Abstract
Despite advances in genomic sequencing and bioinformatics, conservation genomics is still often hindered by a reliance on non-invasive samples. The presence of exogenous DNA and the low quantity and poor quality of DNA in non-invasive samples have been a roadblock to sequencing, thereby limiting the potential for genomic monitoring of endangered species. Recent molecular advances, such as host DNA enrichment, hold promise for facilitating sequencing from non-invasive samples. We used the FecalSeq method to enrich DNA extracted from wild-collected fecal pellets of the imperiled New England cottontail and identified SNPs from 3RAD Sequencing. We obtained SNPs from rabbit pellets, including pellets that were collected in poor environmental conditions and samples that performed poorly with microsatellites. Measures of sequencing success improved with greater amounts of starting DNA and 32% of samples generated SNP genotypes that passed quality control filtering. Genotyping error rates were high, however, and the approach was unable to consistently distinguish unique individuals or matching genotypes, while it was suitable for recovering the expected population structure. Pairing FecalSeq enrichment with RADseq is a promising low-cost method for monitoring wild populations using non-invasive samples in an environmental context, but it may be better suited for informing conservation through population genomics.
Similar content being viewed by others
Introduction
Genetic monitoring is an important tool for conservation1,2,3. In the face of increasing biodiversity loss and environmental change, genetic monitoring is becoming increasingly more important for informing effective conservation and management practices1,4,5. Genetic monitoring refers to a variety of practices to conserve at-risk species, including using molecular markers to identify individuals or species, estimating abundance, range, and survival, and providing estimates of population genetic parameters such as genetic diversity and inbreeding, effective population size or population structure1. Recent advances have dramatically reduced the costs of high-throughput sequencing, and increased computing power has enabled bioinformatic analysis of increasingly larger datasets6,7,8,9,10. The genomic data produced by high-throughput sequencing has promised to transform conservation genetic monitoring2,11,12,13.
So-called second and third generation sequencing data (e.g., from whole genome amplification and reduced representational sequencing approaches) provide higher resolution and different information for genetic monitoring and conservation than traditional molecular markers (e.g., microsatellites, amplified fragment length polymorphisms [AFLPs], and mitochondrial gene sequencing)14. Genomic sequencing data increases the power, accuracy, and precision of population parameters because it provides more markers—100–1000 s of single nucleotide polymorphisms (SNPs) compared to tens of microsatellites2,3,15. For instance, Zimmerman et al.16 compared population genetic analysis using 22 microsatellites from 254 Gunnison sage-grouse (Centrocercus minimus) against 14,091 SNPs from 60 individuals; they found that while both datasets produced concordant results, the SNP dataset provided more precise results, including more fine-grained population structure. Genomic data can also provide better measures of effective population size, an important parameter in wildlife monitoring for understanding the genetic health of a population15. Genomic sequencing data can also answer new questions and provide novel insights3,14. For example, evolutionary responses to pathogens can be tracked and highlight the importance of the role of both specific genes and overall genetic diversity in management decisions, as in the case of the transmissible facial tumor disease in Tasmanian devils (Sarcophjlus harrisii)17,18. Genomic data can enhance understanding of the role of selection on a current population through measuring the adaptive potential of a population, the identification of loci under selection, and linking loci under selection to environmental variables12. Whole genome sequencing also enables new measures important for monitoring the genetic health of a population or species. For instance, whole genome sequencing was used to calculate runs of homozygosity, a measure of inbreeding, and showed that inbreeding depression, not extrinsic factors, such as limited prey abundance, limit the growth of one population of killer whales (Orcinus orca) in the Eastern North Pacific19. This finding informs management decisions, suggesting that translocation to increase genetic diversity in this population may be more beneficial for population growth than dietary supplementation19.
Despite the democratization of genomic data with advances in sequencing, conservation genomics is still often hindered by a reliance on non-invasive samples20,21. This is especially true for many animal species where access to high-quality sources of DNA, such as a tissue or blood samples, is limited due to trapping constraints such as costs, animal body size, cryptic nature of the species, and/or conservation status. Instead, many wildlife studies use non-invasive samples, such as feces, hair, feathers, or eggshells, as a source of DNA22. The poor quality and low quantity of DNA in non-invasive samples22,23,24,25 is a roadblock to sequencing. Additionally, fecal samples contain exogenous DNA from dietary sources and the gut microbiome, which results in wasted sequencing effort and increased cost when targeting the host or endogenous DNA26,27. Due to these limitations, most genetic work with non-invasive samples has relied on AFLPs, mitochondrial DNA, or microsatellites1,28. Methodological advances, however, are beginning to make high-throughput sequencing from non-invasive samples possible26,27,29,30,31,32,33.
Here we obtain genomic sequencing data, specifically SNPs, for an imperiled species from non-invasive fecal samples using FecalSeq enrichment26 followed by restriction site associated DNA sequencing (RAD Seq) with 3RAD library preparation34,35. FecalSeq enrichment decreases the amount of exogenous DNA in a sample, increasing the proportion of host to non-host DNA, which in turn increases the efficiency of sequencing. 3RAD is a modification of the ddRAD approach36, with the advantage that it enables increased individual multiplexing and works well with degraded DNA and low starting amounts of DNA34,35. This is the first study to use FecalSeq with 3RAD and in a wild genetic monitoring context with environmental samples, collect by walking transects, never observing the study species. While FecalSeq and ddRAD have been shown to work with freshly collected fecal samples, where defecation was observed and then collected immediately26, we used fecal samples that were collected in the wild from an imperiled lagomorph, the New England cottontail (Sylvilagus transitionalis), up to 2–5 days after defecation. In this way, here, we illustrate that FecalSeq and 3RAD sequencing produce genomic data from an animal that was never seen by human observers. We also investigated how intrinsic and extrinsic factors—DNA quantity in an extract, environmental sampling conditions, and microsatellite genotyping failures—impacted the success of the FecalSeq enrichment and downstream sequencing. Lastly, we explored applications of this approach for identifying individuals and population structure, and we make recommendations about the applications of this protocol in a genetic monitoring context.
Results
We first optimized the extraction and FecalSeq enrichment protocol using 254 extracts and 102 enrichments (Supplementary Tables S1 and S2). Next, we selected 279 samples to sequence that fulfilled the conditions we sought to test: host and total DNA content, microsatellite failures, and environmental sampling conditions (Supplementary Table S3). These 279 sequenced samples included 137 samples from the extraction and enrichment optimization and another 142 samples that had been previously extracted for occupancy and abundance monitoring which we then enriched prior to sequencing.
Extraction and FecalSeq enrichment optimization
We optimized the extraction protocol to work best with the FecalSeq enrichment (Supplementary Table S1). We found that the Quick-DNA Fecal/Soil Microbe Kits (Zymo Research) with at least a 5-min incubation on the elution step extracted the greatest total DNA (Table 1). Considering only the extracts that were sequenced (and enriched one time only, N = 254) (see Sequencing Success below), those extracted with the Quick-DNA Fecal/Soil Microbe Kits (Zymo Research) with at least a 5-min incubation on the elution step also had the greatest rate of passing our bioinformatics filtering criteria (Fisher’s Exact Test P < 0.001, N = 254), (Fig. 1), indicating they performed the best in sequencing.
The proportion of samples that passed the bioinformatic filtering criteria by extraction kit and protocol. Darker gray represents the proportion that passed filtering and lighter represents the proportion that did not pass filtering. Sample sizes are listed above each column. Zymo (high yield) means that the extraction was allowed to incubate for 5 or more minutes on the elution step. *** = P < 0.001.
To test the efficacy of the FecalSeq Enrichment, we compared total and host DNA both pre- and post-enrichment (Supplementary Table S2) for the 21 samples that had measurable host and total DNA pre- and post-enrichment. The enrichment often removed so much DNA that it was not detectable post-enrichment, reducing our sample size to only 21 for this comparison. We found that there was a significant increase in the proportion of host DNA post-enrichment (median = 0.118, range = 0.004–0.664) compared to pre-enrichment (median = 0. 015, range = 0.002–0.047) (Wilcoxon matched pair test: V = 3, P < 0.001, N = 21) (Fig. 2). Enrichment resulted in an average 15-fold increase in the proportion of host to non-host DNA, with some samples increasing by as much as 70-fold (N = 21). We found that total DNA input for the FecalSeq enrichment was a better predictor (positive relationship, R2 = 0.230, P = 0.028) of the percent increase in the proportion of rabbit DNA than the rabbit DNA input (negative relationship, R2 = 0.066, P = 0.259) or the proportion of host DNA in the input DNA (negative relationship, R2 = 0.151, P = 0.822) (Fig. 3). While none of these R2 values are particularly strong, total DNA in an enrichment reaction was the only factor with a significant and positive effect on enrichment success (Fig. 3).
Matched comparisons of the proportion of rabbit DNA extracted from New England cottontail fecal pellets pre- and post- FecalSeq enrichment. *** = P < 0.001.
The percent increase in the proportion of rabbit DNA due to FecalSeq enrichment against the (a) total DNA (ng) input in the enrichment (b) rabbit DNA (ng) input in the enrichment and (c) the proportion of DNA that is rabbit DNA in the enrichment reaction.
Sequencing success
We were able to generate genomic sequencing data from the enriched fecal sample DNA extractions. After filtering our sequencing data, we retained 88 of the original 279 enriched samples that we sequenced (32%). Across the retained samples, the average alignment rate was 86%, the average number of aligned reads was 759,949, the average coverage was 13.9x, and the average percentage of missing SNPs was 24.3%. 2546 SNPs (of 190,394 total SNPs found) passed filtering. There was a high rate of adapter contamination (partial or complete sequencing of adapter fragments) in our sequencing data; 35–92% of reads contained adapter contamination across the 6 pools. Additionally, many reads (50–85% across 6 pools) had poly-G tails after 90–100 bp, indicating they were short reads likely due to fragmented DNA typical of low quality, non-invasive samples.
Effects of intrinsic and extrinsic factors on FecalSeq enrichment and downstream sequencing
The amount of both total and host DNA input for the FecalSeq enrichment had a positive relationship with measures of sequencing success (Fig. 4). Total DNA input was significantly and positively correlated with the number of reads that aligned (R2 = 0.15, P < 0.001) (Fig. 4A) and was significantly higher in samples that passed filtering (mean = 16.1 ng/uL, N = 88) than those that did not pass filtering (mean = 7.8 ng/uL, N = 189) (Kruskal–Wallis χ2 = 35.378, df = 1, P < 0.0001, N = 277) (Fig. 4C). Rabbit DNA input was also significantly and positively correlated with the number of reads that aligned (R2 = 0.123, P < 0.001) (Fig. 4B), and was significantly higher in samples that passed filtering (mean = 171.4 pg/uL, N = 87) than those that did not pass filtering (mean = 92.3 pg/uL, N = 172) (Kruskal–Wallis χ2 = 38.753, df = 1, P < 0.0001, N = 259) (Fig. 4D).
The effect of total DNA (A, C) and rabbit DNA (B, D) in a sample prior to enrichment on two measures of sequencing success: the number of reads that aligned to the reference genome (A, B) and whether a sample passed filtering (C, D). *** = P < 0.001.
Sequencing success varied across environmental sampling conditions and microsatellite genotyping failure rates. We found that while DNA was recoverable from samples collected in poorer sampling conditions, those samples did not perform as well with the sequencing as samples collected in ideal conditions. The number of reads that aligned varied significantly across environmental conditions (Kruskal–Wallis χ2 = 13.388, df = 2, P = 0.0012, N = 86), driven by lower numbers of aligned reads between the rain category and other 2 categories (see Supplementary Table S6 for P values of the Post-hoc Nemenyi’s Pairwise Comparisons Test) (Fig. 5B). The proportion of samples that passed filtering also varied significantly across environmental conditions (Fisher’s Exact Test P = 0.0240, N = 86), with more samples collected in ideal conditions passing the filtering (Supplementary Table S7). Likewise, we found that while DNA was recoverable from samples with greater rates of microsatellite amplification failures, those samples did not perform as well in the sequencing as samples with fewer microsatellite amplification failures. The number of reads that aligned varied significantly across microsatellite failure rates (Kruskal–Wallis χ2 = 8.6303, df = 2, P = 0.0134, N = 114), but none of the P values for pairwise comparisons reached significance when a Bonferroni correction was applied (see Supplementary Table S8 for P values of the Post-hoc Nemenyi’s Pairwise Comparisons Test) (Fig. 5A). The proportion of samples that passed bioinformatics filtering also varied significantly across microsatellite failure conditions (Fisher’s Exact Test P < 0.0001, N = 114), with more samples lacking microsatellite failures passing the filtering criteria (Supplementary Table S9). Interestingly, the cutoff used to exclude a poor sample from microsatellite genotyping in our 15-locus microsatellite panel is four or more failed microsatellite loci37, and we found that no samples with four or more failed microsatellites passed the bioinformatics filtering criteria either.
The number of sequenced reads that aligned to the reference genome across (A) samples with previously known microsatellite failure and (B) samples collected in different environmental sampling conditions. For the microsatellite failures, the x-axis indicates the number of failures out of 15 autosomal microsatellites. * without a bar represents the significance for the Kruskal–Wallis test comparing the whole model, * = P < 0.05. * with a bar represents the significance values of the Post-hoc Nemenyi’s Pairwise Comparisons Test with the a Bonferroni correction applied, * = P < 0.0167.
FecalSeq applications: genotyping for individual identification and population structure
In our dataset of 37 unique individuals (known from microsatellite genotyping), Colony inferred an average allelic dropout rate of 0.146 and average false allele rate of 0.018 in our SNP genotypes. Due to the high inferred allelic dropout rate, we further filtered out SNPs by three different error rate criteria (Table 2). Even with the strictest SNP filtering scheme, Colony made incorrect matches for duplicate samples and missed many correct duplicates (Supplementary Table S10). Table 2 shows the number of SNPs retained with each of these filtering criteria. While SNPs filtered with the loosest error rate criteria resulted in genotypes for which Colony was able to identify the most duplicated dyads (29–30%), this loose filtering also resulted in the most inaccurate assignments (6–23%) (Table 2). The middle filtering scheme paired with Colony’s best method for inferring duplicates was the only scheme that did not falsely identify a pair of duplicates. However, it was only able to identify 14–18% (11–14 out of 80) of all duplicates. With this filtering scheme, both extracted rabbit DNA (Kruskal–Wallis χ2 = 4.220, df = 1, P = 0.037, N = 54) and total DNA (Kruskal–Wallis χ2 = 0.012, df = 1, P = 0.012, N = 55) were significantly higher in duplicated samples that were correctly identified as duplicates than those that were missed (Fig. 6).
Comparison of extracted (A) rabbit DNA and (B) total DNA in a sample prior to enrichment of known pairs of matching samples between those that Colony was and was not able to infer as duplicates (matches). * = P < 0.05.
Comparison of nine samples from naturally occurring populations revealed population structure consistent with expectation—separating the samples from Connecticut and Maine (Fig. 7). With this smaller dataset, 1996 of the starting 2564 SNPs remained after removing monomorphic loci to construct the PCA. While the sampling sites were clearly separated along the vertical axis, one sample from an individual from Maine separated horizontally along PC1, potentially signaling a genotyping artifact.
Population structure of New England cottontail individuals from naturally occurring populations genotyped at 1996 SNPs, showing the expected separation of the two sampling sites—Connecticut (CT) and Maine (ME).
Discussion
We successfully enriched and sequenced the DNA extracted from New England cottontail fecal pellets collected in the wild 2–5 days post defecation. Using the 3RAD protocol, we obtained SNPs and genotyped individuals to determine their unique identity and for population structure. While our results demonstrate, to our knowledge, for the first time the promise of this approach (FecalSeq) for monitoring wild populations in an environmental context (i.e., without tracking and observing individuals), they also show limitations to widespread application of this approach for genomic monitoring via genotyping. The majority of studies that obtain genomic sequencing data from wild-collected fecal samples use freshly collected fecal samples26,27,31,32,38, but these sampling methods may not be possible for many endangered species that rely on non-invasive sampling. While Taylor et al.33 uses pellets collected off snow 2–5 days post defecation for whole genome sequencing the goal of their analysis is to gain information about population structure, not genetic monitoring through genotyping. Genotyping for genetic monitoring of a population requires processing many more samples and greater accuracy in SNP calls than does population genetics/genomics. Overall, only 32% of the enriched samples in our study yielded sufficient quality sequencing data to pass our quality control bioinformatics filtering, and samples collected in less-than-ideal environmental conditions generally were not retained after filtering. Even among the samples that passed SNP quality filtering, the majority of multi-locus genotypes were of insufficient quality to correctly distinguish unique individuals or identify matching samples. These findings highlight the continued challenge of sequencing non-invasive samples26,27,29,30,38,39,40,41 and suggest genomic monitoring may only be feasible on a subset of field-collected samples that yield the highest possible amounts of total and host DNA.
We found that the FecalSeq enrichment protocol performed better with higher inputs of total DNA. While Chiou and Bergey26 recommend avoiding bead bashing extraction kits due to further fragmentation of DNA, we found that a bead bashing kit, Quick-DNA Fecal/Soil Microbe Kits (Zymo Research), gave us the highest starting total DNA which was associated with improved enrichment, and produced the greatest sequencing success (measured by the proportion of samples that passed filtering). We suggest that labs interested in using the FecalSeq enrichment protocol work to optimize the total DNA of extractions. To increase our total DNA yield from extractions, we tested different extraction kits, optimized our extraction protocol, and concentrated extractions with a vacufuge. These steps ought to be optimized individually for each study species and lab. Another possible option to increase starting DNA is to combine extractions from multiple pellets of the same sample. Combining pellets from the same sample could increase starting DNA and the success of the FecalSeq enrichment. However, this introduces potential error if the two pellets are from two different rabbits (which in our system happens infrequently, but might be more problematic in other systems). Performing multiple extractions and concentrating the larger volume of extract prior to the FecalSeq enrichment increases laboratory effort, increasing both the time and cost per sample. Additionally, for some study species, increasing the number of extractions from a single sample is not feasible.
We found that samples with the highest starting rabbit and total DNA performed the best—they had higher alignment rates to the reference genome and were more likely to be correctly matched as duplicate samples. Unfortunately, due to overlap between successful and unsuccessful categories, we were not able pinpoint a specific cutoff value for the required amount of host and/or total DNA required for successful sequencing. Based on the mean starting total DNA of samples that did and did not pass filtering, a cutoff between 15 and 20 ng of DNA per enrichment reaction is likely a good starting point. Because total DNA was a better indicator of both enrichment efficiency and sequencing success, quantification of host DNA might even be skipped when screening samples for enrichment and sequencing, which would reduce both time and cost. Additionally, we found that the enrichment was not able to fully overcome poor environmental conditions. While we were able to obtain some genomic sequence data from samples collected in poor environmental conditions and samples that had high rates of microsatellite failure, the FecalSeq enrichment was not able to completely overcome the lower quantity and quality of the DNA in these samples and did not typically produce better results than obtained from microsatellite genotyping. If a sample is known to have failed microsatellite genotyping, it is not a good candidate for this enrichment procedure. While some samples from poor environmental conditions were retained after filtering, they had fewer aligned reads. Accordingly, our findings suggest that this protocol is best used on the highest quality fecal-extracted DNA samples, with the highest starting amounts of DNA.
Another important aspect of this methodology to consider when planning sequencing is the high amount of wasted sequencing effort, even with the relative increase in host DNA and accompanying decrease in exogenous DNA. We obtained 85% alignment rates, but we lost many reads due to adapter contamination. The high rate of adapter contamination is mostly likely due to the highly fragmented nature of fecal DNA. It is possible that the use of bead-bashing extraction kits may also have increased the fragmentation of the DNA, but the samples extracted with the bead beading kit (Zymo) were most likely to be retained after filtering. Anticipating higher rates of adapter contamination with fecal samples compared to non-degraded samples is an important step in planning the number of samples to include in a sequencing lane. Size selection for longer fragments may also aid in reducing the number of reads with adapter contamination. We chose the size selection range of 250–800 bp in an effort to increase the amount of DNA present, decreasing the number of cycles necessary in the final PCR to reach the DNA quantity and concentration minimum for sequencing. However, moving forward, we would suggest size selection for longer DNA fragments, since smaller fragments will likely contain adapter contamination. Additionally, the preferential binding of smaller fragments to Illumina sequencers may have contributed to our high levels of adapter contamination. Another option to mitigate the large amount of wasted sequencing is to increase the amount of sequencing effort (e.g., 3 lanes of sequencing, rather than 2), although this obviously comes with a substantial increase in cost.
In terms of testing the applications of the FecalSeq enrichment method paired with 3RAD sequencing for genomic monitoring, the SNP data were of sufficient quality to identify population structure. Because our experimental design focused on exploring the ability of the FecalSeq methodology to genotype samples from a variety of environmental conditions, not on obtaining samples to measure New England cottontail population genetics, we are unable to calculate a variety of population genetic measures and cannot make further evaluation of the performance of these data for those purposes. While we did find the expected population structure, we were unable to produce consistently accurate genotypes for individual identification. When comparing samples from the same individual, we found that there was both a high rate of false positives and false negatives, despite strict filtering of SNPs with high error rates. While filtering to reduce genotyping error rates could substantially or even completely reduce the false positives, it came at the expense of a very low rate of detecting true matches (14–18%). Increased sequencing effort to increase coverage might decrease error rates. However, this will increase cost and wasted sequencing effort. It is important to note that in the original description of the FecalSeq protocol, the authors only discuss its utility for population-level questions and not its use for genotyping26. To our knowledge, prior to this study, the approach had not yet been successfully applied in the context of individual genotype matching.
Other methods for obtaining SNP genotypes from non-invasive samples have had more demonstrated success, including genome-wide bait capture methods39 and Genotyping in Thousands by Sequencing (GT-Seq) panels42. Both of these methods require upfront lab work to design species specific markers, baits or primers, in order to genotype individuals39,42. GT-Seq panels have been shown to be effective for genotyping individuals using poor quality samples: fecal samples40,43 hair samples43, and cloacal swabs44. Burgess et al.43 found a high rate of agreement (99.5%) of SNPs between matched tissue samples and non-invasive fecal and hair samples from the same individuals using a GT-Seq panel. Comparing invasive and non-invasive samples, Hayward et al.40 found that fecal samples had the highest average percent of missing data by locus among sample types and highest rates of individuals with greater that 50% missing loci, but they were still able to successfully genotype 62.9% of non-invasive fecal samples using a GT-Seq panel. Bait capture and GT-seq with non-invasive samples offer promise for SNP genotyping and population genomic measures of the genetic health of populations for genetic monitoring, but still appear to only work for the highest quality non-invasive samples40,45. However, they cannot match the faster speed of FecalSeq and 3RAD for population genetic measures. Alternatively, small-scale sequencing to screen and balance samples prior to final RAD Seq library preparation and sequencing is another promising avenue for reliable SNP genotyping from non-invasive samples38. While there may be room for methodological improvements, it remains uncertain whether FecalSeq with RAD Seq can overcome the limitations of poor sample quality and quantity to be consistently reliable for individual genotyping.
In conclusion, we have shown the FecalSeq enrichment is compatible with 3RAD genotyping and illustrated its potential to inform conservation of rare and cryptic species utilizing samples collected from wild animals that were never observed. While this approach had limited success across the full suite of samples on which it was tested in our study (including samples collected in poor environmental conditions known to promote DNA degradation and which failed prior microsatellite genotyping), it had greater success for the samples with relatively higher amounts of total and host DNA, corresponding to samples collected in ideal sampling conditions. In this way, the method shows promise, although it may only be appropriate for a subset of samples collected in the context of genomic monitoring of a wild population. Modifications of the lab protocols, including multiple extractions to increase starting amounts of DNA, selecting larger fragments for sequencing to avoid adapter contamination, or increasing sequencing effort, may improve the efficacy of the approach. Alternately, combining the fecal enrichment with targeted sequencing, such as a GT-Seq panel, may provide greater success for a broader suite of environmental samples in a monitoring context.
While limited in its application in an environmental context, FecalSeq enrichment combined with 3RAD is a relatively inexpensive method for obtaining reliable genetic sequencing data from non-invasive fecal samples. At this time, further work is needed to generalize the application of this method for genomic monitoring of wild populations more broadly (in particular regarding questions requiring comparison of individual genotypes). Nonetheless, our results indicate this method is well suited for monitoring using samples collected in ideal conditions and for population genetics studies more broadly. In this way, the approach presents an advance in conservation genetics, as obtaining sequence data improves the precision of population genetics measures and enables new measures to be calculated, such as inbreeding estimates, that are important for conservation and management, especially of rare and cryptic species, which require non-invasive monitoring.
Methods
Study system
The New England cottontail is a species of greatest conservation need throughout its range in the northeastern U.S.46,47 and a state-listed endangered species in New Hampshire and Maine. It is considered globally vulnerable to extinction by the IUCN Red List. The species has been managed since 2011 by a Conservation Strategy, developed to restore its rare shrubland habitat and recover the species46,47. Management practices for New England cottontails focus on restoring and creating new early successional forest habitat, monitoring occupancy and abundance of New England cottontails throughout their range, and reintroductions and augmentation of populations from captively bred rabbits46,47. Despite a decade of these conservation efforts and tens of thousands of acres of restored habitat, the species continues to decline48.
The decline of the New England cottontail is tightly associated with the decline of the ephemeral, early successional forests and shrublands on which it strictly depends49,50,51. The species range has severely (> 86%) contracted relative to historical levels52,53 and the range-wide number of rabbits is estimated to be as low as 3000 individuals48. Currently, remnant wild populations of New England cottontail exist in geographically disjunct and genetically distinct metapopulations in eastern New York/western Connecticut, eastern Connecticut, eastern Massachusetts on Cape Cod, and southeastern New Hampshire and Maine48,52,54. Within these areas, metapopulations are small, fragmented, and increasingly isolated, with recent reductions in gene flow leading to critically low effective sizes and low genetic diversity55,56,57. Low population numbers paired with population genetic data showing harmful effects of small, fragmented habitat patches suggest that the genetic health of New England cottontail populations is suffering.
A systematic range-wide monitoring protocol is a critical component of the species’ Conservation Strategy. Due to the cryptic nature of New England cottontails and logistical and ethical limitations of trapping, monitoring occupancy and abundance of cottontails occurs through pellet surveys conducted during the winter months from November to April58. Pellets are collected on freshly fallen snow, facilitating detection of pellets and preservation of DNA52,59. Mitochondrial DNA is used to determine the species identity of a pellet to measure occupancy59 and microsatellite genotyping is used to identify individuals to estimate abundance and track reproduction of founders and patch-level relatedness60,61. Given the limitations of microsatellite data for reconstructing wild pedigrees and making inference about inbreeding and genetic diversity, genome sequencing data would greatly enhance our knowledge about the genetic health of wild populations.
Study sites and samples
Fecal samples were collected during winter pellet surveys conducted in Connecticut, New Hampshire and Maine in 2019–2022. Following New England Cottontail occupancy and abundance monitoring protocols, pellets were collected on snow from November to April while walking along loose transects within delineated habitat patches53,58,62. Pellets were stored in vials on ice during collection until transferred to a − 20 °C freezer for long-term storage. We worked with pellets that were previously confirmed to be from a New England cottontail. Species identification was performed by either sequencing a 350 bp region of the mitochondrial cytochrome b gene63 or microsatellite genotyping where microsatellite length is diagnostic of species identification37 (Kovach, unpublished data). For a subset of the pellet samples (N = 114), individual identify was determined by genotyping with 15 autosomal microsatellites64,65 and 1 sex-specific locus66, following Bauer et al.37. We included some samples that were determined to be from the same individual. Overall, the 279 enriched and sequenced samples represent 170 different pellet collections, from at least 62 individual rabbits and up to 124 rabbits (if each collection where individual identity was not obtained were a different rabbit).
Extraction optimization
We first optimized our extraction and enrichment protocol to determine which extraction method performed best with the FecalSeq enrichment using 254 extracts (Supplementary Tables S1 and S2). Our lab has had the most success in producing microsatellite genotypes with a bead beating extraction kit, Quick-DNA Fecal/Soil Microbe Kits (Zymo Research, Irvine, California), but in their FecalSeq Enrichment protocol, Chiou and Bergey26 caution against using a bead beating kit for extraction since it could further fragment the DNA. Therefore, we tested our current extraction kit against two extraction kits that use chemical lysis, not beat beating. We tested the QIAamp DNA Stool Mini Kit (Qiagen, Valencia, California), which Chiou and Bergey26 used and the NucleoSpin Tissue Mini kit for DNA from cells and tissue (Macherey–Nagel, Düren, Germany). Manufacturers’ protocols were followed for each kit, with a few modifications that were tested to determine optimal protocol (see Supplementary Table S1 for description of all trials). For the Qiagen kit, some samples used the ‘Isolation of DNA from Larger Volumes of Stool’ modification in the manufacturer handbook, while others used 1 mL of Inhibit X regardless of weight; samples were either vortexed for one 1 min, manually shaken for 5 min or left on a shaker overnight with the Inhibit X. For the Macherey–Nagel kit, pellets were incubated on a heat block for 1, 3, 5 h, or overnight and eluted at 55 °C. For the Zymo Research kit, samples were vortexed from 5 to 60 min and eluted in various volumes (50 or 100 uLs), buffers (supplied buffer, low EDTA TE, TE), incubation times (0-, 5-, and 15-min incubation), and steps (double elution, two-part elution). To test each of these conditions, we used two pellets from the same collection or a single pellet divided in half and compared the DNA content of extracts. For each condition, we made 7–9 comparisons. For example, to test elution buffer incubation times of 5 min versus 15 min, we divided 9 pellets in half, did a full extraction with each half pellet, and then used an elution buffer incubation time of 5 min on one of the half-pellet extracts and used an elution buffer incubation time of 15 min the other half-pellet extract from the same pellet. We examined trends between the DNA in the two pellets from the same collection or from the two half-pellet extracts (Supplementary Table S1), but with such small sample sizes (7–9 extracts in each condition) we did not perform statistical tests on each condition. We compared the mean total and rabbit DNA from each extraction kit and proportion of samples that passed our bioinformatics filtering (see Bioinformatics below) for each kit using Fisher’s Exact Tests.
We initially used 102 of the extracts to optimize the FecalSeq Enrichment (see below and Supplementary Table S2). Based on the results of the FecalSeq Enrichment optimization, highlighting the importance of total DNA for enrichment efficiency, we then chose an additional 142 samples that had been previously collected for species monitoring and extracted with the Zymo kit following manufacturers protocols. We chose sample that would fit the experimental design for the intrinsic and extrinsic factors we sought to test (See Effect of Intrinsic and Extrinsic Factors on FecalSeq Enrichment and Downstream Sequencing below). While we were interested in testing the outcome of the different extraction conditions on sequencing performance, we eliminated many of the extracts based on their DNA quantity and enrichment outcome. Only 137 of the 254 extracts used for optimizing our protocol were sequenced (Supplementary Table S1).
Quantification
Total DNA concentration and yield in each extract were quantified using a Qubit dsDNA HS Assay (Life Technologies) on a Qubit 3.0 Fluorometer (Invitrogen, Waltham, Massachusetts). The amount of New England cottontail DNA in each extract was quantified with quantitative PCR (qPCR). We used European rabbit (Oryctolagus cuniculus) primers (Forward: 5′- GCCAGAGGAGAAATGAGCT-3′; Reverse: 5′-GGGCCTTTTCATTGTTTTCCA-3′) for the C-MYC gene (NCBI Reference Sequence: XM_017340986.1) to amplify host DNA present in the extract and compared it to a standard curve created from a New England cottontail tissue sample. The standard curve was run in triplicate (31–2000 pg/uL) and samples were run in duplicate alongside negative controls (sensu67). The amount of MBD2-Fc bound magnetic beads used in the FecalSeq enrichment is determined by the amount of host DNA present in the extract26.
Enrichment
We conducted optimization of the FecalSeq enrichment protocol, following the Chiou and Bergey26 protocol with slight modifications. Using 102 extracts we tested the impact of the extraction buffer, rotation time, age of MBD2-Fc bound magnetic beads, and serial enrichment (enriching the enriched product a second time) on the efficiency of the enrichment (see Supplementary Table S2 for description of all trials). To test the performance of the FecalSeq enrichment, we compared the proportion of host (rabbit) DNA out of total DNA pre- and post-enrichment with a Wilcoxon matched pair test. Host DNA was quantified with qPCR, total DNA was quantified with a Qubit, and the proportion of host DNA was estimated by dividing the rabbit DNA amount obtained from the qPCR by the total DNA amount obtained from the Qubit (sensu26). (See Quantification above for details of qPCR and Qubit methods.) For each condition, we made 7–15 comparisons. For example, we enriched 15 extracts with freshly prepared MBD2-Fc bound magnetic beads and then six days later enriched another portion of the same 15 extracts with MBD2-Fc bound magnetic beads that were six days old. We examined trends between the efficiency of the enrichment between conditions, but with such small sample sizes (7–15 extracts in each condition) we did not perform statistical tests on each condition. We also examined how the amount of host DNA, total DNA, and proportion of host DNA impacted the efficacy of the FecalSeq enrichment by comparing the coefficient of determination for each linear model.
We enriched the host DNA using the NEB Next Microbiome DNA Enrichment Kit (New England Biolabs, Ipswich, Massachusetts)68 following the FecalSeq enrichment protocol26. Briefly, this enrichment protocol uses MBD2-Fc bound magnetic beads, which bind preferentially to eukaryotic DNA to separate the host DNA from non-host DNA. We used a vacufuge to concentrate extracts with low total DNA prior to the FecalSeq enrichment (Supplementary Table S3). If an extract had less than 300 nM of total DNA in the enrichment reaction, it was concentrated with a vacufuge. After optimization, we followed the original Chiou and Bergey FecalSeq protocol26 exactly because none of our optimization trials impacted the efficacy of the FecalSeq enrichment. Only 69 of the samples used for testing the efficacy of the FecalSeq enrichment were sequenced (Supplementary Table S2). The other 33 samples did not fit the criteria for the sequencing because they did not fit the experimental design for the intrinsic and extrinsic factors we sought to test (See Effect of Intrinsic and Extrinsic Factors on FecalSeq Enrichment and Downstream Sequencing below). We enriched another 142 samples that were previously extracted for species monitoring.
3RAD library prep and sequencing
We prepared a library for sequencing with 279 enriched fecal DNA extracts (Supplementary Table S3). This library included 142 samples that were previously extracted for species monitoring and 137 that were used for optimization. These 279 enriched samples represented samples with low and high total DNA (from undetectable up to 75 ng/uL), samples from a variety of environmental sampling conditions, samples with known microsatellite failures, and samples that were extracted with three different extraction kits. We followed a modified double-digest RAD Seq approach, 3RAD34,35 with modifications to accommodate our low starting DNA input (Supplementary Tables S4 and S5). We digested our enriched samples with the restriction enzymes XBaI, EcoRI-HF, and NheI-HF. Due to low input DNA (< 1 ng of rabbit DNA), we increased the volume of reagents and DNA in the digestion (Supplementary Table S5). We prepared pools of samples and size selected for 250–800 bp fragments using a Blue Pippin (Sage Science, Beverly, MA). We then determined the concentration of each pool with a Qubit dsDNA BR Assay (Life Technologies, Carlsbad, California) and average fragment length with a TapeStation 2200 (Agilent, Santa Clara, CA). Following Chiou & Bergey26 for the final PCR amplification, we ran each pool in duplicate to minimize PCR biases and attempted to limit the number of PCR cycles. For the final PCR amplification, we used the Kapa Hifi Hot Start Ready Mix (Roche Diagnostics, Indianapolis, Indiana). Pools required 24–30 cycles to reach the molarity (10 nM) required for sequencing. We normalized and combined all 6 pools into one library that was submitted for sequencing on two lanes of an Illumina NovaSeq SP (800 million reads) with paired-end 150 bp reads and 15% PhiX spike in at the Genomic Sciences Library at North Carolina State University.
Bioinformatics
Sequenced reads were demultiplexed, trimmed to remove adapters sequences, and filtered to remove low-quality reads (average phred score less than 20 across a 23-bp sliding window) using process radtags in Stacks269,70,71. Due to the degraded nature of fecal DNA, we retained reads that lacked RAD cutsites. Next, reads were aligned to the brush rabbit genome (Sylvilagus bachmani) with BWA-MEM72. We used gstacks in Stacks2 to call SNPs69,70,71. Using vcftools for SNP filtering73, we removed samples with less than 100,000 aligned reads and greater than 65% missing data. We removed SNPs with a minor allele count less than 3, less than 5 × coverage, greater than 70 × coverage, and if they appeared in < 70% of samples. These filtering criteria were determined by following standard filtering procedures (e.g.45,74,75) and testing a range of parameters on our dataset76.
Effect of intrinsic and extrinsic factors on FecalSeq enrichment and downstream sequencing
We were interested in how the amount of input host (rabbit) and total DNA affected the success of sequencing. We sought to find a cutoff value at which we were able to reliably obtain sequencing DNA, so that we could use this to screen samples in future lab work with the FecalSeq enrichment. To evaluate the role of intrinsic and extrinsic conditions on the FecalSeq enrichment protocol, we first used linear models to test the relationship between both host (rabbit) DNA and total DNA in a sample and the number of sequenced reads that aligned to the reference genome. We log transformed total DNA, rabbit DNA, and the number of reads that aligned in order to satisfy model assumption and properly fit the linear models. We also used Kruskal–Wallis Rank Sum Tests to compare rabbit and total DNA between samples that did and did not pass filtering.
Poor environmental conditions, such as warmer and wetter weather lead to DNA degradation and are associated with higher failure rates of microsatellite genotyping and species identification from mitochondrial DNA59,77 With increasingly warm winters in New England78,79, poor environmental sampling conditions are becoming an increasing problem. Thus, we sought to test if the FecalSeq enrichment could overcome poor environmental sampling conditions by comparing samples collected in ideal conditions (fresh snow in the last 4 days without thawing temperatures), potentially rainy and thereby poor conditions (rain since last snow, unknown if pellet was rained on due to sampling procedure), and warmer and damp conditions (collected off melty, wet snow). We used Kruskal–Wallis Rank Sum Tests with Post-hoc Nemenyi’s Comparison Test and Fisher’s Exact Tests to compare these categories of environmental conditions with measures of sequencing success (number of sequenced reads that aligned to the reference genome and whether a sample passed filtering).
Additionally, we tested the ability of enrichment to obtain DNA from samples with known microsatellite failures. For sequencing and microsatellite genotyping comparisons, we used the same extract from a single pellet. To examined if the FecalSeq enrichment would work on samples that were known to perform poorly with microsatellite genotyping, we sequenced samples that had previously failed individual identity genotyping with 4 or more microsatellites failing to amplify, samples where 1–3 microsatellites failed to amplify, and samples with no microsatellites failing to amplify. We chose these categories of microsatellite failure rates because when 4 or more microsatellites fail to amplify, we cannot reliably genotype the sample. Thus, the 4 or more category represents a sample that failed microsatellite genotyping, while the 1–3 category represents a sample that can be genotyped but is missing information. We used Kruskal–Wallis Rank Sum Tests with Post-hoc Nemenyi’s Comparison Test and Fisher’s Exact Tests to compare these three categories microsatellite failures with measures of sequencing success (number of sequenced reads that aligned to the reference genome and whether a sample passed filtering).
FecalSeq applications: genotyping for individual identification and population structure
To access the quality of our sequencing data and application of the FecalSeq protocol for genotyping, we assessed the similarity in SNP genotypes of multiple samples from the same individual (see Supplementary Table S10). First, we narrowed our dataset down to only unique rabbits (individual identity known from previous microsatellite genotyping) and calculated error rates and allele frequencies in Colony 2.0.7.080,81. We then further filtered the SNPs with three different sets of error rate criteria (loose, middle, and strictest stringency; Table 1). Next, we used our filtered dataset of 88 samples, including 80 pairwise sets of duplicates (i.e., if there were two samples from the same individual, there was one pairwise duplicate, and if there were three samples from the same individual, there were three pairwise duplicates, etc.) in Colony to infer matching individuals or “duplicates/clones”. Colony uses two different methods to infer duplicates. The ‘Best’ method uses full likelihood, taking allele frequencies and relationships between samples into account, while the ‘Pairwise’ method examines the two genotypes in isolation82. We tallied and compared the number of correct and incorrect duplicate pairs inferred by Colony for each error rate scheme. Because Colony was not able to identify all of the duplicate pairs for any of the filtering criteria, for the error rate and duplicate identification method that produced no incorrect matches (middle stringency error rate scheme and 'Best' method of duplicate identification), we used Kruskal–Wallis rank sum tests to compare the extracted rabbit DNA and total DNA of samples that Colony was able to infer as duplicates to samples that Colony did not infer as duplicates.
Finally, we narrowed down our dataset to only samples from unique rabbits from naturally occurring populations and performed a principal component analysis (PCA) to determine if our dataset showed the expected population structure (see Supplementary Table S10). Many of our samples were from populations with reintroduced rabbits from the captive breeding program and we did not include them in this analysis. These rabbits originated from various source populations, with interbreeding between source populations occurring on reintroduction sites, so we do not expect them to exhibit spatially explicit population structure. We performed the PCA comparing rabbits in naturally occurring populations in Connecticut to naturally occurring populations in Maine with adegenet83 and ade484 packages in R. Based on known population structure54, we expected that rabbits from Maine and Connecticut would separate out in the PCA. All statistical analysis and data visualization were performed in R85.
Data availability
Sample metadata necessary for the analyses in this manuscript are included in the supplemental data. Raw sequencing data is available in the NCBI Short Read Archive (Accession numbers SAMN41003261- SAMN41003539). No custom scripts were used and all statistical packages and bioinformatics pipelines are described in the manuscript methods.
References
Schwartz, M. K., Luikart, G. & Waples, R. S. Genetic monitoring as a promising tool for conservation and management. Trends Ecol. Evol. 22, 25–33. https://doi.org/10.1016/j.tree.2006.08.009 (2007).
Allendorf, F. W., Hohenlohe, P. A. & Luikart, G. Genomics and the future of conservation genetics. Nat. Rev. Genet. 11, 697–709 (2010).
Shafer, A. B. A. et al. Genomics and the challenging translation into conservation practice. Trends Ecol. Evol. 30, 78–87. https://doi.org/10.1016/j.tree.2014.11.009 (2015).
Frankham, R. Where are we in conservation genetics and where do we need to go?. Conserv. Genet. 11, 661–663 (2010).
Frankel, H. O. Genetic conservation: Our evolutionary responsibility. Genetics 78, 53–65 (1974).
Stein, L. D. The case for cloud computing in genome informatics. Genome Biol. 11, 207 (2010).
Metzker, M. L. Sequencing technologies the next generation. Nat. Rev. Genet. 11, 31–46 (2010).
Mardis, E. R. A decade’s perspective on DNA sequencing technology. Nature 470, 198–203 (2011).
Satam, H. et al. Next-generation sequencing technology: Current trends and advancements. Biology 12, 997 (2023).
Goodwin, S., McPherson, J. D. & McCombie, W. R. Coming of age: Ten years of next-generation sequencing technologies. Nat. Rev. Genet. 17, 333–351 (2016).
Theissinger, K. et al. How genomics can help biodiversity conservation. Trends Genet. 39, 545–559 (2023).
Hohenlohe, P. A., Funk, W. C. & Rajora, O. P. Population genomics for wildlife conservation and management. Mol. Ecol. 30, 62–82 (2021).
Segelbacher, G. et al. New developments in the field of genomic technologies and their relevance to conservation management. Conserv. Genet. 23, 217–242. https://doi.org/10.1007/s10592-021-01415-5 (2022).
Luikart, G., England, P. R., Tallmon, D., Jordan, S. & Taberlet, P. The power and promise of population genomics: From genotyping to genome typing. Nat. Rev. Genet. 4, 981–994. https://doi.org/10.1038/nrg1226 (2003).
Hohenlohe, P. A., Funk, W. C. & Rajora, O. P. Population genomics for wildlife conservation and management. Mol. Ecol. 30, 62–82 (2021).
Zimmerman, S. J., Aldridge, C. L. & Oyler-Mccance, S. J. An empirical comparison of population genetic analyses using microsatellite and SNP data for a species of conservation concern. BMC Genomics https://doi.org/10.1186/s12864-020-06783-9 (2020).
Epstein, B. et al. Rapid evolutionary response to a transmissible cancer in Tasmanian devils. Nat. Commun. 7, 12684 (2016).
Stahlke, A. R. et al. Contemporary and historical selection in Tasmanian devils (Sarcophilus harrisii) support novel, polygenic response to transmissible cancer. Proc. R. Soc. B 288, 20210577 (2021).
Kardos, M. et al. Inbreeding depression explains killer whale population dynamics. Nat. Ecol. Evol. 7, 675–686 (2023).
Barbosa, S., Hendricks, S. A., Funk, W. C., Rajora, O. P. & Hohenlohe, P. A. Wildlife population genomics: Applications and approaches. In Population Genomics: Wildlife 3–59 (Springer, 2020).
Andrews, K. R., De Barba, M., Russello, M. A. & Waits, L. P. Advances in using non-invasive, archival, and environmental samples for population genomic studies. In Population Genomics: Wildlife (eds Hohenlohe, P. A. & Rajora, O. P.) 63–99 (Springer International Publishing, 2018).
Beja-Pereira, A., Oliveira, R., Alves, P. C. & Schwartz, M. K. Advancing ecological understandings through technological transformations in noninvasive genetics. Mol. Ecol. Resour. 9, 1279–1301 (2009).
Deuter, R., Pietsch, S., Hertel, S. & Muiler, O. A method for preparation of fecal DNA suitable for PCR. Nucleic Acids Res. 23, 3800–3801 (1995).
Kohn, M. H. & Wayne, R. K. Facts from feces revisited. Trends Ecol. Evol. 12, 223–227 (1997).
Taberlet, P., Waits, L. P. & Luikart, G. Noninvasive genetic sampling: Look before you leap. Trends Ecol. Evol. 14, 323–327 (1999).
Chiou, K. L. & Bergey, C. M. Methylation-based enrichment facilitates low-cost, noninvasive genomic scale sequencing of populations from feces. Sci. Rep. 8, 1975 (2018).
Perry, G. H., Marioni, J. C., Melsted, P. & Gilad, Y. Genomic-scale capture and sequencing of endogenous DNA from feces. Mol. Ecol. 19, 5332–5344 (2010).
Allendorf, F. W. Genetics and the conservation of natural populations: Allozymes to genomes. Mol. Ecol. 26, 420–430 (2017).
Andrews, K. R. & Luikart, G. Recent novel approaches for population genomics data analysis. Mol. Ecol. 23, 1661–1667 (2014).
Carroll, E. L. et al. Genetic and genomic monitoring with minimally invasive sampling methods. Evol. Appl. 11, 1094–1119 (2018).
Snyder-Mackler, N. et al. Efficient genome-wide sequencing and low-coverage pedigree analysis from noninvasively collected samples. Genetics 203, 699–714 (2016).
Gutleb, D. R. et al. Non-invasive genotyping with a massively parallel sequencing panel for the detection of SNPs in HPA-axis genes. Sci. Rep. https://doi.org/10.1038/s41598-018-34223-y (2018).
Taylor, R. S. et al. Whole genome sequences from non-invasively collected caribou faecal samples. Conserv. Genet. Resour. 14, 53–68 (2022).
Graham, C. F. et al. Impacts of degraded DNA on restriction enzyme associated DNA sequencing (RADSeq). Mol. Ecol. Resour. 15, 1304–1315 (2015).
Bayona-Vásquez, N. J. et al. Adapterama III: Quadruple-indexed, double/triple-enzyme RADseq libraries (2RAD/3RAD). PeerJ 7, e7724 (2019).
Peterson, B. K., Weber, J. N., Kay, E. H., Fisher, H. S. & Hoekstra, H. E. Double digest RADseq: An inexpensive method for de novo SNP discovery and genotyping in model and non-model species. PLoS One 7, e37135 (2012).
Bauer, M. L., Ferry, B., Holman, H. & Kovach, A. I. Monitoring a new england cottontail reintroduction with noninvasive genetic sampling. Wildl. Soc. Bull. 44, 110–121 (2020).
Arantes, L. S. et al. Scaling-up RADseq methods for large datasets of non-invasive samples: Lessons for library construction and data preprocessing. Mol. Ecol. Resour. https://doi.org/10.1111/1755-0998.13859 (2023).
Snyder-Mackler, N. et al. Efficient genome-wide sequencing and low-coverage pedigree analysis from noninvasively collected samples. Genetics 203, 699–714 (2016).
Hayward, K. M. et al. Genotyping-in-thousands by sequencing (GT-seq) of noninvasive faecal and degraded samples: A new panel to enable ongoing monitoring of Canadian polar bear populations. Mol. Ecol. Resour. 22, 1906–1918 (2022).
Taylor, R. et al. Whole genome sequences from non-invasively collected caribou faecal samples. Conserv. Genet. Resour. 14, 53–68 (2021).
Campbell, N. R., Harmon, S. A. & Narum, S. R. Genotyping-in-Thousands by sequencing (GT-seq): A cost effective SNP genotyping method based on custom amplicon sequencing. Mol. Ecol. Resour. 15, 855–867 (2015).
Burgess, B. T., Irvine, R. L. & Russello, M. A. A genotyping-in-thousands by sequencing panel to inform invasive deer management using noninvasive fecal and hair samples. Ecol. Evol. https://doi.org/10.1002/ece3.8993 (2022).
Schmidt, D. A., Campbell, N. R., Govindarajulu, P., Larsen, K. W. & Russello, M. A. Genotyping-in-Thousands by sequencing (GT-seq) panel development and application to minimally invasive DNA samples to support studies in molecular ecology. Mol. Ecol. Resour. 20, 114–124 (2020).
Mueller, S. A. et al. Genome-wide diversity loss in reintroduced Eurasian lynx populations urges immediate conservation management. Biol. Conserv. 266, 10944 (2022).
Fuller, S. & Tur, A. Conservation strategy for the New England Cottontail (Sylvilagus transitionalis) (2017).
Fuller, S. & Tur, A. Conservation strategy for the New England Cottontail (Sylvilagus transitionalis) (2012).
Kovach, A. I. et al. Separating proactive conservation from species listing decisions. Environ. Manag. 70, 710–729 (2022).
Litvaitis, J. A. Response of early successional vertebrates to historic changes in land use. Conserv. Biol. 7, 866–873 (1993).
Brooks, R. T. Abundance, distribution, trends, and ownership patterns of early-successional forests in the northeastern United States. In Forest Ecology and Management Vol. 185 65–74 (Elsevier, 2003).
Lorimer, C. G. & White, A. S. Scale and frequency of natural disturbances in the northeastern US: Implications for early successional forest habitats and regional age distributions. For. Ecol. Manag. 185, 41–64 (2003).
Litvaitis, J. A., Tash, J. P., Litvaitis, M. K., Kovach, A. I. & Innes, R. A range-wide survey to determine the current distribution of New England Cottontails. Wildl. Soc. Bull. 34, 1190–1197 (2006).
Rittenhouse, C. D. & Kovach, A. I. Assessment of alternative sampling designs for range-wide monitoring of New England cottontail. Wildl. Soc. Bull. 44, 798–806 (2020).
Fenderson, L. E., Kovach, A. I., Litvaitis, J. A. & Litvaitis, M. K. Population genetic structure and history of fragmented remnant populations of the New England cottontail (Sylvilagus transitionalis). Conserv. Genet. 12, 943–958 (2011).
Fenderson, L. E. et al. A multiscale analysis of gene flow for the New England cottontail, an imperiled habitat specialist in a fragmented landscape. Ecol. Evol. 4, 1853–1875 (2014).
Cheeseman, A. E., Cohen, J. B., Whipps, C. M., Kovach, A. I. & Ryan, S. J. Hierarchical population structure of a rare lagomorph indicates recent fragmentation has disrupted metapopulation function. Conserv. Genet. 20, 1237–1249 (2019).
McGreevy, T. J. et al. Location and species matters: Variable influence of the environment on the gene flow of imperilled, native and invasive cottontails. Front. Genet. 12, 1–16 (2021).
Brubaker, D. R., Kovach, A. I., Ducey, M. J., Jakubas, W. J. & O’Brien, K. M. Factors influencing detection in occupancy surveys of a threatened lagomorph. Wildl. Soc. Bull. 38, 513–523 (2014).
Kovach, A. I., Litvaitis, M. K. & Litvaitis, J. A. Evaluation of fecal mtDNA analysis as a method to determine the geographic distribution of a rare lagomorph. Wildl. Soc. Bull. 31, 1061–1065 (2003).
Kristensen, T. V. & Kovach, A. I. Spatially explicit abundance estimation of a rare habitat specialist: Implications for SECR study design. Ecosphere https://doi.org/10.1002/ecs2.2217 (2018).
Bauer, M. L., O’Brien, K. M. & Kovach, A. I. Can at-risk species serve as effective conservation surrogates? Case study in northeastern US shrublands. Ecosphere https://doi.org/10.1002/ecs2.4081 (2022).
Shea, C. P., Eaton, M. J. & MacKenzie, D. I. Implementation of an occupancy-based monitoring protocol for a widespread and cryptic species, the New England cottontail (Sylvilagus transitionalis). Wildli. Res. 46, 222–235 (2019).
Adams, J. R., Goldberg, C. S., Bosworth, W. R., Rachlow, J. L. & Waits, L. P. Rapid species identification of pygmy rabbits (Brachylagus idahoensis) from faecal pellet DNA. Mol. Ecol. Resour. 11, 808–812 (2011).
King, T. L. et al. Microsatellite marker development from next-generation sequencing in the New England cottontail (Sylvilagus transitionalis) and cross-amplification in the eastern cottontail (S. floridanus). BMC Res. Notes 10, 1–7 (2017).
Berkman, L. K. et al. Hybridization and polymorphic microsatellite markers for two lagomorph species (Genus Sylvilagus): Implications for conservation. Conserv. Genet. Resour. 1, 419–424 (2009).
Vašíček, D., Vašíčková, K., Parkányi, V., Ondruška, Ľ & Vašíček, D. Noninvasive PCR sexing of neonatal rabbits selected for islet cell culture. Slovak J. Anim. Sci. 44, 43–47 (2011).
Morin, P. A., Chambers, K. E., Boesch, C. & Vigilant, L. Quantitative polymerase chain reaction analysis of DNA from noninvasive samples for accurate microsatellite genotyping of wild chimpanzees (Pan troglodytes verus). Mol. Ecol. 10, 1835–1844 (2001).
Feehery, G. R. et al. A method for selectively enriching microbial DNA from contaminating vertebrate host DNA. PLoS One 8, e76096 (2013).
Catchen, J., Hohenlohe, P. A., Bassham, S., Amores, A. & Cresko, W. A. Stacks: An analysis tool set for population genomics. Mol. Ecol. 22, 3124–3140 (2013).
Catchen, J. M., Amores, A., Hohenlohe, P., Cresko, W. & Postlethwait, J. H. Stacks: Building and genotyping loci de novo from short-read sequences. G3 Genes Genomes Genet. 1, 171–182 (2011).
Rochette, N. C., Rivera-Colón, A. G. & Catchen, J. M. STACKS 2: Analytical methods for paired-end sequencing improve RADseq-based population genomics. Mol. Ecol. 28, 4737–4754 (2019).
Li, H. & Durbin, R. Fast and accurate short read alignment with Burrows-Wheeler transform. Bioinformatics 25, 1754–1760 (2009).
Danecek, P. et al. The variant call format and VCFtools. Bioinformatics 27, 2156–2158 (2011).
Rivera-Colón, A. G. & Catchen, J. Population genomics analysis with RAD, reprised: STACKS 2. In Methods in Molecular Biology Vol. 2498 99–149 (Humana Press Inc., 2022).
Wojtusik, K. J., Berlinsky, D. L., Kenter, L. W. & Kovach, A. I. River-of-origin assignment of migratory Striped Bass, with implications for mixed-stock analysis. Trans. Am. Fish. Soc. 152, 15–34 (2023).
Hendricks, S. et al. Recent advances in conservation and population genomics data analysis. Evol. Appl. 11, 1197–1211 (2018).
Whipps, C. M., Cheeseman, A. E., Lindsay, K. A. & Cohen, J. B. Evaluation of cottontail pellets collected in suboptimal conditions for DNA analysis. Wildl. Soc. Bull. 44, 182–190 (2020).
Contosta, A. R. et al. Northern forest winters have lost cold, snowy conditions that are important for ecosystems and human communities. Ecol. Appl. https://doi.org/10.1002/eap.1974 (2019).
Young, S. S. & Young, J. S. Overall warming with reduced seasonality: Temperature change in New England, USA, 1900–2020. Climate 9, 176 (2021).
Wang, J. Estimating genotyping errors from genotype and reconstructed pedigree data. Methods Ecol. Evol. 9, 109–120 (2018).
Jones, O. R. & Wang, J. COLONY: A program for parentage and sibship inference from multilocus genotype data. Mol. Ecol. Resour. 10, 551–555 (2010).
Wang, J. Individual identification from genetic marker data: Developments and accuracy comparisons of methods. Mol. Ecol. Resour. 16, 163–175 (2016).
Jombart, T. Adegenet: A R package for the multivariate analysis of genetic markers. Bioinformatics 24, 1403–1405 (2008).
Dray, S. & Dufour, A.-B. The Ade4 package: Implementing the duality diagram for ecologists. J. Stat. Softw. https://doi.org/10.18637/jss.v022.i04 (2007).
R Core Team. R: A language and environment for statistical computing (2021).
Acknowledgements
AMS was supported by and partial research funding was provided by the New Hampshire Agricultural Experiment Station, through a USDA National Institute of Food and Agriculture McIntire-Stennis Project #1023445. This is manuscript #3022. Additional funding was provided by the National Fish and Wildlife Foundation, the Regional Pittman-Robertson New England Cottontail Initiative grant, United States Fish and Wildlife Service, and Maine Department of Inland Fisheries and Wildlife. We thank wildlife biologists Cory Stearns, Brett Ferry, Heidi Holman and Howard Kilpatrick for sample collection. Melissa Bauer, Emma Spara and Sean Skinner assisted with laboratory work. We are grateful to Kris Wojtusik for technical and bioinformatics support and to the other Kovach lab members and 2 anonymous reviewers for insightful feedback during manuscript development.
Author information
Authors and Affiliations
Contributions
Conceptualization: A.I.K.; Methodology: A.M.S., A.I.K.; Validation: A.M.S.; Formal Analysis: A.M.S.; Investigation: A.M.S.; Resources: A.I.K.; Data Curating: A.M.S., A.I.K.; Writing-Original Draft: A.M.S., A.I.K.; Writing—Review & Editing: A.I.K., A.M.S.; Visualization: A.M.S.; Supervision: A.I.K.; Project Administration: A.I.K., A.M.S.; Funding Acquisition: A.I.K.
Corresponding author
Ethics declarations
Competing interests
The authors declare no competing interests.
Additional information
Publisher's note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Supplementary Information
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.
About this article
Cite this article
Scott, A.M., Kovach, A.I. FecalSeq enrichment with RAD Sequencing from non-invasive environmental samples holds promise for genetic monitoring of an imperiled lagomorph. Sci Rep 14, 17575 (2024). https://doi.org/10.1038/s41598-024-67764-6
Received:
Accepted:
Published:
DOI: https://doi.org/10.1038/s41598-024-67764-6
Comments
By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.
