DNA metabarcoding on roadkill stomach contents reveals the breadth of species present in bobcat diets

TL;DR

This study uses DNA metabarcoding of roadkill bobcat stomachs to reveal their diverse diets in urbanized areas, including a high proportion of small mammals.

Contribution

The study introduces DNA metabarcoding of roadkill stomach contents as a novel method for examining bobcat diets in urbanized environments.

Findings

Cottontail and eastern gray squirrel were found in over 80% of bobcat stomach samples.

Nearly a third of bobcats consumed white-tailed deer.

No evidence of domestic dog or cat consumption was found in bobcat diets.

Abstract

Trophic dynamics can be altered in complex ways as a result of urbanization. Understanding predator diets in these contexts may not only provide insight into these changes, but also in sources of mortality for vulnerable prey species like the New England cottontail (Sylvilagus transitionalis). However, studying the diets of mammalian predators such as bobcats (Lynx rufus) can be challenging because of their elusive behavior. DNA metabarcoding of stomach contents from roadkill is a method which provides a new opportunity to study predator diets when mortality events occur. We used this technique to examine variation in bobcat diet across a range of urbanized environments in Connecticut, USA, as well as determine whether bobcats consume the declining New England cottontail. DNA metabarcoding identified between two and five species in the majority of bobcat stomachs. Cottontail (Sylvilagus spp.) and eastern gray squirrel (Sciurus carolinensis) were each found in over 80% of samples, and most remaining taxa were other small mammals. Nearly a third of the bobcats had consumed white-tailed deer (Odocoileus virginianus). Stomach contents containing cottontail remains were sequenced at an additional species-specific marker, but no samples containing the New England cottontail were identified. Bobcats in Connecticut consumed a wide variety of natural prey species including a relatively high proportion of semi-aquatic mammals, and we found no evidence of domestic dog or cat consumption. DNA metabarcoding of stomach contents is an effective approach for opportunistically examining predator diet, and our use of this tool may provide a more complete picture of bobcat diet where other techniques have failed to do so.

Figures

Neighbor-joining tree comparing snowshoe hare (Lepus americanus), New England cottontail (Sylvilagus transitionalis), and eastern cottontail (Sylvilagus floridanus).Sequences are a portion of a mitochondrial DNA control region from New England individuals sampled in Sullivan et al. (2019). Accession numbers beginning with “la” are snowshoe hare sequences, “ec” represents eastern cottontails, and “ne” represents New England cottontails. Samples from this project are marked with an asterisk (*). Tree made in Geneious Prime, version 2021.2.

Percent of samples in which each diet item was found, i.e., frequency of occurrence.The following representative silhouettes of species were obtained from phylopic.org: Eastern gray squirrel; Cottontail; White-tailed deer; Eastern chipmunk; Muskrat; White-footed mouse; American mink; Chicken; Racoon; Groundhog. The silhouettes for Eastern meadow vole, Southern red-backed vole, Northern short-tailed shrew, and Common shrew were created by author K. Hughes.

Full text

Introduction

Bobcats (Lynx rufus) are true carnivores, and urbanization of their habitat adds another layer of complexity to their influence, as it can lead to altered predator-prey interactions [1] and complex ecological dynamics [2–3]. Connecticut, USA is a mixed landscape of cities, suburbs, and diverse natural land cover, with similarly diverse native wildlife [4–6]. While over 85% of the human population of the state lives in cities, Connecticut is also over 60% covered by forest [7–8]. In this heterogenous environment of natural and urban land, bobcats are now one of the top predators after recovering from population losses in the past century [9–11]. To understand the dynamics of an urbanized ecosystem, predators’ diets must be considered comprehensively [12]. The feeding behavior of a predator population can also provide pertinent information when considering the conservation of vulnerable prey species. As the urbanization of Connecticut progresses, and the already heterogeneous spatial pattern of urbanization evolves, understanding predator-prey interactions in this context will be crucial for anticipating how human disturbance will shape the ecosystem via influences on predation. It is necessary to establish understanding of the behavior of predators directly impacted by urbanization (as roadkill animals were). In particular, characterizing how this altered ecology impacts prey choice for predators like bobcats can be important for conservationists managing prey species or wildlife conflict in the ecosystem.

Bobcats consume a significant proportion of lagomorphs in many ecosystems [13–16], including in Connecticut [17], making the details of their cottontail predation a key question of this study. The introduced eastern cottontail (Sylvilagus floridanus) is increasing in number [18] while the vulnerable, native New England cottontail (Sylvilagus transitionalis) continues to decline in Connecticut, one of the species’ last refuges [19–22]. Two of the five remaining populations of New England cottontails are in Connecticut [20]. However, since 2009, New England cottontails have declined by 43% in the state of Connecticut, and 50% overall [19]. As predators of cottontails rebound, predation is becoming an increasingly important pressure impacting New England cottontails [20]. Distinguishing between the two cottontail species in bobcat diet is necessary because predation by bobcats has historically been an important source of mortality for New England cottontails [20,23]. However, these cottontail species are morphologically similar [24–25], which limits the effectiveness of traditional mechanical sorting for diet analysis. Because of the morphological overlap, wildlife managers use DNA analyses to confirm species identification [24–25]. Therefore, using genetic data to examine the diet of bobcats may serve as a much-needed method for understanding the distribution of and threats to this rare species [23].

In the past century, morphological analysis of contents of the gastrointestinal tract or scat has been a popular method to assess the diets of predators. This method relies on the presence of undigested, identifiable remains of prey, such as bone fragments, hair, or feathers [26–28]. Visual identification may be limited to a short window after feeding [29], and some prey (either larger organisms or those with less digestible parts) may be overrepresented due to ease of identification [26,27,29]. The method can also be limited in terms of taxonomic resolution and accuracy compared to molecular methods such as DNA metabarcoding [26,28–30]. Numerous morphological studies in eastern North America have examined bobcat diets.

Morphological diet studies have historically shown bobcats in the region consume a wide variety of mammalian prey. One study conducted several decades ago on bobcat scats and gastrointestinal contents in New England found snowshoe hare (Lepus americanus), white-tailed deer (Odocoileus virginanus), eastern cottontail, porcupine (Erethizon dorsatum), and “field mouse” (unspecified species) to be the most common prey [31]. By the 1980s, bobcat gastrointestinal tracts in New Hampshire mostly contained cottontails (including New England cottontails), white-tailed deer, snowshoe hares and eastern gray squirrels (Sciurus carolinensis) [32]. Multiple twenty-first century morphological diet studies identified white-tailed deer, cottontails, rats and mice, and squirrels as the most important prey of bobcats [14–16]. Snakes, birds, porcupines, raccoons, muskrats, foxes, minks, and chickens are some of the less commonly reported prey species [16,31,33,34].

Other common methods for characterizing predators’ diets include stable isotope analysis (SIA) and quantitative fatty acid analysis (which is primarily useful for marine predators) [35]. SIA offers the advantage of a choice of timescale for analysis, depending on the tissue chosen. If the researchers have a whole-organism sample, the diet at different life stages can be investigated [36]. Trophic level and ecosystem of origin can also be discerned from isotopic values [27,37]. The availability of both temporal and spatial information make this method attractive for a broad range of questions. Despite the advantages, taxonomic specificity is limited using stable isotopes. Out of a list of potential prey species (for which mixing models must be created), prey with similar diets may be isotopically indistinguishable [17,38]. For a species like bobcats that consume a wide variety of small mammals with similar diets, diet resolution from SIA could be limited. New England and eastern cottontails have considerable overlap in diet [23], and are unlikely to be distinguishable. Therefore, our question of specific taxa consumed by the predator cannot be answered by SIA. Nonetheless, stable isotope analysis has been useful for answering other questions about New England bobcats’ diet. One study reported an observed shift from previous studies to a diet richer in squirrels and large mammals, and away from lagomorph specialization [39].

Molecular DNA-based methods for identifying prey have recently become popular alternatives to morphology and stable isotopes because of their ease of use, taxon specificity, and ability to identify prey from remnant DNA after substantial digestion has occurred [26,29]. DNA metabarcoding has the advantage of being effective with a mixed or degraded sample [28,29,40], and not requiring a priori understanding of what species may be present because of the use of universal markers [40]. This benefit extends to detecting rare prey species [28,29]. However, metabarcoding typically cannot provide direct quantitative measures of prey consumption, because myriad variables such as prey species, tissue consumed, meal size, and the physiological status of the predator can affect the detectability of DNA [41–42]. Nonetheless, this method can be useful for identifying the breadth of predator diets and relative frequency of different prey taxa [43–44]. For example, a DNA metabarcoding study on coyote scats increased the number of prey species detected compared to morphological methods [43]. To what degree DNA metabarcoding increases the detection of prey species compared to morphological methods is highly variable, due to the numerous variables involved in DNA survivability in the gastrointestinal tract [42]. However, studies investigating this have generally found that the advantage of molecular methods is significant [26], in one case increasing prey detection roughly four-fold [45]. Molecular diet approaches may be applied to scat or the contents of the gastrointestinal tract [41,44,46]. DNA metabarcoding of scat has been used to examine bobcat diet in South Carolina, USA, and found that lagomorphs, small mammals, and white-tailed deer were the most common prey items [11].

The full range of bobcats’ diet is of consequence for assessing how they fit into a dynamic anthropogenic ecology after their populations have recovered. In this study, we intended to expand the use of roadkill for wildlife science by using DNA metabarcoding on the stomach contents of roadkill bobcats– detailing their full diets while also specifically monitoring for the presence of the New England cottontail within the stomachs. Roadkill presents a yet-underused, less labor-intensive resource (compared to scat surveys or mark-recapture, for example) for genetic data in wildlife science which can aid in providing this information [14,47]. One other study has applied molecular diet analysis methods to roadkill predators, though it was not next generation sequencing (NGS) based and did require a priori diet information [48]. By taking this novel methodological approach which lacks a priori assumptions about the animals’ diets, we hope not only to advance understanding of an apex predator in a dynamic ecosystem, but also, by proxy, obtain valuable insight into the status of an historic, now vulnerable prey species.

Materials and methods

We collected carcasses of road-killed bobcats from throughout Connecticut opportunistically when encountered, and when people in the public reported roadkill to the Department of Energy and Environmental Protection. Carcasses were stored in a walk-in freezer. During necropsy, carcasses were thawed, stomachs removed and then refrozen in a chest freezer. Additional data were collected on the animals at the time of necropsy, including the sex and weights of the animals. Age data was generated via cementum annuli by extracting and cross sectioning an upper canine tooth for each individual [49]. Each stomach was opened within a biosafety hood. When animal remains were intact and identifiable to species via morphology, the species was noted. Tissue from cottontail remains were collected. The remaining contents were ground using a Weston #8 575W Meat Grinder into a one-quart plastic container, and then homogenized using an All-Clad Stainless Steel immersion blender. When necessary, Qiagen ATL Buffer was added in a small amount to facilitate homogenization. The homogenized samples were placed in up to four labeled 50 mL Falcon tubes and stored in a refrigerator. All stainless-steel parts from the grinder and the blender were soaked for 15 minutes in a 10% bleach solution and then soaked and rinsed in water between the processing of each stomach. Homogenized samples were then stored back in the refrigerator.

For DNA extraction, homogenized tissue was selected from multiple locations in all sample tubes using disposable forceps such that the total amount used in each extraction was approximately between 0.2 to 0.5 cubic centimeters. DNA was extracted from the stomach contents using a Qiagen DNeasy Blood and Tissue Kit using a version of manufacturer instructions modified to include an overnight digest with proteinase K to improve DNA yield. DNA concentrations were quantified with a Qubit Fluorometer 2.0.

Seventeen bobcat stomachs were suspected to contain cottontail remains based on visual inspection of the stomach contents before homogenization. A genetic analysis was used to examine whether these remains were of eastern cottontail or the New England cottontail. After extraction by the protocol outlined above, all seventeen samples were PCR amplified using forward primer L15934 and reverse primer H16442 from Sullivan et al. (2019) [25], which together amplify a 480 bp portion of the mitochondrial DNA control region. This region was determined by Sullivan et al. (2019) to be diagnostic of lagomorph species. PCR was carried out in 25 uL reactions using Promega GoTaq Green MasterMix and products were visualized via gel electrophoresis to confirm amplification. Six of the samples did not amplify after multiple attempts at re-extraction and re-amplification. Regardless of the failure to confirm amplification of those samples, all seventeen samples were sent to Azenta Life Sciences for Sanger sequencing. Of those 17 sent to Azenta, 8 failed to amplify for sequencing.

Sullivan et al. (2019) identified a portion of the mitochondrial control region which is useful for distinguishing lagomorph species. Prior to sequencing, using Geneious Prime [50], we aligned the lagomorph sequences from their study (originating in the New England region) [25](Genbank accession: KC923306-KC923406 for S. floridanus, KC923407-KC923418 for S. transitionalis, KC923298-KC923305 for L. americanus, outgroup) and visually identified short regions within the sequences which were unique to each cottontail species. After sequencing, we aligned the sequences from our samples with the sequences from Sullivan et al. (2019) and used the “Find Motif” tool to search our sequences for the species-specific runs we had previously identified in the published sequence data. To confirm the results, a neighbor-joining tree was also constructed in Geneious Prime using the same data, via the built-in tree builder tool and default settings.

For the overall diet metabarcoding component of our study, the primers (F:5’-TTAGATACCCCACTATGC-3’; R:5’-TAGAACAGGCTCCTCTAG-3’) were used to amplify the 12S-V5 mitochondrial control region of all 76 stomach samples. These primers have been used previously to investigate the diet of brown bears (Ursus arctos), feral cats (Felis catus), and coyotes (Canis latrans) [27,43,51] using DNA extracted from scat. We paired the forward primer with each of the four barcodes, such that every sample in each pool had a barcode F1-F4 (S2 File). The four unique barcodes were chosen from Coissac (2012) [52] to pool the samples for sequencing. These barcodes were used in conjunction with partial Illumina adapter sequences to multiplex four samples at a time during sequencing. Initial PCR amplification was carried out with Promega GoTaq Green 2x Mastermix in 25 uL reactions to test the usability of the samples. The reactions consisted of 12.5 µL of Mastermix, 1 µL of 10 µM forward primer (with barcode), 1 µL of 10 µM of reverse primer, 8.5 µL of sterile water, and 2 µL of DNA. The conditions for the PCR thermocycler were 95° C for 5 minutes, 94° C for 30 seconds, 45° C for 30 seconds, 72° C for 45 seconds, 94° C for 30 seconds as in step two for 49 cycles, before elongation for 5 minutes at 72° C. Sixty-three total samples (of the original 76) were successfully amplified, from 21 female and 42 male bobcats.

Once amplification was verified by visualization of a ~ 200 bp band via gel electrophoresis, each reaction was run again with Thermofisher AmpliTaq Gold 360 Mastermix, with the same reaction mix and thermocycler conditions. A non-pigmented mastermix had to be used for library preparation because pigment would interfere with DNA quantification, which was necessary for normalization. The PCR product was quantified with a Qubit Fluorometer 2.0. Then, after verifying amplification with gel electrophoresis, the PCR product was purified using Thermofisher ExoSap-IT Express PCR Product Cleanup Reagent according to manufacturer’s instructions. After cleanup, purified PCR products were normalized to 20 ng/uL and pooled in groups of four samples with different barcode. Samples were then sent to Azenta’s Amplicon-EZ service to sequence in one lane of Illumina MiSeq 2x250bp sequencing.

Sequence reads were demultiplexed using Cutadapt version 4.1 [53], and the trimmed sequences were imported as fastq files into Geneious Prime version 2021.2.2 [51]. Low quality (quality score <20) and short reads (<80 bp) were filtered out using the BBDuk Trimmer plugin [54], and paired end reads were merged. De novo assembly for each sample was performed in Geneious Prime. Between 23,908 and 155,323 reads were generated per sample (S1 File). Consensus sequences, generated during de novo assembly, were queried on the National Center for Biotechnology Information database (ncbi.nlm.nih.gov) using the Standard Nucleotide BLAST (blastn) tool. A species was identified as a potential diet item if the sum of reads assigned to the taxon constituted at least 250 reads and >0.5% of reads for the sample.

The resolution of the identification varied across samples; some BLAST queries returned a hit to the species level, with only one likely species, and very high percent identity and query cover. Others identified multiple potential species per consensus sequence; for example, every sample had hits that were identified with high certainty as any of a list of felid species: cheetah (Acinonyx jubatus), jaguarundi (Puma yagaouaroundi), etc., and bobcat. In cases like these, the species present in the region was inferred to be the source of the genetic material. This was also common for the lagomorph species, wherein hits returned a list of Sylvilagus species including eastern cottontail, but not the New England cottontail. It appears that Genbank lacks any sequence data for the 12S-V5 region specifically for the New England cottontail, so while these were tentatively assigned the identity of eastern cottontail, uncertainty remains; as such, we will only refer to these as cottontail species. Aside from inferred bobcat sequences and contamination, sequences with over 250 read counts either returned a list of species of the same genus, wherein only one is extant in the region, or only one species.

Frequency of occurrence was calculated for each prey species as a percentage of stomach samples containing the species. We calculated Simpson’s Diversity Index using the R package [55] abdiv [56] to describe the diversity and evenness of bobcat diets [16].

Field research completed under UConn IACUC E18-006 (TAGR) and CT DEEP Permit to Collect Wildlife for Scientific & Educational Purposes 1720006 and 1720006b (TAGR). CT DEEP operates under the authority of the Commissioner of the Connecticut Department of Energy and Environmental Protection (CT DEEP). All animal handling (trapping, drugging, tagging, etc.) protocols used by CT DEEP are under the guidance of, and approved by the Connecticut Department of Agriculture (CTDA) Veterinarians. The work conducted at Fordham University only involved the use of pre-processed homogenized stomach contents, with no live capture, handling, or killing of animals. We were, therefore, exempt from the Fordham University Institutional Animal Care and Use Committee (IACUC) protocols.

Results

For the cottontail analysis, a portion of the mtDNA control region was successfully sequenced for nine samples that were flagged during dissection as suspected to contain cottontail remains. All nine contained short motifs previously associated with eastern cottontail during the Geneious analysis of the Sullivan et al. (2019) Genbank sequences. None contained motifs associated with New England cottontails. Construction of a neighbor-joining tree in Geneious confirmed this result, as all nine sequences grouped with eastern cottontails (Fig 1). This analysis failed to detect any New England cottontail present in the stomach samples set aside by the dissectors.

Neighbor-joining tree comparing snowshoe hare (Lepus americanus), New England cottontail (Sylvilagus transitionalis), and eastern cottontail (Sylvilagus floridanus).Sequences are a portion of a mitochondrial DNA control region from New England individuals sampled in Sullivan et al. (2019). Accession numbers beginning with “la” are snowshoe hare sequences, “ec” represents eastern cottontails, and “ne” represents New England cottontails. Samples from this project are marked with an asterisk (). Tree made in Geneious Prime, version 2021.2.*

Of the 63 bobcat stomachs analyzed using 12S-V5 barcoding, cottontail species and eastern gray squirrel were each present in over 80% of samples (Fig 2). Also prominent in the samples were eastern meadow vole (Microtus pennsylvanicus) at 41% and white-tailed deer at 32% of samples (Fig 2). Eastern chipmunk (Tamias striatus) (19%), American mink (Neovison vison) (11%), and muskrat (Ondatra zibethicus) (16%) were found in at least 10% of samples (Fig 2). Other species that occurred less frequently than 10% included white-footed mouse (Peromyscus leucopus), common shrew (Sorex cinereus), northern short-tailed shrew (Blarina brevicauda), groundhog (Marmota monax), raccoon (Procyon lotor), southern red-backed vole (Myodes gapperi), and chicken (Gallus gallus) (Fig 2). Fourteen species were detected in the samples (Table 1), aside from bobcat, which was present in every sample as we used no blocking primer. The number of prey species per sample ranged from one to seven. Three samples had only one prey species present, 14 had two prey species, 23 had three prey species, 14 had four prey species, 7 had five species, and there was one sample each that had six or seven prey species. No domestic dog or cat DNA was detected. The Simpson’s Diversity Index for the entire dataset was calculated as 0.82.

Table 1: Frequency of occurrence (FOO) of each diet item by bobcat sex.

Percent of samples in which each diet item was found, i.e., frequency of occurrence.The following representative silhouettes of species were obtained from phylopic.org: Eastern gray squirrel; Cottontail; White-tailed deer; Eastern chipmunk; Muskrat; White-footed mouse; American mink; Chicken; Racoon; Groundhog. The silhouettes for Eastern meadow vole, Southern red-backed vole, Northern short-tailed shrew, and Common shrew were created by author K. Hughes.

Discussion

Bobcats are currently a top predator in the Northeast region of the United States and are increasing in abundance as they expand into an urbanized landscape. Broadly, our results support previous research via other methods [14–16] that bobcat diet is generalist and consists of natural prey species, probably based on their availability. Our detailed analysis refines understanding of which natural prey bobcats are encountering and hunting—showing that “natural prey” includes semi-aquatic mammals. Connecticut’s land can be characterized as an urbanized landscape that is largely a mix of forest and development, with a contribution from mostly palustrine wetlands [5,57,58]. We confirmed that bobcat diet is mostly composed of lagomorphs, squirrels, and white-tailed deer. Given that eastern cottontails thrive in edge habitat such as what may be found near suburbs [21], and white-tailed deer and gray squirrels thrive in woodland [58], this squares with what we know about the landscape. The relative proportion of samples which contained cottontails and squirrels (both >80% frequency of occurrence) was the highest yet reported in the literature. Using DNA metabarcoding, we refined the diet composition to a range of specific species. No New England cottontail was confidently detected in the bobcat stomachs analyzed, potentially lending support to the conclusion that the species has declined significantly in recent years [18]. However, significant uncertainties around this conclusion remain and are discussed below. We also confirmed that the squirrel species most common in the diet is eastern gray squirrel, and not red squirrel nor eastern chipmunk. Although we sampled an urbanized landscape where homes with people and pets are embedded within the forest, we found no evidence of bobcats consuming domestic dog nor cat.

Eastern cottontails were more represented in bobcats’ diets than New England cottontails in our study, which makes sense given the current occupancy and relative abundance patterns of the two cottontail species [21–22]. Eastern cottontails are ubiquitous across habitat types and can achieve high local densities. Given New England cottontails only occupy 14% of their original range and prefer dense thicket [59], and the bobcats in this study may have occupied primarily disturbed land during their lives, it is possible that a lack of home range overlap between the animals made predation events rare. Bobcats, widely understood as a generalist predator, would likely consume the cottontails if they encountered them-- so this suggests that at least the bobcats sampled for the Sanger analysis probably did not in their final days. A sampling of bobcat carcasses or scats centered around habitat patches known to contain New England cottontail may be needed to gain an understanding of how bobcat predation may be affecting this rare lagomorph species. Our results confirm that the current bobcat population in Connecticut is being supported by the abundance of cottontails throughout the state, but to what extent they are supported by specific cottontail species (eastern vs. New England cottontail) cannot be confidently concluded given the spatial bias and small sample size in this study.

Furthermore, one of the limitations toward answering the cottontail question was the order in which procedures were undertaken. The analysis differentiating cottontail species using motifs in the mitochondrial control region was completed prior to the metabarcoding analysis. This means that the cottontail analysis relied upon the conclusions of the dissection-based diet analysis, which may not be as robust or accurate as molecular diet determination, as discussed previously. Because the 12S-V5 sequence for New England cottontails is not present on Genbank, we cannot confidently say how likely it is that the cottontails identified in the metabarcoding analysis were New England cottontails vs. eastern cottontails. The morphological analysis flagged far fewer samples as containing cottontail than did the metabarcoding analysis; the dissection only flagged 17 samples, whereas the metabarcoding indicated that 55 of 63 samples contained cottontail genetic material. It is informative that our metabarcoding analysis increased the detection of cottontails roughly threefold, but it invites questions as to whether any of those other 38 samples could have, in fact, contained New England cottontail. All of the samples for which we sequenced the lagomorph-specific region and found only eastern cottontail also showed cottontail in the 12S-V5 analysis.

We identified a number of unusual or notable prey items by using DNA metabarcoding. Both muskrats and minks were detected at relatively high frequencies in bobcat diets compared to previous research [31,33,60,61]. We detected muskrat in nearly 16% of our samples, which is somewhat unusual, particularly in the context of declining muskrat populations in recent decades [62]. Six out of twenty-one females consumed muskrats, compared to four out of forty-two males. Predation is one of several competing hypotheses explaining muskrat declines, with coyotes and bald eagles (Haliaeetus leucocephalus) being the primary predators of muskrats [62]. We also observed a high frequency of American minks, which are only reported rarely in the diet of bobcats. We identified American mink in 11% of bobcat stomachs, three female bobcats and six males.

The relatively high frequencies of occurrence of these two species may point to a behavior of bobcats which may have been previously either underemphasized due to methodology or infrequently occurring due to genuine ecological difference. Bobcats are not widely known to frequently hunt semi-aquatic species such as muskrats or especially minks [15,16,33,61,63]. Their use of these prey resources is usually incidental or opportunistic, but our observed frequency of semi-aquatic prey (nearly 30%) indicates this predation is likely more than incidental. This significant semi-aquatic predation may be in part explained by the prevalence of wetlands in Connecticut, which cover at least 5% of the state [57]. The Clean Water Act protects wetlands in urbanized landscapes, and conserved lands often contain wetlands. If bobcats feed opportunistically in a habitat interspersed with wetlands, regular predation on semi-aquatic species is less surprising.

The prevalence of domestic species in bobcats’ diets is of interest and concern to homeowners within this populous study area. The presence of chicken in multiple samples is unsurprising given how common keeping backyard chickens is in the area and documented observations of bobcats depredating chickens [31,34]. This potential source of conflict with human denizens of the region may be of interest, though bobcats are generally considered to be of low concern for domestic species [64]. Indeed, we did not detect any consumption of domestic dogs or cats. This result supports the idea that bobcats are likely of low concern for conflict with pet owners.

Metabarcoding is a powerful modern tool for wildlife diet analysis. Metabarcoding was recently used to determine bobcat diet from scat samples collected in South Carolina, USA, where the most frequent prey were lagomorphs, small mammals, and white-tailed deer [11]. Using DNA metabarcoding on stomach contents rather than scat for predator diet could potentially offer the advantages of yielding greater concentrations of DNA [41] and better detection of specifically animal matter [46]. However, we did not directly compare multiple methods. A recent study concluded differences in species richness and actual content of the diet may be negligible between stomach and scat [41].

This methodology may have influenced our results in terms of the abundance of prey reported. How long after consumption a diet item is detected in a stomach using metabarcoding is contingent on a variety of factors both specific to the meal(s) consumed and the internal physiological state of the predator. The range of lengths for this period of time between consumption and the end of detectability, or “temporal snapshot”, has been characterized for scat in another felid [42], but understanding how much of a felid’s consumptive life is represented specifically in a stomach sample may be important. Determining the window of prey DNA detectability in bobcat stomachs could potentially clarify whether our unique results are a consequence of methodology. Previous morphological diet studies may not have found such high proportions of squirrels and cottontails only because their temporal snapshots were short compared to metabarcoding. This window of DNA detectability can impact the interpretation of our results-- if prey DNA is hypothetically detectable for only one day by this method, a bobcat would have to eat cottontails nearly daily for them to show up at the > 80% frequency observed in our analysis. However, if prey are detectable for three days, a bobcat eating cottontails only a few times a week could account for our results. An active feeding study in cheetahs found DNA from a meal is detectable in scats for anywhere from 8 hours to three days after feeding, though many variables—satiety, meal size, species, etc.-- influence detectability [42].

The limitations of our methodology also apply to the question of secondary consumption. Because of the inconsistency in DNA survivability across prey species, individuals, meal size, and so on, a quantitative analysis (for example comparing read counts) may not reliably differentiate species present as a consequence of secondary consumption, particularly without blocking primers [65]. Indeed, the confounding variables above would be even greater considering the variability of digestion of the prey animal consumed by the bobcat. In this study in particular, the samples showed evidence of mink consumption and muskrat consumption. Since minks are also predators, it is possible that the muskrats were consumed by the minks, at least in the one case where they are both present in the same sample; however, it is challenging to discern where this may be the case in our study. In the case of secondary consumption of muskrats, for example, it could be reasonably expected that the muskrat DNA would be more degraded, and thus return fewer reads – but, the tissues of the animal consumed, the satiety of the bobcat, the time since the mink consumed the muskrat, the satiety of the mink, the size of the three individual animals, and so on, could all influence DNA survivability [42]. All samples containing mink did contain small prey as well. As in Thuo et al. (2019) [42], secondary consumption focused feeding studies would be an important step towards resolving this uncertainty. Likewise, it is currently very challenging to differentiate prey which were scavenged, except potentially by the presence of blow fly (Calliphoridae) DNA [66].

While, as one may expect of a generalist predator, the diets of individual bobcats may vary to include diverse prey ranging from white-footed mice to white-tailed deer and raccoons, the most popular prey are squirrels and cottontails. The bobcats of our study were likely showing us via their diets what the most abundant prey species on the landscape are. Indeed, other studies employing DNA metabarcoding have characterized generalist predator diets as “biodiversity capsules” [67]. Much in the way that the majority of humanity eats rice daily, these easy meals of squirrels and cottontails are perhaps these bobcats’ dietary staples due to their ubiquity.

Supporting information

S1 File12S-V5 sequences from bobcat stomachs.(XLSX)

S2 FileOligos with Barcodes.(DOCX)