Endogenous Feline Leukemia Virus (FeLV) siRNA Transcription May Interfere with Exogenous FeLV Infection

ABSTRACT

Endogenous retroviruses (ERVs) are increasingly recognized for biological impacts on host cell function and susceptibility to infectious agents, particularly in relation to interactions with exogenous retroviral progenitors (XRVs). ERVs can simultaneously promote and restrict XRV infections using mechanisms that are virus and host specific. The majority of endogenous-exogenous retroviral interactions have been evaluated in experimental mouse or chicken systems, which are limited in their ability to extend findings to naturally infected outbred animals. Feline leukemia virus (FeLV) has a relatively well-characterized endogenous retrovirus with a coexisting virulent exogenous counterpart and is endemic worldwide in domestic cats. We have previously documented an association between endogenous FeLV (enFeLV) long terminal repeat (LTR) copy number and abrogated exogenous FeLV in naturally infected cats and experimental infections in tissue culture. Analyses described here examine limited FeLV replication in experimentally infected peripheral blood mononuclear cells, which correlates with higher enFeLV transcripts in these cells compared to fibroblasts. We further examine NCBI Sequence Read Archive RNA transcripts to evaluate enFeLV transcripts and RNA interference (RNAi) precursors. We find that lymphoid-derived tissues, which are experimentally less permissive to exogenous FeLV infection, transcribe higher levels of enFeLV under basal conditions. Transcription of enFeLV-LTR segments is significantly greater than that of other enFeLV genes. We documented transcription of a 21-nucleotide (nt) microRNA (miRNA) just 3′ to the enFeLV 5′-LTR in the feline miRNAome of all data sets evaluated (n = 27). Our findings point to important biological functions of enFeLV transcription linked to solo LTRs distributed within the domestic cat genome, with potential impacts on domestic cat exogenous FeLV susceptibility and pathogenesis.

IMPORTANCE Endogenous retroviruses (ERVs) are increasingly implicated in host cellular processes and susceptibility to infectious agents, specifically regarding interactions with exogenous retroviral progenitors (XRVs). Exogenous feline leukemia virus (FeLV) and its endogenous counterpart (enFeLV) represent a well-characterized, naturally occurring XRV-ERV dyad. We have previously documented an abrogated FeLV infection in both naturally infected cats and experimental fibroblast infections that harbor higher enFeLV proviral loads. Using an in silico approach, we provide evidence of miRNA transcription that is produced in tissues that are most important for FeLV infection, replication, and transmission. Our findings point to important biological functions of enFeLV transcription linked to solo-LTRs distributed within the feline genome, with potential impacts on domestic cat exogenous FeLV susceptibility and pathogenesis. This body of work provides additional evidence of RNA interference (RNAi) as a mechanism of viral interference and is a demonstration of ERV exaptation by the host to defend against related XRVs.

INTRODUCTION

Endogenous retroviruses (ERV) are scattered throughout vertebrate genomes, representing 8% of genomic content, with documented impacts on normal biologic processes (1, 2). During early stages of endogenization, ERVs accumulate mutations that often render the newly endogenized virus defunct, protecting hosts from potentially deleterious genetic material (3). In addition to accumulating mutations, ERVs can act as retrotransposable elements inserting into novel genomic loci. Because ERVs are initiated by intact retroviruses with palindromic long terminal repeat (LTR) flanking sequences, they can be edited from the genome via homologous recombination and other mechanisms that are incompletely understood. Sometimes this process results in remnant genomic segments in the form of solo LTRs (3, 4). While usually unable to produce infectious virions, many ERVs are still capable of undergoing transcription and may produce functional viral proteins (5, 6). ERVs also function to enhance and/or promote transcription of proximal host genes. Following fixation in the genome, consequently, ERVs have been usurped by vertebrate hosts for essential biological processes, such as placentation, oncogenesis, immune modulation, and infectious disease progression (5, 7–9).

ERVs have also been exapted to participate in antiviral activities against exogenous homologues. Endogenous mouse mammary tumor virus (MMTV)-encoded superantigen negatively selects against self-reacting T-cells, limiting the ability for certain exogenous MMTV strains to infect those T-cells (10). Endogenous jaagsiekte sheep retrovirus (JSRV) produces Gag-like proteins that interfere with the regular trafficking mechanisms of exogenous JSRV, thereby reducing viral budding and maturation (11). Likewise, endogenous JSRV inhibits cell entry of JSRV through hyaluronidase-2 receptor interference by saturating and ultimately limiting the number of receptors that are displayed on the cell surface (12).

Feline leukemia virus (FeLV) endogenization has occurred in Felidae of the Felis genus and has been characterized in the domestic cat (Felis catus). Eight to 12 nearly full-length endogenous FeLV (enFeLV) genomes are present in each genome, with significantly greater numbers of solo LTR remnants (13–15). Full-length enFeLV genomes are 86% similar at the nucleotide level to horizontally transmitted exogenous FeLV (exFeLV) (16). Domestic cat exFeLV infects domestic cats across the globe with an incidence that ranges from 3 to 18% (17–20). exFeLV infection has a variety of clinical outcomes, with approximately 60% of infections resulting in aborted or truncated infection, and the remainder progressing to high levels of viremia resulting in hematologic dyscrasias, cancers, opportunistic infections, and death (21). In a study of a natural FeLV epizootic in a 65-hybrid domestic cat breeding colony, we demonstrated a correlation between higher enFeLV-LTR copy number and cats with regressive or abortive FeLV clinical outcomes. This finding was in contrast to cats with lower LTR copy number, which developed progressive infection and accumulated virulent enFeLV-exFeLV recombinants (14). We experimentally infected domestic cat fibroblasts with FeLV and likewise demonstrated that primary cells from cats with greater enFeLV-LTR copy number were more resistant to FeLV infection and viral replication (13). This relationship was not observed when we examined FeLV infection and replication related to enFeLV-env gene copy number, representing intact full enFeLV genomes, which were found at a considerably lower rate of incorporation than enFeLV-LTR (mean of 11 env copies/cell versus 57 LTR copies/cell) (13).

Cell culture experiments further illustrated a highly significant dose-dependent correlation between enFeLV-LTR copy number and viral antigen production, prompting the hypothesis that enFeLV may directly interfere with exogenous FeLV (13). One potential mechanism for direct interference is transcription of enFeLV small noncoding RNAs that regulate gene expression and viral reproduction by degrading target RNA. Small interfering RNA (siRNA), microRNA (miRNA), and piwi-interacting RNA (piRNA) result in RNA interference (RNAi) via different mechanisms. siRNA and miRNA activate the RNase DICER, which processes siRNA and miRNA and incorporates them into the RNA-induced silencing complex (RISC), which targets complementary mRNA for degradation (22). Once incorporated into an RISC complex, single-stranded RNA (ssRNA) can find its full (siRNA) or partial (miRNA) complementary mRNA strand and signal it for translational repression, mRNA degradation, or mRNA cleavage. A comprehensive review of siRNA and miRNA can be found in reference 23. piRNAs, conversely, are typically longer in length than siRNAs (21 to 35 nucleotides [nt]) and regulate gene expression and viral infection via PIWI-clade Argonautes versus AGO-clade proteins (24). piRNAs have been recently demonstrated to contribute to interruption of exogenous retroviral progenitor (XRV) processes in a newly endogenizing koala retrovirus (KoRV) (25). While RNAi is typically considered a potent antiviral mechanism used by plants and invertebrates, there has been evidence that RNAi is also used in mammalian systems to complement their normal antiviral activities governed by first-line interferon responses (26). siRNA has been used to inhibit influenza RNA transcription in chicken embryos and canine cells (27). siRNAs have also been demonstrated to be capable of silencing hepatitis A viral infections in nonhuman primate and human cells (28). Evidence is mounting that miRNA and siRNAs play a role in both promoting and inhibiting HIV replication (29). As a result, there has been growing interest in research on RNAi as a mechanism of antiviral restriction in humans and other mammals.

To determine mechanisms underlying ERV-XRV interactions in the FeLV system, we used in silico approaches and transcriptome sequencing (RNA-seq) analysis to investigate enFeLV transcripts in domestic cat tissues, evaluate transcript abundance and tissue tropism, and assess the nature of small RNAs that may function to suppress exFeLV infection. Further, we assessed susceptibility of domestic cat peripheral blood mononuclear cells (PBMCs) to examine exFeLV infection compared to fibroblasts. We conclude that enFeLV is transcriptionally active in healthy domestic cats and document significant basal levels of enFeLV siRNA transcription that is tissue specific. enFeLV mRNA and siRNA transcription levels were significantly higher in PBMCs than in other cells, and we noted significant exFeLV replication restriction in primary PBMCs compared to fibroblasts. Our findings provide evidence that enFeLV-LTRs are likely to exert control of FeLV replication via an RNA interference mechanism. We also identified ERV transcripts in domestic cats as well as bobcats (Lynx rufus) and Siberian tigers (Panthera tigris), indicating that transcription of this locus may be linked to an ancient pan-felid retroviral pol remnant with antiretroviral or other functions.

RESULTS

PBMCs are less permissive to FeLV infection than fibroblasts.

Domestic cat PBMCs derived from six cats and challenged with FeLV attained much lower proviral load levels in culture than domestic cat fibroblast infections (Mann-Whitney U test, P = 0.0022) (Fig. 1A). At day 5 postinoculation, the mean proviral load achieved was 7,346 proviral copies of FeLV per million PBMCs (range = 958 to 19,901 proviral copies/million), and only two samples exceeded optical density (OD) thresholds for positive antigen detection (Fig. 1B). In comparison, fibroblast infections yielded a mean of 262,263 (range = 19,376 to 1,851,261) proviral copy numbers/million cells on day 5 (Fig. 1A), and CrFK infections resulted in high levels of antigen production compared to PBMCs (Fig. 1B).

FIG 1.

Domestic cat PBMCs were more resistant to FeLV infection than fibroblasts. (A) At day 5, median proviral load was 6,360 copies per 10⁶ cells in PBMCs compared to 119,000 copies per 10⁶ cells in fibroblasts (Mann-Whitney U test; **, P < 0.01). (B) Domestic cat PBMCs supported low levels of virus replication measured by p27 antigen ELISA. Only two PBMC cultures had transient infections that peaked above the negative cutoff value, established at 3× standard error above average value for negative control replicates (red line). CrFK infections (dotted) were performed as a positive control.

SRA-accessed transcriptome data indicate tissue-specific enFeLV transcription dominated by LTR.

A total of 207 individual animal transcriptomic RNA-seq data sets were retrieved from the SRA data set inquiry. Fifty-six of these data sets were from healthy domestic cat tissues of various origins (e.g., embryonic, lymphoid, neural, etc.). Forty-two data sets were included following quality control analysis, representing two studies (Fig. 2) (99 Lives Cat Genome Sequencing Initiative, unpublished; 30). RNA-seq data sets originating from a jaguar (Panthera onca) and from a bobcat (Lynx rufus) were used as non-Felis controls (see Table S1 in the supplemental material).

FIG 2.

Bioinformatics pipeline used in this analysis. Two hundred and seven RNA-seq data sets were identified during initial database search. Filtering for healthy cats with defined tissue of origin resulted in 56 data sets for our transcriptome analysis and 27 data sets for our miRNAome analysis. Thirty-three of 56 transcriptome data sets satisfied quality controls allowing final analysis. Two non-Felis spp. transcriptome data sets were included as non-domestic cat comparisons.

enFeLV transcript levels were approximately 100 reads normalized per million reads (RPM) for most tissues. Outliers included the following: (i) embryonic tissues (i.e., head, body, whole), (ii) lymphoid tissues, and (iii) a single salivary gland sample. All of these tissues had consistently greater enFeLV transcription levels than neural, skin, reproductive, urinary, lung, digestive, circulatory, and liver tissues (Fig. 3A). Lymphoid (n = 4) and salivary gland (n = 1) tissues had the greatest enFeLV transcription, averaging approximately 10-fold greater transcription than other tissues.

FIG 3.

enFeLV transcription is tissue and gene specific. (A) enFeLV reads are transcribed at greatest levels in lymphoid and salivary gland tissues. Reads per million (RPM) is a measure of comparison to all other available transcripts in the transcriptome data set. (B) enFeLV-LTR was transcribed at greater levels than other enFeLV genes. Multiple comparisons following ANOVA demonstrated an average 10-fold increase in enFeLV-LTR compared to that in gag (P = 0.0079), pol (P = 0.0073), and env (P = 0.0111). FPKM (fragments per kilobase million) is a measure of total RNA normalized by gene fragment length. Data shown here represent SRA accession numbers SRX211594 to SRX211596, SRX211644 to SRX211646, SRX211688 to SRX211690, SRX1610301 to SRX1610326, and SRX1625943 to SRX1625949. Red data points represent negative control data sets (P. tigris altaica, SRX317246; L. rufus, SRR6384483).

Following normalization against gene fragment length, enFeLV gene segments were found to have differential expression profiles (Fig. 3B). LTR, gag, pol, and env transcripts represented 0.439, 0.0302, 0.0265, and 0.0453 fragments per kilobase per million (FPKM) of total transcripts, respectively. Relative expression of size-normalized gene segments as FPKM by tissue supported trends identified for full enFeLV: lymphoid tissues and salivary gland accounted for the greatest level of transcription (Fig. 4), and LTR transcription was approximately 10-fold greater than that of other enFeLV genes (Fig. 4A).

FIG 4.

enFeLV genome elements are transcribed variably between all tissue types. (A) enFeLV-LTR was transcribed 10 times more than gag (B), pol (C), and env (D). Lymphoid tissues and the salivary gland (boxed) harbored the greatest amount of enFeLV transcripts across all genome elements. Gray dotted line at 0.1 FPKM is provided for ease of interpreting differences in the y axis scales. Data shown here represent SRA accession numbers SRX211594 to SRX211596, SRX211644 to SRX211646, SRX211688 to SRX211690, SRX1610301 to SRX1610326, and SRX1625943 to SRX1625949. Red data points represent non-Felis control data sets (P. tigris altaica, SRX317246; L. rufus, SRR6384483).

PBMC enFeLV transcript levels are 2 log higher than fibroblasts.

Total rRNA-depleted RNA was sequenced from PBMCs from six cats and four fibroblast primary cell cultures (as represented in Fig. 1), averaging 50 million reads per sample (Fig. 5; see also Table S2 in the supplemental material). enFeLV transcript level was on average 427 (range = 227 to 899) RPM for PBMCs and on average 4 (range = 3 to 5) RPM for fibroblasts. LTR, gag, pol, and env transcripts represented 0.0180, 0.00343, 0.00274, and 0.00456 FPKM of total transcripts, respectively, for PBMCs and 2.43E−4, 1.31E−5, 4.77E−5, and 4.59E−5 FPKM of total transcripts, respectively, for fibroblasts. All differences between PBMCs and fibroblasts were statistically significant by Mann-Whitney U tests (Fig. 5; **, P < 0.01). Unlike in vivo-derived tissues, LTR transcription was not significantly increased compared to other FeLV gene segments.

FIG 5.

Cultured PBMC enFeLV transcription levels were 2 logs higher than those of cultured fibroblasts across all genes. (A) Average transcription across the full genome was 427 RPM versus 4 RPM. Differences were noted across LTR (259 FPKM versus 3.5 FPKM) (B), gag (127 FPKM versus 0.5 FPKM) (C), pol (236 FPKM versus 4.5 FPKM) (D), and env (221 FPKM versus 2.3 FPKM) (E). Accession number assignment is in process and will be provided.

enFeLV-like RNAs detected in multiple species represent a distinct ERV of felids.

RNA data sets from bobcat and Siberian tiger were analyzed as negative controls, as these species do not have enFeLV present in their genomes (31). We identified two regions that mapped to the enFeLV genome (Fig. 6). One region was found in both bobcat and tiger with short RNA matches driven by a 29-nt polyadenine stretch in the enFeLV 5′-LTR. The second region varied between bobcat and tiger, but both transcripts mapped to an 87-nt region in enFeLV pol in the endonuclease/integrase segment of the genome. The tiger transcript was 187 nt and contained 43 single nucleotide polymorphisms (SNPs) relative to the 87-nt corresponding region of enFeLV pol. The bobcat sequence was 179 nt long and contained 39 SNPs relative to enFeLV pol. Interestingly, an identical 87-nt region with 100% identity was noted in uncharacterized Panthera pardus (Leopard) LOC10927796 mRNA (GenBank accession number XM_019467717), which was previously reported to represent an ERV (32). NCBI’s BLAST tblastn function revealed a shared polymerase gene identity from this region of enFeLV pol to feline endogenous retrovirus gamma4-A1 (GenBank accession number LC176795). Nucleotide similarity between a 90-nt region of LC176795 and enFeLV was 60% with 36 SNPs. Pairwise identity between full-length enFeLV (GenBank accession number AY364319) and feline endogenous retrovirus LC176795 was 49%.

FIG 6.

Genome elements with homology to small segments of enFeLV were identified in bobcat (SRA accession number SRR6384483) and Siberian tiger (SRA accession number SRX317246) transcriptomes. RNA mapped to a poly(A) region of the LTR (nt 275 to 304) and a variable region in pol (nt 5421 to 5608). The poly(A) region only skewed negative control data sets and did not have an impact on Felis catus transcriptome analysis following visual verification. The pol mapped reads represented a conserved region that appears to map to an uncharacterized feline endogenous retrovirus that may be distantly related to enFeLV and is found in both the bobcat and Siberian tiger. The sequences highlighted in green were responsible for driving alignment to enFeLV pol. Nucleotides highlighted in red represent SNPs.

SRA-accessed miRNAome data identifies abundant enFeLV-derived siRNA transcripts that are both positive and negative sense.

Twenty-seven data sets were used to characterize the feline miRNAome from individual animal transcriptomic RNA-seq data sets retrieved from the SRA (see Table S3 in the supplemental material). These consisted of RNA fragments of <30 nt and corresponding to RNAs considered to function as RNA silencing transcripts (33). enFeLV miRNA sequences accounted for 0.0163% of all miRNA in the annotated SRA pool. Approximately 75% of the miRNA sequences mapping to enFeLV originated from the LTR (75.1% ± 21.0%), though gag, pol, and env represented more than 10% of the enFeLV-mapped reads in three to six of the 27 data sets (Fig. 7 and 8A).

FIG 7.

miRNA could be detected for all gene regions in enFeLV but mapped most frequently to enFeLV-LTR (ANOVA; P < 0.001). The contribution of miRNA attributed to gag, pol, or env rarely exceeded 0.01 fragments per kilobase million (FPKM). Relative expression was not different among the three genes with background expression proportional to the size of the gene. Increased LTR expression may be driven by both the increased number of LTRs that exist within the genome relative to the other genes and the increased activity of the LTR. Data shown here represent SRA accession numbers SRR4243109 to SRR4243135.

FIG 8.

enFeLV miRNA maps to four regions of the enFeLV genome. (A) Twenty-seven unique miRNA data sets were evaluated for reads mapping to LTR, gag, pol, and env genome segments. Segments that represented greater than 10% of total mapped reads occurred in 27, 6, 3, and 6 data sets, respectively. Locations of these transcripts are indicated below the genome map, and sequences for these transcripts in LTR (identified in all 27 data sets), gag (identified in 6 of 27 data sets), and env (identified in 6 of 27 data sets) are indicated. Multiple heterogenous miRNA were identified in 3 of 27 data sets, perhaps indicating contributions from various closely related endogenous retroviruses. (B to J) Positive and negative miRNAs identified in three representative data sets illustrate individual cat variation in enFeLV miRNA distribution and polarity. SRA accession number SRR4243132 is represented in panels B, E, and H; accession number SRR4243126 is represented in panels C, F, and I; accession number SRR4243130 is represented in panels D, G, and J. Top row of panels B, C, and D indicates positive-strand transcripts scaled across the enFeLV genome. Schematic below the top row indicates enFeLV genome map correlating to map shown in panel A. Difference in y axis illustrates variation in scaled read depth from one data set to another. Panel D illustrates a high-abundance transcript that maps to enFeLV env in one data set. Middle row of panels E, F, and G indicates scaled read depth of negative-strand miRNA transcripts. The overwhelming abundance of a 21-nt LTR transcript (2 × 10⁵ to 5 × 10⁵ scaled reads/data set) obscures lower abundance negative-strand transcripts. Removal of this transcript from panels H, I, and J allows visualization of lower abundance negative-strand miRNAs (5 to 35,000 scaled reads/data set) that map to enFeLV.

Characterization of abundant miRNAs.

A 21-nt LTR negative-sense miRNA at nucleotide 557 was detected in all 27 individuals and was by far the most abundant enFeLV miRNA identified. This transcript is located just 3′ to the 5′-LTR, 74 nucleotides downstream of the transcription start site (Fig. 8A). The sequence for this LTR miRNA, (5′- ATCCCGGACGAGCCCCCACGC-3′), is identical to enFeLV in the same location, with the exception of the 3 flanking nucleotides and represents a purely negative-stranded population (Fig. 8B, E, and H; see also Fig. S1 in the supplemental material). The 3 mismatched nucleotides have the lowest sequencing quality score, indicating that these are potentially miscalled bases. A 12-nt miRNA segment (5′-TATCTAGCTTA-3′) was identified in env at nucleotide 7045 in 6 of 22 individuals (Fig. 8A). This sequence is positive-stranded and correlates with the gp70 surface protein-like portion of enFeLV Env. Six individuals also had a 14-nt miRNA (5′-CTCCGCGGCGCTGC-3′) at nucleotide 1963 within the virion core peptide p27 portion of enFeLV gag (Fig. 8A). Three individuals also had miRNA segments near the endonuclease/integrase region of pol (nucleotide region 4800 to 5200) (Fig. 8A).

sRNApipe miRNA strand polarity analysis was used to analyze strand specificity for miRNA transcription for both abundant and rare miRNA transcripts that exceeded 18 nt in length. Many low-copy-number miRNA transcripts were found with homology to enFeLV in addition to the primary transcripts noted above. These were generally dispersed across the enFeLV genome. Three representative data sets (SRA accession numbers SRR4243126, SRR4243130, and SRR4243132) that illustrate unique miRNA transcripts are depicted in Fig. 8B to J, and remaining maps are included in Fig. S1. As noted above, the abundant 21-nt 5′-LTR transcript found in all individuals was a negative-sense transcript, while reads mapping to other regions of the genome were overwhelmingly positive stranded (Fig. 8B to D; see also Fig. S1). A unique 20-nt read was exclusively positive-sense mapped to enFeLV env with one deletion and two SNPs in the intervening sequence. miRNA mapping to pol is nonspecific to a specific locus and has both positive-sense and negative-sense strands mapping to the area.

DISCUSSION

The mechanisms underlying the distinct outcomes of domestic cat FeLV infection remain elusive. The majority of FeLV-infected cats overcome infection, and vaccination can successfully protect against disease, suggesting that an adaptive immune response can be protective (34). However, a significant proportion of animals exposed to FeLV are unable to eliminate the infection and ultimately succumb to hematologic dyscrasias, lymphoid tumors, or opportunistic infections (35). FeLV replicates to extraordinarily high titers during progressive infection, and in more than 50% of progressive infections, ERV-XRV recombination occurs, resulting in a switch in receptor usage and more progressive disease (14, 36). Novel observations reported here are highly suggestive that domestic cat enFeLV functions in part to restrict exFeLV infection and provides an explanation for divergent outcomes of FeLV disease.

In silico and RNA-seq analyses demonstrate that basal enFeLV transcripts are abundant in tissues from healthy cats, enFeLV is transcribed in a tissue-specific manner, and transcript level varies by gene segment (Fig. 3B and 4). LTR transcription is approximately 10 times higher than pol, gag, and env (Fig. 4), which may be reflective of the greater number of LTR elements per genome than other segments (13). Lymphoid tissue transcription is 1 to 2 logs greater than transcription in other tissues, and one salivary gland transcriptome available for analysis had higher expression than lymphoid (Fig. 4). miRNA transcripts mapping to enFeLV were also detected, including a 21-nt negative-stranded oligoribonucleotide noted in 27 of 27 miRNA transcriptomes in the SRA database (Fig. 8; see also Fig. S1 in the supplemental material). FeLV is lymphotropic and has replication phases in salivary tissue that result in viral transmission following social or antagonistic contact (37), though here we show that infection in primary PBMCs is highly restricted (Fig. 1). Consequently, enhanced enFeLV transcription and basal miRNA production in these tissues may represent a specific host restriction mechanism.

Conversely, higher basal expression of enFeLV in lymphoid tissues may result in a greater potential for ERV-XRV recombination to occur in these cells following copackaging of endogenous and exogenous transcripts (38). Recombination between enFeLV and exFeLV occurs in the 3′ half of the genome in approximately 50% of progressive infections and is associated with worse clinical outcomes, presumably relating to changes in viral receptor and cell tropism from THTR-1 to PIT-1 (16).

When we examined similar metrics in cultured PBMCs and fibroblasts, we found that similar trends upheld (Fig. 5). There was a 2-log difference in enFeLV transcription between PBMCs of lymphoid origin and fibroblasts derived from the skin, providing more evidence of an impact of enFeLV transcription in FeLV biologically important tissues. Fewer transcripts were identified in cultured cells than in situ transcription reported on GenBank, and conditions of culture (i.e., PBMC stimulation, fibroblast growth arrest at time of harvest for RNA analysis) are potential factors that influenced our findings. Future studies evaluating enFeLV transcript levels in response to stimulation, viral infection, and other conditions will assist in understanding mechanisms and consequences of enFeLV transcription.

We identified a sizable number of short noncoding miRNA transcripts that map to enFeLV sequences. Unlike enFeLV transcription, enFeLV miRNA was present only at specific loci (Fig. 8; see Fig. S1), suggesting a specific miRNA function. Given known function of miRNA to degrade complementary mRNA via RISC degradation, it seems feasible that these loci represent evolutionarily selected transcript sites that provide host defense against virulent FeLV disease. All 27 cats evaluated were positive for a 21-nt negative-sense miRNA transcript that mapped 3 nucleotides downstream from the 5′-LTR U3 region. The length of 21 nt is indicative of a transcript that functions as an siRNA via an RNAi mechanism, as the DICER complex requires a very specific length of RNA (39). Mapping this 21-nt sequence to exogenous FeLV demonstrates that 2 or 3 SNPs occur at the 5′ end, which may impact RNAi functionality. Additional evaluation of this specific miRNA and its role in FeLV infection is warranted to assess the capacity of this miRNA to interfere with FeLV infection.

One potential explanation for regressive versus progressive FeLV outcomes based upon findings presented here is as follows:

1. High levels of lymphoid enFeLV-LTR and miRNA transcription preempt FeLV infection of PBMCs via RNAi-like mechanisms. If miRNA inhibition persists, regression may occur, concurrent with adaptive immune responses that overcome infection.

2. In individuals with lower basal LTR transcription levels (correlating with lower enFeLV-LTR proviral copy number), siRNA restriction mechanisms may fail, resulting in primary infection of lymphoid cells and progressive infection.

3. In other individuals, infection of nonlymphoid tissues with low basal levels of LTR transcription followed by recombination with enFeLV-env transcripts may result in XRV-ERV recombinants with enhanced tropism for PBMCs. This could result in secondary PBMC infection that potentially overwhelms RNA restriction (again correlating with lower enFeLV-LTR proviral copy number), resulting in progressive infection.

It is likely that these mechanisms operate in conjunction with other more well-understood innate and adaptive antiviral mechanisms to drive FeLV infection outcomes in natural systems.

Subsets of individual animals had predominately positive-sense miRNA that aligned to sites in other FeLV gag, pol, and env genes (Fig. 8; Fig. S1). While positive-sense RNA can be directly used as templates for transcription, positive-sense small RNA is unlikely to do so. Mapped miRNA reads to gag and env were highly specific to single loci. The size of these specific miRNA is shorter than the RNA that recognized RNAi mechanisms require (gag – 14 nt; env – 12 nt) and are thus unlikely to participate in RNA silencing using mechanisms that are currently defined. Conversely, the negative-sense 21-nt LTR RNA is complementary to the FeLV sequence and, therefore, may serve as the template for FeLV RNA genomes being produced during infection. Furthermore, the scale of the read coverage at the LTR locus compared to the other small RNAs may indicate its biological importance and give us insight into its relative activity. The diffuse miRNA mapping pattern in pol may indicate that this transcript may act to silence a broad range of retroviruses with conserved sequences in this region.

It is possible that gag, pol, and env transcripts measured are translated into proteins that contribute to normal biology and physiology as they do in other ERV systems (40). In 1994, McDougall et al. reported that truncated enFeLV Env may participate in direct receptor interference inhibiting exogenous FeLV infection (41). However, retroviral LTRs are not protein-encoding regions; rather LTRs harbor both promoter and enhancer regions that induce transcription and read-through fusion transcripts that may be processed into functional proteins or other units (42). As such, enFeLV-LTR may potentially drive cis- or trans-activation of host transcription machinery to encode for antiviral proteins, a process documented in MuLV that relates to ERV-XRV interference (43). This phenomenon has been documented for specific host genes that propagate and inhibit disease processes depending on integration site, including antiviral proteins such as APOBEC3C (43, 44). If enFeLV-LTRs integrate near antiviral restriction factors, it could ostensibly prime cells to be more resistant to viral infection. Solo-LTRs formed following ERV retrotransposition may result in fixation of these loci in sites where transcription provides a survival advantage through positive genetic selection. Examination of LTR integration sites in future studies may prove useful in determining what host genes may be impacted by increased transcription or expression.

We noted the interesting phenomenon of relatively higher enFeLV transcription in embryonic tissues (Fig. 3 and 4). Embryonic ERV transcription has been documented to participate in many normal biological functions. ERV transcription has been inferred as a possible protection mechanism against embryonic viral infections by stimulating innate immunity mechanisms (45). ERV expression has also been shown to be important in placentation through syncytins (a syncytium-forming protein responsible for trophoblast invasion of the uterine wall) (46). During embryonic development of the thymus, host genes (and, by consequence, ERVs) are expressed and recognized by the autoimmune regulator (AIRE) so as not to elicit an autoimmune response against “self” proteins (47–49). This may allow XRV to evade specific adaptive immune responses. Ultimately, discovery of this higher transcription may signal cooption of enFeLV proteins in biological processes.

Transcripts from tiger and bobcat that aligned to the enFeLV genome were also identified. One mRNA location in the LTR mapped to a 30-nt poly(A) stretch in the 5′-LTR (Fig. 6). Felis catus samples did not contain transcripts mapping to this region. A second 90-nt mRNA mapped to nucleotide 5392 to 5488 of pol and shares identity to both enFeLV and another endogenous gammaretroviral element described in domestic cats as feline endogenous retrovirus gamma4-A1 (50). This 90-nt pol endonuclease/integrase region transcript likely represents an ancient conserved motif of retroviral pol from an ERV remnant that arose in Felidae prior to divergence between large and small felids. This transcript may represent a pol remnant from an ancient endogenized retrovirus with homology to FeLV. The associated LTR segment from this ERV may drive transcription via promoter or enhancer function (42). The fact that no other related enFeLV segments were recovered from tiger and bobcat data sets suggests that this fragment represents a highly conserved region of pol that is transcriptionally active. Future studies to identify additional segments of this ERV may reveal a more ancient ERV than FeLV that spans most felid species.

SRA data sets were highly valuable for this analysis, but curation of data sets was noted to be highly variable and low numbers of samples hampered some conclusions from our analysis. Lack of description and quality control required us to discard more than 80% of the 207 data sets initially identified. Further, our inquiry has demonstrated that data from nontraditional animal models that provide highly informative comparative data are sorely lacking from genomic databases. For example, the inquiry “HIV RNA-seq” yielded 1,491 results as of June 2020, whereas “FIV RNA-seq” yielded only 8 responses. The capacity of comparative genomics studies would be greatly expanded by encouraging analyses that will complement the human data sets available.

FeLV represents a naturally occurring retroviral infection of an outbred species with very well-documented clinical and virological outcomes. Here, we present compelling evidence that enFeLV-LTR transcripts in feline lymphoid tissue abrogate exogenous FeLV infections. Additional experiments documenting mechanistic aspects of this system in relation to observed natural disease outcomes represent a significant opportunity to understand the function of ERV in mitigating viral infections and to understand mammalian RNA interference mechanisms, impacts of ERV on host evolution, and LTR enhancer and promoter functions that regulate host innate and adaptive immune responses.

MATERIALS AND METHODS

Peripheral blood mononuclear cell infection.

Blood was drawn from specific-pathogen-free domestic cats housed at Colorado State University (IACUC protocol number 16-6390A). Peripheral blood mononuclear cells were isolated from fresh blood by Ficoll-gradient centrifugation. PBMCs were cultured in 20% fetal bovine serum (FBS)-supplemented RPMI medium supplemented with 100 ng/ml interleukin-2 (IL-2) (Sigma, USA) and 50 ng/ml concanavalin A (Sigma, USA). Primary PBMC cultures were expanded for 2 passages before being directly infected.

PBMCs were plated at 1 × 10⁶ cells/ml and infected with a multiplicity of infection (MOI) of 0.01 FeLV-61E as was previously described (13). Briefly, supernatant was sampled at days 0, 1, 3, 5, and 7 and tested for viral antigen using p27 enzyme-linked immunosorbent assay (ELISA) as previously described. PBMCs were harvested at day 5 to enumerate cell viability and proviral copy number. PBMC proviral and antigen load were compared to fibroblast infection proviral and antigen load conducted simultaneously and reported previously (13). Briefly, primary fibroblasts and control Crandell-Rees feline kidney control cells (CrFK) were plated at a density of 50,000 cells per 2 cm² in a 24-well plate and infected with an MOI of 0.01 FeLV-61E. Supernatant was sampled at days 0, 1, 3, 5, 7, and 10, and cells were harvested at days 5 and 10 to enumerate cell viability and proviral copy number.

Endogenous FeLV transcriptomic analysis.

Domestic cat transcriptome and miRNAome data sets were acquired through the search function in the NCBI Sequence Read Archive (SRA) using the search key words “felis” and “rna-seq.” Data sets were included in the study if they were derived from healthy cats, identified the tissue of origin, and represented transcriptome (excluding miRNAome) data sets (see Table S1 in the supplemental material). Two additional non-Felis spp. felid transcriptome data sets (Lynx rufus, Panthera tigris; SRA accession numbers SRR924676 and SRR6384483/SRR6384484) were included as negative controls that would not be expected to harbor enFeLV transcripts (Table S1). Tissues analyzed included embryonic (fetus, embryo body, embryo head), neural (cerebellum, parietal lobe, occipital lobe, temporal lobe, hippocampus, spinal cord, retina), skin (skin, ear tip, ear cartilage), lymphoid (spleen, lymph node, bone marrow, thymus), and other organ (muscle, liver, uterus, kidney, testes, pancreas, heart, salivary gland) tissues. All data processing and analysis was completed using the Colorado State University College of Veterinary Medicine and Biomedical Science server. Transcriptome data sets were analyzed using a custom bioinformatics pipeline (Fig. 2). Reads were trimmed for appropriate adapters and by quality (q = 20) using cutadapt (version 1.18). The first 600 nt and last 600 nt of the full-length enFeLV were discarded prior to creating the index due to the potential for transcripts to map to other host genomic elements that surrounded the enFeLV integration site. Indices were first generated for full-length domestic cat enFeLV (accession number AY364319) and individual enFeLV gene regions, including LTR, gag, pol, and env separately using the Bowtie2-build function (version 2.3.4.1). Transcriptome sequences were mapped to indices using “--sensitive” settings in local mode in Bowtie2 to allow for heterogeneity among different enFeLV genotypes. Alignments were visually inspected by importing mapped .sam files into Geneious 11.1.2. Exogenous FeLV was ruled out as the source of mapped reads by looking for exogenous FeLV-specific DNA segments. Any transcripts mapped to the negative controls were manually inspected and their identities confirmed using NCBI’s BLAST tblastn function. Transcriptome reads mapping to full-length FeLV were reported as reads per million reads (RPM). While the reads were mapped as paired-end reads, reported RPM was calculated as unpaired reads. Individual genome elements were reported as fragments per kilobase million (FPKM) to normalize for size of the respective gene region (full-length enFeLV, 8,448 nt; LTR, 592 nt; gag, 1,512 nt; pol, 3,630 nt; env, 2,002 nt). Trimmed means of M (TMM)-based normalization was not attempted because enFeLV LTR, gag, pol, and env copy numbers vary per individual and were unknown for these samples. Percent transcription for full-length FeLV was analyzed by analysis of variance (ANOVA) in Prism (version 7.0). Custom scripts are available at https://github.com/VandeWoude-Laboratory/Florida-panther-virus.

Additional transcriptome data sets were generated specifically for cultured fibroblasts and PBMCs (described above) following culture. RNA was extracted using a RNeasy kit (Qiagen, USA) per manufacturer’s direction. Ribo-depleted total RNA library preparation was performed by the Genomic Shared Resource core at the University of Colorado Anschutz Medical Campus and run on an Illumina NovaSeq 6000. High-throughput sequencing data were then put through our custom bioinformatics pipeline described above and deposited in NCBI’s Sequence Read Archive.

Feline miRNAome analysis.

miRNAome data sets were accessed as described above and by querying “felis” and “rna-seq” that also identified miRNA-seq in the library preparation strategy (Fig. 2). Tissues analyzed included neural (cerebellum, cerebral cortex, brain stem), skin (skin, lip, tongue), lymphoid (spleen, lymph node), and various organ (pancreas, kidney, liver, lung, testis, ovary) tissues. Using the full-length enFeLV index described above constructed with Bowtie2-build, miRNAome reads were mapped on local mode with a minimum threshold score set at 20 in Bowtie2 to account for the miRNA intrinsically short length. miRNA reads mapping to full-length enFeLV were reported as percent reads mapped to each genome element compared to total mapped reads to the full-length FeLV genome. Percent miRNA mapped to enFeLV was analyzed by ANOVA in Prism (version 7.0). Custom scripts are available at https://github.com/VandeWoude-Laboratory/Florida-panther-virus.

miRNAome data sets were visualized in strand-specific orientation using sRNAPipe on the Galaxy platform (51). In addition to default settings, three genome mismatches were allowed to accommodate identified SNPs at the 3′ end of a 21-nt LTR siRNA sequence. miRNA was mapped against the enFeLV genome (GenBank accession number AY364319), with LTR signifying the only transmissible element input file and protein-encoding genes as the transcripts and mRNA input files. All 27 data sets were analyzed to determine positive- and negative-strand miRNAs of a minimum of 18 nt in length. A second separate negative-sense analysis was conducted that excluded a highly transcribed 21-nt LTR siRNA in all data sets so that less-abundant transcripts could be readily visualized.

Data availability.

High-throughput sequencing data generated in our study are publicly available in NCBI’s Sequence Read Archive (https://www.ncbi.nlm.nih.gov/sra) under BioProject accession number PRJNA750528. High-throughput sequencing data that were put through our custom bioinformatics pipeline described above were deposited in NCBI’s Sequence Read Archive under accession numbers SAMN20462815 to SAMN20462824.