Recombination analysis of human-tropic porcine endogenous retroviruses

Xenotransplantation of genetically modified pig tissue aims to compensate for the shortage of human donor organs. Cross-species transmission of pathogens will be a major obstacle. Use of specific-pathogen-free animals focuses the potential infectious risk to porcine endogenous retroviruses (PERV). Endogenous retroviruses are copies of exogenous retroviral genomes integrated into the germ line of the host and have been found in multiple copy number in all mammals. Intact proviruses harbour the genes gag, pro/pol and env enclosed by long terminal repeats (LTR) at both ends. Most endogenous retroviruses are defective due to deleterious mutations (Beckmann et al., 2000). Production and cross-species infection of functional PERV have been observed in in vivo experiments (Deng et al., 2000; van der Laan et al., 2000). In addition, PERV have been shown to infect human cells in vitro (Boneva et al., 2001). Xenotransplantation of tissues derived from pigs that have been genetically modified to reduce rejection of the donor organ has been suggested to enhance the risk of infection with PERV (Weiss, 1998).

PERV are classified into the retroviral β (B- or D-type) and γ (C-type) genera (van Regenmortel et al., 2000). All known human-tropic infectious PERV have been assigned to the PERV γ1 family, consisting of the subfamilies A, B and C (Patience et al., 2001). Examination of porcine cell lines and pig breeds resulted in the detection of about 50 PERV γ1 sequences, including several intact copies (Akiyoshi et al., 1998; Bartosch et al., 2002; Czauderna et al., 2000; Krach et al., 2001; Niebert et al., 2002). PERV γ1A, -B and -C are highly homologous in their gag and pro/pol retroviral genes, whereas significant differences in the envelope (env) gene explain their different host tropism (Akiyoshi et al., 1998; Le Tissier et al., 1997; Takeuchi et al., 1998). Recently, chimeric PERV γ1 sequences have been observed (Klymiuk et al., 2002; Lee et al., 2002; Oldmixon et al., 2002; Wilson et al., 2000).

To assign proviral genomic sequences to different host tropism and to evaluate the potential infectious risk of recombinant clones in xenotransplantation, we analysed full-length PERV γ1 genomes as well as complete PERV γ1 env gene sequences.

Eighteen full-length PERV γ1 genomes and 82 complete PERV γ1 env gene sequences were identified in GenBank using BLAST searches. The GenBank accession numbers of the sequences are given in the legend to Fig. 2. Comparative sequence analysis was done using CLUSTALW (Jeanmougin et al., 1998), MACCLADE () and SEQAPP (). Complete polymorphism pattern analysis was carried out by the alignment of the full-length PERV γ1 gag, pro/pol and env genes. In the data set, the gag, pro/pol and env genes span nt 11577, nt 15785167 and nt 50407074, respectively. After the removal of invariant nucleotide positions, potential recombination sites were identified in individual sequences by a change in the pattern of nucleotide polymorphism. Phylogenetic trees of the respective genome fragments were created using PHYLIP (). Recombination analysis subsequently included the nucleotide positions where only one single clone differed from the other proviruses.

(40K): Fig. 2. Schematic depiction of recombination patterns in the PERV γ1 env genes. Representative clones are given for each pattern. Asterisks (*) indicate that at least the representative harboured an ORF. Filled boxes represent PERV γ1A (shaded), -B (grey) and -C (black) sequences. Hatched boxes show fragments that were not classified clearly to PERV γ1A, -B or -C. Fragments that were not classified due to their low identity to any of the subfamilies are depicted by boxed question marks. The functional sites of env are shown at the top. PRR, proline-rich region; MSD, membrane-spanning domain. Nucleotide positions are included below the sequences. The numbers of sequences classified to the recombination patterns are given. AF417227 is representative for AF417228; AX052633 for AF130444; AF426922 for AF426920 and AF426929; AY099323 (differing in 6 of 1700 nt from AF417223) for AF038601, AF435966, AJ133817, AX002802 and Y12238 and the 19 sequences AF426917 (differing in 40 of 1700 nt from AF417223), AF426918, AF426919, AF426921, AF426923, AF426924, AF426926, AF426927, AF426928, AF426930, AF426931, AF426934, AF426941, AF426942, AF426943, AF426944, AF426945, AF507940, AJ288584; AJ288590 for AJ288586. Non-recombined PERV γ1A (AF417223) env genes are as follows: AF417222, AF417224, AF417226, AF435967, AJ279056, AJ288585 and AJ293656. Non-recombinant PERV γ1B (Y12239) env sequences are A66552, A66553, AF426916, AF426932, AF426933, AF426935, AF426937, AF426938, AF426939, AF426940, AF426946, AJ133816, AJ133818, AJ279057, AJ288588, AJ288589, AJ288592, AJ293657, AY056024, AY056025, AY056026, AY056027, AY056028, AY056035, AY099324, AX002804 and Y17013. AF038599 is a non-recombinant PERV γ1C (AF038600) env gene. Origin of sequences: MS, miniature swine; We, Westran; PK15, porcine kidney fibroblasts (ATCC CCL-33); LW, Large White; CMS, Chinese miniature swine; ns, not stated.

PERV γ1 env gene analysis started with the assignment of the different host tropisms to the specific nucleotide sequences representing the three subfamilies A, B and C. Subsequently, chimeric sequences were compared to these subfamilies. env gene fragments that showed significant sequence diversity to PERV γ1A, -B and -C were not classified to the known subfamilies. Recombinations in PERV γ1 full-length sequences
For comparative analysis, 16 full-length PERV γ1 nucleotide sequences harbouring the complete gag, pro/pol and env genes were taken from GenBank (Fig. 1). Alignments started with the first ATG codon of gag and ended with the env stop codon at nt 7059 and nt 7074 for PERV γ1A and PERV γ1B and -C, respectively. For the detection of similarities between the 16 sequences, common nucleotides were deleted and nucleotide positions where only one of the sequences showed a polymorphism (n=108) remained unconsidered. As a result, we obtained 972 polymorphic nucleotide positions (13·7 %) for further analysis. Of these, 904 (representing 83·7 % of all polymorphic nucleotides) were found to be involved in the definition of three distinct subfamilies. Patterns were assigned to PERV γ1A, -B and -C, which were defined by their different host tropism. Subsequent comparison of the polymorphic nucleotide patterns revealed the appearance of recombination events in individual sequences. Four obvious recombination sites were observed in the alignment between nt 40524113, nt 44755026, nt 50595073 and nt 67546764 (Fig. 1A). Separate phylogenetic analyses were carried out for the five fragments between the four recombination sites by the most parsimony method and strictly confirmed the classification of the clones (Fig. 1B). In addition, the same result was found in genetic distance trees showing slightly lower bootstrap values (data not shown).

(64K): Fig. 1. (A) Alignment of full-length PERV γ1 gag, pro/pol and env genes. After removal of invariant nucleotides, nucleotide positions were eliminated where only one single clone differed from the other proviruses. The number of these unique nucleotide positions is given for every clone. Polymorphic nucleotides (n=972, 13·7 % of the total nucleotides) were used for further analysis. The 904 nucleotide positions discriminating the three subfamilies PERV γ1A, -B and -C (separated by horizontal lines) are depicted in bold. Four recombination sites (rec 14) of individual sequences (vertical lines) were observed. Boxes with broken lines include the retrovirus region where the exact assignment of the recombination site was not possible. To condense the alignment, columns of identical nucleotide polymorphism patterns are merged in the region within the recombination sites. The number of the respective pattern occurring in the sequence comparison as well as the first and last nucleotide position in the alignment are shown at the top. The columns show the nucleotides that appeared first for the respective pattern. Polymorphic nucleotides are depicted in shaded boxes. ORFs for gag, pro/pol and env and the host tropism, if described, are indicated. For A66553, no infectivity assay has been described (nd). The genes gag, pro/pol and env span nt 11577, nt 15785167 and nt 50407074, respectively. AF038601 had two large gaps (nt 29293738 and nt 50175105), which are indicated by asterisks (*); question marks (?) indicate nucleotides where the GenBank entries were not defined exactly. Origin of the sequences: MS, miniature swine; Ts, Tsukuba-1 produced from the porcine malignant lymphoma-derived cell line Shimozuma-1; PK15, porcine kidney fibroblasts (ATCC CCL-33); LW, Large White; ns, not stated. (B) Separate most parsimony trees of the genomic fragments between the four recombination sites. The trees are the consensus of 100 bootstrap replicates for the first, fourth and fifth tree, and of 500 replicates for the second and third tree (due to their shorter sequence length). Bootstrap values greater than 50 % are indicated. Recombinant clones are depicted in bold and changes between subfamilies are indicated by dotted arrows.

From the five recombinant proviruses (A66552, A66553, AF038601, AJ133817 and AY099323), AF038601 and AJ133817 were cloned from intact genomes (Akiyoshi et al., 1998; Czauderna et al., 2000). AY099323 was derived from overlapping PCR fragments; however, the recombination sites observed matched within single PCR fragments (Bartosch et al., 2002). The origins of A66552 and A66553 were not investigated further. All five recombinant proviruses were classified to PERV γ1B, in both the 5' and the 3' end, whereas the intermediate sequences were PERV γ1A. The recombinant PERV γ1A fragments included the partial pro/pol gene (A66552 and A66553) as well as the 3' end of pro/pol and the major part of the env gene (AF038601, AJ133817 and AY099323), resulting in the γ1A host tropism for the replication-competent clones AJ133817 and AY099323. AJ133817 showed the same nucleotide polymorphism pattern as AF038601 and AY099323 3' of nt 5027. However, the exact assignment of the recombination site of the 5' end was not possible due to sequence variations in nt 44984975. Two of the five recombined proviruses (AJ133817 and AY099323) have been shown to be human tropic and replication competent (Bartosch et al., 2002; Krach et al., 2001), whereas another recombined locus (A66553) harboured intact ORFs for all three genes; this was not tested for its infectivity to human cells. Inclusion of the unique nucleotide positions in the recombination analysis revealed that AJ279056 showed 12 nucleotide differences in the fragment of nt 18021895 to the closest relative AJ293656 (data not shown). Due to the absence of a potential donor sequence, a recombination could not be confirmed confidently as cause for this divergence.

Two additional full-length PERV γ1A sequences (AF435966 and AF435967), which were derived from BAC clones of Large White pig genomic DNA and described previously to be replication competent upon transfection in human cells (Niebert et al., 2002), were not included in Fig. 1 due to multiple nucleotide polymorphisms in the gag and/or pro/pol genes. In addition to the high number of unique nucleotide positions (n=163 and 105, respectively, when aligned to the sequences of Fig. 1), pro/pol gene fragments of AF435966 differed from all PERV γ1 sequences on both nucleotide and amino acid sequences due to multiple frame-shift mutations. In the separate comparative analysis of both sequences, AF435967 was found to be γ1A throughout the whole sequence and AF435966 was classified to the recombined clones AF038601, AJ133817 and AY099323 with the 5' end of the intermediate γ1A fragment located in the gag gene. Due to its high polymorphism, the exact 5' end of the recombination was not investigated further (data not shown). The env genes of AF435966 and AF435967 did not show increased sequence polymorphism and, therefore, were included in the subsequent study (see below).

Recombination patterns in PERV γ1 env sequences
As the env gene is crucial for retrovirus host tropism and, therefore, determines which PERV are capable of infecting human cells, we focused on recombination events of this gene. In total, we screened 82 complete PERV γ1 env genes (Table 1), which have been submitted to GenBank by Bosch et al. (2000) (n=9), Herring et al. (2001) (n=6), Lee et al. (2002) (n=31), Oldmixon et al. (2002) (n=11) and additional groups (n=25). The env genes of the 18 PERV γ1 full-length sequences described above were included. A total of 58 fragments harboured an ORF. Designation of the env sequences to PERV γ1A, -B and -C was carried out by comparison to AF417223, Y12239 and AF038600, respectively (Akiyoshi et al., 1998; Le Tissier et al., 1997; Oldmixon et al., 2002), which showed maximal sequence diversity in the nucleotide polymorphism patterns. Of these sequences, 38 env sequences (46·3 %) were classified completely to one of the three original subfamilies (Fig. 2), whereas 44 (53·7 %) were hybrid sequences (Table 1). The hybrid sequences were classified to 15 distinct recombination patterns (Fig. 2). The recombined clones included Y12238, which has been assigned previously to γ1A.

Table 1. Comparative analysis of PERV γ1 env genes

Five env genes (AF296168, AJ288586, AJ288587, AJ288590 and AJ288591) showed sequence fragments with differences to the known PERV γ1 subfamilies (Table 2). The first 837 nt of AF296168 showed low identity to PERV γ1A, -B and -C, whereas the 3' end was classified to γ1A. Four additional sequences described to be γ1B (Bosch et al., 2000) harboured regions differing from the known PERV γ1 subfamilies. These env sequences were suggested to be derived from recombinations with retroviral genomes not yet known.

Table 2. Fragments of PERV γ1 env genes differing from γ1A, -B and -C

In conclusion, recombination events were found in full-length PERV γ1 sequences derived from in vitro studies with PK15 cells (Fig. 1) as well as in the genome of different breeds (Fig. 2). In addition to the recombined env sequences, hybrid sequences in genomic pig DNA have been observed recently for pro/pol (Klymiuk et al., 2002). Misincorporation of nucleotides during reverse transcription of the retroviral genome has been shown as the cause as well as the result of recombination processes (Mikkelsen & Pedersen, 2000). However, we found no evidence for the accumulation of polymorphic nucleotide positions around the recombination sites of the five recombinant full-length PERV γ1 clones (data not shown).

The PERV γ1 sequences analysed in this study have been derived from genomic pig DNA of cell lines and different breeds as well as from retroviruses after infection experiments. Due to the particular conditions of the in vitro experiments and the putative preference of detecting individual sequences by the different techniques used, the data may not represent exactly the real genomic PERV γ1 load in the pigs. Compared to the proposed number of 50 PERV γ1 loci in the pig genome, the high number of env genes examined in this study indicated breed-specific and/or individual sequence polymorphisms of the genomic PERV γ1 load. Concise examination of additional pig breeds may lead to the detection of further PERV recombination patterns. The appearance of a low number of PERV γ1C sequences is in accordance with previous reports (Akiyoshi et al., 1998; Bosch et al., 2000; Klymiuk et al., 2002; Le Tissier et al., 1997; Mang et al., 2001).

In the env genes, we observed a high number of different recombination patterns. Compared to the C-region of the surface subunit and to the transmembrane subunit, we found a smaller number of recombination events in the receptor-binding domain (RBD). This may be caused by increased sequence polymorphisms in the RBD between the subfamilies, as low sequence similarity reduces the rate of recombination (Negroni & Buc, 2001). On the other side, only a minor part of the clones harbouring recombinant RBD may give rise to infectious viruses and/or to developmental advantages under invariant environmental conditions. Three retroviruses with hybrid sequences in the RBD (AF417227, AF417228 and AF417229) have been shown to be human tropic and replication competent (Oldmixon et al., 2002); however, it is not clear if the recombination has influenced host tropism. Additional data on host tropism are also not available for AF296168 and AJ288587.

Having carried out the polymorphism pattern comparison, we assigned here the proviral nucleotide sequences to the different host tropism that has been described previously for the PERV γ1 proviruses. In addition, the PERV γ1 env sequences of the three subfamilies A, B and C were defined showing maximal sequence diversity in the polymorphism patterns. These results will contribute to subsequent approaches to protect the recipient from infections with PERV in xenotransplantation. Chimeric env sequences containing fragments with low identity to PERV γ1A, -B and -C indicated the potential of retroviral genomes not yet known to get involved in recombination events with unknown consequences for host tropism and pathogenicity of recombinant PERV γ1 proviruses. Recombinational patch repair resulting in new retroviral genomes has been described previously in defective retroviral genomes (Mikkelsen & Pedersen, 2000; Negroni & Buc, 2001). Although this has not been determined yet for mutant PERV γ1 sequences, the potential infectious risk cannot be ruled out for defective PERV γ1 loci.

Footnotes

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