The present study was designed to confirm the report of Newkirk et al
. They reported that, among the sera of 100 normal healthy adults (50 CMV seropositive and 50 CMV seronegative), 54% contained antibodies to RNP, 50% contained antibodies to Sm, and 33% contained antibodies to U1-70 kD [14
Newkirk et al
. also observed that the frequency of autoantibodies to each of the antigens occurred more frequently among CMV seropositive subjects than among CMV seronegative subjects[14
]. For CMV seropositive subjects, they observed that 42 (84%) subjects had antibodies to RNP, 32 (64%) had antibodies to Sm, and 23 (46%) had antibodies to U1-70 kD [14
]. If Newkirk et al
. used a negative/positive cutoff value of the mean plus three standard deviations then, overall, less than 10% of their sera contained autoantibodies.
We could not reproduce the data of Newkirk et al
. The subjects in the study of Newkirk et al
. were similar to our subjects; 80% female and 98% Caucasian. Although there are only a few published reports on the frequency of these antibodies in normal populations, those published reports all find a frequency of between 0 and 3%, similar to those reported in the present study [23
]. One study of over 1000 healthy pregnant and nonpregnant Israeli women found that none had IgG antibodies to either Sm or RNP. IgM antibodies, however, were detected in 4% or less of subjects. Patients with autoimmune disease have predominantly IgG antibodies to Sm and to RNP, and to a lesser extent IgM antibodies, whereas patients with inactive autoimmune disease are most likely to have IgM antibodies to these antigens [28
]. Both the present study and that of Newkirk et al
. measured IgG antibodies to these nuclear antigens.
Several factors may account for the difference between our results and those of Newkirk et al
. Differences in assay methods or antigens could be important. This is suggested by the fact that the mean OD (>0.5) observed by Newkirk et al
. in their Sm and RNP EIA assays was significantly higher than the mean OD (<0.15) observed in the present study. Another possibility relates to the negative/positive cutoff value used. For all three antigens, Newkirk et al
. used EIA assays and established their negative/positive cutoff value using the mean plus two standard deviations of 15 CMV seronegative sera [14
]. This appears to have resulted in a negative/positive cutoff value significantly lower than that observed in the present study using either the manufacturer's recommended cutoff value or our own cutoff value established with the 40 seronegative sera. To detect antibodies to U1-70 kD, Newkirk et al
. used only an EIA assay. Using the EIA assay, we found only one of 104 sera contained antibodies to this protein.
Another factor that may account for the difference between our results and those of Newkirk et al
. is the prevalence of the HLA antigen DR4. This HLA type occurs among 60% of patients with autoimmune disease and antibodies to U1-70 kD, but its prevalence in the normal healthy individuals is only about 25% [16
]. Hence, if the association between HLA DR4 and the presence of antibodies to U1-70 kD exists for healthy individuals and if, due to selection bias, our population contained very few (<4%) DR4-positive individuals and the population of Newkirk et al
. contained a very high (≥ 50%) prevalence of DR4-positive subjects, this could account for the observed differences. This possibility, however, seems very improbable.
In another study, Newkirk and coworkers also observed that a recombinant gB vaccine, which expressed the gB protein of the Towne vaccine, induced antibodies to CMV gB when administered to mice, suggesting that CMV gB induces antibodies crossreactive to U1-70 kD [13
]. If this is the case, it predicts a correlation between levels of antibodies to gB and U1-70 kD in sera. In humans, neither the present study or that of Newkirk and colleagues [13
] found such a correlation. This indicates that either there is no such crossreactivity or that, if it exists, it occurs very infrequently or only to a few epitopes. It is also possible that the mice Newkirk and coworkers used were genetically primed to produce autoantibodies in response to this antigen.
Whether viruses cause autoimmune disease is controversial. If they do cause disease, several mechanisms may explain the association between viruses and autoimmune disease. To stimulate a complete autoimmune response, two signals (one antigen specific and one not antigen specific), are necessary [31
]. The best described antigen-specific mechanism is molecular mimicry, whereby some component of the offending virus resembles the host structure on a molecular level, thus providing the template for antibody formation that may crossreact with self-antigen. Several of the nonantigen-specific signals include costimulatory cell surface markers as well as the generation of a multitude of cytokines. Theoretically, viruses may play a role in eliciting either or both of these signals.
Infection with CMV is ubiquitous within the human population, and nearly 100% of humans eventually acquire a CMV infection. On the contrary, autoimmune disease is relatively rare, occurring in less than 5% of the population. If CMV was a frequent inducer of autoantibodies, and by implication an autoimmune disease, both the frequency of autoantibodies in disease-free individuals and the incidence of autoimmune disease in the general population would be much higher than observed by other workers and ourselves. It is not excluded, however, that a low frequency of these three autoantibodies may be infrequently but significantly associated with CMV infection. To establish this will require testing of a large number of sera. For example, testing of nearly 700 sera will be required to determine whether an autoantibody frequency of 5% among CMV seropositive individuals and of 1% among CMV seronegative individuals is a significant difference.