A barcoding‐based scat‐analysis assessment of Eurasian otter Lutra lutra diet on Kinmen Island

Abstract While it is well known that Eurasian otters principally feed on fishes and crustaceans, their detailed diet taxonomies are not fully understood. This is partly due to their nocturnal behavior and the limited resolving power of traditional morphological identification from scat. A suitable, reliable molecular method for diet studies is therefore needed. I performed a series of Sanger‐sequencing reactions, utilizing nine primer sets for Eurasian otter diet research. These are mainly based on the barcoding concept to determine the taxonomic composition of spraints. The primer sets target different types of animals, amplifying each separately. This procedure was used to detect the prey contents of 64 spraint samples collected from Kinmen Island. Through high‐resolution gel electrophoresis and sequencing, it was evident that PCR products could be successfully amplified by the different primer sets and from spraint samples comprising multiple prey species. Extracted DNA from all spraint samples was PCR‐amplified with 9 primer sets. In total, 16 prey types were identified across all 64 samples. Fourteen were identified at the species level. The aim of this study was to develop and apply a novel diet research method to Eurasian otters. Eight of the primers are universal primers designed for COI segments of different animal groups, and one primer set was designed specifically for tilapia groups. This method can be applied to study the diets of not only Kinmen Eurasian otter populations, but also other Eurasian otter populations and other small carnivorous animals.

its battlefront position in the Taiwan Strait and strictly limited land use by local people in the decades prior to 1992 (You et al., 2013).
Though some surveys of Kinmen's otter population structure and dynamics have been performed (Lee, 1996), ecological data on this population are otherwise very limited.
Diet analysis is a precondition to understanding the biology of a species and their interactions with others, as well as the functioning of ecosystems. Such studies therefore provide important data for understanding animal ecology, evolution, and conservation (Buglione et al., 2020;Jedlicka et al., 2016;Krahn et al., 2007;Rolfe et al., 2014;Shehzad et al., 2012;Symondson, 2002;Tournayre et al., 2021;Zhong et al., 2019). In prior studies, diets were mainly determined by direct observation of feeding, or by microscopic examination of gut contents or feces. Such traditional diet analyses have provided an abundance of useful data (Almeida et al., 2012;Carss, 1995;Heggberget & Moseid, 1994;Liu et al., 2018;Pierce & Boyle, 1991;Wasser et al., 1997). Nevertheless, they also have known biases and limitations. Direct observation approaches preclude working on tiny animals, most nocturnal species, anything beneath the soil, under water, hidden or elusive, while microscopic examination is labor-intensive and relies on the researchers' skill in identifying species from masticated, semidigested pieces of food (Liu et al., 2018;Moreby, 1988;Pierce & Boyle, 1991). Most of all, identification at the species level is difficult to achieve with these traditional diagnostic approaches (Carss, 1995). An accurate technique for determining the taxonomic composition of a species' diet is therefore greatly needed.
When prey are too thoroughly digested for recognition, or when food species cannot otherwise be diagnosed from fecal remains (mollusks without bones, for example, or part of individuals such as soft muscle tissue), molecular identification of prey may be the only practical means of procuring data on trophic interactions that are difficult-if not impossible-to obtain in any other way (Liu et al., 2018;Symondson, 2002). Consequently, there is potential for applying such molecular approaches-and specifically, following the DNA barcoding concept-for otter diet analyses (Marcolin et al., 2020). However, prey DNA in feces is often highly degraded, preventing the amplification of long fragments for analysis (Lanszki & Molnár, 2003;Sittenthaler et al., 2019;Wasser et al., 1997). In early molecular studies, most attempts to analyze diet were performed by cloning PCR products and through subsequent Sanger sequencing of these clones by capillary electrophoresis (Deagle et al., 2005(Deagle et al., , 2007Guillaud et al., 2017;Jarman et al., 2004;Valentini et al., 2009).
These approaches are both time-consuming and expensive (Pegard et al., 2009;Shehzad et al., 2012). Notably, Hong et al. (2019) used a Sanger sequencing-based approach to identify vertebrate species from individual bones isolated from otters' feces (spraints). This approach is laborious and requires technical expertise that limits the capacity of data generation, ignoring all information from boneless food items.
At present, next-generation sequencing (NGS)-based diet analysis of complex DNA mixtures such as feces, for example, scat DNA metabarcoding (sDNA metabarcoding), is becoming increasingly useful. This approach facilitates the generation of abundant sequence data from very large numbers of individual DNA molecules, deriving from a complex mixture and without the need for cloning (Schuster, 2008;Valentini et al., 2009). Such sDNA metabarcoding has already been applied to study several animal species from highly diverse taxa, as well as Eurasian otters (Buglione et al., 2020;Kumari et al., 2019;Pertoldi et al., 2021).
Nevertheless, there are several limitations to DNA metabarcoding-based diet analyses of Eurasian otters. First, NGS is still prohibitively expensive for many smaller labs. Second, NGS data analysis can be time-consuming and requires special knowledge of bioinformatics to garner accurate information from sequence data (Grada & Weinbrecht, 2013). Third, while the sensitivity of NGS is vastly superior to Sanger sequencing and is capable of detecting very low DNA concentrations, this high sensitivity is F I G U R E 1 A Eurasian otter eating a tilapia in Tai Lake, Kinmen (site 11 of this study). The photograph was taken by Fu-Sheng Huang in June 2020 a double-edged sword: It also facilitates the amplification of minute quantities of contaminating DNA (King et al., 2008), as well as secondary prey. Such organisms are potentially ingested by and/ or attached to larger organisms predated by the otters, else were eaten by predatory fishes or other animals that contained them in their guts. The huge sequence output of NGS will thus include a high number of species derived from contamination or secondary predation, confusing our understanding of the real predation behavior of these otters.
Here, I present a barcoding-based spraint-analysis procedure, which I use to assess the diet of the Eurasian otter on Kinmen Island. This approach is based on Sanger sequencing with 9 primer sets, each targeted toward different prey taxa, and thus allowing for greater resolution than morphological studies. The aim of this work was to (a) provide an easy and affordable molecular method for detecting the species contained in spraint samples of Eurasian otter on Kinmen; (b) to demonstrate the performance of each of the primer sets; (c) to develop an efficient, custom-designed primer set for the most common prey species group, tilapia, of otters in Kinmen; (d) to discuss the results of the sequencing and the limitations of this method, where applicable; and (e) to provide new best practices for studying the diets of Eurasian otters and those of other obligate carnivores that feed on similar prey.

| Study area
Kinmen Island is located 10 km (6.2 mi) off the southeastern coast of mainland China. It was originally a military reserve and a frequent battlefront between 1949 and 1979, before it was returned to the civilian government in the mid-1990s. For agricultural and military needs, many reservoirs, artificial lakes, and ponds were constructed for storing water, raising fish, and irrigation on Kinmen. My colleagues and I collected spraint samples from 22 sites on Kinmen from April 2017 to November 2018. These collection sites can be catalogued as 6 types: freshwater stream (7 sites), freshwater pond (5), freshwater and brackish reservoirs (5 and 1, respectively), rocky coast (2), sand beach (1), and brackish wetland (1) (Figure 2).

| Spraint samples used in this work
Scat freshness affects the proportion of detectable food DNA (McInnes et al., 2017). In this study, I used the freshest spraint samples (including jelly-like and mucosal spraints) as possible. Furthermore, all spraint samples used in this work passed the DNA prescreening quality control procedure suggested by F I G U R E 2 Locations of sampling sites in this study. Detailed location information is listed in Table 1 TA B L E 1 Sample information and results of barcoding identification Note: "Locality": site of spraint collection. "Sample": the codes of spraint samples. "Food Species no.": diet species number of each spraint sample. "Valid sequence no.": the number of all readable sequences from each sample. "Remark": morphologic notes of unusual spraint samples.
The abbreviation codes of food species (in blue color) refer to Table 3, and superscript codes (a, b, c, LL) of nonfood species (in red color) refer to the "Nonfood species" column of this table for each sample. Primer set details refer to Table 2. *: the highest similarity of sequence and compared data is higher 90% but less than 98%; ** is higher 80% but less than 90%. Species without star mark indicate that the similarity is higher than 98%.
TA B L E 1 (Continued) Hung et al. (2004) with few modifications. This procedure was applied to check the qualities of extracted DNA for subsequent individual identification procedures using microsatellite methods.
Though such tests are time-consuming, they facilitate exclusion of poor-quality spraints and minimize the occurrence of false negatives in diet analysis.
Finally, 64 "very fresh" spraint samples, as catalogued by Lerone et al. (2014) were collected. Some samples contained special materials beside fish remains, such as hairs or feathers, broken shells, and bird bones which were observed by eye or via microscope prior to DNA extraction, as recorded in the "Remark" column (see Table 1). Fresh spraints were collected and preserved in 99% alcohol individually and kept frozen at −80°C until examination.

| DNA extraction
The DNeasy Blood & Tissue Extraction Kit (QIAGEN, Germany) was used according to the manufacturer's instructions, with few modifications as detailed in Appendix S1. The extracted DNA was suspended in 80 μl AE buffer.

| Selection and design of primer sets
Each DNA sample extracted from spraints was PCR-amplified 9 times and with 9 primer sets. I browsed the published universal COI primers and chose eight sets to estimate the diet contents of spraints collected in Kinmen. Besides the COI primers, a group of Brown jelly COIII primers was designed for this study for the most abundant prey species in Kinmen, the introduced tilapia. They are forward primer Til9020F and reverse primer cocktails Mos9516R+Nil946 4R+Esc9305R+Zil9212R (TilMR). The details of primers used are listed in Table 2.
The primer cocktails are more effective than conventional primers, facilitating barcode work on taxonomically diverse samples (Ivanova et al., 2007). In this study, the reverse primer cocktails demonstrate dif-

| PCR amplification
The PCR was performed using a Taq

| Checking the PCR products with high resolution methods
The Sanger method requires a single amplicon, comprising a single target, in order to produce a sequence. To ensure that PCR products were suitable for downstream sequencing, I ran each PCR product on the QIAxcel Advanced automated electrophoresis system (QIAGEN, Germany), which affords a resolution down to 3-5 bp. For the COI gene segments, I chose the PCR products with a single signal peak to sequence (Figure 3a). For the PCR products with primer set Til9020F/RM (COIII gene), multiple peaks are acceptable to sequence as they are caused by different 5' end primers (see Figure 3d).

| DNA sequencing
PCR products that clearly showed amplification of the appropriate number of base pairs (i.e., close to 700 bp) without multiple peaks ( Figure 3) were purified using a clean-up reagent (HT ExoSAP-IT High Throughout  The highest similarity of sequence and compared data is higher 90% but less than 98%.

| Results from spraint samples
b Higher 80% but less than 90%. Species without superscript a or b indicate that the similarity is higher than 98%.

F I G U R E 4
The food species (a) and nonfood species (b) sequenced in this study. Numbers following the species/ catalogue are the repeat counts detected from the 64 tested spraint samples. More detailed information is recorded in Table 3 in 3.13% (2 samples). No amphibians or mammals were found in this study as food species.
Further, three ambiguous food species, the small mosquito fish Gambusia affinis (Baird & Girard, 1853), and two small gobies Mugilogobius chulae (Smith, 1932) and Microphilypnus sp. had been considered as "nonfood species" (4 sequences, 2.58% of all successful sequences) for their tiny size and coexistence with potential predators as the food species in the same fecal samples (see Section 4). They also were considered as "indirect food species," that is, not hunted by otters for food. All prey and nonfood species are listed in

| Prey species number in spraint samples
In total, 39 spraint samples contained 1 prey species, which is 60.94% of all 64 tested samples. Eight samples contained 2 species (12.5%), and only 2 samples contained 3 prey species (3.13%). There were 15 samples with no food species inside. However, half of them were host valid species sequences, which were identified as nonfood species (from 1 to 4 species). Only 8 samples hosted no valid sequence, given the absence of PCR products using any of the 9 primer sets, or poor sequencing results in some PCR products, despite them passing the electrophoresis check and being selected for sequencing ( Figure 5).

F I G U R E 5
Number of prey species detected in spraint samples. Spraints containing only a single prey were most frequent, that is, up to 39 samples among the 64 tested. The orange color indicates 7 spraints containing nonfood species but prey

| Food species identified in this study
The Eurasian otter is a generalist predator and thus displays foraging strategies as adaptive responses adjusted to food availability (Almeida et al., 2012;Barrientos et al., 2014;Krawczyk et al., 2016).
The results of this work suggest that the Kinmen otters displayed clear piscivorous foraging, especially on the introduced tilapia species. While the dominance of invasive tilapia has become a great threat to local endangered fishes , they appear to provide plentiful food resources for predators like otters. Not surprisingly, tilapia were the most frequent food species identified in Kinmen, present in 55% of the analyzed 64 spraint samples, and in Other native nonfish food species were also rarely detected in this study. It is possible that this is an effect of the small sample size (n = 64), or that there is simply no need for Kinmen otters to hunt alternative prey, given that tilapia constitute a sufficient food resource.

| Nonfood species identified in this study
In total, 59 sequences were found in this work that derived from nonfood species, including 4 bacteria groups (genera Aeromonas, Vibrio, Shewanella and one similar to family Idiomarinaceae with 80.86% similarity; 18 sequences), 1 amoeba (1), 1 amphipoda (1), 1 diatom (1), 3 fishes (4), 3 insects (4), some rotifers (4) and their eggs). Besides, King et al. (2008) suggested that fecal matter may be contaminated with planktonic organisms while collected in the sea and potentially even in fresh water. In the case of otters, contamination with water is inevitable, as they typically defecate soon after leaving the water, and water from their bodies wets the spraints. Tiny aquatic creatures such as rotifers and diatoms were similarly found in this study (Table 3b).
Secondary predation may also be evident in otter diets (Kumari et al., 2019). Some organisms were probably incidentally consumed by larger organisms predated by the otters. For example, the Kinmen otters feed on tilapia, euryphagous fishes, which in turn feed on smaller, dead organisms, or even the organic matter of other animals (e.g., feces) that are thus later present in the otters' guts. Here, I have the opportunity to check the relationships between the predator (the donor of the spraint) and potential "indirect food species," given that I list the diet analysis results of each spraint sample one by one. In doing so, I suggest that some tiny fishes such as mosquito fish and small size gobies should be considered as indirect food species, as they were observed alongside their possible predators, the little grebe, Nile tilapia Oreochromis niloticus and Japanese tiger prawn in this work (see samples #4, #8, #20, #63 in Table 3). I also suggest that organisms unintentionally ingested by means of being attached to larger organisms (e.g., parasites) predated by the otters should be considered as nonfood or indirect food species.
There were 26 valid sequences of Lutra lutra, comprising 16.77% of all successful sequences among this work. When barcoding with mtDNA sequences, it is difficult to distinguish whether these otter DNA fragments were from donor or prey species, that is, if cannibalism occurred. In the absence of any evidence suggesting cannibalism, I consider these to be the DNA sequences of spraint donors and thus did not include the in the diet species list.

| The quality of analyzed spraint samples
Sample freshness is very important in molecular scatology studies.
In this study, I used the highest quality spraint-derived DNA as possible, having prescreened the extracts with a panel of microsatellites (see Hung et al., 2004;Park & Cho, 2017). Such genotypes could be used in a later study to connect individual identifications to diet information, if needed.
In general, only the single-band PCR products were selected for sequencing, except for those amplified with the Til9020F/RM primer set. Most of the PCR products amplified with these primers showed multiple bands (e.g., Figure 3d), but they could always be sequenced successfully and with good quality (sequencing using the forward primer Til9020F only). It is interesting that the Til9020F/RM primer set worked very well in this work by detecting prey successfully in 29 spraint samples, not only on tilapia (27 samples) but also on other fish (1) and shrimp (1) species. The success of this primer set can be attributed to its custom design to target tilapia specifically, on the understanding that these species are abundant in Kinmen and probably constitute prey. Reviewing potential prey species, and designing appropriate custom primers, is therefore recommended in future studies where possible prey can be identified. This improves on the use of published universal primers, which are better used to "discovery" novel prey than to amplify known prey species.
Fifteen spraint samples showed the absence of food species inside. It is possible that they did contain nonfood species, and that they were false negative results. The production of such false negatives (i.e., failure of amplification when target food DNA is or was present in the sample) could be due to degradation of the DNA present in the sample, failure of the DNA extraction, or failure of the PCR amplification (Deagle et al., 2005;Kalle et al., 2014;King et al., 2008). In studies where the real diet is unknown, such as in this study, which focused on wild individuals, monitoring the incidence of false negatives is extremely difficult and their complete elimination is unlikely to be possible. My solution was to collect spraint samples as fresh as possible for DNA extraction, to maximally reduce DNA degradation.
For adult Eurasian otters, defecation is also an important behavior to mark their territories. Beside food remains, spraints include the fairly inconspicuous secretions of two anal glands, plus a jelly-like substance secreted somewhere in the intestine itself. Occasionally, a spraint consists of nothing but this jelly. When Eurasian otters produce this jelly spraint, the defecation is not for purposes of elimination, but for scent communication (Kruuk, 2010). Among the nonfood species spraints, 7 samples were noted as jelly or mucus samples (#17, #39, #44, #45, #47, #60 and #64), which is probably the reason they contained no food materials. However, it is not inevitable that jelly or mucus will contain no diet information. In other jelly or mucus samples (#19 and #43), I detected tilapia sequences, though no scales or spiny bones were observed. Aside from jelly or mucus, 5 spraints were greenish, small in size, soft and wet, and no scales or bones (or very few) were found inside (#7, #21, #31, #32, and #38). Such spraints were considered to be feces belonging to unweaned cubs and were easily distinguished from their mother's (or other adults') spraints. When the cubs are small and feeding only by suckling inside the natal holt, their mother typically eats their spraints. Later, during the days of the cubs' life outside the natal holt, when they are about 2 months old and before weaning, it is possible for the cubs to leave spraints beside their mother's ones outside the natal holt. Occasionally, such spraints comprised no food species DNA.
In two of the spraint samples, no food species were detected, but scales and spiny bones of fish were visually apparent inside them (#36 and #37; see Figure 6). Both spraint samples were very small, but had passed the prescreening test and could even be identified at the individual level. In these cases, an abundance of nonfood species' DNA may have caused their preferential amplification in the PCR.

| A recommended modified procedure
It is better to have an idea of the possible prey fauna in advance, 4a. If more than half the sequencing results (>3) refer to one single prey species, note the result and end the experiment.
4b. If two or more species were identified, or multiple bands occurred in any PCR products but no valid sequence was produced, execute PCRs with the remaining primer sets (i.e., up to 9 PCRs in total).
In the first step, the first four recommended primer sets are the most universal in barcoding analysis, and can amplify most animals' COI genes assuming they are not degraded. When some bands are observed in the step but prove difficult to Sanger sequence, I suggest running all primer sets before noting that the sample produced no sequencing data. Repeated PCR tests provide several opportunities to detect the diet species in a spraint sample; in this case, such efforts did yield data eventually.
This approach cannot be applied in species that consume many prey individuals each day (such as middle and large size carnivores and piscivores, especially those that feed on countless F I G U R E 6 Spraint samples of the various forms and conditions produced by Eurasian otters on Kinmen Island. (a) A standard-looking spraint collected in Tai Lake, Tai-140 (sample #28 in Table 1 small invertebrates in a day), as they are likely to have too many prey species in their scats. The sDNA metabarcoding methods are much more suitable in such cases if funds permit their use.
Those with few prey individuals/species are otherwise ideal subjects for the series of Sanger-sequencing reactions proposed herein.

| CON CLUS IONS
Given that Eurasian otters generally consume a small number of prey species, this procedure is ideal for discerning those appar-

CO N FLI C T O F I NTE R E S T
The author declares no conflicts of interest.