School of Public Health, University of Tampere, Department of Paediatrics, Tampere University Hospital and Department of Epidemiology and Health Promotion, National Institute of Public Health, Helsinki, and
Dr Heikki Hyöty, University of Tampere, Medical School, Department of Virology, PO Box 607, FIN-33101, Tampere, Finland. E-mail: email@example.com
This study evaluated the possible role of enterovirus infections in the pathogenesis of type I (insulin-dependent) diabetes in a prospective dietary intervention trial. Children participated in the second pilot of the Trial to Reduce IDDM in Genetically at Risk (TRIGR) project. They were randomized into two groups receiving either a casein hydrolysed formula (Nutramigen®) or a regular formula, whenever breast milk was not available over the first 6–8 months of life. Altogether 19 children who turned positive for autoantibodies associated with type I diabetes by 2 years of age and 84 matched control children were analysed for enterovirus antibodies and enterovirus RNA in serum. Enterovirus infections were common during the first 2 years of life and more frequent among boys than girls (P = 0·02). Autoantibody-positive children had more enterovirus infections than autoantibody-negative children before the appearance of autoantibodies (0·83 versus 0·29 infection per child, P = 0·01). The average levels of IgG antibodies to echovirus antigen were also higher in autoantibody-positive than in autoantibody-negative children (P = 0·0009). No difference was found in the frequency of enterovirus infections between children receiving the casein hydrolysed formula or regular formula. These results suggest that enterovirus infections are associated with the induction of β-cell autoimmunity in young children with increased genetic susceptibility to type I diabetes.
The aetiology of type I (insulin-dependent) diabetes mellitus is multi-factorial, comprising both genetic and environmental factors. Environmental risk factors include certain virus infections and dietary factors, but the mechanisms by which these may trigger β-cell damage are not known. Cow's milk proteins are among the strongest dietary risk factor candidates [1–3], while among all potentially diabetogenic viruses, enteroviruses have been the most suspected ones. Serological studies have indicated increased enterovirus antibody levels in patients with type I diabetes , and enterovirus RNA has been detected in the blood of subjects with type I diabetes more frequently than in control subjects [5–7]. Recent prospective studies have suggested that enterovirus infections can initiate and accelerate the β-cell damaging process years before the manifestation of clinical type I diabetes [8–12], and that in some cases this may happen already in utero[8,9,13].
In the present study we had the opportunity to evaluate the risk effect of enterovirus infections in children who were followed prospectively from birth, also covering the prenatal period. Adenovirus infections were analysed in the same children as a control. These children participated in the second pilot phase of the Trial to Reduce IDDM in Genetically at Risk (TRIGR) project, which evaluates the possible effect of dietary elimination of cow's milk proteins in early infancy on the subsequent risk to develop type I diabetes in subjects with increased genetic disease susceptibility . This provided a possibility to also test the feasibility of this type of prospective birth-cohort studies in the evaluation of possible interactions between two potentially important environmental risk factors: enterovirus infections and cow's milk proteins. Theoretically, enterovirus infections could increase the possible risk effect of dietary antigens, e.g. by altering the balance of the gut immune system or by increasing the gut permeability to dietary antigens.
MATERIALS AND METHODS
The study population comprised 103 children and their mothers from families with at least one member with type I diabetes, the newborn infants having an HLA-DQB1 genotype indicating increased genetic risk. The families participated in the second pilot study of the TRIGR project in Finland. This pilot study was aimed at evaluating the possible effect of the elimination of cow's milk proteins in early infancy on the appearance of diabetes-associated autoantibodies by the age of 2 years. The study design was double-blinded, and the newborn infants were randomized into two groups. When supplementary milk feeding was started, those in the intervention group received a casein hydrolysate formula (Nutramigen®, Mead Johnson & Co., Evansville, IN, USA) until the age of 6–8 months, while the infants in the control group were given a regular formula (Enfamil®, Mead Johnson & Co.). The control formula included 20% casein hydrolysate formula in order to eliminate the taste and smell difference between the two study formulas. The infants were born between April 1995 and November 1997 in 15 hospitals all around Finland. In addition to the cord blood sample, serum samples were obtained from the children at follow-up visits at the ages of 3, 6, 9, 12, 18 and 24 months. Maternal serum samples had been taken at the end of the first trimester of pregnancy and at the time of delivery. All sera were stored at − 20°C until analysed. Written informed consent was obtained from the mother before enrolment. The study was approved by the Joint Ethics Committees of the participating hospitals.
The103 children of the present study included 19 index cases and 84 control subjects. The 19 index cases were the subjects in the whole TRIGR cohort comprising 208 children who developed signs of progressive β-cell autoimmunity, i.e. positivity for type I diabetes-associated autoantibodies by the age of 2 years. Eight of them progressed to clinical type I diabetes by the end of March 2002. Three to five control infants were chosen for each index case (mean four, altogether 84) according to a nested case–control design matching for the intervention group, gender and HLA-DQB1 genotype. The control infants were observed according to the same protocol as the index cases but remained constantly negative during the observation for all four autoantibodies analysed. All children in TRIGR study have now been screened for autoantibodies until the age of 3 years and none of the control children had developed autoantibodies or diabetes by that age. A total of 141 samples were available from the case children (mean 7·4 samples per child) and 616 samples from control children (mean 7·3 samples per child).
IgG and IgA class antibodies against purified coxsackievirus B4 (CBV4), purified echovirus 11 (EV11), a synthetic peptide antigen (amino acid sequence KEVPALTAVETGAT-C) and adenovirus hexon protein were measured using enzyme immunoassay (EIA) as described previously . Echovirus 11 and coxsackievirus B4 antigens were heat-treated to make them broadly reactive with antibodies induced by another enterovirus serotypes as well. IgM class enterovirus antibodies were measured against a mixture of three enterovirus antigens [coxsackievirus B3 (CBV3), coxsackievirus A16 (CAV16) and EV11] using a capture EIA method as described previously . Neutralizing antibodies were measured against all coxsackie B serotypes 1–6 (ATCC reference strains) by a standard plaque neutralization assay .
Detection of enterovirus RNA
Extraction of viral RNA from serum samples and the subsequent reverse transcription and polymerase chain reaction (RT-PCR) and hybridization methods have been described in detail previously . Separate rooms were used in each step of the RT-PCR work and each analysis included positive and negative controls. All positive samples were confirmed to be positive by repeated RT-PCR and subsequent hybridization assay.
Diagnostic criteria of an infection
In the EIA analyses a twofold or greater increase in the antibody level against any of the antigens and exceeding the cut-off level for seropositivity (15 EIU for IgG and IgA assays) was considered significant indicating infection in that sample interval . In the capture-IgM assays similar criteria were used (the cut-off level for seropositivity was three times the level which was obtained with PBS instead of the antigen). In the neutralization assay a fourfold or greater increase in the titre of antibodies was considered as significant. Presence of enterovirus RNA in serum was taken as a marker of current infection. Diagnosis of infections during pregnancy was based on the detection of antiviral IgM antibodies in maternal serum samples or on a twofold or greater increase in IgG or IgA antibody levels during pregnancy (comparison between the maternal sample taken at the end of the first trimester of pregnancy and the maternal sample taken at delivery). In addition, presence of enterovirus RNA or enterovirus IgM antibodies in cord blood was taken as a marker of fetal infection. All samples were analysed blindly without information about case/control status.
Islet cell antibodies (ICA) were determined by a standard immunofluorescence method  with a detection limit of 2·5 Juvenile Diabetes Foundation units (JDF-U). Insulin autoantibodies (IAA) were quantified with a microassay  modified from the method described by Williams et al. , with a cut-off limit of 1·56 relative units (RU) for antibody positivity. Antibodies to the 65 kilodalton isoform of glutamic acid decarboxylase (GADA) were measured with a radioligand assay as described earlier [20,21]. The cut-off limit for GADA positivity was 5·35 RU. Antibodies to the protein tyrosine phosphatase-related IA-2 molecule (IA-2 A) were analysed with a radiobinding assay as described in detail elsewhere [22,23]. The cut-off limit for IA-2 A positivity was set at 0·43 RU. The cut-off limits for IAA, GADA and IA-2 A positivity were set at the 99th percentile in more than 370 Finnish non-diabetic subjects. All samples with antibody levels between the 95·5th and 99·5th percentiles were reanalysed to confirm the antibody positivity or negativity.
The HLA DQB1 genotype was defined from cord blood specimens by a method based on the PCR for amplifying the gene segment and subsequent hybridization with lanthanide-labelled sequence-specific oligonucleotide probes . Alleles conferring disease susceptibility (HLA DQB1 *0302 and *02) and those associated with protection (HLA DQB1 *0301, *0602 and *0603) were defined. Children carrying DQB1 *02/0302, DQB1 *0302/X (X ≠*0301, *0602 or *0603) and DQB1 *02/X genotypes were included into the follow-up cohort.
Differences in the infection frequency observed during the follow-up between the case and matched control children were tested using the Score-test (Stata statistical software). Differences in the occurrence of at least one infection during a given follow-up period between the case and control children were assessed using the Mantel–Haenszel odds ratio (multiple and varying number of control subjects per case, Stata statistical software). Differences in the infection frequency between the case and control children in the two formula groups were tested using conditional logistic regression. The area under the curve method was used to compare the virus antibody levels between the case and control groups . A P-value less than 0·05 was considered statistically significant.
Frequency of infections
Enterovirus infections were detected in 16% of all sample intervals (107 infections per 653 sample intervals) during the follow-up of the 103 children. Infections were most frequent between the age of 12 and 18 months, while only 13% of the infections occurred before the age of 6 months. Enterovirus RNA positivity was most common before the age of 12 months and after that it declined rapidly (Fig. 1). Forty-six per cent of the infections were diagnosed by significant increases in IgG levels against EV11, 20% by increases in IgG levels against CBV4 and 23% by RNA positivity in RT-PCR, while the other tests applied indicated infections with lower sensitivity (Table 1).
Table 1. Proportion of enterovirus infections diagnosed by different tests and their combinations
Type of test
Proportion of infections(%)
Synthetic enterovirus peptide antigen (amino acid sequence KEVPALTAVETGAT-C, see Methods).
2 Enterovirus IgM was measured using a capture IgM EIA against a cocktail of coxsackievirus B3, echovirus 11 and coxsackievirus A16 antigens.
3 Neutralizing antibodies were measured against six coxsackievirus B serotypes.
The duration of the IgG responses varied substantially. In some cases these responses persisted for more than 1 year, whereas in others the IgG levels decreased very rapidly and were detectable in only one sample (Fig. 2). Long-lasting response was seen if the infection was caused by the same serotype as that used as the antigen in the EIA test, and a short-lasting response if the infection was caused by a different serotype to that used as the antigen in the EIA test (Fig. 2b).
The study population comprised 63 boys and 40 girls. Enterovirus infections were diagnosed more frequently in boys than in girls (1·21 versus 0·80 infections per child, P = 0·02). Twenty-seven per cent of the boys had enterovirus RNA in serum compared to 17% of the girls (P = 0·1).
Thirty-four children carried the HLA-DQB1*02 allele, 41 children the DQB1*0302 allele and 28 children both these susceptibility alleles. Children with the DQB1*02 allele without the DQB1*0302 allele had less enterovirus infections (0·91 infection per child) compared to other children in this study (those with the DQB1*0302 allele had 1·15 and those with both risk alleles had 1·11 enterovirus infections per child) (P = 0·02). Children with the DQB1*02 allele had enterovirus RNA in 4% of their samples, children with the DQB1*0302 allele in 7% of the samples and children with both risk alleles in 3% of the samples (n.s.).
Relationship between infections and β-cell autoimmunity
The average IgG levels to all enterovirus antigens were high in cord blood in both case and control children, reflecting the presence of maternal antibodies (Fig. 3). IgG levels in maternal serum taken at delivery correlated with those in the cord blood (r = 0·94 for CBV4 IgG). After birth IgG levels decreased and reached a nadir at the age of 6 months. They then started to increase peaking at the age of 18 months. IgG antibody levels to echovirus 11 antigen were higher in case than control children (P = 0·0009 with the AUC method; Fig. 3) while the levels of coxsackievirus B4 IgG did not differ between the groups. An enterovirus infection was observed in 19% of the sample intervals in autoantibody-positive children compared to 15% in the control children (n.s.). Enterovirus RNA was detected in 6·8% of the samples taken from the autoantibody-positive children compared to 3·2% of the samples from the control children (P = 0·03).
The time period preceding the appearance of autoantibodies was analysed separately. Enterovirus infections were more frequent during this period in autoantibody-positive children compared to control children analysed until the same age (0·83 versus 0·29 infections per child, P = 0·01). Enterovirus RNA was found in 14·0% of the case children's samples and in 8·4% of the control children's samples, respectively (P = 0·07). Only two children in the whole cohort were positive for enterovirus RNA in two consecutive samples and both of them were case children. One of them was repeatedly RNA-positive at the age of 9 and 12 months (the first autoantibody-positive sample was taken at the age of 12 months). The other was RNA-positive in cord blood and in the next sample taken at the age of 3 months, 12 months before autoantibodies appeared, and the child again had enterovirus infection. No difference was observed in adenovirus infections between autoantibody-positive and -negative children (0·26 versus 0·20 infections per child, n.s.).
Enterovirus infections were diagnosed in 22% of the sample intervals in male cases compared to 17% of the sample intervals in male controls (P = 0·1). The corresponding figures among girls were 15%versus 13% (n.s.). Enterovirus RNA was detected in 67% of male cases compared to 24% of male controls (P = 0·009) and in 14%versus 18% of female case and control children (n.s.), respectively. The case children with the DQB1*0302 risk allele had slightly more enterovirus infections than their controls (1·71 compared to 1·03 infections per child, P = 0·04), while no significant difference was observed between the case and control children in other HLA groups.
None of the observed enterovirus infections in case children was caused by CBV serotypes based on the observation that case children had no neutralizing antibodies against any of the six CBV serotypes. However, CBV infection was detected in 15% of the control children (CBV1 in one child, CBV2 in two children, CBV3 in three children, CBV4 in three children and CBV5 in four children, and CBV6 in none of the children).
Seven mothers of the case children (37%) and 17 mothers of the control children (20%) had experienced an enterovirus infection during pregnancy [odds ratio (OR) 2·4, 95% CI 0·9–7·0]. One case child and two control children tested positive for enterovirus RNA in cord blood, while none of the children had enterovirus IgM in cord blood. Adenovirus infections were diagnosed during pregnancy in the mothers of two case children and three control children.
Effect of infant formula on enterovirus infections
Five of the case children and 15 of the control children were in the intervention group receiving a casein hydrolysate formula. The remaining 14 case children and 69 control children were in the group receiving regular formula. In the casein hydrolysate group case children had 1·4 enterovirus infections per child during the follow-up compared to 1·16 infections per child in the corresponding control subjects receiving the same formula (n.s.). Case children in the regular formula group had 1·14 enterovirus infections per child compared to 0·96 infections per child in their control children (n.s.). Conditional logistic regression analysis showed no effect of intervention on enterovirus infection frequencies (P = 0·3).
According to current knowledge the β-cell damaging process is initiated by yet unidentified environmental factors in genetically susceptible individuals. However, none of these factors has been so far identified, although the possible role of dietary factors and virus infections has been studied intensively. In the 1990s an extensive international primary prevention trial (TRIGR) evaluating the impact of early elimination of cow's milk proteins in infancy in subjects with increased genetic risk has been initiated. The second pilot phase has recently been completed, suggesting that the elimination of intact cow's milk proteins in early infancy may indeed decrease the emergence of diabetes-associated autoantibodies . Our aim was to evaluate whether enterovirus infections, among the most strongly implemented environmental triggers of type I diabetes, are associated with the induction of β-cell autoimmunity in this trial. This allowed us to also test the feasibility of this type of prospective study design in the analyses of possible interactions between enterovirus infections and cow's milk proteins. Theoretically, enterovirus infections may either be an independent risk factor, which could initiate the process even in the absence of cow's milk proteins, or increase the risk effect of cow's milk proteins, e.g. by altering the functional balance of the mucosal and gut immune system.
Enterovirus infections were found to be frequent in these young children, which is in line with our previous findings in another birth-cohort study (Diabetes Prediction and Prevention trial, DIPP) . The incidence of enterovirus RNA in serum peaked earlier (before the age of 12 months) than infections diagnosed by antibody assays (peaking at the age of 12–18 months), suggesting that the risk of viraemia is particularly high at an early age when the child's own immune system is still immature and unable to clear the virus efficiently. In addition, this may reflect the fact that early enterovirus infections are difficult to diagnose by antibody assays due to the interfering effect of maternal antibodies. Enterovirus infections were also relatively common during pregnancy (24% of the mothers had an infection), but fetal infection could be confirmed in only three children who were positive for enterovirus RNA in cord blood. We found an excess of enterovirus infections in case children compared to control children, while the amount of adenovirus infections did not differ between the groups. This suggests that the excess of enterovirus infections is specific and not common for all viruses.
The duration of enterovirus antibody responses varied greatly depending on the serotype causing the infection, the immunoglobulin isotype measured and the virus antigen used in the EIA test. On some occasions the antibody response was detectable only in a single sample during the follow-up. Accordingly, a wide panel of virological tests and short sample intervals (no longer than 3–6 months) are essential for reliable diagnosis of enterovirus infections in this kind of prospective birth-cohort studies.
In the present study boys had more enterovirus infections than girls. This may reflect higher susceptibility to enterovirus infections in boys compared to girls as complications of enterovirus infections, such as meningitis or myocarditis, are known to be more frequent among boys . This is also in line with our previous findings suggesting that boys may have higher risk for enterovirus-induced beta-cell damage than girls [10,12].
The frequency of enterovirus infections and the average levels of IgG antibodies to echovirus antigen were higher in autoantibody-positive children than in control children matched for age, sex and HLA-DQB1 genotype. This finding, although based on a relatively small number of cases, supports our previous observations in other prospective studies suggesting an association between enterovirus infections and induction of β-cell autoimmunity [8,10,12,29]. According to the low frequency of neutralizing antibodies to CBV serotypes the infections in case children were mainly caused by other than CBV serotypes. In previous studies particularly coxsackievirus B serotypes 4 and 5 have been linked to type I diabetes, but some studies have indicated that other enterovirus serotypes might also be diabetogenic [10,15,30]. In the present study the exact identification of diabetogenic serotypes was not possible, as only some of the 64 different enterovirus serotypes were included in neutralization assay. In addition, we used enterovirus prototype strains in this assay, and these strains may not detect antibodies against all antigenic variants. Enterovirus infections were more frequent in children with the HLA-DR4 associated DQ*0302 allele than in children without this allele. The possible effect of HLA on immune responsiveness could not have biased the comparisons between the case and control subjects as they were matched for these HLA-DQ alleles.
The mothers of the case children had more enterovirus infections during pregnancy than the mothers of control children. This difference was small, reaching only a borderline statistical significance, but it is in line with some previous reports suggesting that enterovirus infections during pregnancy may in certain conditions predispose the child to later development of β-cell damage [8,9,13]. Three cord blood samples were enterovirus RNA-positive. The mothers of two of these children had an enterovirus infection during the pregnancy according to the antibody analyses. Viral RNA in cord blood probably reflects fetal infection. However, we cannot exclude the possibility that it may also derive from contaminating maternal blood.
Possible interactions between enterovirus infections and cow's milk proteins are of interest because enterovirus infections modulate the cytokine milieu and activate immune cells in gut-associated lymphoid tissue (GALT), which may potentiate immune responses to dietary antigens. On the other hand, cow's milk proteins may have an influence on the course of enterovirus infection in the gut. In the present study the dietary intervention had no significant effect on the frequency of enterovirus infections. However, the number of study subjects was too small to analyse if these factors could potentiate each other's diabetogenic effect. This will be possible to analyse by comparing the risk effect of enterovirus infections separately in the casein hydrolysate and regular formula groups in the large international study proper phase of the TRIGR trial, which has been initiated recently in several countries.
Altogether, the results suggest that enterovirus infections are associated with the induction of β-cell damage. This study also shows that by carrying out virological studies in the context of dietary intervention trials it is possible to evaluate the role of enterovirus infections in the initiation of the β-cell damaging process and gain new information on possible interactions between viral and dietary risk factors.
The Finnish Trial to Reduce IDDM in Genetically at Risk (TRIGR) Study Group is composed of the following members. Principal investigator: H.K. Åkerblom. Local investigators: V. Eskola, H. Haavisto, A-M. Hämäläinen, R. Jokisalo, A-L. Järvenpää, U. Kaski, J. Komulainen, P. Korpela, M-L. Käär, P. Lautala, K. Niemi, A. Nuuja, M. Renlund, M. Salo, T. Talvitie, T. Uotila, G. Wetterstrand. Special investigators: J. Ilonen, P. Klemetti, M. Knip, P. Kulmala, J. Paronen, A. Reunanen, T. Saukkonen, E. Savilahti, K. Savola, K. Teramo, O. Vaarala, S.M. Virtanen. This research was supported by the European Commission DGXII, Contract no. BMH4-CT96-0233, the Sigrid Jusélius Foundation, the Juvenile Diabetes Foundation International (grants 192612 and 195003 to H.K.Å., grants 197114 and 395019 to H.H.), the Liv and Hälsa Foundation, the Novo Nordisk Foundation, the Paediatric Research Foundation in Finland and the Finnish Diabetes Association. We gratefully acknowledge the technical assistance of Heini Huhtala, Eeva Jokela, Maarit Patrikainen, Anne Karjalainen, Inkeri Lehtimäki, Sari Valorinta and Tiina Toivonen. Special thanks go to the children and families who participated in the trial.