The mountain lion (Puma concolor) is an elusive, common large predator native to California, USA. Monitoring of wild felids often requires non-invasive genetic tools (Ernest et al. 2000). In California, population monitoring has increasingly relied on canine surveys for feces followed by fecal genoty** and mark-recapture analyses to estimate population abundance. Although microsatellites have previously been relied upon for such noninvasive surveys, they have several weaknesses that must be overcome to prevent errors (Taberlet et al. 1999), including low amplification rates (Brinkman et al. 2010; Murphy et al. 2007; Piggott 2004), challenges calibrating between laboratories (von Thaden et al. 2017), allelic dropout, and false alleles (Creel et al. 2003; Pompanon et al. 2005).

There is great need for genetic methods that can be applied to low quality DNA samples while maintaining statistical power to differentiate individuals in low-diversity populations. In southern California, many mountain lion populations are impacted by urban development, and exist in small numbers with low genetic diversity (Ernest et al. 2014; Gustafson et al. 2017; Riley et al. 2014), making it difficult to apply genetic methods without high probability of monomorphic loci and identification errors. A previous SNP panel (PumaPlex100; Erwin et al. 2021) demonstrated the utility of SNPs to differentiate among mountain lion individuals. However, this panel was not optimized for low-diversity populations and requires up to four multiplexes. Therefore, we developed a single multiplex of 95 SNPs for the Fluidigm platform (Fluidigm, San Francisco USA) that can be used across laboratories for surveys of California and potentially other mountain lion populations. These SNPs can identify individuals for abundance monitoring, assign sex, and discriminate mountain lion feces from bobcat (Lynx rufus), the most common non-target species detected during canine surveys.

We identified SNPs using existing RADseq data (Illumina HiSeq, paired-end 150 bp reads) from mountain lion tissues sampled as part of a California statewide population genetic study (Gustafson et al. 2021; raw data available at https://doi.org/10.17605/OSF.IO/HUF4K) and mapped reads to the Puma concolor scaffold-level genome assembly (PumCon1.0), using 104 individuals, including up to 15 individuals from each of 10 populations identified by Gustafson et al. (2019). We used vcftools to filter biallelic autosomal SNPs to a PHRED score ≥ 30 (99.9% accuracy) with ≥ 80% of individuals genotyped at each SNP. To maximize heterozygosity and resolution to differentiate individuals, we used plink v1.90b6.10 to filter the minor allele frequency to 0.40–0.45. We retained 134 SNPs distributed across 28 scaffolds (mean within-scaffold distance of 11.3 Mbp) for further validation (Supporting File 1). We identified species-diagnostic SNPs using published mitogenome sequences for mountain lion and bobcat (Supporting information Table S1). We also identified sex-linked SNPs from both the SRY locus on the non-recombining region of the Y chromosome and the zinc finger orthologs of the X and Y chromosomes (Supporting information, Table S2).

We used pysam v0.15.0 to retrieve 100 base pairs of up- and down-stream flanking sequence for the 134 SNPs. We used Fluidigm’s D3 Assay Design Tool (https://d3.fluidigm.com) in conjunction with the flanking sequences (Supporting Information, Table S3) to design primers, which we ordered from Fluidigm (Supporting information, Table S4).

To test the performance of the SNP panel on fecal DNA, we used 19 pairs of tissue and fecal samples collected during mountain lion necropsies conducted by the California Department of Fish and Wildlife from 14 different counties. Feces were experimentally weathered outdoors in direct sunlight for seven days. Tissue and fecal DNA were extracted using the Genomic Tip 20G kit and the QIAamp Fast DNA Stool Mini Kit (Qiagen, Valencia, CA), respectively. To assess the performance of the SNP panel on low-diversity populations, we extracted tissues for 26 mountain lions from the highly inbred Santa Ana Mountains population in southern California. We also extracted 16 bobcat tissues to test the species-diagnostic SNPs.

We used Fluidigm’s protocol for 96.96 Dynamic Arrays with integrated fluidic circuits run on the Juno thermocycler paired with the Biomark HD system. We modified the pre-amplification PCR reaction to align with optimizations for degraded samples (see Table S2 in von Thaden et al. 2020). We also increased the manufacturer’s PCR cycling profile from 34 to 45 in the final allele specific reaction. We analyzed genoty** results using Fluidigm SNP Genoty** Analysis Software version 4.5.1. We selected a final panel of 91 autosomal SNPs based on consistent cluster separation and high call rates. We also selected two species-diagnostic and two sex-linked SNPs. For the Santa Ana population, we tested for linkage disequilibrium LD using the r package poppr (Kamvar et al. 2014) to calculate the index of association among pairs of loci and corrected for a false detection rate FDR of 5%. We estimated expected heterozygosity HE using the adegenet package (Jombart and Ahmed 2011) and calculated the probability of identity PID and probability of identity in siblings PIDsibs using the popgenutils package (Tourvas 2021). All packages were implemented in R 4.1.1 (R Core Team 2021).

When genoty** the 19 paired tissue and fecal samples at the 91 autosomal SNPs, we obtained > 95% call rates on 18 of the 19 fecal samples. The overall agreement between the paired samples was 99.9% (SD = 0.6%), indicating a genoty** error rate < 0.1%. For the low-diversity Santa Ana population, no loci were in LD following FDR correction. The average HE for the 91 autosomal SNPs was 0.404, an estimate higher than previously measured using 42 microsatellites (HE = 0.33; Gustafson et al. 2019), suggesting our SNP filtering protocol successfully inflated heterozygosity and increased power for individual identification. In fact, values for PID and PIDsibs across all SNPs were < 1 × 10−10 (Supporting information, Table S5) and individual identifications could be made with high confidence from as few as 15 SNPs (Fig. 1). For 33 samples where both sex-linked SNPs (2FelidSRYSNP-GT, mlbcZfy-680) yielded genotypes, 30 pairs (91%) agreed with known sex, suggesting < 5% sex-ty** error rate. The species-diagnostic SNPs (mtdna_658, mtdna_2089) were 100% successful at differentiating mountain lions from bobcats. Further, bobcat reference genotypes were monomorphic at all autosomal SNPs.

Fig. 1
figure 1

Decrease in PID and PIDsibs with increasing number of loci for 26 mountain lions from the low-diversity Santa Ana Mountains population in California, USA. Figure generated using the pid_permute function for the R package popgenutils (Tourvas 2021) and based on 1000 permutations across different combinations of loci. Only 1 to 15 of 91 total autosomal loci are shown for clarity

Full size image

This high-resolution panel of 95 SNPs was optimized for the Fluidigm system but can be genotyped in any SNP-ty** platform. The panel was designed from mountain lions throughout California and may be subject to ascertainment bias when applied to other geographic regions. As a result, the utility of this panel for individual identification throughout the species range requires further verification. The high minor allele frequency screening step for SNP selection proved useful for differentiating individuals but also inflates estimates of heterozygosity. Therefore, certain analyses, such as estimating and comparing heterozygosity or identifying population structure and gene flow would not be appropriate uses of these markers.