Enterovirus RNA Is Found in Peripheral Blood Mononuclear Cells in a Majority of Type 1 Diabetic Children at Onset

The study comprises all 24 children below the age of 16.0 years with newly diagnosed diabetes during 1 year in Uppsala county, which has slightly less than 300,000 inhabitants. Six of the type 1 diabetic children were female and 18 male, with a mean age of 8.4 years (range 1.6–15.7). They had a mean history of polydipsia and polyuria of 18 days, with a wide variation (1–60 days). Mean HbA_1c was 9.8% (range 6.5–16.5; upper reference limit 5.0), and the first blood glucose was between 15.5 and 40.7 mmol/l (mean 25.0). They were all in relatively good condition, all were conscious, and only three had a pH <7.30. The first blood was taken within 1 week in all cases, except in one boy who fell ill in another part of the country (day 26). The second blood sample was taken 2–6 months after the diagnosis. The control group consisted of 24 age- and sex-matched control subjects from the same county or neighboring counties, with the exception of two control subjects who were from the north of Sweden. The control subjects were recruited within 2 months after the proband, from among patients without evidence of ongoing infection. Samples were also taken from 20 siblings of the type 1 diabetic children who were willing to participate (mean age at the first sample 12.4 ± 6.0 years). The first blood sample from each sibling was obtained at or close (within 6 weeks) to the diagnosis of the index case. The study was approved by the ethics committee of the Medical Faculty at Uppsala University.

Three strains of CBV-4 were used, V89-4557 and VD2921 (two plaque-purified strains isolated from patients suffering from aseptic meningitis), and the E2-Yoon strain (shown to be diabetogenic in mice). All three strains are readily neutralized with a standard polyclonal neutralizing anti-CBV-4 serum (American Type Culture Collection). Two of the strains have been sequenced and compared phylogenetically (9,24), and they have earlier been used for the study of viral involvement in type 1 diabetes (5,7–9). Green monkey kidney (GMK) cells were used for the neutralization test. They were cultured in Eagle’s minimum essential medium (EMEM) supplemented with 10% newborn bovine serum.

For analyses of neutralizing antibodies, 5 μl of the serum specimens were serially diluted with EMEM in twofold steps from 1:20 to 1:2,560. The diluted sera and EMEM containing 100 of the 50% tissue culture infective dose (TCID₅₀) of the virus/0.1 ml were mixed and incubated for 90 min at 37°C. The mixtures were then transferred to GMK cells cultured in 96-well plates. The titers were recorded after 4–8 days. Antibodies against GAD65 in serum samples from the type 1 diabetic group, the siblings, and the control group were measured with Diamyd’s Anti-GAD65 RIA (Mercodia, Uppsala, Sweden). Using a cutoff of 9.5 units/l, the specificity was 99% and the sensitivity was 74%. IgM antibodies against EV were measured at the Karolinska Institute, Stockholm (25).

PBMCs were isolated from the type 1 diabetic children, their siblings, and healthy control subjects by centrifugation of blood samples with Lymphoprep (Nycomed Pharma, Oslo). They were stored at −70°C until the DNA/RNA extraction. Total RNA was extracted from PMBCs using a QIAamp Viral RNA Mini Kit (QIAGEN, Hilden, Germany). The RNA was stored at −70°C. DNA from patients and control subjects was extracted from PMBCs using a QIAamp DNA Mini Blood Kit (QIAGEN).

HLA typing of DRB1 and DQB1 genes was achieved with Dynal Classic SSP (Branborough, U.K.), a PCR-based method using sequence-specific primers for each HLA allele and visualization of the PCR products on a 1.2% agarose gel. All of the type 1 diabetic children, 23 of the 24 control subjects, and 11 of the 12 EV-PCR-positive siblings were HLA typed.

The primer pairs ECBV-1 to -5 correspond to highly conserved regions within the 5′-NCR of the enteroviral genome (Table 1). The external primers ECBV5 and ECBV1 generate a 600-bp fragment, whereas the use of an internal primer for PCR group A ECBV4 (downstream primer) and ECBV1 (upstream primer) generates a 539-bp PCR product. The external primers used in group B were ECBV4 and ECBV1. The internal primers for the PCR group B ECBV4 (downstream primer) and ECBV2 (upstream primer) generated a 436-bp PCR product.

The reverse transcription (RT) mixture (12 μl) containing 11 μl extracted RNA with 1 μl of negative-strand primer (E1) (Table 1) was heated for 10 min at 70°C and put on ice. Then, 7 μl of master mix containing 4 μl of 5 × first-strand buffer (250 mmol/l Tris-HCI, 375 mmol/l KCI, and 15 mmol/l MgCI₂), 2 μl of 0.1 mol/l dithiothreitol, and 1 μl of 10 mmol/l concentration of each dNTP (Life Technologies) was added, and the mixture was incubated for 2 min at 42°C. This was followed by the addition of 200 units of Superscript II (Life Technologies) and incubation for 60 min at 42°C. The reverse-transcribed samples were then denatured for 15 min at 70°C before storage on ice.

The PCR was performed in two steps. First, a PCR was carried out by using the two primer pairs ECBV5/ECBV1 (group A) and ECBV4/ECBV1 (group B). cDNA (2 μl) was amplified in a volume of 50 μl containing 50 mmol/l KCl, 20 mmol/l Tris-HCl (pH 8.4), 2.5 mmol/l MgCl₂, 0.1 mg/ml BSA, 0.2 mmol/l of the dNTPs, 50 ng of each primer, and 2 units of Taq DNA polymerase (Amersham Pharmacia Biotech, Piscataway, NJ) in a DNA thermal cycler-Touchgen (Techne, Cambridge, U.K.) with the following profile: 94°C for 4 min, followed by 40 cycles of 94°C for 30 s, 45°C for 30 s, and 72°C for 1 min, followed by a final extension step at 72°C for 5 min. A second-run PCR was performed using primer pairs ECBV4/ECBV1 in group A (ECBV5/ECBV1) and ECBV4/ECBV2 in group B (ECBV4/ECBV1). The PCR was performed as described above. Briefly, 2 μl of the amplified products were added to the PCR mixture, consisting of 10 μl of 10 × PCR buffer, 4 units Taq DNA polymerase, and 100 ng each of primers ECBV-1 to -4. The mixture was adjusted to a volume of 100 μl with water. The PCR was performed as in the first PCR, but using 50°C instead of 45°C for the annealing temperature. Precautions were taken to avoid any kind of contamination with extraneous nucleic acid. A negative control (no RNA) was used in each PCR. RT and the first and second PCRs were performed in separate locations in the laboratory.

PCR products from the second PCR were identified by 1.5% agarose gel. The positive PCR products with correct size (539 bp with the primers ECBV4/ECBV1 and 436 bp with the primers ECBV4/ECBV2) were purified using QIAquick PCR Purification Kit (QIAGEN). Both strands of the DNA fragments were sequenced using an automatic ABI Prism 310 sequencer by the Big Dye-labeled terminator method using Amplitaq DNA polymerase FS (Perkin-Elmer). The sequencing primers were the same as those used in the second run of PCR amplification (ECBV4, ECBV1, and ECBV2).

To estimate genetic relationship, the nucleotide sequences from EV-positive siblings of patients, from control subjects, and from known sequences of other EV serotypes (deposited in GenBank) were aligned using the CLUSTAL W (1.5) multiple sequence alignment program (26). The analyses were performed using the programs DNADIST and NEIGHBOR-joining included in the PHYLIP package (PHYLIP Phylogeny Inference Package, Version 3.5p; Department of Genetics, University of Washington, Seattle, WA) (27). To check the reliability of the branches defined by the phylogenetic tree, we used SEQBOOT (100 replicates), which was also included in the PHYLIP package. Boot-strapping of 1,000 replicates was performed.

Statistical analyses of four field tables were performed using Fisher’s exact two-tailed test. A P value <0.05 was considered to indicate a statistical significance. The statistical package SPSS for Windows, Version 10.0, was used in statistical analyses.

As shown in Table 2, EV-RNA was found in PBMCs from 12 (50%) of the 24 newly diagnosed type 1 diabetic patients and 5 (26%) of 19 siblings, but in none of the control subjects (Table 3) with the use of primers ECBV4/ECBV1 (group A). With the use of primers ECBV4/ECBV2 (group B), EV-RNA positivity was detected in PBMCs from 11 (46%) of the 24 type 1 diabetic children, 11 (58%) of 19 siblings, and 7 (29%) of the 24 control subjects (Tables 2 and 3). One of the EV-PCR-positive siblings (P20S) belonged to an EV-PCR-negative family. The total number of EV-RNA-positive patients was 18 (75%), among the siblings the corresponding figure was 12 (63%) of 19, and in the control group the figure was 7 (29%) of 24. With the use of group A primers, 50% of the type 1 diabetic children revealed EV-PCR positivity, compared with none (0%) of the 24 control subjects, a statistically significant difference (P < 0.001). The difference in group A EV-PCR positivity between the siblings (5 of 19) and the type 1 diabetic children (12 of 24) was not significant, but siblings had a higher frequency of EV-PCR positivity compared with the control subjects (P < 0.05).

Two phylogenetic trees were constructed by comparing the sequences from EV-PCR amplicons obtained from type 1 diabetic children, siblings, and control subjects and previously published EVs (Figs. 1A and B). Figure 1A shows the phylogenetic analysis of the 5′-NCR sequences obtained with the ECBV4/ECBV1 primers (group A), which revealed clustering of the sequences into five major branches. CBV-1, CBV-3, and the other EV formed their own cluster (bootstrap 99%). Six of the patients and one of the type 1 diabetic children’s siblings (subject P5B) clustered with the CBV-4 strain VD2921 (VD2921-like), and one patient (subject P5) formed a sub-branch in the VD2921-like cluster. P2B, P23S, P3, P4, and P6 were found in the same cluster as the CBV-4/E2 (CBV-4-like). The sequence from P13S was clustered with CBV-2, echovirus-6, and SVDV (EV-like). P9, P10, and P16B1 sequences were more closely related to CBV-5 and -6 (CBV-5/CBV-6-like, respectively). The sequence of one patient (subject P1) formed a separate cluster with poliovirus, although this patient was clearly in the same branch as P9, P10, and P16B1 (PV-like). The sequence alignment showed that the nucleotide sequence of patient 1 was 79.1% similar to poliovirus type 1. The CBV-4/E2-like, CBV-4/VD2921-like, and CBV-5/CBV-6-like groups were strongly supported by bootstrap values of 94–100%, whereas the bootstrap score for the echovirus-like group was 61%.

Figure 1A shows the phylogenetic analyses of the 5′-NCR sequences obtained with the group B primers. They segregated into three major clusters: CBV-5/CBV-6-like, echovirus-like, and CBV-4/E2-like. In this tree, unlike the tree obtained with the group A primers, VD2921 was clustered together with CBV-4/E2. Eight positive amplifications (subjects P7, P16, P21, P7S, P10B, P16B2, P20S, and P21S) were closely grouped together in one of the sub-branches formed with the CBV-4/E2 strain and VD2921 with 100% bootstrap support. Subjects P13, P9S, and P22B were clustered with VD2921 with 95.4–97.7% nucleotide identity. Six of the control children formed the other sub-branch of the CBV-4/E2-like cluster. One patient and one control subject (subjects P24 and C9) clustered with CBV-5/CBV-6-like, and the sequence of one patient (subject P2) was related to echovirus (EV-like cluster). Comparison of these two trees revealed some variation in the branching order of some subgroups, some variation in bootstrap values for certain nodes, and also a variation in branch lengths. In addition, the branching order of the CBV-5/CBV-6-like and echovirus-like clusters differed between the two trees.

IgM antibodies against EV were found in 4 of 24 patients (subjects P4, P6, P12, and P24). All of these were EV-PCR-positive. Two of the EV-IgM-positive patients (subjects P6 and P24) also revealed antibodies against GAD65. No analyses of EV-IgM antibodies were performed in serum samples from siblings or control subjects. The neutralizing antibody (NT) titers against CBV-4 differed somewhat, depending on which strain was used in the tests (Tables 2–4). The NT titer in sera from one patient could be negative against one of the CBV-4 strains and as high as 2,560 against one of the other strains. The number of NT-positive samples against the CBV-4 strains did not vary significantly among the patients, siblings, and control subjects. No significant difference in titer levels against the three CBV-4 strains was obtained when the three groups were compared. In 15 of the patients (subjects P1–3, P5, P8–10, P13, P16, P17, and P19–23), a convalescent serum sample was taken 2–3 months after the onset of type 1 diabetes and, as shown in Table 5, 7 of the type 1 diabetic children (subjects P2, P5, P9, P13, and P21–23) had a significant rise in NT titer between the acute and convalescent sera against V89-4557. In total, 7 (41%) of 17 revealed a significant rise in NT titer, and 1 of these children was EV-PCR-negative (subject P23). Five of the type 1 diabetic children were too young (age ≤5 years) for a second serum sample to be obtained (they were all EV-PCR-positive), and we were not able to get a second sample from three of the patients for other reasons. Among the siblings, three revealed significant NT titer rises against CBV-4/CBV-2 against the E2-Yoon strain and one revealed a rise against the V89-4557 strain. All but one of these siblings (subject P1S) were EV-PCR-positive. In total, 10 children revealed significant rises in NT titer against CBV-4, and in 8 of them, the finding could be confirmed by a positive EV-PCR. Comparison of sequencing results with the presence of NT titer against the different CBV-4 strains revealed that of the 13 patients with CBV-4-like sequences, 2 showed a significant rise in NT titer against the V89-4557 strain and 1 had a rise in titer against the E2-Yoon strain (Table 2). We could not obtain convalescent serum samples from five of these patients. In the remaining five type 1 diabetic children, no confirmation of the observed sequence homology with CBV-4 could be obtained with the use of NT; however, some of these patients showed high NT titers against one or more of the CBV-4 strains already in the acute serum. No NT tests were performed against other serotypes of CBV or against other EVs.

Antibodies against GAD65 were found in 16 (67%) of the 24 acute sera from the type 1 diabetic children, compared with in 2 (10%) of the 20 siblings (P < 0.01) and 1 (4%) of the 23 control subjects (P < 0.001) (Tables 2 and 3). These antibodies were also detected in 13 of the 17 convalescent sera (subjects P1, P3, P5, P7, P8, P10, P11, P13, P16, P17, P19, P21, and P23). In total, 18 (75%) of the 24 type 1 diabetic children revealed these antibodies at the onset or 2 months later. Among the siblings, 2 (10%) of 20 had these antibodies in the acute sera, and both of these subjects were negative when the convalescent sera (taken 2–3 months later) were analyzed. One of the control subjects (subject C19) had a low level of antibodies against this autoantigen. Of the two anti-GAD65-positive siblings, one was EV-PCR-positive and one had a significant rise in NT titer against V89-4557. Group B-positive individuals (type 1 diabetic children, siblings, and control subjects) more often than group B-negative individuals revealed antibodies to GAD65 (P < 0.05). Group A positivity, however, was not significantly associated with antibodies against this autoantigen. Individuals positive in both group A and group B EV-PCR also had a higher frequency of antibodies against GAD65 than individuals positive in just group A or group B or those who were EV-PCR-negative (P < 0.001).

HLA-DRB1 and -DQB1 alleles for each patient are presented together with EV-PCR results in Table 2. The allele HLA-DRB1*04, which has been linked to increased risk for type 1 diabetes in earlier studies, was overrepresented in the type 1 diabetic group, being present 71% of the type 1 diabetic children compared with 30% of the control subjects (P < 0.01), whereas the alleles DR and HLA-DQB1*06 were more common among control subjects. None of the type 1 diabetic children, compared with 39% of the control subjects, carried the DRB1*15 allele (P < 0.01), and 17% of the patients, compared with 70% of the control subjects, carried the DQB1*06 (P < 0.01) allele.

EV-PCR positivity among newly diagnosed type 1 diabetic children and adults has been shown before (14–16), but to our knowledge, this is the first time that primers that detect EV that might be diabetogenic have been used. In the earlier studies, the frequencies of EV-PCR positivity varied (27, 67, and 37%, respectively) and, compared with the total EV-PCR positivity shown in this study (75%), they were all lower. Chehadeh et al. (15) also detected EV-RNA in patients previously diagnosed with type 1 diabetes (both children and adults), indicating a persistent infection or an increased susceptibility in this group of patients for EV infections. A possible explanation for the high frequency of EV-RNA positivity in our study might be that RNA was extracted from peripheral blood cells instead of from plasma. It has been reported that peripheral blood cells can harbor EVs during and after a viremic phase (28). During viremia, EVs are transported both intracellularly by mononuclear cells (lymphocytes or monocytes) and extracellularly in serum, indicating better chances to detect viral RNA in the PBMC fraction compared with serum. Another possible explanation could be the selection of primers used in this study. As shown in the results, different pairs of primers (groups A and B) resulted in different frequencies of EV-RNA positivity among the type 1 diabetic children, their siblings, and the control subjects. It is noteworthy that by the use of the group A primers, 12 of the type 1 diabetic children revealed EV-PCR positivity, compared with none of the control subjects (P < 0.001) and only five of the siblings. This difference in detection might indicate important differences related to diabetogenicity in the nucleotide sequences of the specific annealing sites of the primers. Other differences in this study compared with those above (14–16) are that samples from the siblings of the type 1 diabetic children were also analyzed and that we were able to compare the viral sequences from the three groups of children with each other and with the sequence of a strain of CBV-4 (VD2921) that is able to establish persistent infection in human islet cells (9).

Phylogenetic trees revealed that the amplicons from the infected siblings shared a high degree of homology with that of the respective type 1 diabetic children. Two of the branches of the tree obtained with the group A primers consist of E2- or VD2921-like sequences. Both of these CBV-4 strains have been shown to be “diabetogenic” (5,8,9,23). The similarity with CBV-5 and swine vesicular disease virus (SVDV) are new findings, although it has recently been shown that SVDV utilizes the same receptor as CBV (29), indicating that the tropism of these viruses is similar. The phylogenetic tree of group B, as in the tree obtained with the group A primers, revealed that the EV sequences in these type 1 diabetic children and their siblings also shared a high degree of homology. An interesting finding was that six of seven control subjects were clustered on a separate sub-branch of CBV-4, compared with the sequences of the type 1 diabetic children and their siblings. In addition, with the use of these primers, the VD2921 and the CBV-4/E2 strain had branches that clustered together. The high degree of similarity between the sequences from the type 1 diabetic children and their brothers or sisters indicates that both were infected with the same or a closely related strain of a serotype of EV.

Because the 5′-NCR is very conserved among all serotypes of EVs, sequencing of this part of the viral genome means that comparisons of the sequences only vaguely implicate a specific serotype. The most variable regions of the EV genome are within the genes coding for the capsid proteins VP1, VP2, and VP3, which are partially exposed on the virus surface. VP1 codes for the major antigenic sites and most type-specific neutralization determinants (30). Molecular typing of EVs should ideally focus on regions encoding determinants for neutralization. Mapping with monoclonal neutralizing antibodies has shown that, in particular, the NH₂-terminal part of VP1, encompassing the B–C loop, is a major antigenic region. It has been shown that amplification of the VP1-P2C region, using a single primer pair, enabled serotyping of >90% of the EVs (31,32). This has yet to be done with the samples in this study, but such an approach would make it possible to determine which serotype caused the infection and, possibly, the type 1 diabetes.

In line with the idea that VP1 is the major antigenic region, differences between NT titers obtained with the VD2921 strain and the “diabetogenic” E2 strain might be explained by differences in amino acid sequences in this region. Sequencing of the VD2921 strain has revealed the substitution of 11 amino acids in the structural protein VP1, near the fivefold axis, compared with the E2 strain (9). In this study all but two of the significant rises in NT titer were obtained against the V89-4557 strain. It has been shown earlier that the highest frequencies of IgM antibodies among newly diagnosed type 1 diabetic children were obtained when this strain was used as an antigen (13). This might indicate that infection with strains that have a high degree of homology with the VP1 region of this strain or related strains is more common among children at the onset of type 1 diabetes. Antibodies able to neutralize a virus are considered to be of high affinity, and neutralizing antibodies produced after EV infections are usually type-specific. However, if the child has the immunological memory of an EV infection from the past and is then infected by another EV, neutralizing antibodies to the first infection could be produced together with such antibodies against the second infection (33). This might also explain possible rises in NT titers against one EV and sequence homologies to another.

EV infections have been reported to be temporally associated with the appearance of islet-specific autoantibodies (17). The frequencies of antibodies against GAD65 obtained in this study are in line with results obtained by other authors (18). The finding in this study that EV-PCR positivity obtained with the group A primers alone was not associated with the presence of antibodies against GAD65 is not in line with other groups results (17). However, EV positivity with both groups of primers (A and B) was positively associated with GAD65 antibodies, which is in line with the findings of others and might also indicate that multiple EV infections are needed to induce antibodies to autoantigens (34). The mechanism by which EV infections may induce diabetes in humans is unknown (35–39); these viruses are known to induce pancreatitis, especially the CBVs, and it has also been shown that these viruses can infect human β-cells in vitro (7–10).

If the role of EV infections in the etiology of type 1 diabetes holds true, it will be important to resolve the diabetogenic mechanism of the virus. If the virus itself causes the cell destruction by direct cell lysis, a putative EV vaccine might prevent some of the cases of type 1 diabetes. In contrast, if immunological cross-reactions or exposure to ICAs, such as GAD65 or heat shock protein 60, causes inflammation, by the molecular mimicry hypothesis, immune responses induced by EV vaccines might also facilitate autoimmune processes in type 1 diabetes. Further studies should investigate whether single, multiple, or persistent infections are needed for type 1 diabetes development, or whether people with autoimmunity are more prone to EV infections. The PCR technology used in this study holds further promise of enabling researchers to finally type the candidate diabetogenic strains amplified from newly diagnosed type 1 diabetic patients.

In conclusion, EV-RNA was detected in the PBMCs of a majority of newly diagnosed type 1 diabetic children, and the sequences of the PCR amplicons from these children revealed homology with CBV as well as other EVs. These results support the involvement of EV infections in the development of type 1 diabetes.

This study was supported in part by the European Community (BMH4.CT-95-1561, BMH4-CT98-3952), the Medical Research Council (K2001-32X1435-01A), the Swedish Diabetes Association, the Gillbergska Foundation, the family Ernfors fund, the Novo Nordisk Foundation, and Ronald McDonald Barnfond.

Barbro Westerberg, Christina Pilström, and the diabetes team of the Children’s hospital are acknowledged for help with the sample collection and with informing the patients and their families. Charlotta Wolgast is gratefully acknowledged for skilful technical assistance.