the relative abundances of the different fragments are caused by their susceptibility

71

to environmental decay. Furthermore, using the eDNA degradation data and two different

72

modelling approaches we aimed to gain a deeper understanding of the eDNA decay process in

73

freshwater environments. Finally, we highlight the potential implications of our findings for

74

advancing eDNA-based monitoring.

75

MATERIALS AND METHODS

76

Primer design and testing

77

We designed primers to amplify different sized fragments (ca. 100, 300 and 500 bp) of the

78

mitochondrial cytochrome c oxidase subunit I (COI) gene and a short (ca. 100 bp) fragment of the

79

nuclear internal transcribed spacer (ITS) region of common goldfish. Target regions were selected

80

based on high genetic variability between species and the presence of multiple copies within a

81

cell.6,41,42 Fragment sizes for the mitochondrial DNA were chosen based on current size

82

recommendations for eDNA studies (i.e. length of 50 to 200 bp) and the conditions currently

83

preferred for High-Throughput Sequencing (HTS) platforms (i.e. 550 to 570 bp using 2 x 300 bp

84

paired-end sequencing on the Illumina MiSeq allowing for a 50 to 30 bp overlap for sequence

85

merging).3,43 The primer pairs were designed and tested both in silico and in vitro (Supporting

86

Information). All primer pairs used for further analyses are given in Table 1. Hereafter, different

87

DNA fragments will be referred to by the eDNA fragment ID’s in Table 1 which consist of the

88

target region (COI or ITS) followed by the total length of the amplicon in brackets (i.e. COI

ACS Paragon Plus Environment

5

Environmental Science & Technology

Page 6 of 31

89

(096bp), COI (285bp) and COI (515bp) for the short, medium and long mitochondrial fragments

90

respectively and ITS (095bp) for the short nuclear fragment).

91

Experimental protocol

92

We sterilised all experimental and sampling equipment before use using a 10% bleach solution

93

followed by thorough rinsing with UV-sterilized tap water. All tanks were set-up with a gravel

94

substrate and a ca. 10 cm piece of PVC pipe for habitat enrichment. A constant airflow was

95

provided to each tank through a cylindrical air stone (15 x 25 mm) and tanks were kept a constant

96

temperature of 20 °C and a 12 h daylight cycle for the duration of the experiment. Goldfish (mean

97

mass = 5.7 g, range = 3.3-9.5 g) purchased from a local pet shop were stocked individually in 10

98

L, 30 L and 60 L tanks filled with UV-sterilized tap water to simulate high density (HD), medium

99

density (MD) and low density (LD) conditions, respectively in three replicate treatments (n = 9

100

tanks). A negative control tank (NCT), consisting of a 60 L tank without fish, was included to test

101

for cross-contamination. Fish were fed half a cube of frozen artemia (ca. 1.4 g) every second day

102

and no food was added to the NCT. Prior to stocking the tanks with fish, a 50 mL water sample

103

was collected from each tank to test that they were initially free from goldfish DNA, with the time

104

of this sample set to zero hours (0 h). Additional 50 mL water samples were collected at 2, 6, 24,

105

72, 168 and 336 h after stocking. Immediately after collecting the 336 h sample, all fish were

106

removed from the tanks and additional water samples were collected at 338, 342, 360, 408, 504,

107

672 and 1008 h. Water volumes within tanks were kept constant by adding 50 mL of UV-sterilized

108

tap water after each sample collection. Fin clips were collected from each fish after they were

109

removed from the tanks. Genomic DNA was extracted from the fin clips and the COI and ITS

110

target regions were amplified using PCR. Amplicons were sequenced for all (COI) or two (ITS)

ACS Paragon Plus Environment

6

Page 7 of 31

Environmental Science & Technology

111

of the experimental animals to test for the presence of individual primer-template mismatches

112

(Supporting Information).

113

Sample processing and qPCR analyses

114

Processing and analyses of eDNA samples were performed in a trace DNA laboratory at the

115

University of Canberra (Australia). This laboratory is spatially separated from other laboratories

116

and is UV-radiated nightly to reduce the risk of contamination. Spatially separated room were used

117

for eDNA extractions and PCR set-up with the latter having a positive air pressure.

118

From each 50 ml water sample, eDNA was captured by filtering onto a glass fibre filter with a

119

nominal pore size of 1.2 µm (MicroScience, Taren Point, Australia). While this approach may

120

inherently bias the results due to a reduced likelihood of capturing free floating DNA, a 1.2 µm

121

pore size is more suitable for field applications and would thus better reflect the accessible eDNA

122

during field surveys. Negative equipment controls (NEC) were obtained for each sample by

123

filtering 500 mL of UV-sterilized water through the sterilized equipment prior to filtering samples.

124

Environmental DNA was extracted using the PowerWater DNA Extraction Kit (MoBio

125

Laboratories, Carlsbad, USA) following manufacturing instructions (i.e. 100 µL of eDNA eluent).

126

For each timepoint, one NEC was extracted to test for cross-contamination occurring during either

127

the filtrations or DNA extraction phase. DNA extracts were transferred to 2 mL screw-cap tubes

128

and stored at -20 °C for downstream analyses.

129

Quantitative real-time PCR (qPCR) was used to determine eDNA concentrations. The absence

130

of goldfish DNA from all negative control samples (i.e. water samples collected from the

131

experimental tanks prior to introducing goldfish, water samples from the NCT and NEC’s) was

132

confirmed by performing six qPCR replicates for the short COI and ITS fragments. For all

133

experimental samples, six qPCR replicates were performed with each primer set. Due to the large

ACS Paragon Plus Environment

7

Environmental Science & Technology

Page 8 of 31

134

number of qPCR analyses and the costs of probe-based qPCR assays, a SYBR® Green chemistry

135

with previously published reaction conditions scaled down to a final reaction volume of 15 µL

136

containing 1.5 µL of DNA extract was used.37 All qPCR reactions were set-up in a dedicated room

137

within the trace DNA laboratory in 384-well plates using the epMotion® 5075 Liquid Handling

138

Workstation (Eppendorf, Hamburg, Germany). Six non-template controls (NTC) and a standard

139

curve, consisting of six qPCR replicates for each concentration (i.e. 3 x 101 to 3 x 106 molecules

140

per reaction), were included on each plate. Synthetic gBlock® fragments (IDT, Coralville, USA)

141

were used to construct standard curves. Synthetic fragments contained bp mismatches with the

142

sequences of the experimental animals (Supporting Information) and Sangers sequencing of qPCR

143

amplicons obtained from eDNA samples was used to evaluate potential cross-contamination (i.e.

144

amplicon sequences should only match the sequences from the experimental animals).

145

All qPCR analyses were performed using the Viia7 Real-Time PCR System (Applied

146

Biosystems, Foster City, USA) under previously published cycling conditions.37 Amplification

147

and melt curves were visually inspected and replicates were excluded from further analyses if the

148

amplification curve did not show a clear exponential phase and/or the melt curve deviated from

149

those obtained with genomic goldfish DNA (Supporting Information). Finally, for each fragment

150

a subset of the PCR amplicons (ca. 10% of all samples showing a positive amplification) were

151

purified using the MinElute PCR Purification Kit (Qiagen, Hilden, Germany) and Sanger

152

sequenced at the ARCF Biomolecular Resource Facility (Australian National University) using an

153

Applied Biosystems 3730xl DNA Analyzer (Thermo Fisher Scientific) following the

154

manufacturer’s protocol.

155

ACS Paragon Plus Environment

8

Page 9 of 31

Environmental Science & Technology

156

Data analyses

157

Environmental DNA concentrations ([eDNA]) were expressed as copy numbers per mL of tank

158

water and qPCR replicates that failed to amplify were assumed to have a concentration of zero

159

copies per mL. Concentrations were calculated using eDNA copies per qPCR reaction (eDNA),

160

the volume of template DNA used per qPCR reaction (1.5 x 10-3 mL) and the sample volume (50

161

mL). A correction for the dilution effect, caused by adding UV-sterilized water to the experimental

162

tanks, was included using the sample volume (50 mL) and the volume of the experimental tank

163

expressed in mL (VT) (Equation 1). Nonetheless, eDNA measurements obtained after removing

164

fish from the experimental tanks will be influenced by both eDNA decay and the dilution effect.

165

As only small sample volumes were used, the dilution effect is unlikely to influence the overall

166

eDNA decay trends observed.

167

[𝑒𝐷𝑁𝐴] =

(𝑒𝐷𝑁𝐴 × (100 × 10−3 𝑚𝐿)) ((1.5 × 10−3 𝑚𝐿) × 50 𝑚𝐿)

×

(𝑉𝑇 −50 𝑚𝐿) 𝑉𝑇

(1)

168

For further statistical analyses, eDNA concentrations were transformed using the natural logarithm

169

(i.e. loge([eDNA] + 1)) and all analyses were conducted using R version 3.4.1 with the packages

170

tidyverse and nlme.44–46 The data obtained for two experimental tanks was excluded from the final

171

analyses as bp mismatches were observed between the primers and the template sequences for two

172

experimental animals (see the “RESULTS AND DISCUSSION” section for a more detailed

173

description).

174

Inspection of the data prior to fish removal revealed that all eDNA fragments reached

175

equilibrium concentrations after approximately 24 h (Supporting Information). We therefore

176

estimated equilibrium eDNA concentrations using the data from 24 h until fish were removed (i.e.

177

sampling time ≥ 24 h and sampling time ≤ 336 h). We fitted a linear mixed-effect model to the

178

data with loge-transformed eDNA concentrations as the response variable. Fish density and

ACS Paragon Plus Environment

9

Environmental Science & Technology

Page 10 of 31

179

fragment type were included as categorical fixed effects, and nested random effects were included

180

to correct for pseudo-replication (i.e. individual samples nested within individual tanks). We

181

estimated differences in eDNA equilibrium concentrations between density treatments and

182

fragment types relative to a reference class set to zero (COI (096bp) in the HD treatment).

183

The decay of aquatic eDNA has previously been modelled assuming a first-order exponential

184

decay function.1,36,47 The concentration of a particular eDNA fragment at time (t), C(t), can thus

185

be modelled using Equation 2 with C0 representing the equilibrium eDNA concentration prior to

186

fish removal and k being the decay constant.

187

𝐶(𝑡) = 𝐶0 × 𝑒 −𝑘𝑡

(2)

188

While this first-order decay model may be a simplification of the true decay process, it allows us

189

to quantify the decay constants for each fragment by density treatment combination, k, and to

190

compare how decay constants vary under different conditions.31,40 Decay constants were estimated

191

using the data obtained at least 6 hours after fish removal (i.e. sampling time ≥ 342 h). This was

192

done to avoid a spike in eDNA concentrations immediately after fish removal in one tank, possibly

193

caused by increased stress levels in the experimental animal and/or the resuspension of faecal

194

matter. We also excluded sampling times for which all qPCR replicates failed to amplify, which

195

implied that eDNA abundance had decayed below the detection limits (i.e. here assumed to be zero

196

copies per mL). We estimated the parameters of the exponential decay model by fitting linear

197

models using loge-transformed eDNA concentrations as the response variable. A linear mixed-

198

effect model was used to evaluate the impact of fish density and fragment type on the decay

199

constant. Sampling time was included as a continuous fixed effect along with categorical fixed

200

effects for fish density and fragment type. Two-way interactions between sampling time and fish

201

density, and sampling time and fragment type were also included (significant interactions imply

ACS Paragon Plus Environment

10

Page 11 of 31

Environmental Science & Technology

202

differing slopes and hence differing values of the decay constant, k, for density and fragment type).

203

A nested random effect was included to correct for pseudo-replication (i.e. individual samples

204

nested within individual tanks). Differences in decay constants between density treatments and

205

fragment types were estimated relative to a reference class set to zero (COI (096bp) in the HD

206

treatment). To evaluate the fit of the exponential decay model we used separate linear models for

207

each fragment by density treatment combination and calculated the Akaike Information Criterion

208

(AIC).

209

The assumption that the eDNA decay process can be summarized into a single decay constant

210

may not apply if some fragments were more prone to decay than others (e.g. fragments present

211

within cells versus free floating DNA). This would result in the decay constant decreasing over

212

time as fragments prone to rapid decay (i.e. having a higher decay constant) were removed first

213

from the population, leaving behind fragments with lower decay constants. To test if the eDNA

214

decay constant varied over time for each treatment by fragment combination, we fitted a Weibull

215

decay model to the data (Equation 3). The Weibull parameter (β) in this model allows for the decay

216

constant to vary with time, with values of β less than one indicating a decreasing decay constant

217

and values larger than one indicating an increasing decay constant over time. When β equals one

218

the model reduces to the first-order exponential decay function.

219

𝐶(𝑡) = 𝐶0 × 𝑒 (−𝑘 × 𝑡

𝛽)

(3)

220

We estimated parameters for the Weibull model for each fragment by density treatment

221

combination as above (i.e. using the data obtained at least 6 hours after fish removal and sampling

222

times for which all replicates tested negative for eDNA excluded). We used the AIC to evaluate

223

the fit of the model for each treatment by fragment combination and compared AIC values from

ACS Paragon Plus Environment

11

Environmental Science & Technology

Page 12 of 31

224

the Weibull model with those obtained from the first-order exponential decay model to assess

225

which model best described the data.

226

RESULTS AND DISCUSSION

227

Primer design and testing

228

All primer pairs successfully passed the in silico and in vitro evaluation. The performance of the

229

different primer pairs was evaluated based on the standard curves included in the qPCR analyses.

230

The results show that the amplification efficiency decreases with increasing amplicon length and

231

that the efficiency of the primers targeting the ITS (096bp) fragment is lower compared to those

232

for the COI (095bp) fragment (Table 1). Nonetheless, this will not affect the validity of the eDNA

233

quantifications as both standard curve samples and experimental samples will be amplified with a

234

comparable efficiency.48

235

The COI barcode sequences (660bp in length) from the experimental animals revealed that two

236

mitochondrial haplotypes were present. Two out of the nine sequences had a 98% match with C.

237

auratus voucher specimens, which were used as a basis for primer development. This genetic

238

variation in the mitochondrial genome resulted in a reduced or failed amplification in two of the

239

experimental tanks when using the goldfish specific COI primers. The data obtained from these

240

replicate tanks was thus excluded from further analysis (i.e. removal of one LD and one HD

241

replicate tank). In contrast, identical sequences were obtained for the ITS regions of two

242

experimental animals with different COI haplotypes and these were used as a basis for primer

243

development (Supporting Information). Primers targeting the ITS (096bp) fragment showed

244

consistent amplification in all replicate tanks. While this suggests that the ITS region can be a

245

valuable target for the development of primers specific to species with a high mitochondrial

ACS Paragon Plus Environment

12

Page 13 of 31

Environmental Science & Technology

246

genetic variability (e.g. goldfish), the relative abundance and the decay rate of the ITS (096bp)

247

fragment also needs to be considered.

248

Effect of fragment size

249

Equilibrium concentrations of mitochondrial eDNA fragments decreased with decreasing fish

250

density, which is consistent with previously published results.28,29 The model estimates of the mean

251

eDNA concentrations in the MD and LD treatments were on average 0.43 and 0.16 times that of

252

the HD treatment, respectively (Figure 1). The 95% confidence intervals around the model

253

estimates revealed a significant decrease in equilibrium concentrations for longer mitochondrial

254

fragments, which is consistent with the prevailing view that shorter mitochondrial DNA fragments

255

are more abundant in environmental samples (Figure 1).4–6 The mean estimates for the equilibrium

256

concentrations of the COI (285bp) and COI (515bp) fragments were 0.76 and 0.31 times that of

257

the COI (096bp) fragment, respectively. In contrast, a previous study by Deagle et al.4 revealed

258

that a 327 bp sea lion DNA fragment is on average 0.099 (± 0.057) less abundant than a 91 bp

259

fragment in sea lion faecal samples. Furthermore, a study by Dell’Anno et al.34 found that

260

degradation rates for extracellular DNA in marine sediments were 7-100 times higher than those

261

observed in the water column, and Deiner et al.38 recently showed that even complete

262

mitochondrial genomes can be extracted and amplified from aquatic eDNA samples.

263

Consequently, the aquatic matrix seems to preserve longer DNA fragment relatively well.

264

Environmental DNA decay constants estimated from the first-order exponential decay models

265

ranged from 0.11 to 0.77 day-1, which are towards the lower end of previously published records

266

(i.e. 0.05 to 17.9 day-1).19,26,27,30,31 This could be a consequence of the UV-sterilized tap water used

267

in the experimental tanks, which is likely to reduce the influence of microbial organisms on the

268

decay process. However, similar decay constants have been obtained by Thomsen et al.19 (i.e. 0.32

ACS Paragon Plus Environment

13

Environmental Science & Technology

Page 14 of 31

269

and 0.70 day-1) when using natural seawater and by Strickler et al.31 (i.e. 0.05-0.34 day-1) when

270

using tap water inoculated with micro-organisms. More importantly, we expected to find higher

271

decay constants for longer fragments if DNA fragmentation (i.e. events such as strand breaks that

272

prevent DNA replication during PCR) is a random process and a major cause of eDNA decay.4,9

273

However, given the strongly overlapping confidence intervals around the estimates of the decay

274

constants we can conclude that there is no clear evidence that larger eDNA fragments have higher

275

decay constants (Figure 2). In contrast to our findings, a recent study did report a significantly

276

higher decay constant for a 719 bp fragment relative to a 127 bp fragment.36 The smaller

277

differences in length between fragments in our study may partly explain why we did not observe

278

an effect. Alternatively, the short decay time used by Jo et al.36 (i.e. up to 48 hours after fish

279

removal) may have resulted in an over-estimation of decay constant given that we found that decay

280

rates can be high initially but decrease over time (see “Comparison of eDNA decay models”

281

section). Nevertheless, the absence of a clear length dependent decay constant in our study

282

provides evidence that DNA fragmentation may not be the rate determining step in the decay

283

process of aquatic eDNA. In the current study, the rate of decay may be more strongly affected by

284

the time taken to break down phospholipid bilayers as aquatic eDNA is mainly present in

285

mitochondria or whole cells and the methods used here will primarily capture this eDNA

286

fraction.39,40

287

Equilibrium concentrations of aquatic eDNA are a function of both eDNA production and decay

288

rates. As decay rates did not vary consistently by fragment length, our results imply that the

289

differences in equilibrium concentrations were caused primarily by differences in the production

290

rates of eDNA fragments of different length. Intestinal cells present in faecal matter are an

291

important source of aquatic eDNA.25,49 Given that DNA in faeces is often fragmented, some

ACS Paragon Plus Environment

14

Page 15 of 31

Environmental Science & Technology

292

proportion of the eDNA released by animals may already be fragmented.4 Therefore, the eDNA

293

released into the water column through fish excrements might explain why production rates differ

294

for fragments of different length. This inherent fragmentation of eDNA released by aquatic

295

organisms will need to be considered in future studies aimed at evaluating the relative contribution

296

of DNA fragmentation on the overall decay process.

297

Effect of fragment origin

298

A comparison of the relative abundance of nuclear and mitochondrial eDNA fragments of similar

299

size reveals that nuclear eDNA appears less abundant than the mitochondrial fragment. The

300

estimates of the equilibrium eDNA concentrations showed that the ITS (095bp) fragment was 0.29

301

times less abundant than the COI (096bp) fragment (Figure 1). While some previous studies have

302

shown that nuclear eDNA may be equally or even more abundant than mitochondrial fragments,

303

species-specific differences in the number of ribosomal operons are well known.37,41,49,50

304

Counter to our expectations, the first-order decay constant estimated from the nuclear data was

305

lower for the ITS (095bp) fragment than the COI (096bp) fragment (Figure 2). When considering

306

the absolute estimates for the decay constant for each fragment by treatment combination, we only

307

find a significant difference between the ITS (095bp) and COI (096bp) decay constants in the HD

308

treatment (Supporting Information). Within the HD treatment, the build-up of fish excrements is

309

likely to influence both the pH and microbial activity, which in turn can increase the decay

310

constant.30,31 While we did observe increased decay constants for the mitochondrial fragments in

311

the HD treatments, no significant increase was observed for the ITS (095bp) fragment

312

(Supplementary Information). Consequently, the impact of environmental conditions on the decay

313

process may differ depending on the origin of the DNA fragment.

ACS Paragon Plus Environment

15

Environmental Science & Technology

Page 16 of 31

314

Comparison of eDNA decay models

315

Modelling of aquatic eDNA has commonly assumed an exponential rate of decay.19,30,31 However

316

we found that a Weibull decay model with a decay constant that decreases over time better

317

describes the eDNA decay process. The mean estimate of log(β) was less than zero (implying β

318