Phylogeny, recombination and expression of porcine endogenous retrovirus {gamma}2 nucleotide sequences

Xenotransplantation of genetically modified pig tissues aims to compensate for the shortage of human donor organs. A major obstacle is the cross-species transmission of pathogens. The proposed use of specific-pathogen-free animals has focused the potential infectious risk on porcine endogenous retroviruses (PERVs). Endogenous retroviruses (ERVs) are proviruses of exogenous retroviral genomes that have been integrated into the germ line of the host and inherited by the offspring. ERVs have been found in multiple copy numbers in all mammals, with most being defective due to a lack of selective pressure. Intact ERVs harbour the genes gag, pro/pol and env enclosed by long terminal repeats (LTRs) at both ends (Beckmann et al., 2000). Phylogenetic analyses have classified ERVs into the retroviral β (B- or D-type) and γ (C-type) genera, which consist of various families (van Regenmortel et al., 2000). The PERV γ sequences described to date have been classified into ten families, γ1 to γ10. The well-characterized PERV γ1 family and an additional family originally named PERV E are present at about 50 copies per haploid genome, whereas the other families are present with less than ten copies each (Klymiuk et al., 2002; Mang et al., 2001; Patience et al., 2001). Breed-specific and/or individual variations have been revealed for the PERV γ1 load (Magre et al., 2003).

The known infectious, human-tropic PERVs have been assigned to the PERV γ1 family consisting of the subfamilies A, B and C. Differences in the env genes of the three subfamilies explain their different host tropisms (Patience et al., 2001). Functional hybrid viruses derived from recombination events of PERV γ1 genomes have been found (Klymiuk et al., 2003a; Lee et al., 2002; Oldmixon et al., 2002).

PERV E was named after its phylogenetic relationship to the human ERV (HERV) E sequences (Mang et al., 2001). Independently, sequences of this family were also designated PERV γ2 and γ6 (Klymiuk et al., 2002; Patience et al., 2001). Following the nomenclature of PERV γ sequences proposed by Patience et al. (2001), in this study we have classified the sequences of this family as PERV γ2.

Few PERV γ2 sequences have been published to date. Two clones have been derived from a domestic pig genomic DNA phage λ library. One clone (GenBank no. AF356697) showed a complete 8 kb retroviral sequence, whereas the other 7 kb clone (GenBank no. AF356698) lacked the 5' LTR and gag untranslated regions. Multiple mutations have been observed in both sequences (Mang et al., 2001). In another study, a non-functional 1 kb pro/pol gene sequence (GenBank no. AF274706) was found (Patience et al., 2001). Recently, we described 17 additional pro/pol clones (including GenBank nos AF511100AF511110) derived from at least 11 loci amplified from the Landrace genome. None of the clones was found to contain an open reading frame (ORF) (Klymiuk et al., 2002).

Here, we examined PERV γ2 in various pig breeds at the genomic and expression level and carried out phylogenetic analyses by using pro/pol and env gene sequences.

Pig breeds examined.
PERV γ2 pro/pol sequences were examined in three animals each of the six pig breeds Duroc, Landrace, Large White, Mangaliza, Pietrain and Turopolja. The genomic organization of the PERV γ2 sequences was analysed in Duroc, Landrace and Mangaliza animals, as well as in Munich Troll mini pigs and SchwaebischHaellisch pigs. Expression analysis was carried out in Landrace animals.

Detection of PERV γ2 sequences.
Genomic pig DNA was used to amplify PERV γ2 pro/pol sequences according to published protocols (Klymiuk et al., 2002). Briefly, five degenerate primers (two 5' primers and three 3' primers) were mixed in a PCR with an annealing temperature of 38 °C. PCR products were separated by agarose-gel electrophoresis and 0·51·2 kb fragments were isolated and cloned into the pGEM-T Easy vector (Promega). After bacterial transformation into TOP10F' cells (Invitrogen) and cultivation on agar plates, bacterial colonies were transferred to Hybond-N⁺ membranes (Amersham Biosciences). After treatment with bacterial cell-lysis solution (0·5 M NaOH, 1·5 M NaCl), the filters were neutralized [0·5 M Tris/HCl (pH 7·5), 1·5 M NaCl] and washed (5x SSC). Fixation of DNA was done by UV cross-linking.

Radiolabelling of the probe and hybridization were carried out according to standard protocols. The 941 nt PERV γ2-specific PCR product from clone m211 (GenBank no. AF511105) showed no cross-hybridization to the closest relative, PERV γ10 (Klymiuk et al., 2002), and was used as a probe. Plasmid DNA from positive clones was prepared and sequenced with an Amersham DNA sequencing kit and an ABI PRISM 377 automated DNA sequencer. Clones were sequenced bidirectionally by using additional primers that annealed within the sequences.

Bioinformatic tools.
Sequences were analysed for homology, intact ORFs and endonuclease restriction sites by using Gene Jockey II (Biosoft). Additional PERV γ2 sequences were obtained from GenBank by using a BLAST search. For multiple alignments, sequences were prepared in SeqApp () or BioEdit (Hall, 1999), aligned by CLUSTAL_W (Jeanmougin et al., 1998) and post-edited manually. MacClade () and PAUP* () were used to prepare the data for the phylogenetic studies. Most-parsimony, neighbour-joining and maximum-likelihood phylogenetic trees were calculated with the PHYLIP package () and trees were generated by TreeView ().

Southern blot hybridization.
For analysis of the genomic organization of PERV γ2, porcine DNA was isolated from tissue samples by standard protocols. Ten micrograms of genomic DNA was digested with EcoRV or PstI and separated on a 1 % agarose gel. After transfer of the DNA to Hybond-N⁺ membranes (Amersham Pharmacia Biotech), hybridization was carried out using the 941 nt PERV γ2-specific probe described for the colony hybridization (see above).

PERV γ2 expression analysis.
Total RNA was extracted from eight tissues (blood, heart, kidney, liver, lung, placenta, spleen and thymus) by using TRIzol reagent (Invitrogen). Two pregnant Landrace sows and two or three of their fetuses each were examined in the last third of the pregnancy. Potential DNA contamination was removed by DNase I treatment [2 U (µg RNA)¹; Roche] in the presence of RNasin ribonuclease inhibitor [1 U (µg RNA)¹; Promega]. Reverse transcription was carried out with Superscript II RNase H reverse transcriptase (Invitrogen) and oligo(dT)₁₂₁₈ primers (Invitrogen), the primer rna-tag [5'-(GA)₇-CTCGAGCGGCCGC(T)₁₆V-3'] or random-primer hexamers (Promega) following the protocol supplied by the manufacturer.

RNA integrity was assessed with the primers act1 (5'-ACCACACCTTCTACAACGAGC-3') and act2 (5'-TAGTTTCGTGAATGCCGCAGG-3'), producing a 573 nt fragment from β-actin cDNA. PERV γ2-specific pro/pol sequences were amplified with the primers f37 (5'-CAACCTGKGGCTCCCCTCTYG-3') and r840 (5'-TTYCTTTGTCCCTGTATGTGK-3') resulting in an 804 nt product, and the primers f191 (5'-TRATGGGAARAGAYCTGCTGG-3') and r840 giving rise to a 650 nt product. Absence of genomic DNA contamination was shown by using the PERV γ2-specific primers and RNA that had not been reverse-transcribed as template.

RNA integrity, DNA contamination and PERV γ2 expression were analysed by PCR with an initial denaturation step at 95 °C for 2 min, followed by 35 cycles of 95 °C for 30 s, 56 °C for 30 s and 72 °C for 1 min. For quantification of PERV γ2 expression, the DNA concentration of the 941 nt PERV γ2-specific probe m211 (GenBank no. AF511105) was measured by spectrophotometry at 260 nm and also determined visually on an agarose gel. Dilutions of the probe were used as controls in the semi-quantitative RT-PCR.

The amplification products of the RT-PCR were purified, cloned into the pGEM-T Easy vector and sequenced as described above to confirm their PERV γ2 origin. Long-range PCR was done on cDNA from liver and thymus generated with the primer rna-tag. The primers f37 and cdna-tag [5'-(GA)₇-CTCGAGCGGCCGCTT-3'] were used with an initial denaturation step of 95 °C for 4 min, followed by 20 cycles of 95 °C for 1 min, 56 °C for 1 min and 72 °C for 6 min, and 15 cycles of 95 °C for 1 min, 56 °C for 1 min and 72 °C for 9 min. The PCR products were diluted 1 : 10 000 and used as a template for the semi-nested PCR with the primers f191 and cdna-tag. The PCR products were separated on a 1 % agarose gel and transferred to membranes. PERV γ2-specific hybridization was carried out as described above.

Nucleotide sequence accession numbers.
The sequences determined in this study have been deposited in GenBank with accession numbers DQ084470 (clone g63), DQ084471 (g620), DQ084472 (g61), DQ084473 (g68x), DQ084474 (g611), DQ084475 (g617), DQ084476 (g619), DQ084477 (m2116), DQ084478 (m2118) and DQ084479 (g618x). ORF-containing clones are indicated with an x.

Our recently described non-functional PERV γ2 sequences (Klymiuk et al., 2002) were included in the phylogenetic analysis of this study and have been published with GenBank accession numbers AF511100 (m116), AF511101 (m2110), AF511102 (m2119), AF511103 (m2113), AF511104 (m218), AF511105 (m211), AF511106 (m2117), AF511107 (m217), AF511108 (m1116), AF511109 (p1117) and AF511110 (m216).

Detection and sequence analysis of PERV γ2 pro/pol clones
The porcine genome has been described as containing up to 50 PERV γ2 proviruses. In previous studies, analysis of three pigs of the Landrace, and LandracexDuroc breeds and an unspecified breed observed no intact pro/pol sequences (Klymiuk et al., 2002; Mang et al., 2001; Patience et al., 2001). For evaluation of PERV γ2 diversity, genomic DNA was pooled from 18 animals of six different breeds to obtain a broad spectrum of γ2 proviruses. After amplification and cloning of the 1 kb fragments, PERV γ2-specific colony hybridization revealed 19 positive clones. Sequence analysis of these clones found that 18 showed >90 % identity to published γ2 sequences and had a length of 810950 nt. Five of the 18 clones revealed an ORF [g67x, g68x (GenBank no. DQ084473), g69x, g612x and g618x (GenBank no. DQ084479)] and four [g67x, g69x, g612x, g618x (GenBank no. DQ084479)] differed in not more than four out of 941 nt (0·4 %). One clone (g610) had a 132 nt gap, corresponding to five previously described clones [m216 (GenBank no. AF511110), m2111, m2118, p118, p1117 (GenBank no. AF511109)].

Phylogeny and recombination analysis of the PERV γ2 pro/pol clones
In addition to our clones and the published γ2 sequences AF356697, AF356698 (Mang et al., 2001) and AF274706 (Patience et al., 2001), a BLAST search revealed five 710 kb retroviral sequences (GenBank nos AX111980, AX175459, AX175460, AX175461 and NC_003059) and the 120 kb BAC clone PigE-108A11 (GenBank no. CR450381) that were highly homologous to the PERV γ2 family. Thus, a total dataset of 44 pro/pol fragments (808950 nt) was used in the phylogeny and recombination analysis (Table 1).

Table 1. Origin of the PERV γ2 pro/pol sequences used in the phylogeny analysis DC, Duroc; LR, Landrace; LW, Large White; MG, Mangaliza; MS, Meishan; PT, Pietrain; TP, Turopolja; NS, not stated. The clones were isolatedby using PCR, porcine BAC clones or porcine genomic DNA by undefined methods (gDNA).

To prevent the multiple presence of a given sequence in the dataset, haplotypes were defined by distribution of the genetic distances. The distances divided into two clearly separated sections of distance values from 0·000 to 0·005 and distance values higher than 0·015 (Fig. 1a). The genetic distances of between 0·000 and 0·005 were proposed to occur within a given haplotype, whereas genetic distances between different haplotypes resulted in higher values. Thus, we defined 22 haplotypes and chose representative clones for each haplotype by the lowest number of single nucleotide exchanges in an alignment of all 44 sequences (not shown). Subsequently, phylogenetic trees were generated from the 22 haplotypes and showed weak resolutions (Fig. 1b). Eleven haplotypes representing 16 sequences were separated by bootstrap values higher than 50 %. All 16 sequences were analysed in a separate nucleotide alignment, which was condensed by PAUP* as described previously (Klymiuk & Aigner, 2005). The alignment showed 198 polymorphic nucleotides (20 %) and clearly revealed four different recombination patterns (Fig. 1c). The recombined sequences were separated in homologous and different fragments in the alignment, and phylogenetic trees were generated from this manipulated dataset. The trees revealed a bush-like evolution of the PERV γ2 family with low bootstrap values (Fig. 1d). Independent evolution of the γ2 proviruses was confirmed by their nucleotide alignment, as 273 out of 282 polymorphic sites (97 %) were either single nucleotide exchanges or patterns occurring only once in the alignment and therefore did not contribute to the evaluation of the phylogenetic relationship between the haplotypes (data not shown).

(36K): Fig. 1. Phylogenetic analysis of PERV γ2 pro/pol sequences. The sequences from this study and our previous study (Klymiuk et al., 2002) are depicted by their clone number, whilst sequences obtained from GenBank are indicated by accession numbers. (a) Definition of haplotypes. A genetic-distance matrix was generated for the 44 sequences by using the PHYLIP package. The 44 sequences were classified by their genetic-distance values. x axis, segments of genetic-distance values; y axis, number of genetic distances in the respective segment. Values higher than 0·005 were assumed to occur from different haplotypes. (b) Phylogeny of the 22 haplotypes. The consensus of 100 neighbour-joining trees was generated with the representative sequences chosen by the lowest number of single-nucleotide exchanges. Additional sequences of a given haplotype are shown in parentheses. Bootstrap values higher than 50 % are shown. (c) The 16 haplotypes separated by bootstrap values higher than 50 % were reduced to 198 polymorphic nucleotides (20 % of 971 nt). Shaded boxes indicate identical nucleotide polymorphisms of the sequences. Recombination events were observed in the first four groups depicted and are indicated by solid lines within the boxes. Unclassified regions are indicated by dashed lines. (d) Recombinant sequences were separated subsequently into homologous (Hom) and differing (Diff) fragments. m116 was excluded from the dataset, as it was represented by the g68x haplotype in the 590 nt 5' end and by m2110 3' of nt 478. m2113 and m218 were reduced to the nt 1205 and 413471 regions and a consensus sequence m2113/m218 was generated for the remaining regions. AF274706 was reduced to the 3' end starting with nt 436. Twelve nucleotides in this region separated AF274706 from m212, whereas only two mismatches were found in the 5' end. The m1116/g617 consensus sequence consisted of nt 1562, whereas the 3' end of m1116 and g617 both remained in the dataset. The consensus of 100 most-parsimony trees from the manipulated dataset revealed a bush-like evolution of the PERV γ2 family with low bootstrap values. Additional sequences represented by the haplotypes are shown in parentheses. All bootstrap values are indicated.

PERV γ2 proviral load in different pig breeds
The PERV γ2 proviral load was examined by Southern blot analysis and was estimated to be 50 copies per haploid genome for the five breeds Duroc, Landrace, Large White, Mangaliza and Pietrain, and about 25 copies in the Turopolja breed (Klymiuk et al., 2002). BamHI restriction analysis resulted in a γ2-specific 3 kb fragment for most of the γ2 loci due to conserved restriction sites among the proviruses (e.g. in GenBank no. AF356697 at nt 926 and 3943). To evaluate the genomic distribution of the proviruses among the breeds, we produced γ2 locus-specific signal patterns. PERV γ2 env and LTR sequences have not been well defined to date and therefore the γ2-specific pro/pol fragment was chosen as a probe. EcoRV and PstI showed different restriction sites in AF356697, AX111980 and CR450381γ2 and were used for the digestion of genomic DNA. Analysis of the five breeds Duroc, Landrace, Mangaliza, Munich Troll and SchwaebischHaellisch showed similar signal patterns among the breeds (not shown).

PERV γ2 expression
Expression analysis of PERV γ2 pro/pol sequences was carried out in various tissues of pregnant Landrace animals and their fetuses. PERV γ2-specific primers were designed to amplify the cDNA of the ORF containing sequences g67x, g68x (GenBank no. DQ084473), g69x, g612x and g618x (GenBank no. DQ084479). Use of the primer pair f37 and r840 resulted in detection of the expected 804 nt product (Fig. 2). Sequencing of 24 RT-PCR clones confirmed the PERV γ2 origin. In addition, smaller fragments were obtained in several tissues. These 682 nt transcripts were found to be identical to p118 by sequence analysis. Use of the primer pair f191 and r840 gave analogous results (not shown). The results of the semi-quantitative PCR were validated by repeating the cDNA preparation three times.

(37K): Fig. 2. RT-PCR analysis of PERV γ2 pro/pol sequences. Dilutions of PERV γ2 PCR products were used for the semi-quantitative determination of expression levels (g2, Dil.). M, Pregnant Landrace pig; M3, fetus of M; W1, fetus of Landrace pig W. cDNA was generated with the primer rna-tag and the subsequent PCR was carried out by using the primers f37 and r840 (g2, cDNA). Absence of contaminating genomic DNA was shown by the use of non-transcribed RNA as template (g2, RNA). Amplification of an actin gene fragment indicated RNA integrity (act, cDNA). H2O was used as a negative control. Amplification products for γ2 and actin were isolated, cloned and sequenced to verify the origin of the RT-PCR products. The lengths of the γ2-specific amplification products are indicated. A 1 kb DNA ladder (Invitrogen) was used as a molecular mass marker.

The mode of γ2 expression was examined in various tissues of seven animals. Differences were observed among non-related as well as among related animals. Compared with fetal tissues, lower expression levels were observed in pregnant pigs. No tissue was found to be negative in all animals; however, low abundance generally occurred in the heart and blood. A more diverse expression pattern was found in other organs, where the expression level was not consistent in the tested animals. The maximum expression level of about 10 000 copies (µg total RNA)¹ was found in the spleen and liver (Table 2).

Table 2. Tissue-specific PERV γ2 expression levels Samples were examined from two pregnant Landrace sows (M, W) and their fetuses (M1, M2, M3, W1, W2). Semi-quantitative RT-PCR was done with primers f37 and r840, resulting in an 804 nt pro/pol product.Minimal detection level: 100 copies (µg total RNA)1. Positive PERV γ2 results were classified to three expression levels (+ to +++). The analysis for M, M3 and W1 was carried out in triplicate and gave analogous results. ND, Not determined.

To examine the presence of full-length transcripts, we carried out a long-range PCR on cDNA generated by using the primer rna-tag. We analysed expression in the liver of fetus M1 and in the thymus of fetus M2, which showed high PERV γ2 expression. The primer pairs f37/cdna-tag and f191/cdna-tag failed to produce a visible product. However, a semi-nested PCR using dilutions of the f37/cdna-tag PCR products as template for a subsequent PCR with the primers f191 and cdna-tag resulted in multiple bands (Fig. 3a). The separated PCR products were blotted on to a nylon membrane and hybridized with the PERV γ2-specific probe used for colony and Southern blot hybridization. In the liver and thymus, several signals occurred. In addition to strong 12 kb signals, a faint 5 kb signal was observed (Fig. 3b).

(74K): Fig. 3. Expression analysis of PERV γ2 full-length transcripts. (a) cDNA from liver of fetus M1 (Li) and thymus of fetus M2 (Th) was generated by using primer rna-tag and amplified by PCR using the primer pairs f37/cdna-tag or f191/cdna-tag. In addition, semi-nested PCR using dilutions of the f37/cdna-tag PCR products as template for the subsequent PCR with primers f191 and cdna-tag was carried out. A 1 kb DNA ladder (Invitrogen) was used as a molecular mass marker. (b) Electrophoresed RT-PCR products were blotted and hybridized with the radiolabelled 941 nt PERV γ2-specific probe from clone m211. Arrows indicate the approximate lengths of the fragments, including a faint 5 kb signal.

PERV γ2 full-length sequences
Eight entries from GenBank were found to contain PERV γ2 sequences outside the investigated pro/pol region. In the phylogenetic analysis of the 1 kb pro/pol fragment, four (GenBank nos AF356698, AX111980, AX175459 and AX175460), three (GenBank nos AF356697, AX175461 and NC_003059) and one (GenBank no. CR450381) of these were classified as haplotypes. In the adjoining parts of the sequence, they differed in not more than 3 nt out of 7 kb within a haplotype. Compared with GenBank no. AF356698, the 10 753 nt sequence AX111980 had two mismatches in the overlapping 7153 nt region and extended the PERV γ2 sequence at the 5' end by 84 nt. The upstream sequence was not of PERV γ2 origin; thus, the AX111980 provirus lacked 967 nt at the 5' end. The 120 kb clone PigE-108A11 (GenBank no. CR450381) was derived from a Large WhitexMeishan offspring. The 6298 nt 5' end of CR450381 appeared to be a PERV γ2 sequence in reverse orientation and was therefore designated CR450381γ2. The ORF containing the 941 nt clone g618x was identical to the corresponding CR450381γ2 sequence. The BAC clone contained the 5' LTR and the gag and pro/pol genes, but lacked most of the env gene. Thus, we found three different haplotypes of PERV γ2 revealing large parts of the respective provirus. The alignment of the three sequences with GenBank nos AF356697, AX111980 and CR450381γ2 showed deleterious mutations (gaps, insertions and stop codons) in each sequence, but indicated a lower number of mutations in the gag and pro/pol genes of CR450381γ2. The largest ORF-containing fragment included the region from nt 2683 to 3900 (Fig. 4).

(8K): Fig. 4. Alignment of the PERV γ2 sequences AF356697, AX111980 and CR450381γ2. LTRs and retroviral genes are depicted according to Mang et al. (2001). AX111980 lacked the LTRs and most of the untranslated part of gag at the 5' end. The env gene of CR450381γ2 has not been determined (dashed line). The comparison revealed multiple deleterious mutations in the three proviruses. In-frame ATG codons for methionine (M), stop codons (*) and frame-shift mutations as a result of gaps (G) or insertions (I) are indicated. AF356697 and AX111980 lacked the stop codon separating gag and pro/pol due to an identical gap at nt 2689.

Based on the molecular-clock hypothesis, the age of proviruses in the host genome was estimated by sequence comparison of the proviral LTRs. Both LTRs of a provirus evolve independently after reverse transcription and integration into the host genome. Until now, GenBank no. AF356697 is the only PERV γ2 provirus containing both LTRs differing in 2 out of 434 nt. In the absence of selective pressure, the age of proviruses depends on the neutral mutation rate in the genome and the generation time. For PERV γ1, a mutation rate of 2·3x10⁹5·0x10⁹ substitutions nt¹ year¹ has been calculated (Tönjes & Niebert, 2003). This proposed mutation rate resulted in an age of 100 000200 000 years for the sequence with GenBank no. AF356697.

Analysis of PERV γ2 Env sequences
GenBank sequences AF356697 and AX111980 contained the complete env gene, which is crucial for retroviral host tropism. We compared the published PERV γ2 env sequences with other retroviral env genes at the amino acid level. A BLAST conserved-domain database search (Marchler-Bauer et al., 2005) revealed the pfam00429.11 database (Bateman et al., 2004), which contains the most diverse retroviral Env proteins. We included the Env protein of the PERV γ2 proviruses AF356697 and AX111980, as well as those from representatives of the three subfamilies A, B and C of PERV γ1 (GenBank nos AJ279056, AY099324 and AF038600; Klymiuk et al., 2003a) in the alignment of the dataset. Furthermore, the Env protein of HERV E (GenBank no. M10976), HERV R (GenBank no. M12140), human syncytin (GenBank no. NG_004112), the endogenous human chronic myeloid leukemia virus (HCML; GenBank no. AF499232) and the murine leukemia virus (MLV; GenBank no. NC_001501) were included. M10976, a multiple defective HERV E (HERV 4-1), was used for phylogenetic analysis by Mang et al. (2001). M12140 was obtained from a disrupted provirus named HERV R or ERV3 in different studies (Andersson et al., 2002; de Parseval et al., 2003) and NG_004112 encodes human syncytin, a gene of proviral origin also named HERV W, which is involved in placental morphogenesis (Blond et al., 2000; Mi et al., 2000). The intact HERV sequence AF499232 was found to be the closest relative of PERV γ2 in the BLAST search and MLV (NC_001501) has been used previously to assign the Env domains (Bénit et al., 2001; Rothenberg et al., 2001).

The 19 sequences were aligned and resulted in an 806 aa alignment. Phylogenetic trees were based on all conserved regions spanning aa 70107, 194220, 327357, 398451 and 537782 and revealed the PERV γ2 sequences to be related most closely to HCML, HERV E and HERV R. Three other branches contained the other retroviruses. They were clearly separated from the γ2-containing branch by the most-parsimony method, as well as by the genetic-distance method. In addition, five separate phylogenetic trees based on defined highly conserved regions (aa 70107, 327357, 612642, 656702 and 754782) of the Env protein confirmed the same branching (Fig. 5a). However, the γ2 sequences had lengths of 489 and 511 aa, whereas the HERV sequences contained 664672 aa. To understand this difference, we aligned γ2 Env with its closest relatives. Alignment of the five sequences resulted in a dataset of 686 positions. Identity as well as similarity analysis of the amino acids was carried out by defining similarity as positive scores in the PAM120 scoring matrix. A large 123 aa gap from aa 136 to 258 in the alignment explained the different lengths, whereas the rest of the alignment was conserved among the sequences (Fig. 5b). The additional alignment of the thoroughly investigated MLV Env (Rothenberg et al., 2001) localized the gap in the receptor-binding domain (RBD) of the PERV γ2 env gene.

(59K): Fig. 5. Comparative analysis of the PERV γ2 Env sequence of AF356697 and AX111980 with those of retroviral Env proteins. (a) The organization of the Env protein with the surface unit (SU), transmembrane region (TM) and receptor-binding domain (RBD, aa 108253) is shown at the top and the conserved regions (aa 70107, 194220, 327357, 398451 and 537782) are depicted in shaded boxes. One hundred bootstrapped most-parsimony trees branched PERV γ2 with HCML, HERV E and HERV R. Separate trees were generated from the five highly conserved regions IV (aa 70107, 327357, 612642, 656702, 754782) and confirmed the branching of PERV γ2 Env with HCML, HERV E and HERV R throughout the env gene. The other sequences of the alignment were used as the outgroup. Bootstrap values higher than 50 % are depicted. The viruses and accession numbers of the sequences are as follows: BoLV, Bovine leukemia virus (GenBank no. M35242); FeERV, feline ERV (GenBank no. X51929); FSFFV, Friend spleen focus-forming virus (GenBank no. K00021); GaLV, Gibbon ape leukemia virus (GenBank no. M26927); HCML, human chronic myeloid leukemia virus (GenBank no. AF499232); HERV E (GenBank no. M10976); HERV R (GenBank no. M12140); HERV W (GenBank no. NG_004112); HTLV 1, human T-lymphotropic virus 1 (GenBank no. J02029); HTLV 2, (GenBank no. M10060); MLV, Murine leukemia virus (GenBank no. NC_001501); MPMV, MasonPfizer monkey virus (GenBank no. M12349); PERV γ1A (GenBank no. AJ279056); PERV γ1B (GenBank no. AY099324); PERV γ1C (GenBank no. AF038600); REV, reticuloendotheliosis virus (GenBank no. X01455); SMRV, simian sarcoma virus (GenBank no. M23385). (b) Alignment of686 aa for the closely related sequences AF356697, AX111980, HCML, HERV E and HERV R. Identical amino acids (dark-shaded boxes) and similar amino acids (light-shaded boxes, based on the PAM120 similarity matrix) are highlighted.

Analysis of PERV γ2 pro/pol sequences revealed a number of ORF-containing clones, as well as recombined clones. We found five ORF-containing clones, which could be divided into two γ2 haplotypes. One haplotype matched the 6298 nt sequence CR450381γ2, showing multiple mutations outside the investigated pro/pol region. In total, we identified 21 haplotypes as defective, which was almost half of the number of PERV γ2 loci in the porcine genome. Thus, the presence of intact PERV γ2 proviruses in the pig genome cannot be ruled out, although their number certainly is low.

With regard to retroviral recombination, the observation of hybrid PERV γ2 pro/pol clones (14 %, n=5) was in accordance with our findings in BAC clones of the PERV γ1 family (Klymiuk et al., 2003a). Our previous study on ovine ERVs using the same PCR amplification technique failed to detect retroviral recombination in sheep, showing that the method does not result in artificial recombination per se (Klymiuk et al., 2003b). Retroviral recombination occurs favourably between similar sequences, but has also been found between distantly related sequences (Mikkelsen & Pedersen, 2000; Negroni & Buc, 2001). Chimeric ERVs consisting of B/D-type pro/pol and C-type env sequences have been described (Huder et al., 2002).

Contrary to the establishment of the three subfamilies A, B and C in the PERV γ1 family, which are discriminated by the phylogeny of env as well as of the pro/pol gene (Klymiuk et al., 2003a), PERV γ2 pro/pol revealed a bush-like evolution without defined subfamilies. Whether the PERV γ2 env genes encode different host tropisms will need to be addressed in further analyses. Comparison of the two published PERV γ2 Env protein sequences with related retroviral Env sequences showed the absence in both clones of large parts of the RBD, which is essential for host-cell infection. These results strongly indicate diminished infectious potential of PERV γ2, although transcriptional activity was shown for several proviral loci.

Phylogenetic analysis of PERV γ2 included the definition of retroviral haplotypes and the identification of recombined sequences. This method may also help our understanding of the evolution of other multi-copy ERV families, such as the HERV families, which are present in the human genome at up to 1000 copies each (de Parseval et al., 2003; Tristem, 2000). The phylogeny of the PERV γ2 pro/pol sequences supported the propagation of the proviruses in the porcine genome by a single round of (re)infection. A significantly lower γ2 copy number has been revealed in the ancestor of the domestic pig, the wild boar (Mang et al., 2001). The age of γ2 was estimated to be 100 000200 000 years by using the LTR of GenBank accession no. AF356697 and a similar γ2 signal pattern was found in the tested breeds. Therefore, the data indicate an accumulation event in the time period following pig domestication, but before the establishment of modern pig breeds. Based on comparisons of the LTRs, integration of PERV γ2 presumably took place after the establishment of the PERV γ1 family. However, this is in contrast to the occurrence of multiple mutations in the proviruses. The additional finding of an identical 132 nt gap in three independent haplotypes, as well as the partial absence of the RBD in the two env genes examined, may suggest the hypothesis of co-infection, but leaves the mechanism of the accumulation of PERV γ2 in the porcine genome unclear.

Expression analysis of PERV γ2 in seven individuals showed an inconsistent pattern. Under the influence of environmental factors, the diverse genetic background of the animals may explain the varying presence of γ2 transcripts in our samples. This aspect also needs to be examined further in PERV γ1. Overall, our analysis of the PERV γ2 family at the genetic and expression levels revealed novel findings for PERV biology. Our data did not indicate an obvious infectious risk of PERV γ2 for the implementation of xenotransplantation.

Footnotes

†Present address: Lehrstuhl für Molekulare Tierzucht und Biotechnologie, Moorversuchsgut, Hackerstraße 27, D-85764 Oberschleißheim, Germany.