When studying HERV RNA expression in human diseases, it seems important to clearly dissect from which genomic HERV loci the detected HERV RNA transcripts originate . Consistent with previous findings suggesting that expression of HERV transcripts is a ubiquitious phenomenon occurring in every human tissue [34, 45, 46], we herein show that at least seven HERV-W env loci are transcribed in PBMC from patients with MS and healthy controls. Since the primers used in this investigation only amplify a limited number of genomic HERV-W env elements our study is not exhaustive, and it seems rather likely that more than seven HERV-W env elements are transcriptionally active in human PBMC. Additionally, HERV-W env loci that are transcribed at very low rates could be missed in the cloning procedure unless much higher numbers of clones are generated. Three of the transcribed HERV-W env elements (15q21.3, Xq22.3, 17q12) identified in this study were previously found to be expressed in human PBMC by a cloning and sequencing approach [38, 39]. Assignment of cDNAs to genomic HERV-W env loci in the former investigations was based on rather short sequences (30 bp, excluding primers), containing only few informative nucleotides, that is, nucleotides that are exclusively present in a single genomic HERV-W env locus and thereby allow unambiguous assignment of cDNAs. Usage of a ~600 bp sequence (excluding primers) in the present work resulted in a higher number of informative nucleotides and thus strengthened the accuracy of the assignment. Our finding of ERVWE1 transcripts in human PBMC is consistent with previous observations [19, 20] and corroborates that although ERVWE1 expression is most abundant in placenta this locus is transcribed in non-placental tissues as well .
Several studies have analyzed expression of HERV-W env RNA in PBMC or brain tissue from patients with MS [18–21]. Lack of systematic cloning, sequencing, and assignment of cDNA sequences to genomic HERV-W env loci have impaired the exact identification of transcriptionally active genomic HERV-W env loci responsible for the observed HERV-W env RNA expression in these investigations. Whereas in the present detailed analysis we could identify distinct transcriptionally active HERV-W env loci, we did not observe significant differences in the transcriptional activity of those loci in PBMC from patients with MS versus healthy controls. Although the number of individuals studied was rather small, these data argue against a dysregulated transcription pattern of HERV-W env in PBMC from patients with MS. In contrast, a consistent finding of former investigations was a significantly higher global HERV-W env RNA expression in brain tissue from patients with MS as compared to brain tissue from patients with other neurological diseases or normal brain tissue [18–21]. Using the methodological approach of the present work, it will therefore be interesting to identify the HERV-W env elements underlying upregulated HERV-W env RNA expression in MS brain tissue.
Antony and coworkers addressed this question by designing primers that specifically amplify HERV-W env 7q21.2 (ERVWE1) and the MSRV env clone AF123882 , which, as shown by our analyses, corresponds to a HERV-W env element on chromosome 15q21.3. These authors also employed a pair of degenerate (HERV-Wdeg) env primers that were based on the MSRV env clone AF331500, which, again as shown in this work, corresponds to a recombined cDNA originating from HERV-W env loci on Xq22.3 and 5p12. According to the Antony et al. study, elevated HERV-W env RNA expression in MS brain tissue originates mainly from the HERV-W env elements amplified by the HERV-Wdeg
env primers and (somewhat less) from HERV-W env 7q21.2, while HERV-W env 15q21.3 expression was similar in patients and controls . A BLAT-PCR search showed that the HERV-Wdeg
env primer pair potentially amplifies at least three genomic HERV-W env loci, among them HERV-W env Xq22.3. It is thus tempting to speculate that HERV-W env Xq22.3 may significantly contribute to increased HERV-W env RNA expression in MS brain tissue. Again, using the methods described herein, this issue could be resolved in a straightforward manner.
Remarkably, we observed a high number (29.8%) of recombined sequences among the analyzed HERV-W env cDNAs. As detailed in a previous study on transcribed HERV-K(HML-2) sequences , those recombinant cDNA sequences very likely resulted from in vitro recombinations that were due to template switches of reverse transcriptase during cDNA synthesis and/or PCR-mediated recombinations. Both of these mechanisms are well-recognized and have been proven experimentally to produce chimeric sequences [41, 47–53]. The percentage of recombined sequences detected in the present study was higher than that in the study on HERV-K(HML-2) in which ~5% of recombined clones were observed . This is most likely explained by the fact that in the HERV-K(HML-2) study only cDNA sequences with more than 17 nucleotide mismatches to the best matching locus were analyzed for recombinations, whereas in the present work all cDNA sequences were scrutinized for recombinations. Altogether, our data indicate that during experimental studies of repetitive elements by RT-PCR, in vitro recombinations are relatively common and almost inevitable complications.
An important result of this investigation is that previously published MSRV env and gag sequences appear to either be derived from transcripts of specific genomic HERV-W elements or to result from recombinations among such transcripts (Table 3). Given the high frequency of in vitro recombinations between transcripts from different HERV loci observed in this and the study by Flockerzi et al. , and given that MSRV clones were generated by methodologically similar approaches, it seems possible that the recombined MSRV env sequences (AF127229, AF331500) have resulted from in vitro recombinations as well.
An alternative explanation is that the recombined MSRV env sequences, and the recombined HERV-W env sequences isolated in this study, originated from novel, recombined, genomic HERV-W insertions. Hypothetically, such insertions could have formed in vivo after recombination of RNA transcripts from different HERV-W env loci through template switches during reverse transcription. Although we cannot formally exclude this possibility, a number of points argue against it. First, all known HERV-W elements are defective and replication-incompetent [8, 14]. Therefore, HERV-W is a priori rather unlikely to have the capacity to form new insertions in human DNA. Second, if there were novel recombined HERV-W loci in human DNA, one would expect to repeatedly observe defined recombined sequences originating from such insertions. However, this is neither the case with the 99 recombined HERV-W env cDNA sequences analyzed in this study nor with the published MSRV env clones. Third, given that about 30% of HERV-W env and 33% (2 of 6) of the investigated MSRV sequences represent recombinants, if all these recombinant MSRV/HERV-W env sequences were derived from novel proviral insertions, formation of such novel insertions would be an astonishingly frequent event. It seems very unlikely that as many recombined HERV-W loci should have been overlooked in previous genome sequencing projects.
Collectively, the most plausible and simplest explanation for the origin of MSRV env and gag sequences seems to be that those sequences originate from RNA transcripts from various endogenous HERV-W loci, or from in vitro recombinations among them. All of the HERV-W loci from which MSRV sequences are derived are defective and except for the 5p12 HERV-W env element, all of those loci resemble processed HERV-W pseudogenes. The human genome sequence was not yet available when MSRV was described, which hampered the identification of the precise origin of MSRV sequences at that time. It was, however, noted that those sequences cannot be attributed to a single replication-competent genome . Nevertheless, the nature of MSRV was subsequently controversial, and it has been speculated that MSRV could be an exogenous, replication-competent retrovirus [6, 30–32]. In contrast, our present data clearly suggest that the published MSRV env and gag RNA sequences are not derived from the genome of a currently replication-competent exogenous retrovirus. In the light of these results and previous observations of an increased prevalence of MSRV pol transcripts in plasma from patients with MS as compared to healthy controls [54, 55], it may similarly be interesting to analyze which HERV-W pol elements those MSRV pol transcripts could be derived from.
Although our findings argue against MSRV being an autonomous retroviral entity, they do by no means rule out that individual HERV-W env loci that correspond to MSRV sequences, or the Syncytin-1 (ERVWE1) gene, could be of relevance in MS. Indeed, we show that two MSRV env clones (AF331500, AF127228), which have been instrumental for the characterization of proinflammatory effects of MSRV Env  and the generation of a monoclonal anti-MSRV/HERV-W Env antibody (6A2B2) , are derived from a HERV-W env locus on chromosome Xq22.3. This locus is highly remarkable as it is interrupted by only a single premature stop at codon position 39 and otherwise harbors a long ORF for a N-terminally truncated 475 amino acid HERV-W Env protein (Figure 6). Bonnaud and colleagues described frameshift insertions/deletions (indels), that is, indels whose length is not a multiple of three, in 33 out of 36 analyzed genomic HERV-W env loci. Interestingly, among the three loci without frameshift indels were the ERVWE1 gene and the Xq22.3 HERV-W env element . We further note that the 475 amino acid Xq22.3 HERV-W env ORF is also present in the orthologous locus in chimpanzees (data not shown). These data may be taken as hints that selective pressure could act on the Xq22.3 HERV-W env locus, raising the possibility that Xq22.3 HERV-W env could exert a biological function. Our finding that the Xq22.3 HERV-W env locus is transcriptionally active in human cells indicates that it fulfills at least one essential prerequisite for a protein expression capacity in vivo.
Neuropathological studies revealed that the 6A2B2 anti-HERV-W Env antibody reacts with an antigen that is strongly expressed by glial cells in MS brain lesions, but not in normal control brain tissue [18, 21, 22]. Because Syncytin-1 has been thought to be the only HERV-W env locus capable of producing a HERV-W Env protein, and because 6A2B2 may crossreact with Syncytin-1 [16, 43], the antigen detected by 6A2B2 in MS brain lesions was considered to be Syncytin-1. However, our analyses show that the protein against which the 6A2B2 antibody was raised is practically identical to the Xq22.3 HERV-W Env protein (Figure 6) . We meanwhile cloned Xq22.3 HERV-W env into a eukaryotic expression vector. Transient transfection of HeLa cells with this clone showed that the Xq22.3 HERV-W env has retained a coding capacity and can produce a HERV-W Env protein in vitro which is detected by the 6A2B2 antibody in immunocytochemistry and immunoblots (C. Crusius, S. Wahl, K. Ruprecht, manuscript in preparation). These data suggest that the antigen recognized by 6A2B2 in MS lesions could likewise originate from the Xq22.3 HERV-W env locus, provided that this locus has a protein expression capacity in vivo. More elaborate studies will be required to clarify the exact nature of the HERV-W Env protein detected in MS lesions. Further characterization of the putative Xq22.3-encoded HERV-W Env protein, especially in comparison to Syncytin-1, will be necessary for such clarification.