Insulin-dependent diabetes mellitus (IDDM) is a β cell–specific, T cell–mediated autoimmune disease. Last year, Bernard Conrad and coworkers proposed that the β cell destruction is caused by the systemic activation of autoreactive T cells due to the expression of a superantigen encoded by an endogenous retrovirus (HERV-K) on the surface of antigen-presenting cells (2Conrad B. Weissmahr R.N. Böni J. Arcari R. Schüpbach J. Mach B Cell. 1997; 90: 303-313Abstract Full Text Full Text PDF PubMed Scopus (355) Google Scholar). The search for this putative superantigen had been initiated by the observation of a pancreatic enrichment of the Vβ7 family of T cell receptor chains in two patients. Here we present findings contradicting this intriguing hypothesis. HERV-K is an endogenous retrovirus family of old world monkeys and humans with homologies to type B and D retroviruses. Within the approximately 50 proviral copies of this HERV family only very few of the full-length genomes (designated type 2) have retained colinear coding capacity. As this HERV family codes for the human teratocarcinoma-derived retrovirus (HTDV), it was tentatively designated HTDV/HERV-K to discriminate it from other MMTV-related HERV families with primer-binding sites for tRNA lysine (HERV-K). A subclass of HTDV/HERV-K proviruses (type 1) is characterized by a deletion affecting the C terminus of the integrase domain and the amino terminus of the envelope (Env), the signal peptide region (reviewed by 6Löwer R. Löwer J. Kurth R Proc. Natl. Acad. Sci. USA. 1996; 93: 5177-5184Crossref PubMed Scopus (611) Google Scholar). The deletion also eliminates the first exon of the corf gene, an accessory gene that enhances the expression of structural proteins. This function is lacking in cell types in which only type 1 proviruses are expressed. In addition, many type 1 proviruses do not exhibit env splicing (5Löwer R. Tönjes R.R. Korbmacher C. Kurth R. Löwer J Virology. 1995; 69: 141-149Google Scholar). Most of the HTDV/HERV-K proviruses carry stop codons in their env genes preventing expression of functional Env proteins. Sequence comparison clearly shows that the truncated IDDMK1,222 env sequence identified by Conrad et al. stems from one of these proviruses. It is a type 1 element similar to the proviruses K10 and K18 described previously (7Ono M J. Virol. 1986; 58: 937-944PubMed Google Scholar). The sequence information available for IDDMK1,2 shows a 3% difference to K10 in env and in the long terminal repeats (LTRs), the most variable regions of a virus, and a near identity to the LTRs of K18. In summary Conrad et al. have identified one of the proviruses of a well-studied HERV family. In human teratocarcinoma cell lines, it is easy to detect virus particles by electron microscopy or viral transcripts by Northern blots (Figure 1) although the level of expression is significantly lower than that seen with exogenous retroviruses. To measure reverse transcriptase activity, concentration of particles by ultracentrifugation is required for even the most sensitive of assays. A simple explanation may be that many HTDV/HERV-K proviral RT genes do not code for enzymatically active proteins when studied in recombinant expression systems (11Tönjes R.R. Löwer R. Boller K. Denner J. Hasenmaier B. Kirsch H. König H. Korbmacher C. Limbach C. Lugert R. et al.J. Acquir. Immune Defic. Syndr. Hum. Retrovirol. 1996; 13: S261-S267Crossref PubMed Scopus (78) Google Scholar). HTDV/HERV-K sequences only are detected in RNA isolated from HTDV particle preparations (10Simpson G.R. Patience C. Löwer R. Tönjes R.R. Moore H.D.M. Weiss R.A. Boyd M.T Virology. 1996; 222: 451-456Crossref PubMed Scopus (67) Google Scholar). Although no evidence for the physical presence of particles or of virus transcripts is presented, Conrad et al. also describe a low level of RT activity in the supernatant of cell lines derived from IDDM islets with a similar extremely sensitive test system (8Pyra H. Böni J. Schüpbach J Proc. Natl. Acad. Sci. USA. 1994; 91: 1544-1548Crossref PubMed Scopus (184) Google Scholar). However, using a modified version (14Weissmahr R.N. Böni J. Schüpbach J J. Virol. 1997; 71: 3005-3012PubMed Google Scholar) of the RU5-PCR approach published previously (4Löwer R. Löwer J. Tondera-Koch C. Kurth R Virology. 1993; 192: 501-511Crossref PubMed Scopus (69) Google Scholar), they surprisingly amplified not only HTDV/HERV-K sequences in the IDDM particle fractions but also a variety of other HERV-K families such as an element identified in the complement (C4) locus. If the 100-fold increase in RT activity in the islet supernatant described by Conrad et al. had originated from the presence of viral particles, one would have expected a similar accumulation of the matching viral genome as in HTDV particles. The lack of such a specific enrichment may indicate that the collected material, although banding at the density of a retrovirus, merely contains polyribosomes or nucleoproteins associated with a variety of diverse endogenous retrovirus transcripts. In addition, there is no real proof that an IDDMK1,222 provirus exists at all. In the multistep, nested PCR approach to clone IDDMK1,222, only the primer in the R region of the 3′LTR used to amplify the 3′ part of the genome is specific for IDDMK1,222- or K18-like proviruses. The 5′ part may be derived from any other provirus as the primer will amplify many HTDV/HERV-K sequences. It is not clear from the publication whether a full-length cDNA had been the template for IDDMK1,222 or whether this sequence was generated using the overlapping 5′ and 3′ amplicons as templates. The latter procedure most probably would have resulted in a mosaic element. As no controls for the RT-PCR reactions are shown, the template for the entire procedure may also have been contaminating DNA. Thus, IDDMK1,222 may be identical to the K18 provirus or it may turn out to be a composite sequence due to the cloning procedure. At the moment, only portions of the sequence are published or deposited in the databases. Even if the IDDMK1,222 provirus would code for an RT protein, an enzymatic activity remains speculative until demonstrated. In summary, the data presented by no means exclude that the RT activity detected in the IDDM supernatants originated from HTDV/HERV-K proviruses other than IDDMK1,222 or from other HERV families. LINE elements expressed in T cell lines are another likely source for the RT activity. Taking precautions not to amplify contaminating DNA, we repeated the epidemiological study of plasma RNA levels in a panel of diabetes patients and healthy blood donors with the primer described by Conrad et al. (located exclusively in the 3′LTR) and with HERV-K-specific primer designed to amplify other regions of the genome, especially the env gene. In this study no difference was observed in the prevalence of HERV-K plasma RNA in IDDM patients (n = 14, of which 9 were of recent onset) versus healthy blood donors . The failure to assign specific HERV-K expression to diabetes patients is not unexpected as we have found that full-length and even spliced transcripts of this endogenous retrovirus family are present at low level in a variety of normal tissues including lymphocytes (see Figure 1). Taking into account that during the onset of diabetes autoreactive T cells are activated, an early rise in plasma RNA level could merely reflect increased numbers of decaying cells. However, because no differences in patients versus healthy controls were detected, an additional quantitative RT-PCR approach was not performed. The primer used by Conrad et al. for the epidemiological study are located in highly conserved parts of the U3 and R regions as judged from the many LTR sequences we have studied so far. Thus, they are not specific for the IDDMK1,222 sequence in contrast to the R primer used to clone the 3′ part of the genome. Although the action of a superantigen may not necessarily correlate with the induction of a humoral immune response to viral structural proteins, the presence of virus particles as indicated by Conrad et al. certainly would do so. More than 40 sera collected from patients with IDDM have been tested for HERV-K-specific antibodies using the following test systems: (a) immunofluorescence staining of a virus-producing teratocarcinoma cell line, (b) Western blot analyses using virus particle preparations or recombinant Env protein as antigen, and (c) ELISA reactivity against diagnostic synthetic Gag and Env peptides. All these test systems are routinely used to screen panels of sera from diverse groups of patients and healthy blood donors. Among these we usually see 3% reactivity in normal blood donors with tests (a) and (c) (1Boller K. Janssen O. Schuldes H. Tönjes R.R. Kurth R J. Virol. 1997; 71: 4581-4588PubMed Google Scholar). The same rate of reactivity is detected in IDDM patients, yet there is no concordance of results between the test systems. Using recombinant Env protein as antigen in Western blots (12Tönjes R.R. Limbach C. Löwer R. Kurth R J. Virol. 1997; 71: 2747-2756PubMed Google Scholar), we detect a higher percentage of positive reactions (20%–25%) in both IDDM patients and healthy controls. Similar results have been reported with different panels of patient sera (13Vogetseder W. Dumfahrt A. Mayersbach P. Schonitzer D. Dierich M.P K. AIDS Res. Hum. Retroviruses. 1993; 9: 687-694Crossref PubMed Scopus (42) Google Scholar). In summary, these findings are in agreement with the observation that HERV-K plasma RNA levels are similar in IDDM and healthy blood donors and raise the question whether the antibody reactions are based on the expression of HERV-K proteins or are induced by a cross-reacting antigen of unknown origin. Remarkably, the anti-Env Western blot reactivity detected in IDDM patients and healthy controls is not associated with a corresponding reactivity to Gag proteins, again arguing against the presence of virus particles. In contrast, with all the test systems used an antibody response with high titers is consistently detected in approximately 80% of patients with testicular tumors. This is probably the consequence of upregulation of HTDV/HERV-K proteins, which are easily detected in testicular tumor cell lines and in thin sections of germ cell tumors stained with polyclonal and monoclonal antisera specific for recombinant Gag, Env, and cORF proteins and patient's sera. Patients with testicular tumors always develop a strong anti-Gag immune response in addition to anti-Env antibodies (6Löwer R. Löwer J. Kurth R Proc. Natl. Acad. Sci. USA. 1996; 93: 5177-5184Crossref PubMed Scopus (611) Google Scholar; Sauter et al., 1966). Conrad et al. stated that the sequence of the truncated IDDMK1,222 env gene encoded a superantigen. Sequence comparison with a variety of additional HTDV/HERV-K env sequences revealed extensive homologies with only a few nucleic acid substitutions. The superantigen sequences encoded by mouse mammary tumor virus (MMTV) also differ between virus strains, in particular within the COOH-terminal residues that may contribute to their Vβ specificity. As the IDDMK1,222 expression clone was not available to test for superantigen function, we amplified and cloned an HTDV/HERV-K sequence using a 5′ primer corresponding to the IDDMK1,222 superantigen sequence and the 3′ primer R-poly(A) described by Conrad et al. As the 3′ R-poly(A) primer was not RNA specific, we used genomic DNA as the template. The Env sequence obtained differed from IDDMK1,222 by six amino acids. One exchange abrogated the premature Env termination signal leading to a longer putative superantigen protein; the other changes were G46→V, H48→R, T115→P, G123→E, and V140→A. Preliminary attempts to use the THP1 cell line for expression of the Env protein failed due to insufficient transfection efficiency (less than 1% transfectants with a reporter gene using different transfection methods). We therefore used a well-established murine system (3Günzburg W.H. Heinemann F. Wintersberger S. Miethke T. Wagner H. Erfle V. Salmons B Nature. 1993; 364: 154-158Crossref PubMed Scopus (41) Google Scholar) that allows for higher transfection efficiencies (30%–40% using electroporation). There is as yet no evidence that superantigens are species specific and, indeed, Conrad et al. recently reported similar results using human and murine test systems (Symposium on "Retroviruses and Autoimmunity," 1998, London). However, murine A20 cells transiently transfected with the IDDMK1,222-related expression clone did not stimulate corresponding BALB/c T cells in a Vβ-specific fashion while A20 cells transfected with an MMTV superantigen induced a Vβ14-selective expansion of T cells. Truncating the longer Env sequence to the size of IDDMK1,222 did not alter the negative result. Clearly, further studies including more HTDV/HERV-K sequences as well as the expression clone established by Conrad et al. are needed to either confirm or refute the hypothesis that a superantigen is encoded by this HERV family. If the superantigen exists, its involvement in the onset or progression of insulin-dependent diabetes mellitus needs to be examined. With regard to other bacterial and especially viral superantigens, it seems very unlikely that the sequence published by Conrad et al. would be unique amongst all HTDV/HERV-K proviruses in encoding a superantigen.

The human genome contains a wide variety of endogenous retrovirus-like sequences. The human endogenous retrovirus type K (HERV-K) family comprises 30-50 members per haploid genome in humans and is highly conserved in Old World monkeys and apes. Some proviruses are displaying open reading frames (ORF) with coding capacity for viral particles. HERV-K sequences most likely code for the previously described human teratocarcinoma-derived virus (HTDV) and correlated expression of HERV-K Gag has been demonstrated by immunoelectron microscopy studies. Protease, but not yet reverse transcriptase (RT), enzymatic activity was demonstrated for recombinant HERV-K proteins. However, an ultrasensitive RT assay revealed specific polymerase activity associated with the HTDV particles. HERV-K transcription is specifically regulated by viral long terminal repeats and RNA is expressed at low steady-state levels in a variety of human tissues and rumours. In teratocarcinoma cell lines, HERV-K is highly expressed in a complex pattern showing full-length as well as subgenomic envelope (env) and two alternatively spliced small transcripts. The doubly spliced 1.8-kb mRNA codes for cORF protein which resembles Rev of HIV-I and is located in the nucleolus. In addition, the cORF sequence acts as a leader and is essential for effective expression of glycosylated HERV-K Env protein. Although HERV-K sequences code for all necessary retroviral proteins, infectious particles could not yet be demonstrated. The putative implication of HERV sequences in pathophysiological processes, for example, testicular malignancies, remains to be elucidated.