ARTICLE

Enterovirus RNA in longitudinal blood samples and risk of isletautoimmunity in children with a high genetic riskof type 1 diabetes: the MIDIA study

Ondrej Cinek & Lars C. Stene & Lenka Kramna & German Tapia &

Sami Oikarinen & Elisabet Witsø & Trond Rasmussen & Peter A. Torjesen &

Heikki Hyöty & Kjersti S. Rønningen

Received: 26 February 2014 /Accepted: 18 June 2014 /Published online: 23 July 2014# Springer-Verlag Berlin Heidelberg 2014

AbstractAims/hypothesis Only a few longitudinal molecular studies ofenterovirus and islet autoimmunity have been reported, andpositive results seem to be limited to Finland. We aimed toinvestigate an association between enterovirus RNA in bloodand islet autoimmunity in the MIDIA study from Norway, acountry which largely shares environmental and economicfeatures with Finland.Methods We analysed serial blood samples collected at ages3, 6, and 9 months and then annually from 45 children whodeveloped confirmed positivity for at least two autoantibodies(against insulin, GAD65 and IA-2) and 92 matched con-trols, all from a cohort of children with a single high-risk

HLA-DQ-DR genotype. Enterovirus was tested in RNAextracted from frozen blood cell pellets, using real-timeRT-PCR with stringent performance control.Results Out of 807 blood samples, 72 (8.9%) were positivefor enterovirus. There was no association between enterovirusRNA and islet autoimmunity in samples obtained strictlybefore (7.6% cases, 10.0% controls, OR 0.75 [95% CI 0.36,1.57]), or strictly after the first detection of islet autoantibodies(10.5% case, 5.8% controls, OR 2.00 [95% CI 0.64, 6.27]).However, there was a tendency towards a higher frequency ofenterovirus detection in the first islet autoantibody-positivesample (15.8%) compared with the corresponding time pointin matched controls (3.2%, OR 8.7 [95% CI 0.97, 77]).Neither of these results was changed by adjusting for potentialconfounders, restricting to various time intervals or employingvarious definitions of enterovirus positivity.Conclusions/interpretation Positivity for enterovirus RNA inblood did not predict the later induction of islet autoanti-bodies, but enterovirus tended to be detected more often atthe islet autoantibody seroconversion stage.

Keywords Autoimmunity . Enterovirus . Infancy .

Longitudinal Study . RNA . RT-PCR . Virus-associatedaetiology

AbbreviationDAISY Diabetes Autoimmunity Study in the Young

Introduction

Enterovirus infection has long been suspected to trigger oraccelerate islet autoimmunity, which precedes type 1 diabetes

Electronic supplementary material The online version of this article(doi:10.1007/s00125-014-3327-4) contains peer-reviewed but uneditedsupplementary material, which is available to authorised users.

O. Cinek (*) : L. KramnaDepartment of Paediatrics, 2nd Faculty of Medicine, CharlesUniversity in Prague and University Hospital Motol, V Uvalu 84,15006 Prague 5, Czech Republice-mail: Ondrej.Cinek@Lfmotol.cuni.cz

L. C. Stene :G. Tapia : E. Witsø : T. RasmussenDivision of Epidemiology, Norwegian Institute of Public Health,Oslo, Norway

S. Oikarinen :H. HyötyDepartment of Virology, University of Tampere, Tampere, Finland

P. A. TorjesenHormone Laboratory, Oslo University Hospital, Oslo, Norway

H. HyötyFimlab Laboratories, Pirkanmaa Hospital District, Tampere, Finland

K. S. RønningenDepartment of Pediatric Research, Oslo University Hospital, Oslo,Norway

Diabetologia (2014) 57:2193–2200DOI 10.1007/s00125-014-3327-4

[1, 2]. Although several serological patient–control studiesobserved an association between enterovirus antibodies andtype 1 diabetes, their validity has been disputed [3], mostlybecause of insufficient matching or unreliable antibody as-says. The advent of molecular testing opened new avenues ofresearch into enteroviruses as potential causative agents fortype 1 diabetes. Infections detected in the blood and itscomponents seem to be more strongly associated withan autoimmune response compared with gut infectionsdetected in faeces [1, 4].

Published prospective studies on islet autoimmunity thatutilise molecular detection of enterovirus from the blood aresurprisingly scarce. A recent review [5] and our literaturesearch identified six publications utilising molecular detectionfrom blood in case–control sets nested within prospectivecohorts. Indications of an association between enterovirusinfection and islet autoimmunity in these prospective studiescame from Finland [6–9], whereas another prospective mo-lecular study, Diabetes Autoimmunity Study in the Young(DAISY), from the USA [10] did not support such an associ-ation. Moreover, significant heterogeneity exists in the defini-tions of islet autoimmunity, in sampling frequency and inenterovirus detection methods.

The present study used the NorwegianMIDIA (Norwegianacronym for ‘Environmental Triggers of Type 1 Diabetes’)birth cohort. Norway shares many characteristics with itsneighbour Finland: among these are a very high incidence oftype 1 diabetes, latitude, shape of the territory, climatic con-ditions, a heterogeneous population density and socioeconom-ic conditions that include a high gross domestic product. Toreduce differences in detection methods, this study used anassay very similar to the one used in the Finnish studies. Incontrast to other studies, the participating children carried asingle high-risk genotype, and whole blood rather than serumwas used for enterovirus testing.

Our aim was to investigate a putative association betweenthe presence of enterovirus in the blood and islet auto-immunity in the Norwegian MIDIA study.

Methods

Enterovirus frequency (i.e. members of Enterovirus A, B, CandD species) was compared between cases (participants withislet autoimmunity) andmatched controls nestedwithin a birthcohort of Norwegian children with the highest-risk HLAgenotype.

Study cohort After genetic screening of 46,939 newbornsfrom the general Norwegian population during 2001–2007,the MIDIA study identified 1,047 newborns with the HLAclass II genotype conferring the highest risk of type 1 diabetes,DRB1*04:01-DQA1*03-DQB1*03:02/DRB1*03-DQA1*05-

DQB1*02. Of these participants, 911 were recruited forfollow-up. Three families later withdrew and their data weredeleted. Of the remaining 908 children, blood samples andquestionnaires were collected at ages 3, 6, 9 and 12 months,and annually thereafter, and plasma samples were tested forthe presence of islet autoantibodies, indicating isletautoimmunity.

Written parental consent was obtained. The study wasapproved by the Regional Committee for Medical ResearchEthics (Office for Human Research Protections IRB name‘Regional Med Resch Ethics Comm South IRB #2—South-East A’, IRB00001871) and the Norwegian Data ProtectionAuthority.

Nested case–control dataset By October 2011, 48 out of 908children in the observed cohort [11, 12] had developed isletautoimmunity (as defined below) and were included as cases.For each case, we randomly selected two controls from thecohort, matched for age (with a maximum tolerance of1 month wherever possible; this difference is narrow enoughto account for the known enterovirus seasonality becausesampling schedules of case and control participants wereidentical up to the first occurrence of islet autoantibodies)and county of residence (including the closest neighbouringcounty, if necessary). Children were ineligible as controls ifthey had one or more autoantibodies (as described below).Controls were followed at least until the time point of con-firmed islet autoimmunity in the corresponding case. To en-sure comparability, blood samples from cases that did nothave a matching control sample were removed from theanalysis. For three case and four control participants, therewas not enoughmaterial for enterovirus testing, and they weretherefore excluded from the analysis. The final case–controldataset included 45 cases and 92 controls: their characteristicsare listed in Table 1 and generation of the case–control datasetis shown in Fig. 1.

Sample handling, islet autoantibody assays and casedefinition Capillary blood samples were collected into tubescontaining EDTA and sent to the Norwegian Institute ofPublic Health by mail. Upon arrival, samples were centri-fuged: plasma was separated and tested for islet autoanti-bodies, while cell pellets containing residual plasma werestored at −80°C until further processing. All frozen sampleswere thawed once only.

Autoantibodies to insulin, 65 kDa glutamic acid de-carboxylase (GAD-65) and tyrosine phosphatase-likeprotein IA-2 were measured in duplicate by RIA in the hor-mone laboratory at Aker University Hospital, as previouslydescribed [12]. We defined islet autoimmunity as detection ofeither (1) two or more islet autoantibodies in two or moreconsecutive samples, or (2) at least one islet autoantibodytwice or two antibodies once followed by progression to

2194 Diabetologia (2014) 57:2193–2200

type 1 diabetes during the current follow-up, or (3) atleast one islet autoantibody in at least three consecutivesamples, including the most recent sample.

Nucleic acid extraction Frozen cell pellets containing red andwhite blood cells with a small amount of residual plasma werethawed at 4–8°C, and RNAwas co-purified with DNA usingthe TRIzol Plus RNA Purification System (Invitrogen, Carls-bad, CA, USA) according to the manufacturer’s protocol withminor modifications. Specifically, 0.1 ml of the blood cellpellet was added to 1 ml Trizol containing 3.7 μg carrierRNA (Qiagen, Hilden, Germany) and 0.5 μl 1% (vol/vol inPCR-grade water) Armored RNAWest Nile Virus (AsuragenDiagnostics, Austin, TX, USA) as an exogenous internalcontrol. The extraction was then performed as indicated inthe manufacturer’s protocol until binding to PureLink columnsand washing: these steps were not performed by centrifugation,but instead used a Qiagen vacuummanifold and vacuum pumpset to −20 to −30 kPa, with columns connected to the manifoldthrough Qiagen VacConnectors. Before the final elution steps,columns were centrifuged for 2 min at 12,000g to dry themembranes. Elution was performed into 120 μl AVE buffer(Qiagen) and used for downstream applications after an over-night incubation at 4°C.

Quantitative RT-PCR of enterovirus RNA Samples were test-ed for enterovirus, an exogenous internal control and humanRNA content. Enterovirus RNA was assayed using theQuantiTect Probe RT-PCR Kit (Qiagen), using 900 nMprimers and 300 nM probes (as reported in a previous Finnishstudy [13]). The combination of primers and probes reactswith an equal sensitivity to human Enterovirus A—D species(i.e. members of species Enterovirus A, Enterovirus B, En-terovirus C and Enterovirus D of the genus Enterovirus,family Picornaviridae, order Picornavirales, according tothe latest nomenclature of the International Committee onTaxonomy of Viruses, www.ictvonline.org/, accessed 9 May2014). The assay does not cross-react with Rhinovirus A–Cspecies.

The analysis was carried out using ABI7300 or ABI7700real-time thermocyclers and Sequence Detection Software(Applied Biosystems, Foster City, CA, USA). PCR conditionswere 30min reverse transcription at 50°C, followed by 15mininitial denaturation at 95°C, and 50 cycles of 15 s denaturationat 94°C and 1 min combined annealing–synthesis at 60°C.Absolute quantification was performed using an eight-pointstandard curve spanning the range from 375,000 to 1 copy permicrolitre Armored RNAWest Nile Virus (quantitated spec-trophotometrically by the manufacturer). Samples were run intriplicate. The assay was sufficiently sensitive to consistentlydetect a single copy per microlitre source RNA; of note, verylow virus quantities showed strong stochastic effects, withvarying number of positive replicates. To ensure robust

Table 1 Characteristics of case and control participants in the present study

Characteristic Participants

Case (n=45) Control (n=92)

Age at onset of islet autoimmunity(months)a

24.5 (6.2–75) NA

Female sex 28 (62.2%) 50 (54.4%)

Number of other children in the householdb

0 12 (26.7%) 30 (32.6%)

1 15 (33.3%) 45 (48.9%)

≥2 18 (40.0%) 17 (18.5%)

First-degree relatives with type 1 diabetesc

None 35 (77.8%) 87 (94.5%)

Yes

Siblings only 3 0

Father only 3 2

Mother only 2 3

Multiple family members 2 0

Progression to type 1 diabetes

Yes 20 (44.4%) n.a.

Age (years) at diabetes onset 4.2 (0.7–7.4)

Blood samples in the matched analysis

Totald 258 417

Before the development of isletautoimmunitya,e

144 251

Islet autoantibody seroconversionsamplef

38 62

After seroconversiong 76 104

Data are median (range), n (%) and n

NA, not availablea Age at the time when the first islet autoantibody-positive sample wascollectedb Including full siblings, half-siblings and step-siblings. The number of chil-dren was counted when the index child was 3 months old. Case participantstend to have more children in the family compared with controls (p=0.018)c Only full siblings were counted. Of the two cases with multiple affectedfirst-degree relatives, the father, mother and one sibling were affected inone case, and two siblings and the father (but not the mother) wereaffected in the other case. First-degree relatives with diabetes wereobserved more often in the families of cases than controls, p=0.009d Of the 807 samples tested for enterovirus and passing the qualitycontrol, 675 were included in the matched analysis. The remaining 132were removed because their matching sample was missinge These include the samples before the first islet autoantibody-positive sample,and the corresponding samples from controls. Several initial blood samplestaken in 2003 and 2004 were not available for enterovirus testing because ofthe lack of a cell pellet sample as the RNA source. Three cases and fourcontrols were affected; thus, in total, 45 cases and 92 controls were analysedf The first islet autoantibody-positive sample and its matched controlsample; several case samples could not be matched with a sufficientlysimilar control sampleg For association analysis with islet autoimmunity, samples without prop-erly age-matched controls (primarily after islet autoantibody seroconver-sion, owing to study design in which these participants were followed withmore frequent blood samples) were removed from analysis, and not allcases and controls contributed samples to all three periods (before, duringand after seroconversion)

Diabetologia (2014) 57:2193–2200 2195

detection, we defined samples as positive for enterovirus onlyif at least two out of three triplicate wells were positive.Negative controls were included in every extraction (at leastone per every batch of 15 samples, at a random location withinthe batch) and independently in the PCR. Neither of thesenegative controls tested positive.

In parallel with enterovirus and using an identical protocol,we also tested an exogenous internal control (a small amountofWest Nile Virus fragment added to the first extraction step),and the B2M gene transcript, a commonly used control that isabundant in human blood [14]. Threshold cycles wereinspected to identify those strongly deviating from the mean(signifying a lack of cells or possible failure of extraction).Samples that exceeded two SDs of the threshold cycle fromthe average were marked as suspect of a low RNA content,and this fact was taken into account in downstream analyses.Forty-one samples (4.8%) had low RNA content; however,three of these were still enterovirus positive. A second set ofnucleic acid extractions could not be performed because allblood sample material was used up in the first extraction.Therefore, we excluded these 41 suspect samples in the pri-mary analysis; in the sensitivity analysis, we re-ran all statis-tical analyses including these 41 samples, and results wereessentially unchanged.

Statistical analysis The association between enterovirus in-fections and islet autoimmunity was estimated using mixedeffects logistic regression models with enterovirus positivityin a blood sample as the dependent variable, and case or

control status (and other covariates) as independent variables.To account for the matched design and repeated measure-ments of enterovirus within individuals, random interceptswere specified for each matched set and for each individualwithin the matched set using the xtmelogit command withdefault settings in Stata, version 12 (StataCorp, College Sta-tion, TX, USA). The primary analysis used data from samplescollected strictly before islet autoantibody seroconversion.

Additional analyses included samples from the seroconver-sion period and after seroconversion. Case samples without aproperly age-matched control (primarily after islet autoanti-body seroconversion, owing to the study design where theseparticipants were followed with more frequent blood sam-ples), were removed from the analysis.

We performed two sensitivity analyses: (1) we assessed theinfluence of also defining as enterovirus positive those sam-ples that tested positive in a single replicate, i.e. we extendedthe positivity definition to samples likely to be affected by thestochastic character of low copy number templates; (2) weperformed a further sensitivity analysis including samplespresumed to contain low levels of RNA, as indicated byassessment of control templates.

Results

Enterovirus in blood and islet autoimmunity A total of 807samples were tested, of which 72 (8.9%) were positive forenterovirus. Positivity for enterovirus was not associated with

46,939 newborns screened for the highest-risk HLA genotype

1,047 infants carrying the high-risk genotype

Cohort of 908 high-risk children in the follow-up

48 cases in the present study (islet autoantibody positive or type 1 diabetic children)

96 controls matched 2:1 by:(1) age (± 1 month where possible)(2) county of residence (tolerating the closest neighbouring county if necessary) (3) length of follow-up (equal to or longer than the date of confirmed islet autoimmunity in index case).

48 matching trios

45 cases + 92 controls807 blood samples with enterovirus data

Lack of cell fraction for enterovirus testingof blood samples taken in 2003 and 2004

Fig. 1 Generation of thecase–control set used in thepresent study

2196 Diabetologia (2014) 57:2193–2200

the subsequent development of islet autoimmunity in samplestaken prior to islet autoantibody seroconversion (Table 2; OR0.75 [95% CI 0.36, 1.57]). There was a tendency towards anassociation with enterovirus detection in samples taken in theislet autoantibody seroconversion interval of case participants(6/38, 15.8%), and the corresponding time point in controls(2/62, 3.2%, OR 8.7 [95% CI 0.97, 77]). No significantassociation was observed after islet autoantibody seroconver-sion. The results are summarised in Table 2, where unadjustedmeasures of association are presented along with results frommodels adjusted for potential confounders and modifiers (sex,first-degree relative with diabetes [yes or no] and number ofother children in the household [coded as 0, 1 and 2 or more]).The results were essentially unchanged upon adjustment. Agraphical representation of enterovirus testing and isletautoimmunity data is provided in electronic supplementarymaterial (ESM) Fig. 1.

The cumulative probability of a first enterovirus infectionin blood was 26% (95%CI 19, 34) by the second birthday and38% (95% CI 30, 48) by 3 years of age (ESM Fig. 2), with nosignificant difference between case and control participants inthe timing of the first infection (data not shown).

The estimated virus quantity in blood was generally low:only three samples had more than 100 copies per microlitreRNA extract, 13 had 10–99 copies per microlitre, and allremaining positive signals contained fewer than 10 copiesper microlitre, where an accurate quantification is difficult.

As our definition of enterovirus positivity required two ormore positive replicates, we included a sensitivity analysiscontaining all low-quantity samples that were positive in onlyone of the three replicates (samples indicated by half-filledcircles in ESM Fig. 1; the total number of positive samplesincluded in this relaxed definition of positivity was 127/807,15.7%). This analysis did not alter either of the above conclu-sions (data not shown).

Enterovirus detection over the whole observation period Themedian number of samples tested for enterovirus per child wassix (range 1–14), and 55 participants (40.1%) had at least onepositive sample during follow-up. Forty participants (29.2%)had one enterovirus RNA-positive blood sample duringfollow-up, 13 (9.5%) participants had two positive samplesand two (1.5%) participants had three positive samples.Among the 15 participants for whom two or three enterovirusRNA-positive samples were observed during follow-up, fivehad consecutive samples positive for enterovirus RNA,whereas 10 had a second enterovirus RNA-positive sampleafter an intervening negative sample. If we assume that pos-itive consecutive samples formed part of a single infectiousepisode and the intermittent remaining positive samples rep-resented separate infection episodes, then there was a total of67 infectious episodes, of which 5 (7.5%) were ‘prolonged’(group 7, case; group 11, case and the first control; group 19,case; group 36, case; ESM Fig. 1). There was no detectable

Table 2 Frequency of human enterovirus in blood, and islet autoimmunity

Samples Participants OR (95% CI)a

Case Control Unadjusted Adjustedb

Collected before islet autoantibody seroconversionc

EV− 133 226 1.0 (reference) 1.0 (reference)

EV+ 11 (7.6%) 25 (10.0%) 0.75 (0.36, 1.57), p=0.44 0.60 (0.27, 1.32), p=0.20

Collected during seroconversiond

EV− 32 60 1.0 (reference) 1.0 (reference)

EV+ 6 (15.8%) 2 (3.2%) 8.7 (0.97, 77), p=0.053 9.08 (0.95, 86), p=0.055

Collected after seroconversione

EV− 68 98 1.0 (reference) 1.0 (reference)

EV+ 8 (10.5%) 6 (5.8%) 2.00 (0.64, 6.27), p=0.23 2.76 (0.87, 8.77), p=0.086

Data are n, n (%) and OR (95% CI)

EV− , enterovirus negative samples; EV+ , enterovirus positive samplesa Estimated from a mixed effects logistic regression model, with enterovirus positivity as the dependent variable and random intercepts for matched setand (within each matched set) individualsb Adjusted for sex, first-degree relative with type 1 diabetes (yes or no) and number of other children in the household (0, 1 or ≥2). The final adjustedmodel does not include other tested covariates such as age, time of blood sampling and introduction of cow’s milk before 3 months of age since they didnot influence the observed OR and did not improve the modelc Strictly excluding the sample first showing islet autoantibody positivityd Seroconversion time point defined as the first islet autoantibody-positive sample from case participants and corresponding matched controls. Note thatfor these samples we cannot know the temporal sequence of events, i.e. whether enterovirus RNA or islet autoantibody seroconversion occurred firste Seroconversion time point defined as first islet autoantibody-positive case sample and the control sample from the corresponding period

Diabetologia (2014) 57:2193–2200 2197

association between prolonged infections and seroconversionof islet autoantibodies.

Modifiers of enterovirus positivity The highest frequency ofenterovirus positivity was seen in samples obtained at 2 yearsof age (ESM Fig. 3a), and the frequency decreased thereafter.There was an appreciable seasonal pattern, with a peak in latesummer (p=0.019 in the Walter and Elwood test [15]; ESMFig. 3b). Enterovirus detection rates did not differ amongcalendar years.

The frequency of enterovirus positivity was similar inchildren exposed to cow’s milk products before vs after3 months of age (OR 0.68 [95% CI 0.36, 1.24]) and did notvary with sex (OR 1.05 [95% CI 0.64, 1.72] for males) norwith having first-degree relatives with diabetes (OR 1.21[95% CI 0.61, 2.39]). In contrast, enterovirus positivity wasmarginally increased when one or more other childrenwere present in the household compared with none (OR1.82 [95% CI 1.01, 3.28]).

We explored whether islet autoimmunity was associatedwith enterovirus RNA positivity in subgroups defined by theaforementioned factors (sex, children exposed before vs after3 months of age to cow’s milk products, first-degree relativeswith diabetes, presence of other children in the household andcalendar season of birth) using a stratified analysis and testingfor interactions. None of these tests indicated any subgroupeffects.

Discussion

Our results do not support the hypothesis that enterovirusRNA in longitudinally collected blood samples predicts thesubsequent development of islet autoimmunity, despite a highoverall enterovirus frequency and inclusion of the highestnumber of cases among the prospective molecular studiespublished so far.

Interestingly, in samples showing the first occurrence ofislet autoantibodies we observed an almost significantincrease in enterovirus RNA frequency. It is currently impos-sible to determine the sequence of events, i.e. whether entero-virus positivity occurred before islet autoimmunity or viceversa. If our enterovirus-positive samples reflected only acuteviraemia (which is usually short), then only infections towardsthe end of the sample interval would be detected; thus, ourdata would support reverse causation (increased susceptibilityto enterovirus in the blood as a result of islet autoimmunity).However, all but 16 positive samples contained only minutequantities of the virus, which is inconsistent with viraemicpeaks. Our source of RNA (i.e. packed blood cells) may bear acertain ‘memory’ of past viral infections because peripheralwhite blood cells are permissive to enterovirus infection [16]

and may be a better source of enterovirus RNA than plasma orserum [17, 18]. Thus, we speculate that the enterovirus posi-tivity we observed may reflect infection occurring over longerperiods of time, leaving the direction of the causalityunresolved.

Study strengths Using a strict definition of positivity, weobtained a high enterovirus frequency of 8.9% in blood sam-ples. This frequency is similar to that found in the DAISYstudy [10] (although in that study a visit was deemed positiveif serum, saliva or stool samples were positive) and is twofoldhigher than that of the Finnish DIPP (Diabetes Prediction andPrevention) study (4% and 3% sera positivity in case andcontrol participants) [7]. Our high enterovirus frequencymay thus also support the notion of the non-inferiority orsuperiority of whole blood compared with other sources ofnucleic acid, as previously observed in testing for otherviruses [19, 20].

Our study stems from the most homogeneous prospectivecohort yet published, which reduces the possible variation inenterovirus frequency resulting from HLA genotypes [21]. Astrict matching protocol was employed, demonstrated by theclustering of infections across the matching groups (ESMFig. 1). Furthermore, factors modifying enterovirus positivitywere either taken into consideration in the matched studydesign (age, place, season) or adjusted for in the analysis.

Comparison with other studies The prospective studies pub-lished so far on enterovirus and islet autoimmunity showvariations in several aspects, including sampling intervalsand definitions of positivity [1]. We specifically aimed tominimise methodological differences with prospective molec-ular studies from neighbouring Finland [6–9].

In the Finnish studies, a significantly higher enterovirusRNA frequency was noted in sera taken in the 6-month periodbefore the first detection of islet autoantibodies, thus reflectingan excess of acute viraemic infections during that period[7, 8], whereas no such association was observed in thepresent study. The differences in enterovirus frequency maypartly explain this discrepancy. There is an unusually lowenterovirus prevalence in Finland, which is several fold lowerthan in other populations [7, 22–24]. Assuming that a putativediabetogenic strain or serotype is hiding among a hundred ormore other, innocuous strains, its effect may be easilyovershadowed if these background strains are overly preva-lent. This may also be the case for the present Norwegianstudy, which shows the highest frequency so far of enterovirusRNA in the blood of prediabetic children.

Study limitations One limitation shared by this study, as wellas all other studies using pan-enterovirus RT-PCR analysis ofblood or serum [7, 8, 10], is the lack of sequence information.Although detection using the 5′-untranslated region is very

2198 Diabetologia (2014) 57:2193–2200

sensitive, the amplicon contains no information on serotypespecificity. Amplification of the VP1 (serotype-specific) re-gion is unfeasible in low positivity samples such as oursbecause degenerate PCR analysis requires relatively hightemplate concentrations. Second, if a hit-and-run infectionmechanism is involved in the pathogenesis of islet autoimmu-nity, then all currently available molecular studies—includingours—have an insufficient blood sampling frequency andwould therefore detect only a small fraction of the totalinfections.

In summary, we were not able to see a statistically signif-icant association between enterovirus RNA positivity in bloodand the subsequent development of islet autoantibodies inchildren with a high genetic risk of type 1 diabetes. Althoughinduction of islet autoantibodies tended to be associated withenterovirus detection in blood, we cannot conclude a causalrelationship between the two events.

Acknowledgements We thank the public healthcare nurses for theirrecruitment efforts for the MIDIA study and the follow-up of high-riskchildren, and the staff at the Biobank Department at the NorwegianInstitute of Public Health. In particular, we would like to express ourgratitude to all of the parents for their efforts in handling their children’stype 1 diabetes risk, for allowing samples to be taken from their childrenand for completing questionnaires.

Funding This study and the MIDIA project were funded by theNorwegian Organization for Health and Rehabilitation (2008/0182), theMinistry of Health of the CzechRepublic (IGAMZ11465–5), the ResearchCouncil of Norway (grants 135893/330, 155300/320, 156477/730,205086/F20 and 166515/V50), the Norwegian Diabetes Association, theAcademy of Finland (grant to HH) and the Project for the ConceptualDevelopment of Research Organisation 00064203 (University HospitalMotol, Prague, Czech Republic). OC’s sabbatical at HH’s laboratory in2012 was supported by an ISPAD Research Fellowship.

Duality of interest HH is a minor shareholder (<5%) of Vactech Ltd.,which develops picornavirus vaccines. All other authors declare that thereis no duality of interest associated with their contribution to thismanuscript.

Contribution statement KSR and LCS conceived and designed thestudy. OC, LK, SO and HH collected data (performed or supervised theenterovirus testing), PAT collected data (tested islet autoantibodies), andTR managed the databases. LCS, GT, EW, TR and OC analysed andinterpreted data. All authors drafted the manuscript and/or revised it forimportant intellectual content, and have approved the final version.

LCS is the guarantor of this work and, as such, had full access to allthe data in the study and takes responsibility for the integrity of the dataand the accuracy of the data analysis.

References

1. Stene LC, Rewers M (2012) Immunology in the clinic review series;focus on type 1 diabetes and viruses: the enterovirus link to type 1diabetes: critical review of human studies. Clin Exp Immunol 168:12–23

2. Hyoty H, Taylor KW (2002) The role of viruses in human diabetes.Diabetologia 45:1353–1361

3. Green J, Casabonne D, Newton R (2004) Coxsackie B virus serologyand Type 1 diabetes mellitus: a systematic review of published case-control studies. Diabet Med 21:507–514

4. Tauriainen S, Oikarinen S, Oikarinen M, Hyoty H (2011) Enterovirusesin the pathogenesis of type 1 diabetes. Semin Immunopathol33:45–55

5. Yeung WC, Rawlinson WD, Craig ME (2011) Enterovirus infectionand type 1 diabetes mellitus: systematic review and meta-analysis ofobservational molecular studies. BMJ 342:d35

6. Salminen KK, Vuorinen T, Oikarinen S et al (2004) Isolation ofenterovirus strains from children with preclinical Type 1 diabetes.Diabet Med 21:156–164

7. Salminen K, Sadeharju K, Lonnrot M et al (2003) Enterovirusinfections are associated with the induction of beta-cell autoimmunityin a prospective birth cohort study. J Med Virol 69:91–98

8. Oikarinen S, Martiskainen M, Tauriainen S et al (2011) EnterovirusRNA in blood is linked to the development of type 1 diabetes.Diabetes 60:276–279

9. Sadeharju K, Hamalainen AM, Knip M et al (2003) Enterovirusinfections as a risk factor for type I diabetes: virus analyses in adietary intervention trial. Clin Exp Immunol 132:271–277

10. Graves PM, Rotbart HA, Nix WA et al (2003) Prospective studyof enteroviral infections and development of beta-cell autoim-munity. Diabetes autoimmunity study in the young (DAISY).Diabetes Res Clin Pract 59:51–61

11. Tapia G, Cinek O, Rasmussen Tet al (2011) Human enterovirus RNAin monthly fecal samples and islet autoimmunity in Norwegianchildren with high genetic risk for type 1 diabetes: the MIDIA study.Diabetes Care 34:151–155

12. Stene LC, Witso E, Torjesen PA et al (2007) Islet autoantibodydevelopment during follow-up of high-risk children from the generalNorwegian population from three months of age: design and earlyresults from the MIDIA study. J Autoimmun 29:44–51

13. Honkanen H, Oikarinen S, Pakkanen O et al (2013) Human entero-virus 71 strains in the background population and in hospital patientsin Finland. J Clin Virol 56:348–353

14. Beillard E, Pallisgaard N, van der Velden VH et al (2003) Evaluationof candidate control genes for diagnosis and residual disease detec-tion in leukemic patients using 'real-time' quantitative reverse-transcriptase polymerase chain reaction (RQ-PCR) - a Europe againstcancer program. Leukemia 17:2474–2486

15. Walter SD, Elwood JM (1975) A test for seasonality of events with avariable population at risk. Br J Prev Soc Med 29:18–21

16. Hwang JY, Jun EJ, Seo I et al (2012) Characterization of infections ofhuman leukocytes by non-polio enteroviruses. Intervirology55:333–341

17. Yin H, Berg AK, Tuvemo T, Frisk G (2002) EnterovirusRNA is found in peripheral blood mononuclear cells in amajority of type 1 diabetic children at onset. Diabetes 51:1964–1971

18. Schulte BM, Bakkers J, Lanke KH et al (2010) Detection ofenterovirus RNA in peripheral blood mononuclear cells of type 1diabetic patients beyond the stage of acute infection. Viral Immunol23:99–104

19. Perlman J, Gibson C, Pounds SB, Gu Z, Bankowski MJ, Hayden RT(2007) Quantitative real-time PCR detection of adenovirus inclinical blood specimens: a comparison of plasma, wholeblood and peripheral blood mononuclear cells. J Clin Virol40:295–300

20. Ruf S, Behnke-Hall K, Gruhn B et al (2012) Comparison of sixdifferent specimen types for Epstein-Barr viral load quantificationin peripheral blood of pediatric patients after heart transplantation orafter allogeneic hematopoietic stem cell transplantation. J Clin Virol53:186–194

Diabetologia (2014) 57:2193–2200 2199

21. Sadeharju K, Knip M, Hiltunen M, Akerblom HK, Hyoty H (2003)The HLA-DR phenotype modulates the humoral immune response toenterovirus antigens. Diabetologia 46:1100–1105

22. Oikarinen S, Tauriainen S, Hober D et al (2014) Virus antibodysurvey in different European populations indicates risk asso-ciation between Coxsackievirus B1 and Type 1 diabetes.Diabetes 63:655–662

23. Viskari H, Ludvigsson J, Uibo R et al (2004) Relationship betweenthe incidence of type 1 diabetes and enterovirus infections in differentEuropean populations: results from the EPIVIR project. J Med Virol72:610–617

24. Lonnrot M, Knip M, Marciulionyte D et al (1999) Enterovirus anti-bodies in relation to islet cell antibodies in two populations with highand low incidence of type 1 diabetes. Diabetes Care 22:2086–2088

2200 Diabetologia (2014) 57:2193–2200