Koala Samples and DNA Extraction

Archival and modern samples are described in Table 1. All archival samples were extracted in a dedicated ancient DNA laboratory in the Department of Wildlife Diseases of the Leibniz Institute for Zoo and Wildlife Research under plexiglass UV hoods dedicated to DNA extraction. The ancient DNA laboratory was never used for molecular or genetic work on modern samples, and followed procedures designed to minimize the possibility of contamination, such as wearing protective clothing during extractions to avoid contamination from the researchers. Each extraction involved approximately 250 mg of dried skin, and used the Geneclean Ancient DNA extraction kit from MP Biomedicals, USA, following the manufacturer’s protocol. Mock extractions were performed for each set of museum specimens as controls for potential contamination during the extraction process. Each DNA extract was further purified using Qiaquick spin columns (Qiagen) as described previously [16]. DNA extraction from a blood sample of modern koala Pci-SN265 (zoo koalas in North America and Europe are included in the North American regional studbook, and are here designated by studbook number, “SN”) was performed in a separate laboratory in a different floor of the Leibniz Institute for Zoo and Wildlife Research. This extraction was performed using the Qiagen DNeasy Blood & Tissue Kit following the manufacturer’s protocol. The extracted DNA was then fragmented using a Covaris-S220 to generate 150 bp fragments.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Koala sample information.

https://doi.org/10.1371/journal.pone.0095633.t001

Blood samples of San Diego Zoo koalas were collected during routine physical exams and genomic DNA was isolated from buffy coat using the Qiagen DNeasy Blood & Tissue Kit following the manufacturer’s protocol. DNA from blood samples of wild koalas had been extracted using a phenol-chloroform method. These samples were used to generate baits.

Ethics Statement

All experiments involving koala tissues were approved by the Internal Ethics Committee of the Leibniz Institute for Zoo and Wildlife Research, approval number 01-01-2013. Work involving other modern koala samples was conducted at the University of Illinois at Urbana-Champaign (UIUC), under IACUC approval number 12040.

Polymerase Chain Reaction

All museum specimen were initially screened for a KoRV pol fragment by PCR (Table 1) performed in a volume of 34 µl using 5.5 µl of extract, 10 nm of primers, 0.5 U Platinum HiFi supermix (Invitrogen), 1µl of bovine serum albumin (Fermentas), and 1 µl of primers P1aF ‘5-TTGGAGGAGGAATACCGATTACAC-3′ with P1aR ‘5-GCCAGTCCCATACCTGCCTT-3′ [8]. Cycling conditions were: 94°C for 4 min; 60 cycles at 94°C for 30 s, 55°C for 30 s, 72°C for 30 s; and 72°C for 10 min, with the samples then held at 4°C [17]. The high cycle number (60) PCR was only used for screening museum koala samples for the presence of KoRV and not for polymorphism analyses. The modern sample was screened by PCR amplification performed in a volume of 34 µl using 1 µl (26.7 ng/µl) of extract, 10 nM of each primer, 0.5 U of Platinum HiFi supermix (Invitrogen). Cycling condition were: 94°C for 4 min; 35 cycles at 94°C for 30 s, 55°C for 30 s, 72°C for 30 s; and 72°C for 5 min, with the samples then held at 4°C. PCR products were visualized on a 3% gel. All gels used GelRed nucleic acid gel stain by Biotium. PCR products were purified using the NucleoSpin Gel and PCR Clean up kit (Macharey-Nagel). PCR products were commercially sequenced by the Sanger method using the forward and reverse PCR primers (StarSeq, Germany). Primers used in this study are listed in Table S1. The Sanger sequences were not included in the hybridization capture alignments but were only used to establish the presence of KoRV in museum and modern samples.

Illumina Library Preparation

Aliquots from each DNA extract were used in generating Illumina libraries. Archival extract libraries were generated in the ancient DNA facility in a library-dedicated plexiglass PCR UV hood, while the modern koala library was generated in a modern DNA laboratory in a different part of the Institute. Libraries were generated as described in Mayer et al. [18]. Each library contained a unique index adapter to allow for subsequent discrimination among samples after the sequencing of pooled libraries. A negative control extraction library was also prepared and indexed separately to monitor any contamination introduced during the experiment. Indexes were added by PCR using Amplitaq Gold DNA polymerase (Applied Biosystems [ABI]) in 100 µl reactions. Cycling condition were: 94°C for 5 min; 10 cycles at 94°C for 30 s, 55°C for 30 s, 72°C for 30 s; and 72°C for 5 min; the samples were then held at 4°C. After indexing, the samples would effectively be at little or no risk from cross contamination either from the other libraries or from laboratory DNA. Quantitative PCR (qPCR) was performed after index PCR with a standard that was developed using 100 bp PCR product with Illumina primer binding sites ligated at the 5′ and 3′ ends as described in Mayer et al. [18]. The qPCR standard curve was obtained using a series dilution of the standard. The assay was performed in a Stratagene MxPro 3000p qPCR system using Brilliant III Ultra-Fast SyBr Green qPCR master mix (Agilent) with Illumina bridge primers P5 and P7 [18] to determine the number of molecules in each sample. Additional amplification was followed using Herculase II DNA polymerase (Agilent) with P5 and P7 Illumina library outer primers with the same cycling conditions. DNA products were purified using Minelute columns (Qiagen) after each amplification step. Final quantification was performed on an Agilent 2200 tape station D1K tape.

Primer Design and Preparation of Baits

PCR products used as “bait” for capturing sequences from the Illumina libraries were generated at the University of Illinois to limit the amount of koala and KoRV amplicons present in the laboratories in Berlin. DNA of one northern koala, Pci-SN404 (see above) and three southern koalas (PCI-157 and PCI-142 from the Stony Rises and PCI-106 from the Brisbane Ranges of southern Australia) were used in preparing the bait. Primers were newly designed to cover the complete KoRV genome outside the envelope region. For the envelope region, previously designed primers were used [6] but with primer combinations that would yield amplicon sizes of approximately 500 bp. For the other KoRV regions, novel primers based on the published KoRV sequence (GenBank: AF151794) [7] were designed using Primer3 (http://fokker.wi.mit.edu/primer3/input.htm) [19] to yield amplicons of approximately 500 bp. The KoRV genome was amplified in thirty-eight 500 bp overlapping products using the primers shown in Table S1. The PCR mix consisted of 1 X PCR Buffer II (ABI), 1.5 mM MgCl₂ (ABI), 0.4 µM of final concentration of each primer, 200 µM of each dNTP (ABI), with 0.04 units/µl final concentration of AmpliTaq Gold DNA Polymerase (ABI). The PCR algorithm consisted of an initial 95°C for 10 min; with cycles of 15 sec at 95°C; followed by 30 sec at 60°C, 58°C, 56°C, 54°C, 52°C (2 cycles at each temperature) or 50°C (last 30 cycles); and 1 min at 72°C; with a final extension of 7 min at 72°C. An aliquot of each PCR product was visualized on a 1% agarose gel with ethidium bromide. PCR products were enzyme-purified [20] and Sanger-sequenced to verify that the target region had been amplified. The PCR products were purified using Qiaquick columns (Qiagen) and then quantified using a NanoDrop ND-1000 (Thermo-Scientific). KoRV amplicons were then blunt-ended, ligated to a biotin adapter, and immobilized on streptavidin magnetic beads in equimolar amounts of 1.3 µg as described previously [14].

Hybridization Capture

Mixtures of blocking agent, blocking oligos, and indexed koala libraries were heated to 95°C to separate the DNA strands [14]. One aliquot from each index library was mixed with streptavidin beads bound with biotinylated KoRV PCR products. Samples were incubated for 48 hours at 65°C under rotation in a Labnet mini incubator. After 48 hours the beads were washed and the hybridized libraries eluted by heating. The DNA concentration was measured by quantitative PCR (qPCR), and the eluted libraries were further amplified accordingly using P5 and P7 Illumina outer primers. The products were then pooled at equimolar concentrations for paired-end sequencing on an Illumina MiSeq platform at the National High-Throughput DNA Sequencing Center, University of Copenhagen.

Sequence Assembly, Identification of Polymorphisms and Integration Site Analysis

Sequences were separated based on their index sequence at the National High-Throughput DNA Sequencing Center, University of Copenhagen, Denmark. The programs cutadapt v1.2 and trimmomatic [21], [22], respectively, were used to remove adaptor sequences and poorly sequenced reads. After trimming, reads that were shorter than 20 bp were excluded from further analyses. Reads were mapped to the KoRV full genome reference sequence (NCBI: AF151794) using BWA version 0.6.2 [23] with default parameters. The resulting SAM files were further processed with samtools [24] and picard (http://picard.sourceforge.net) for sorting and removal of clonality, respectively. The Perl script mapDamage was run on the museum data using the default settings to determine the percentage of DNA damage present, before SNP calling [25]. Variant call analysis was performed using VarScan 2.2.3 with the following settings -min-coverage 8, -min-var-freq 0.01, and -p-value 5e-02 [26]. The resulting variants were further curated using Geneious 6.0.4 for visualization. Negative control reads were also compared to the reference KoRV sequence. The 5′ and 3′ LTRs were distinguished from each other by examining sequences adjacent to the LTR sequence for genomic flank sequences or for KoRV sequence (gag leader or env). The 5′LTR is preceded by a koala genomic flank and followed by a KoRV gag leader, while the 3′LTR is preceded by KoRV env and followed by a koala genomic flank. Where possible, LTR sequences that also included a KoRV non-LTR sequence or a koala genomic flanking sequence were used to distinguish between 5′ and 3′ LTR polymorphisms (Figure 1). Consensus sequences generated were deposited in GenBank (accession numbers KF786280–KF786286). Illumina reads mapping to KoRV for each koala were deposited in the NIH Short Read Archive (SRP03960187947).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 1. Alignment of modern and museum koala retrovirus sequences, showing positions of proviral genes and proteins.

Upper Panel: Character states matching the reference sequence (AF151794) are indicated in light grey, while mismatches (position 312) or polymorphisms (all other positions) are shown as black hatch marks. The infectious clone KV522 (AB721500) is the first sequence below the reference. The aligned sequences all display open reading frames for viral gag, pol, and env regions, except that polymorphisms at three positions in the museum samples code for a stop codon that would disrupt an open reading frame; these are indicated by red hatch marks. Green lines represent polymorphisms that could be placed in phase in overlapping sequence reads. Lower Panel: The coded proteins are indicated, following the divisions proposed by Hanger et al. (2000) relative to the polymorphism alignment. The positions of the SRLPIY domain potentially involved in viral infectivity of the GAG protein and p10 domain (Gag assembly and nuclear export signal, respectively) are indicated. Likewise, the Env motif CETTG, and p15E transmembrane envelope protein are indicated. Regions known to be divergent in Japanese isolates, KoRV-B and KoRV- C, and KoRV-D [12], [40], are indicated by orange, purple, and green arrows, respectively.

https://doi.org/10.1371/journal.pone.0095633.g001

Integration sites were identified in sequence reads that contained 5′ or 3′ LTR sequences extending into non-KoRV sequences. To examine whether proviral integration sites identified in the ancient samples were present among modern koalas, koala genomic sequences flanking the integration sites found by hybridization were queried against sequences flanking the integration sites of six modern koalas (three northern and three southern koalas, Table 1) that had been generated using a different method (Ishida et al. manuscript in preparation). Integration site sequences were also queried against a koala (Pci-SN404) whole genome sequence generated using one-sixteenth of a PicoTiterPlate of 454 GS-FLX+ Technology (Roche Applied Science) following standard protocols.

Statistical Analyses and Tests of Selection

For non-synonymous polymorphisms, a “radical” change was defined as a mutation that produces a negative score in both BLOSUM62 and BLOSUM90 substitution matrices. Associations between variables were examined using a 2×2 contingency table, testing for significance using Fisher’s exact test implemented in GraphPad (graphpad.com/quickcalcs/contingency1.cfm). The number of synonymous and non-synonymous substitutions was determined, and the Nei-Gojobori method [27] was used to determine the proportion of synonymous substitutions per synonymous sites and the proportion of non-synonymous substitutions per non-synonymous sites. MEGA, version 5.2 [28] was used to estimate Tajima’s D, and to implement the codon-based Z-test for selection and the codon-based Fisher’s exact test of selection. These were determined for the concatenated KoRV codons of gag, pol, and env, and for each of the three separately. Bonferroni correction for multiple hypothesis testing divided a p value of 0.05 by the number of hypotheses tested.

The dN/dS ratio provides an indicator of the selective pressures that acted upon a gene, with low values indicating purifying selection and increases in values indicating relaxation of constraint or positive selection. To account for the different phase of polymorphisms at the same site we generated an individual sequence for each different phase of a polymorphism and analyzed all of the individual sequences. For the modern koala, available sequences were long enough for phase to be determined for many (but not all) polymorphisms (Figure 1, positions underlined in green). For historical samples, sequence lengths were short, and the phase of polymorphisms could only rarely be determined. To test for this signature of selection in this dataset, we calculated dN/dS using two different approaches: the GA-Branch and FUBAR methods [29]. In the first case estimates were obtained using a fixed tree topology generated by the Neighbor-joining method. The nucleotide model was specified as GTR; otherwise, the default GA-Branch configuration was used. This dataset, which compares all identified polymorphisms (even if they are not present in the consensus sequence) against the modern sequence, was also analyzed using the Z-test for selection and Tajima’s D.

Identification of Protein Domains and Functional Residues, and Protein Modeling

The corresponding amino acid sequences were subjected to domain identification analysis using the Conserved Domains Database (CDD) from NCBI. We also examined whether any of the observed polymorphisms alter amino acid residues of known function using the Conserved Features/Sites option of the CDD database.

To examine the structural characteristics of KoRV variants, we predicted their three-dimensional structures using the SWISS-MODEL server [30]. Only models with high statistical support (high reliability score as defined by QMEAN4 values) [31] were considered for further analyses. Using this strategy we were able to reliably model several regions corresponding to different domains of all three viral polypeptides (Gag, Pol, and Env). Pairwise structural alignments and structural superimposition were performed using the DaliLite server [32]. Models and Figures were drawn using Pymol (DeLano Scientific).

Transcription Binding Factor Site Analysis

The long terminal repeat polymorphic sequences for each koala were analyzed for putative transcription binding domains using MatInspectror software (Genomatix, Munich). The default core similarity and matrix similarity greater than 0.8 were employed as the selection criteria.

Hybridization Capture and Sequencing of KoRV

DNA was extracted in an ancient DNA dedicated facility from 10 museum skins from southern (n = 2) or northern (n = 8) Australian koalas, which had been collected as long as 130 years ago (Table 1). In separate facilities, modern DNA was extracted from blood samples of zoo and free-ranging wild koalas (Table 1). Illumina libraries were prepared from all museum koala DNA extracts (in an ancient DNA facility), and from one sample of modern DNA from an adult northern koala 14 years old, Pci-SN265 (“Mirra-Li”, studbook number 265 from the Zoo Vienna, Austria). In order to process all samples in a single next-generation sequencing run, each library was tagged with a distinct index sequence. Baits for hybridization were generated covering the entire KoRV genome, which was amplified in thirty-eight fragments, each ca. 500 bp, from four koalas representing northern (Pci-SN404) and southern (Pci-SN106, 142, 157) koala KoRV diversity (Table 1). Equimolar amounts of index samples were pooled and applied to KoRV baits bound to streptavidin beads for in-solution hybridization capture. The enriched koala libraries were then sequenced using an Illumina MiSeq. After sequencing, a bioinformatics routine used the distinct index sequence tags to separate sequences by individual. Sequences were screened for quality and reliability before being aligned to KoRV reference genomes (GenBank accession number AF151794, AB721500).

Full coverage of the KoRV genome was obtained from six of the northern Australian museum specimens and for Pci-SN265 (Figure 1); museum specimens of two additional northern and two southern koalas were not successful. Among the six successful historical samples, KoRV-specific sequences represented 2.5% to 41% of the total number of reads, comparable to enrichment rates previously reported for ancient DNA. Negative controls demonstrated only sporadic matches to KoRV (Figure 2). Such sporadic reads are observed in hybridization capture experiments [33] and may reflect index sequence errors (misassignment) or PCR jumping causing exchange between sample index and control [34]. However, the profile of coverage was randomly dispersed and the number of reads marginal in the negative control. Coverage was consistently far higher at every position for the modern koala (Pci-SN265) than for any of the museum specimens (Figure 2). The historical samples yielded only 20–80% of the coverage of the KoRV genome obtained for the modern sample. Among the museum koalas, the earliest collected sample (Pci-maex1738) had the poorest coverage whereas the most recently collected sample (Pci-QMJ6480) had the highest relative coverage. There was otherwise no obvious correlation between the number of reads obtained and the year the sample was collected.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 2. Hybridization capture sequence coverage across the KoRV genome for modern and museum koala samples.

The sequence coverage is shown for each nucleotide position numbered as in the KoRV reference genome (AB721500). Results are shown for 1 modern (Pci-SN265) and 6 museum koala samples. Mapping of results for a negative control (NC) are also shown. Each sample is color-coded.

https://doi.org/10.1371/journal.pone.0095633.g002

KoRV Polymorphisms

For the museum samples, the average read length was ca. 90 bp. This is similar to read lengths reported previously for DNA from archival specimens, which may be degraded as a result of environmental, bacterial, and enzymatic damage [35]–[37], and was shorter than the ca. 135 bp read length for modern sample Pci-SN265. Although similar analyses were conducted on the historical and on the modern koala sequences, prior to assembly the museum specimen datasets were processed using the mapDamage Perl script [25], to account for DNA damage present in ancient DNA. The mapDamage results identified the expected nucleotide misincorporation patterns of cytosine to thymine and guanine to adenine on the 5′ and 3′ end termini, respectively (not shown). However, damage occurred only at a very low frequency of 0.02 to 0.08%, indicating that the damage present would have negligible effects on polymorphism scoring or other analyses. After assembling the reads to the KoRV reference sequence AF151794, polymorphisms were scored if they occurred at a position in 8% or more of the reads for an individual koala [38]. For the env gene, four of 20 env polymorphisms that had been previously detected by PCR from museum samples were also found in the current dataset [6]. Of the remaining 16, seven could be identified but were not present above the cutoff employed when identifying polymorphisms by the current study. The other nine could not be identified from the data, likely due to insufficient coverage in some koalas for those regions of env (Figure 2). Fourteen novel polymorphic sites in the env region were identified by hybridization capture that had not been identified in the same museum koalas when previously examined by PCR.

At position 312 a fixed difference as opposed to a polymorphism was present in all koalas relative to the reference AF151794 (Figure 1). Across the modern and archival koalas, a total of 138 KoRV polymorphisms were detected. At each of these polymorphisms, one of the character states matched that of the KoRV reference sequence AF151794. Considering only the character states that differed from the reference, seventy-one of the polymorphic sites were detected in two or more koalas (shared alleles) and sixty-seven were detected only in one koala (private alleles) (Table 2, Table S2). Of 92 polymorphisms in the coding regions, 3 would result in stop codons, of which one was shared across individuals (Figure 1, Table 2 KoRV). Of the remaining coding region polymorphisms, 35 were synonymous and 54 were non-synonymous (Table 2, Table S2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Types of KoRV polymorphisms detected across 6 museum specimens of koalas.

https://doi.org/10.1371/journal.pone.0095633.t002

Functional regions in the viral sequence reported to reduce the infectivity of KoRV when compared to that of GALV were also examined [3]. The CETAG motif in KoRV (CETTG in other gammaretroviruses) is believed to be responsible for viral fusion activity while the gag L-domain is believed to affect the release of mature virus. Across KoRVs in the newly sequenced koalas, there were no polymorphic sites in either of these regions (Figure 1). The immunosuppressive domain of the p15E transmembrane protein of retroviral Env exhibited only a single polymorphism, present in Pci-SN265 and Pci-MCZ12454 (Table S2). In the museum specimens, multiple non-synonymous polymorphisms were detected in the nucleocapsid protein region (p10) of the gag gene. Using the Conserved Features/Sites function of the CDD database we also determined that none of the amino acid residues of known or inferred function, e.g., DNA binding site of the reverse transcriptase domain in POL or the homotrimer interface in ENV, are polymorphic (data not shown).

Amino acid substitution matrices have been generated by comparing large numbers of proteins to identify non-synonymous mutations that are only rarely observed empirically. These rare amino acid substitutions, termed “radical”, typically involve major physiochemical differences between the two amino acids. We defined a radical change as a mutation that produces a negative score in both BLOSUM62 and BLOSUM90 substitution matrices. Among the non-synonymous polymorphisms observed in the koala, 48% (26/54) of substitutions were defined as radical. The proportion of radical non-synonymous mutations appeared to be higher in env than in gag or pol (Table 2), and this difference was confirmed as significant using Fisher’s exact test (p = 0.0397) comparing radical vs. non-radical non-synonymous substitutions in env to those in gag-pol. This suggested that selective constraints on env may differ from those affecting the other two KoRV coding regions. Across the three coding regions, no other pattern suggestive of an association across variables was evident in the dataset for private vs. shared polymorphisms, non-synonymous vs. synonymous polymorphisms, or radical vs. non-radical amino acid changes (Table 2).

The selective pressure variation among all branches of the KoRV tree estimated by the GA-Branch method suggested that several branches in gag (more than 70%) and fewer in pol and env (60 and 17%, respectively) are under purifying selection (not shown). FUBAR implemented in HyPhy also suggested that only a few codons deviate neutrality (not shown). Similarly, the total distance estimates of dN/dS using the Nei-Gojobori method suggested stronger purifying selection in gag than in pol and env (Table S3). The same trend was observed by the Z-tests for selection and the Tajima’s test of neutrality, with gag showing multiple significantly negative dN-dS values and the lowest negative values, respectively (Table S4 and Table S5).

Comparisons of KoRV Consensus Sequences

The nucleotide consensus sequence (majority character state at every position in an alignment of sequences) was generated for each of the seven successful KoRV-positive koalas. These were compared to the first reported KoRV sequence (AF151794) [7] and to the infectious clone KV522 (AB721500) [39], which themselves are 0.5% divergent, generating an alignment of 9 sequences. Each of the newly generated consensus sequences was more similar to the infectious clone KV522 (99.2–99.5% similarity) versus for AF151794 (99.0–99.2% similarity). All of the koala retroviral consensus sequences from the current study included a 6-bp insertion in the non-coding gag leader region, position 651, which is also present in KV522 (Figure 1, Table S2). The six archival sample consensus sequences also shared a 3-bp insertion at position 905 in the gag leader region (Table S2) also present in KV522 (AB721500) [39]. The 3 bp insertion could be found in the modern koala (Pci-SN265) as a minority sequence, thus the consensus for this animal lacked the 3 bp insertion (Figure 1). Thus, in contrast to the museum samples, the modern koala had an underrepresentation (16.8%) of the 3 bp insertion variant. In addition, the deletion itself is polymorphic representing 1–3 bp deletions though the 3 bp deletion is the most common and therefore represented in the consensus sequence generated.

The alignment of nine sequences was also examined for signatures of selection (or neutrality), for gag, pol, and env, and for all three codon sets concatenated. The Nei-Gojobori method was used to estimate synonymous and non-synonymous mutation rates for each pair of sequences. Codon-based Fisher’s exact tests of selection found no evidence of positive selection in any of the pairwise comparisons for any of the coding regions (not shown). Codon based Z-tests of selection (Table 3) suggested that among the coding regions purifying selection may be more pronounced in gag, with significant purifying selection detected in eight of the pairwise comparisons (although these would not be significant after Bonferroni correction). The pol coding region appeared to be under weaker purifying selection, with negative values significant (before Bonferroni correction) for only 2 comparisons, while env comparisons yielded both positive and negative estimates consistent with neutrality (none significant). Tajima’s D was calculated using the same nine KoRV sequences (Table 4), for each coding region separately or all three combined. A negative value would be consistent with purifying selection. However, although gag had the most negative value, none of the values for Tajima’s D were extreme, thus there were no significant deviations from neutrality. The consensus sequence and total polymorphism data yielded consistent results with respect to selective pressures on KoRV.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Codon based Z tests of selection.

https://doi.org/10.1371/journal.pone.0095633.t003

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Estimates of Tajima’s D^*.

https://doi.org/10.1371/journal.pone.0095633.t004

KoRV-B and J

A recent study of koalas from Los Angeles Zoos identified a KoRV variant, designated KoRV-B, which has greater virulence than previously characterized KoRV, and which was present in a subset of zoo koalas [12]. Hybridization capture should enrich sequences, such as those of KoRV-B, that are somewhat divergent from the KoRV sequence used as bait. We therefore screened the novel next generation sequencing data, searching for the sequence of KoRV-B at the junction where KoRV-B diverges from the KoRV reference sequences in the env region. The divergent region of KoRV-B env was not detected in any of the ancient koalas, suggesting that KoRV-B may have evolved recently as a variant of KoRV. However in the modern koala Pci-SN265, sequences matching KoRV-B were detected for some (but not all) of the regions within the env gene that characterize KoRV-B (Figure S1).

Three other variants of KoRV have recently been described among zoo koalas in Japan, tagged as clones 11-1, 11-2, and 11-4 [11], [40]. Clone 11-1 has been designated KoRV-D, clone 11-2 has been designated KoRV-C, and clone 11-4 has been designated KoRV-J. The three recently identified KoRV clones differ mainly in variable region A of the env gene that is involved in retroviral receptor determination and recognition [11], [41]. Our novel sequences were screened for each of the KoRV variants reported in the Japanese zoo koalas. Sequences similar to those of KoRV-C and KoRV-D were identified in the modern koala Pci-SN265 but not in any of the museum samples (Figure S2 A and B) [40]. Sequences related to KoRV-J were not identified in any of the novel reads, whether from the modern or museum samples.

Potential Effects of KoRV Polymorphisms on Protein Structure

Variants present below a cutoff of 8% of relevant Illumina reads were not considered to represent confirmable polymorphisms. Those that appeared at a higher frequency than this cutoff likely represent common variants rather than mutations within a single provirus. We examined the effects of the non-synonymous polymorphisms on the protein structure of KoRV by generating three-dimensional models for Gag, Pol, and Env protein fragments. First, we sought to identify whether amino acid differences present across modern sequences of KoRV led to major structural differences. Sequences included in the comparisons of modern KoRV were the original KoRV isolate AF151794 and infectious clone KV522 (AB721500), which differ by 0.5% at the nucleotide. In these comparisons, our consensus KoRV sequence from the modern koala Pci-SN265 served as the reference (the amino acid residue in Pci-SN265 is always the first listed in each substitution). When superimposed on the structure of Pci-SN265, the structures of AF151794 and KV522 showed minor localized changes affecting the polarity, charge, or local protein conformation (Figure 3). Specifically, in the Gag protein mutations K47E and S464F alter the local charge and the local protein conformation, respectively (Figure 3). In the Pol protein, mutations P6S, A124V, K764R, R771G, and N924D between the Pci-SN265 and AF151794 had only minor effects on the overall topology of the structure. In Pol, mutations I19V, A822T, and S829P altered the local conformation of the Pci-SN265 and KV522 relative to the AF151794 structure by changing a surface residue to a buried one (I19V) and changing two partially buried residues to surface ones (A822T and S829P) (Figure 3). Only two positions in the Env protein could be structurally modeled (P147S, D187G). Both of these were radical substitutions that changed buried amino acids to exposed ones (Figure 3). Both of these changes were located away from the putative receptor-binding region, as this has been defined in [6].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 3. Structural superimpositions of Pci-SN265 (green), AF151794 (red), and KV522 (gold) KoRV Gag, Pol, and Env protein structures, demonstrating the overall similarity of the structures.

Amino acid variations across these three sequences are mapped on the protein models (arrows). The structural differences predicted are attributed to changes in the polarity, charge, and atom conformations. The models are shown in cartoon (left panels) and semi-transparent surface (right panels) representations. The atoms of the variable amino acid residues are shown in line representations to view the side chains. In all comparisons the Pci-SN265 consensus sequence was used as the reference.

https://doi.org/10.1371/journal.pone.0095633.g003

Second, non-synonymous polymorphisms in the historical KoRV sequences were examined for predicted changes to the protein structure as compared to the modern consensus sequence Pci-SN265, which again served as a reference sequence (the amino acid residue in Pci-SN265 is always the first listed in each substitution). The effects on polypeptide structures of the non-synonymous polymorphisms present in KoRV for each koala were examined using a composite sequence that contained all of the amino acid differences versus Pci-SN265. This composite sequence would necessarily combine polymorphisms present on different proviral loci. Nonetheless, this composite sequence would be useful in identifying all of the potential disruptions to predicted structure, when the effects of each mutation are considered individually. Several ancient variants were predicted to cause small local fluctuations of the KoRV structure. Specifically positions G33E, K421E, Q429K, and S464Y are predicted to alter the local charge at the Gag surface resulting in deviated conformations (Figure 4A, Figure S3). In the Pol protein, positions S514R, F396Y, A685S, and Y676N exchange a buried residue for a surface one, resulting in topological differences (Figure 4B, Figure S4). Additionally, in the Pol protein major conformational changes were predicted to occur at ancient variants R853Q, P933T, V939E, and T1014I. Lastly, two ancient variants S75F and R214W found in the Env protein were predicted to have major structural effects (Figure 4C, Figure S5). Both of these changes are located away from the putative receptor binding region as this has been defined in [6]. It is important to note that the other character state found at each of the polymorphic sites in the ancient koalas matched the character state present in the reference sequence. Thus, despite the presence of these polymorphisms and their modeled effects on proteins, KoRV overall has remained stable in sequence and structure over time.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 4. The effects of historical KoRV polymorphisms on protein structure.

Superimpositions are shown between the present day consensus KoRV (Pci-SN265) protein structure and ancient KoRV variants. Amino acid variations between these sequences mapped on the protein models are shown in red and with arrows. The models are shown in cartoon ribbon representations (left panels) and as semi-transparent surfaces (right panels). The atoms of the variable amino acid residues are in line representations to view the side chains. In all comparisons the Pci-SN265 consensus was used as the reference sequence. (A) The model of the Pci-SN265 Gag protein is superimposed with the models of variants found in archival koalas um3435 and maex1738. (B) The model of the Pci-SN265 Pol protein is superimposed with variants found in QMJ6480, 582119, MCZ8574, Um3435, and maex1738. (C) The model of the Pci-SN265 Env protein is superimposed with the model of variants found in MCZ_12454 and um3435. For all three polypeptides, the structural differences predicted are attributed to changes in the polarity, charge, and atom conformations and are largely localized onto flexible loop regions.

https://doi.org/10.1371/journal.pone.0095633.g004

LTR and Integration Site Diversity

Multiple polymorphisms were observed in the long terminal repeats (LTRs) that serve as promoter and terminator of the retroviral transcription process (Figure 1). MatInspector (Genomatix) was used to examine U3, R, and U5 regions of the LTRs for sequences matching transcription factor binding sites (TFBs) and for disruptions of TFBs by polymorphisms. Polymorphic sites from the next generation sequencing data were placed in phase when possible (Figure 1). This analysis of the LTRs revealed the presence of 82 putative TFBs (Table S3). Six of these had been referred to by Hanger et al. (2000), including a TATA box, CCAAT retroviral signal, and C-type poly-A signal (Table S6). None of the polymorphisms detected would have disrupted the previously predicted TFBs. However 20 additional predicted TFBs would be generated by the various polymorphisms.

Sequences at the 5′ and 3′ end of the KoRV genome often extended beyond the proviral integration sites into the host genomic flank. Shared integration sites among koalas would be strong evidence that a given locus represents an endogenized retrovirus, since the chance that two proviruses would independently integrate into the same locus is minuscule [42]. Four hundred twenty nine 5′ flanks and three hundred thirty one 3′ flanks were identified across all the koalas tested (Table S7). Thirty-two 5′ and twenty-three 3′ flanking sequences were shared by two or more koalas, representing 7.5% and 7% of the total respectively (Table S4).

The sequences flanking the integration sites were queried against flanks found in six modern koalas, which had been detected by other methods (Ishida et al., in preparation). Eight of the integration sites found by hybridization capture were also identified in one or more of six modern koalas tested (Table S7). The sequences flanking KoRV were also queried against whole genome sequences from a single koala, generated from one-sixteenth plate GS-FLX shotgun sequence (Ishida et al., in preparation). Two flanking sequences had matches among the GS-FLX results. One of the flanking sequences matched 45 of the next-generation sequences, suggesting that this KoRV provirus had integrated in a repetitive element of the koala genome. The other matching flank sequence was detected only once. This sequence was KoRV negative at the locus, with the host genomic flank on the other side of the provirus evident in the sequence.