Altered lentiviral infection dynamics follow genetic rescue of the Florida panther

Abstract

Wildlife translocations are a commonly used strategy in endangered species recovery programmes. Although translocations require detailed assessment of risk, their impact on parasite distribution has not been thoroughly assessed. This is despite the observation that actions that alter host–parasite distributions can drive evolution or introduce new parasites to previously sequestered populations. Here, we use a contemporary approach to amplify viral sequences from archived biological samples to characterize a previously undocumented impact of the successful genetic rescue of the Florida panther (Puma concolor coryi). Our efforts reveal transmission of feline immunodeficiency virus (FIV) during translocation of pumas from Texas to Florida, resulting in extirpation of a historic Florida panther FIV subtype and expansion of a genetically stable subtype that is highly conserved in Texas and Florida. We used coalescent theory to estimate viral demography across time and show an exponential increase in the effective population size of FIV coincident with expansion of the panther population. Additionally, we show that FIV isolates from Texas are basal to isolates from Florida. Interestingly, FIV genomes recovered from Florida and Texas demonstrate exceptionally low interhost divergence. Low host genomic diversity and lack of additional introgressions may underlie the surprising lack of FIV evolution over 2 decades. We conclude that modern FIV in the Florida panther disseminated following genetic rescue and rapid population expansion, and that infectious disease risks should be carefully considered during conservation efforts involving translocations. Further, viral evolutionary dynamics may be significantly altered by ecological niche, host diversity and connectivity between host populations.

1. Significance

Florida panthers historically harboured a unique subtype of feline immunodeficiency virus (FIV) typically found in bobcats and distinct from the common FIV found in all other puma populations studied to date. We investigated the consequences of genetic rescue of the panther on infection dynamics of FIV. Our findings reveal extirpation of the historical panther FIV and replacement by a virus more typical of puma infections elsewhere. This ‘modern’ panther FIV shares remarkably high sequence homology with FIV from translocated Texas pumas and exhibits extremely low genetic variation relative to other lentiviruses across a 2-decade sampling period. These findings provide evidence that translocations impact microparasite distribution and evolution in unpredictable ways and highlight risk of inadvertent disease emergence in threatened populations.

2. Introduction

Anthropogenic introduction of a parasite to a geographical region or host species outside of its natural range has been termed ‘pathogen pollution’ and is increasingly recognized as an important driver of disease emergence [1–3]. Multiple mechanisms underlie pathogen pollution, including wildlife conservation efforts involving translocation [4,5]. In the face of progressive habitat loss and fragmentation, active management strategies such as translocation have become pervasive tools to prevent extinction of threatened species [6,7]. Common goals of translocations include establishment of new populations, reintroduction of a species to a historical range, movement of species from regions of human development or conflict, and supplementation of existing populations to enhance population size, genetic diversity or both [8]. Translocations have historic precedent, and infectious disease has only recently been recognized as an important associated threat [4]. Important examples of host–parasite co-introduction include the translocation of rabid raccoons from Florida to Virginia in the late 1970s and early 1980s [9], the release of captive-bred plains bison harbouring bovine tuberculosis and brucellosis into Wood Bison National Park [10], and human-mediated introduction and dissemination of the causative agent of salmonid whirling disease across the USA [11]. Such instances with obvious implications for disease emergence or reemergence have been studied in detail. Yet the evolution of a microparasite in response to selection pressures imposed by a naive host in a new geographical region can have negative, neutral or positive outcomes on community dynamics that may become evident only with time and concurrent ecological change [5]. Long-term, broad assessments of the influence of subclinical microparasite translocation on ecosystem health are important yet rare.

The Florida panther (Puma concolor coryi) represents a widely documented translocation success story [12]. Once widely ranging throughout the southeastern United States, this iconic subspecies of puma was isolated to a rapidly urbanizing region of South Florida following decades of persecution and habitat destruction [13–15]. Despite state declaration of the panther as a protected species in 1958 [16], and federal listing as an endangered species in 1967 [17,18], panthers were reduced to an estimated isolated population of less than 30 adults by the early 1990s [19], with a projected time to extinction of less than 2 decades [12,20]. This rapid and progressive decline prompted several studies aimed at rescue and restoration [21–23]. Habitat suitability studies were conducted in North Florida (figure 1) in the 1980s and early 1990s prior to the well-known genetic rescue effort of 1995 in South Florida [21,22]. During two separate assessments in North Florida, conservation biologists released a total of 26 pumas that were either wild-caught in Texas or were offspring of wild-caught, captive-held Texas pumas that were part of a captive breeding programme [21,22]. The release site was sufficiently distant from South Florida to preclude the possibility of mixing with the canonical panther population (figure 1) [22]. In addition, males were vasectomized, all released pumas were radiocollared and closely monitored, and all 26 animals were removed from North Florida following completion of the project—all safeguards to prevent exposure to the South Florida panthers [21,22].

Figure 1.

FIVpco is distributed across South Florida panther habitat following genetic rescue translocation. Each circular point represents a capture location of an infected Florida panther, while stars represent the first known locations of Texas pumas following release. Texas pumas 105 and 106 tested positive for FIVpco at the time of release in 1995, while puma 104 tested positive on recapture in 1997. The release site for the historic reintroduction feasibility studies in North Florida is distant from the occupied habitat in South Florida. Primary, secondary and dispersal habitat zones were designated by Kautz et al. [24]. RFSRS, reintroduction feasibility study release site; ENP, Everglades National Park; BCNP, Big Cypress National Park; FSPSP, Fakahatchee Strand Preserve State Park. (Online version in colour.)

As the reintroduction feasibility studies were taking place in North Florida, the panther population in South Florida was concurrently reaching its nadir in terms of size and genetic diversity [13,15,19]. In 1995, with extinction a distinct possibility, the decision was made to pursue genetic introgression via the translocation of eight female pumas from Texas to Florida. Subsequently, significant improvements were documented in fitness parameters, heterozygosity and survival rates of all age groups; the population has since expanded to approximately 120–230 adult and subadult panthers [25–27]. Intensive management has undoubtedly improved long-term prospects for persistence of the Florida panther population [12,26], and this outcome has provided one of the most heralded examples of translocation as a useful and necessary conservation tool [28,29].

While many studies have examined the impact of genetic rescue on the health, viability and ecology of the Florida panther [12,23,27], the influence of puma translocations on microparasite dynamics has not been critically evaluated. Pumas are host to several retroviruses, including the highly pathogenic gammaretrovirus feline leukaemia virus (FeLV), which has resulted in significant morbidity and mortality in Florida [30,31], and the putatively apathogenic lentivirus feline immunodeficiency virus (FIV), which is prevalent in most puma populations but not associated with overt clinical disease [32–34]. Three subtypes of FIV have been reported in the puma: the host-adapted FIVpco (also known as puma lentivirus B (PLVB)), the bobcat-adapted virus known as FIVlru (PLVA) and rare infections originating from domestic cats (FIVfca) [34–36]. Most puma populations are host to endemic FIVpco, with few accounts of FIVlru as a spillover infection from bobcats [33,34,37,38]. By contrast, the Florida panther has historically harboured a preponderance of FIVlru based on previous studies [33–35].

In this study, we investigate the dynamics of FIV infection in the Florida panther pre- and post-Texas puma translocation and document in detail: (i) FIVpco infection in several translocated Texas pumas; (ii) extirpation of historic subtype FIVlru from the panther following translocation; (iii) an exponential increase in FIVpco infections concurrent with population expansion; and (iv) low genetic variation of FIVpco among and between Florida panthers and Texas pumas across a 23-year sampling period. The genetic homogeneity of this viral lineage is highly unusual among lentiviruses, and may relate to the unique demographics and isolated nature of this population. This work provides rare empirical evidence of altered viral ecology as a direct effect of wildlife translocation and is the first to thoroughly characterize viral evolutionary processes concurrent with genetic introgression of the host. As such, it offers a unique opportunity to investigate the impacts of human-mediated dissemination of a subclinical parasite in a naive host population.

3. Results

(a). Historic translocations impacted feline immunodeficiency virus subtype dynamics in the Florida panther

To investigate FIV in the Florida panther pre- and post-Texas puma translocation, we adapted a recently described approach to whole-genome sequencing of low-copy-number viruses from variably degraded biological samples [39]. Our analysis revealed a remarkable shift in the occurrence of circulating FIV subtypes that coincided temporally with the translocation of Texas pumas to Florida (figure 2). Previously published FIV isolates from Florida panthers were exclusively of a FIVlru subtype recovered from samples collected in the late 1980s and early 1990s [34]. We conducted an exhaustive investigation to search for isolates of FIVpco in free-ranging canonical Florida panthers and pumas translocated to Florida from Texas. In this study, samples collected between 1984 and 2011 (n = 264) were assessed by PCR and NGS to identify 56 FIVpco infections. We identified: (i) the first and two earliest FIVpco isolates recovered from Florida to date (TX16, May 1988, and TX15, June 1988), which both originated from Texas pumas released in North Florida during the reintroduction feasibility studies prior to genetic rescue [21,22]; (ii) FIVpco from three of the eight Texas pumas that were translocated and released into South Florida as part of the genetic rescue project in 1995 (TX104, TX105 and TX 106); and (iii) rare co-infection of FIVlru and FIVpco in two panthers (FP37 January 1990 and FP36 January 1992), which also represent the earliest FIVpco isolates from panthers in South Florida.

Figure 2.

FIV dynamics in Florida are altered by introductions of infected Texas pumas. Following translocation of FIVpco-infected Texas pumas in the 1980s and 1990s, the number of FIVpco-infected Florida panthers increased dramatically, while FIVlru infections fell below detectable levels. Texas puma 104 tested negative in 1995 but was positive on recapture in 1997 (curved arrow). Cumulative period prevalence was calculated throughout the study period. (Online version in colour.)

We estimated the cumulative period prevalence of FIVpco and revealed a marked increase in infection following genetic rescue (figure 2). The relative risk of a Florida panther sample testing positive for FIVpco after 1995 (following genetic rescue translocation) was 3.8 times higher (95% CI 1.8–8.0, p = 0.0004) compared with panthers sampled prior to 1995. We used coalescent theory to estimate the effective number of FIVpco infections across time based on whole-genome sequence data [40] and revealed an exponential increase in infections between approximately 2000 and 2002, corresponding to one of several periods of accelerated expansion of the Florida panther population (figure 3a,b) [19]. Surprisingly, FIVlru infections were not detected after 1992 (figure 2), despite previous findings that intraspecific panther transmission of FIVlru was strongly supported by phylogenetic analysis [35]. Collectively, these results demonstrate a sharp increase in FIVpco infections concurrent with the apparent disappearance of FIVlru in the Florida panther and provide evidence for the genetic rescue translocation and subsequent population expansion as a driver of change in infection dynamics.

Figure 3.

An exponential increase in the effective number of FIVpco infections corresponds to a period of demographic panther recovery. (a) The number of samples screened for FIVpco and FIVlru is plotted alongside the minimum panther count for each year. (b) Bayesian skyline plot shows a sharp rise in the effective population size of FIVpco (effective number of infections) between approximately 2000 and 2002. The rise in the effective number of infections coincides temporally with one of several periods of accelerated expansion of the Florida panther population as estimated by minimum panther count [19]. Dashed lines highlight the correlation. (Online version in colour.)

(b). Genetic variation of FIVpco in the florida panther is remarkably low

To investigate the evolution of FIVpco in the Florida panther in the context of Texas translocations, we constructed FIVpco consensus sequences (n = 56) from overlapping fragments within a tiled amplicon framework [39]. We recorded strikingly low interhost diversity of FIVpco in the panther population, demonstrated by a mean pairwise nucleotide identity of approximately 99% for nearly whole-genome sequences spanning over 2 decades of sample collection (table 1). Across the 28-year sampling period (1983–2011), divergence between isolates from translocated Texas pumas and Florida panthers was extraordinarily low (pairwise identity 98.8–100% excluding missing data sites; table 1). For comparison, sequence homology of FIVpco genomes from other populations, including those in California and Colorado, was approximately 88% across a shorter sampling period (less than 10 years) [33]. Similarly, low divergence was reported across isolates from historic Florida panthers and those from contemporary Florida–Texas admixed animals (table 1). The genetic stability of FIVpco and previously reported relative fitness advantage over FIVlru [35] probably contributed to an increase in highly homologous FIVpco infections and a circulating FIV subtype shift mediated by multiple ecological and molecular determinants of infection dynamics through time.

Table 1.

All FIVpco isolates from Texas and Florida share high sequence homology across the genome. Whole-genome pairwise identity averages greater than 99%. The number of individual samples for each complete open reading frame is provided, with average pairwise identify for each gene (missing data excluded).

	gag	pol	vif	orfA	env
no. of sequences with no missing data	51	18	55	39	23
mean pairwise identity (%)	99.5	99.7	99.4	98.8	99.4

(c). FIVpco isolates from Texas and Florida comprise a single lineage

To further investigate relationships between and among FIVpco isolates from Texas and Florida, consensus sequences were subjected to phylogenetic analyses. Due to short branch lengths derived from highly homologous sequences, a Bayesian maximum clade credibility tree was converted to a cladogram for improved visualization (figure 4). Two isolates from translocated Texas pumas (TX104 and TX106) are basal to all other sequences from Florida. TX104 and TX106 were translocated to Fakahatchee Strand Preserve State Park in 1995 (figure 1). TX106 tested positive for FIVpco in 1995, while TX104 samples from 1995 consistently tested negative. A sample collected during recapture of TX104 in 1997, however, yielded a positive result with a consensus sequence highly homologous (99% pairwise identity) to that from TX106. FIVpco was additionally recovered from a third breeding female translocated to Everglades National Park (TX105) in 1995 (figure 1). Early isolates (1988–1994) recovered from Texas pumas released in North Florida as part of the reintroduction feasibility studies shared high sequence homology with those of Florida origin and did not comprise a separate lineage (figure 4).

Figure 4.

Two Texas FIVpco isolates are basal to all isolates recovered from Florida panthers. All sampled isolates comprise a single FIVpco lineage with low genetic variation, independent of sample collection date. This maximum clade credibility tree was constructed from nearly whole-genome sequences and converted to a cladogram for improved visualization due to short branch lengths. Host ancestry was assigned based on relationships reported by Johnson et al. [12]. FIVpco infections do not form separate clades based on host ancestry. Two instances of presumed direct maternal transmission are identified, along with indirect familial transmissions, as based on sequence relatedness. Temporal structuring is largely absent in the phylogeny. FIVpco and FIVlru co-infections are denoted by stars. (Online version in colour.)

The genetics of the Florida panther population have been impacted by several anthropogenic introductions in addition to those from Texas, including escaped and/or released captive pumas from private collections [12,13,23]. We therefore sought to identify unique FIVpco lineages corresponding to the genetic ancestry of the host; however, distinct clades of FIVpco corresponding to panther ancestry were not identified (figure 4). Rather, all FIVpco sequences were highly homologous, consistent with a point source introduction of FIVpco and a founder effect resulting in a highly conserved virus throughout Florida.

To investigate the frequency of inferred vertical transmission, we sought to identify phylogenetic relatedness of viruses isolated from dam–offspring pairs. Putative maternal transmission was identified in two cases (FP124 to FP126 and FP161 to K279); transmission divergence was greater for dam–offspring pair FP113 and FP177, suggesting intermediate transmission(s) rather than direct dam to offspring infection (figure 4).

(d). FIVpco from Florida and Texas is genetically distinct from all other puma feline immunodeficiency viruses

Following phylogenetic analysis of FIVpco isolates from Florida, we compared this unique FIV lineage to other puma FIVs across North America. Using sequences obtained from GenBank and a representative subset of the sequences from this study, we constructed a Bayesian maximum clade credibility tree from approximately 480 bp within the reverse transcriptase (RT) region of pol, a highly conserved region that has classically been used to characterize FIV and other lentiviruses (figure 5). The RT-pol phylogeny illustrates that all previously published sequences of FIV from Florida panthers prior to the 1995 translocation are of subtype FIVlru. Sequences from the current study share high homology with two isolates derived from wild-caught Texas pumas held in Texas zoos (Pco-28 and Pco-733), comprising a single lineage distinct from others in the Americas.

Figure 5.

Florida panther FIVlru is highly divergent from Texas-derived FIVpco. Early FIV isolates (subtype FIVlru) recovered from Florida panthers are highlighted at top (dark grey box). Contemporary FIVpco isolates from Florida panthers (light grey box) cluster with those from Texas pumas (italicized). This maximum clade credibility tree is based on approximately 480 bp within the conserved region encoding the viral RT (RT-pol). Isolate names are coloured according to puma geographical region. (Online version in colour.)

4. Discussion

The Florida panther represents a rare opportunity to document the role of translocation in dissemination, persistence and evolution of an apparently non-pathogenic virus [21,22,34]. We used a PCR amplification protocol that provided enhanced sensitivity to detect FIVpco in highly degraded field samples collected 2–3 decades ago. This technique allowed us to document FIVpco in at least two of eight pumas translocated from Texas to Florida in 1995 for genetic rescue. We unequivocally show that the introduction of two FIVpco carriers during genetic rescue is associated with a dramatic increase in FIVpco prevalence, and that following genetic rescue, FIVpco replaced FIVlru as the dominant FIV subtype in the panther (figure 2).

We previously reported strong evidence for intraspecific transmission of FIVlru in the Florida panther in the late 1980s and early 1990s [35]. Here, we show that the transmission chain of bobcat-adapted FIVlru in the panther was transient (figure 2). FIVlru in the puma has low fitness compared with FIVpco [35], and co-infections are uncommon (n = 2; figures 2 and 4). A competitive fitness advantage of FIVpco is therefore likely to have contributed to the extirpation of FIVlru from the Florida panther. Further, the disappearance of FIVlru signals an absence of contemporary spillover events from bobcats to panthers, which is probably related to altered interspecific contact rates stemming from changes in behaviour, habitat use and/or prey preference in contemporary panthers.

In addition to the two infections in animals translocated for genetic rescue, we additionally identified four FIVpco infections in samples collected from Florida prior to 1995. These include two infections from the North Florida reintroduction feasibility studies [21,22], separate from the genetic rescue, and two infections in canonical panthers, FP36 (1990) and FP37 (1992); these animals were co-infected with FIVlru (figure 4). All FIVpco sequences recovered from Florida panthers are highly homologous to those from Texas pumas (figures 4 and 5), supporting a Texas origin of FIVpco in panthers. The detection of FIVpco in two panthers prior to genetic rescue may indicate either pre-existent low-level FIVpco infection that was not readily transmitted or pre-supplementation contact with animals of Texas origin. Although at least two of the Texas pumas released in North Florida were infected with FIVpco, these animals were closely monitored and did not overlap in space with panthers in South Florida, suggesting these animals are not the source of this FIVpco in the contemporary panther population [22]. Outside of this study, only two isolates from Texas have been partially sequenced: Pco-28 (wild-caught in Texas and held at a zoo in San Antonio) and Pco-733/TX106 (wild-caught in Texas and held at a zoo in Houston prior to release in South Florida) [34]. Additional analysis of Texas FIVpco is warranted to determine the phylogenetic relationships between contemporary Texas isolates and Florida FIVpco reported here.

Whole-genome sequencing of FIVpco revealed exceptionally low diversity between and among isolates from Florida and Texas (table 1), a surprising finding given that a previous study reported diversification of eight FIVpco lineages in the northern Rocky Mountains occurring within the last 20–80 years [41]. Moreover, diversity among other FIVpco lineages is similar to that reported for other FIVs, and evidence for ongoing positive selection and adaptation was reported in a recent genome-wide analysis of FIVpco in the puma [33]. The limited genetic variation of FIVpco in Florida is thus unusual and could be related in part to the introduction of only a single homologous FIV lineage and the isolation of the panther population. Among ‘open’ populations with contiguous habitat, FIVpco genetic variation can be partially attributed to recombination events arising from co-infections following dispersal of infected animals [42]. Because new members do not naturally enter and disperse from Florida, viral recombination as a source of genetic variation is limited by the absence of divergent FIV lineages from other geographical regions. This premise is supported by a study of FIVpco in the Snowy Range of Wyoming, which concluded that endemic FIV in pumas evolves slowly in the absence of co-infections and recombination [37].

The relatively large proportion of infected panthers suggests frequent vertical transmission, which could in part explain the observed exponential increase in infections in the absence of evidence for epidemic parameters such as an enhanced replication rate. High rates of vertical transmission have been previously reported for endemic FIVpco; one study reported infection of more than 50% of cubs born to infected dams [37]. Vertical transmission provides a simple putative mechanism underlying the observed increase in FIVpco infections concurrent with demographic recovery of the panther, a pattern that has also been reported for pumas in the Rocky Mountains [41]. While the mechanistic details remain theoretical, the relative risk associated with translocation is striking and strongly supports a role for infected Texas pumas as primary drivers of FIVpco spread through Florida.

Collectively, our findings document a shift in circulating FIV subtype following translocation of Texas pumas infected with FIVpco. This work represents rare documentation of human-mediated alterations in viral ecology and is the first study to thoroughly investigate the evolution of a virus concurrent with genetic introgression of the host. Further, we report on evolutionary patterns that diverge from the expected adaptive processes typical of lentiviruses. This contribution provides a distinct example of the potential impacts of translocation on extant parasite communities and highlights the deterministic influence of intensive management on species interactions and persistence. While the benefits of genetic rescue clearly outweigh the costs in the case of the Florida panther, dynamic alterations in host–pathogen relationships should be expected in response to translocation, as illustrated by the cautionary tale of Florida panther FIV.

5. Methods

(a). Sample collection and nuclei acid extraction

Blood and tissue samples analysed in this study were collected from Florida panthers and translocated Texas pumas between 1988 and 2011. Capture of free-ranging animals involved trained tracking hounds provided by Livestock Protection Company (Alpine, TX, USA), as previously described [25]. Translocated Texas pumas were captured in southwest Texas, radiocollared for regular monitoring, quarantined for collection of biological data and health screening, and held in captivity, released for reintroduction feasibility studies or released immediately following assessment for purposes of genetic introgression, as described previously [21,22,25]. Biological samples collected at the time of initial capture in Texas were analysed in this study, as well as samples from subsequent re-captures. Additional biological samples were opportunistically collected during routine postmortem examinations. Aliquots of blood and tissue samples were sent to the VandeWoude lab, Colorado State University, for analyses described below. Multiple tissue types were analysed as available with priority assigned to lymphoid organs (i.e. spleen, followed by lymph node). DNA was extracted from tissue, whole blood or peripheral blood mononuclear cells (PBMCs) using an adapted version of the DNeasy Blood and Tissue protocol (Qiagen Inc., Valencia, CA, USA) Tissues were homogenized using the benchtop FastPrep-24 cell and tissue homogenizer (MP Biomedicals, LLC., Santa Ana, CA, USA). Blood samples were incubated in lysis buffer at 56°C overnight. The 264 archival samples were screened for FIV using a multiplex PCR protocol as described below. Collection date, location and host demographic information for each sample selected for viral genotyping are provided in electronic supplementary material, table S1. Capture locations for FIV-infected panthers and translocated Texas pumas are displayed in figure 1, along with the Texas puma release site for North Florida reintroduction feasibilities studies [21,22].

(b). Detection and sequencing of FIVpco

FIVpco was sequenced from 50 Florida panthers and 6 Texas pumas using a method adapted from Quick et al. [39] (https://github.com/VandeWoude-Laboratory). Briefly, 62 primers (31 pairs) spanning the coding regions of the FIVpco genome were designed for amplification of approximately 400 bp amplicons using Primal Scheme [39]. Primer sequences are provided in electronic supplementary material, table S2. Extracted DNA was subjected to two multiplex PCR reactions using Q5 High-Fidelity DNA Polymerase Enzyme (New England Biolabs Inc., Ipswich, MA, USA) and touch-down cycling conditions. Specifically, annealing was initiated at 68°C and decreased by 0.5°C for 6 cycles, followed by 34 additional cycles at 65°C for a total of 40 cycles. Negative samples and no template reactions were included as controls and were consistently negative. Amplicons were labelled using Nextflex Dual-Indexed Barcodes and a library was prepared using Nextflex Rapid DNA-Seq Library Prep Kit (Bioo Scientific Inc., Austin, TX, USA). Products were then sequenced on an Illumina MiSeq (Illumina Inc., San Diego, CA, USA) using the MiSeq reagent kit v. 2 (500 cycles). Paired fastq reads (SRA accession PRJNA566447) of approximately 250 bp with approximately 50 bp overlap were analysed as follows: (i) trimming of indexes, primers, low quality (phred less than 20) and short reads (less than 50 bp) using Cutadapt [43], (ii) mapping of trimmed reads to a multiple reference index using Bowtie2 [44], (iii) conversion of .sam files to .bam files using Samtools [45], and (iv) viewing of sorted .bam files in Geneious [46]. The multiple reference index included all FIVpco whole-genome sequences generated in our laboratory previously and/or all those available from GenBank (42 total sequences including seven unpublished and GenBank accession numbers EF455603–EF455615, DQ192583, KF906185–KF906174). All reads were then mapped to the single ‘best-fit’ reference with the highest number of mapped reads to maximize genome-wide coverage. Reads with low mapping coverage (less than 75%) were additionally mapped to an FIVlru reference sequence derived from a Florida bobcat. In all cases, mapping coverage to the FIVlru index was poor (less than 20%). Consensus sequences (GenBank accession numbers MN531083–MN531106 and MN531112–MN531143) were generated from mapped reads using the highest quality parameter in Geneious as a threshold. ‘N’ was assigned to sites with coverage less than 2 to represent missing data. In addition to the 56 FIVpco consensus sequences constructed for this study, an additional five previously generated, unpublished FIVpco sequences recovered from Florida panthers were included in analyses. These sequences (FP69, FP79, FP127, FP133 and FP134) were generated according to Lee et al. [33] (GenBank accession numbers MN531107–MN531111).

(c). Detection of FIVlru

Primers spanning the coding regions of the FIVlru genome (58 total/29 pairs) were additionally generated using Primal Scheme with parameters as described above (electronic supplementary material, table S2). Puma blood and tissue samples (n = 264) were subjected to at least one of two multiplex PCR reactions to screen for FIVlru. Previously sequenced positive control samples from Florida bobcats were used to confirm assay detection. Positive control samples consistently generated products of the expected size (approx. 400 bp), as confirmed by gel electrophoresis. Negative samples and no template control reactions were consistently negative. Reactions produced weak bands for two known positive, partially sequenced isolates from Florida panthers (GenBank accession numbers KX899918 and KX899922) [35]. However, of 264 panther samples screened, no newly discovered FIVlru infections were detected.

(d). Phylogenetic analysis

Codon alignments of consensus sequences were constructed using the ClustalW algorithm and adjusted manually in MEGA [47]. Multiple sequence alignments were partitioned into open reading frames and screened for recombination using GARD through the Datamonkey interface of the HyPhy package [48]. Model selection was performed independently for each alignment subject to phylogenetic analysis using jModelTest [49]. To examine ancestral relationships within Florida, a genome alignment of all newly sequenced Florida and Texas FIVpco isolates was subjected to Bayesian analysis performed using the MrBayes 3.2.6 Geneious plug-in [50] with gamma-distributed rate variation and the HKY85 + G substitution model. Missing data (N) were included to maximize the number of sites sampled. Four heated chains of 1 100 000 chain length were run with a subsampling frequency of 200 with the initial 10% discarded as burn-in. An isolate recovered from Vancouver Island (PLV-1695) was used an outgroup (accession number DQ192583). To determine if unique lineages of FIV correspond to panther ancestry, genetic and field observation data from Johnson et al. [12] were extracted and assigned to isolates as tree annotations based on colour. Additionally, isolates from parent–offspring pairs were identified based on Johnson et al. [12] to examine putative maternal transmission.

To estimate changes in the effective number of infections through time as approximated by effective population size of the virus [41], we examined the distribution of coalescent events using a Bayesian skyline plot constructed in BEAST 2 [51]. Effective population size was inferred for five intervals based on coalescence under the HKY + G model with four rate categories and an uncorrelated relaxed molecular clock with lognormal rate distribution. Two independent Markov chain Monte Carlo (MCMC) runs were performed for 10 000 000 generations, sampled every 500th generation. Examination of the MCMC samples revealed convergence and adequate mixing of the chain with estimated sample sizes greater than 200.

Because most studies of FIV in pumas have been based on a conserved approximately 480 bp region encoding the RT enzyme within pol (RT-pol), a representative subset of the sequences derived from Florida and Texas in this study were subjected to further phylogenetic analysis, along with additional sequences obtained from GenBank (accession numbers KX899918–KX899922, KF906167–KF906170, KF906163–KF906174, EF455603–EF455615, U53718–U53766, U03982, DQ19258). Bayesian analysis was used to infer phylogenetic relationships for RT-pol under the GTR + G substitution model with four gamma-distributed rate categories and uncorrelated branch lengths, again using the MrBayes Geneious plug-in [50] with chain length and burn-in as described above. HIV-1 was used as an outgroup (accession number LT726763). All trees were annotated in iTOL v. 3 [52].

(e). Estimate of prevalence and assessment of translocation risk

Panther minimum count was obtained from McBride et al. [19]. To examine the impact of genetic rescue on the proportion of panthers infected with FIVpco, we estimated cumulative period prevalence as the cumulative number of infections detected divided by the total number of samples screened to date. FIVpco prevalence estimates were calculated for years 1984–2011. To further investigate Texas translocations as a risk factor for FIVpco infection, we calculated the relative risk [53] of a sample testing positive after 1995 when compared with previous years. The 1995 translocation was treated as the exposure and relative risk (RR) was calculated as follows:

$$R=\frac{a/(a+b)}{c/(c+d)},$$

where a is the number of exposed testing positive, b is the number of exposed testing negative, c is the number of controls (not exposed) testing positive and d is the number of controls testing negative. Controls were defined as those samples collected prior to 1995. The standard error of the log relative risk is

$$\text{s}.\mathrm{e}.\{\text{ln}\hspace{0.17em}\text{RR}\}=\hspace{0.17em}\sqrt{\left(\frac{1}{a}+\frac{1}{c}-\frac{1}{a+b}-\frac{1}{c+d}\right)},$$

and the 95% confidence interval is

$$\begin{array}{ccc}95\mathrm{\%}\hfill & =\hfill & \exp \hspace{0.17em}(\text{ln}\hspace{0.17em}(\text{RR})-1.96\times \mathrm{s}.\mathrm{e}\{\text{ln}\hspace{0.17em}(\text{RR})\})\hfill \\ \hspace{0.17em}\hfill & \hspace{0.17em}\hfill & \mathrm{to}\hspace{0.17em}\exp \hspace{0.17em}(\text{ln}\hspace{0.17em}(\text{RR}))+1.96\hspace{0.17em}\times \mathrm{s}.\mathrm{e}\{\text{ln}\hspace{0.17em}(\text{RR})\}).\hfill \end{array}$$

Data accessibility

Primer sequences are provided in electronic supplementary material, table S2. Raw sequence data have been deposited in the NCBI SRA database (SRA accession: PRJNA566447). Consensus sequences have been deposited in GenBank (accession numbers MN531083–MN531143). The protocol used to generate the data is available via GitHub at https://github.com/VandeWoude-Laboratory.

Authors' contributions

J.L.M., J.S.L., R.B.G. and S.Kr. carried out the molecular laboratory work and sequence alignments, and participated in data analysis and the design of the study and drafted the manuscript; S.Ke. carried out the molecular laboratory work, participated in data analysis and carried out sequence alignments; M.C., D.O., M.R. and R.M. collected field data and critically revised the manuscript; S.V. and K.R.C. conceived of the study, designed the study, coordinated the study and helped draft the manuscript. All authors gave final approval for publication and agree to be held accountable for the work performed therein.

Competing interests

We declare we have no competing interests

Funding

This study was funded by American College of Veterinary Pathologists/Society of Toxicologic Pathologists Coaltion for Veterinary Pathology Fellows (Linda Munson Fellowship for Wildlife Pathology Research) and National Science Foundation (grant nos. NSF-EID 1413925 and NSF-EID 723676).