Virally induced inflammatory responses, beta cell destruction and release of beta cell autoantigens may lead to autoimmune reactions culminating in type 1 diabetes. Therefore, viral capability to induce beta cell death and the nature of virus-induced immune responses are among key determinants of diabetogenic viruses. We hypothesised that enterovirus infection induces a specific gene expression pattern that results in islet destruction and that such a host response pattern is not shared among all enterovirus infections but varies between virus strains.
The changes in global gene expression and secreted cytokine profiles induced by lytic or benign enterovirus infections were studied in primary human pancreatic islet using DNA microarrays and viral strains either isolated at the clinical onset of type 1 diabetes or capable of causing a diabetes-like condition in mice.
The expression of pro-inflammatory cytokine genes (IL-1-α, IL-1-β and TNF-α) that also mediate cytokine-induced beta cell dysfunction correlated with the lytic potential of a virus. Temporally increasing gene expression levels of double-stranded RNA recognition receptors, antiviral molecules, cytokines and chemokines were detected for all studied virus strains. Lytic coxsackievirus B5 (CBV-5)-DS infection also downregulated genes involved in glycolysis and insulin secretion.
The results suggest a distinct, virus-strain-specific, gene expression pattern leading to pancreatic islet destruction and pro-inflammatory effects after enterovirus infection. However, neither viral replication nor cytotoxic cytokine production alone are sufficient to induce necrotic cell death. More likely the combined effect of these and possibly cellular energy depletion lie behind the enterovirus-induced necrosis of islets.
KeywordsCoxsackievirus Echovirus Enterovirus Interleukin 1 Microarray analysis Pancreatic beta cells Pancreatic islets Tumour necrosis factor Type 1 diabetes
Activator protein 1
Gene set enrichment analysis
Inducible form of nitric oxide synthase
c-Jun N-terminal kinase
Mitogen-activated protein kinase
Nucleotide-binding oligomerisation domain containing
Nuclear factor of kappa light polypeptide gene enhancer in B cells
P38 mitogen-activated protein kinases
Pattern recognition receptor
Reactive oxygen species
TNF-related apoptosis-inducing ligand
The rapid worldwide increase in incidence of type 1 diabetes cannot be explained by a change in genetic predisposition, suggesting that environmental factors play an important role [1, 2]. In this respect viral infections have been mentioned and, according to cross-sectional and prospective studies, human enteroviruses (HEVs) are the strongest candidates (reviewed in ). Furthermore, HEV genomes, proteins and/or infections have been reported in pancreatic islets of deceased patients with type 1 diabetes [4–6], indicating that HEVs are capable of reaching the islets during infection.
Direct cytolysis and/or virus-targeted innate immune responses are held responsible for some cases of fulminant type 1 diabetes without evidence for autoimmunity [7–10]. However, it is possible that virus-induced host responses culminate in autoimmune reactions that are involved in type 1 diabetes in the majority of type 1 diabetes patients. In one possible scenario, virus-induced beta cell damage would result in the presentation of autoantigens to the immune system in a local inflammatory milieu with upregulation of MHC molecules and secretion of pro-inflammatory cytokines. Uptake and presentation of autoantigens by antigen-presenting cells would promote further beta cell damage by augmenting the innate antiviral responses and by recruiting and activating autoreactive T cells [11, 12]. Thus, beta cell tropism and virus-induced cytolysis, combined with host-cell responses to infection (e.g. inflammatory responses), are likely to determine the diabetogenic potential of a virus.
HEVs are a large and genetically diverse group of RNA viruses with over 100 known serotypes . Although HEVs are often implicated as triggers of type 1 diabetes, it is not known precisely which of the HEVs are involved in type 1 diabetes development. In-vitro studies with human islets suggest that enterovirus strains differ in their capability to induce islet destruction [14–17]. Some HEV strains are highly cytolytic in vitro whereas others replicate without apparent islet destruction [15–17]. The ability to induce islet destruction is not determined solely by the virus serotype, but strain-specific differences exist within some serotypes (e.g. echovirus 9 [E-9] and 30 [E-30]) . This suggests a role for viral genetic factors in beta cell dysfunction and death. Host-cell factors have a role in the process since there is a significant delay between viral growth and islet destruction, even for highly destructive virus strains [14, 15]. Therefore, pancreatic beta cell tropism and the ability to induce beta cell dysfunction and death depends both on the genetic properties of the virus and on the host-cell response to the infection.
The outcome of islet infection (i.e. destructive vs benign) may have implications on type 1 diabetes pathogenesis, since lytic infection (with necrotic cell death) is likely to induce a strong inflammatory response that, together with presentation of beta cell autoantigens, could induce autoimmunity. Benign infection, on the other hand, may induce persistent infection  and thus lead to sustained stimulation of autoimmunity.
In a previous study, a complex network of hundreds of genes participating in islet responses to a viral infection was identified . In addition to upregulation of multiple inflammatory mediators probably contributing to the homing and activation of immune cells in the islets during infection and/or autoimmune response, HEV infection is capable of changing expression of several cytokines that directly contribute to islet destruction . However, a prolonged exposure of human beta cells to cytokines and/or double-stranded RNA (dsRNA) culminates in apoptosis [19–21] whereas most coxsackievirus B5 (CVB-5)-infected human pancreatic islet cells die by necrosis . The pathways leading to beta cell apoptosis have been studied extensively , but the molecular networks leading to necrosis after HEV infection are not known.
We hypothesised that: (1) HEV infection induces a specific gene expression pattern that results in islet destruction and (2) that the ability to induce such a pattern is not common for all HEVs, but that different HEV strains may trigger distinct gene expression profiles leading to either a destructive or a merely benign outcome.
To identify a ‘destructive’ gene expression pattern, human pancreatic islets were infected in vitro with beta cell tropic HEV strains that induce varying degrees of necrosis, ranging from highly lytic to benign. Since the outcome of infection may depend on the timing of the expression changes, the dynamics of human pancreatic islet responses to infection were tested for a highly cytolytic CVB-5-DS strain leading to islet necrosis using two time points, representing the time of maximum virus concentration and the time before apparent cellular damage. A ‘destructive gene expression profile’ was further assessed by comparing the gene expression patterns induced by beta cell destructive and benign E-9 clones, one leading to necrosis, the other replicating with no apparent damage to islets. These E-9 clones differed only by two amino acids, thereby minimising the effect of viral genetics.
Human pancreatic islets
Pancreatic islets were isolated and purified, as described previously , at Uppsala University Hospital, Sweden, or Islet Cell Resource (ICR) Center Basic Science Islet Distribution Program, with the consent of corresponding ethics committees.
Pancreatic islets from four donors (three men, one woman; age 50–64 years) were used for microarray analysis (electronic supplementary material [ESM] Table 1). The cold ischaemia time was between 5 and 11 h. The mean purity of the islets was 80 ± 8% (mean ± SD). For viral replication and viability analyses six additional preparations (four men, two women; age 19–63 years; purity 50–85%) were used.
The islets were maintained in non-adherent culture plates (Falcon; Becton Dickinson Lincoln Park, NJ, USA) in Hams F-10 cell culture medium (Sigma-Aldrich, St Louis, MO, USA) containing 2% FBS, 25 mmol/l HEPES (pH 7.4), penicillin (100 U/ml) and streptomycin (0.1 mg/ml). The culture medium was changed twice a week.
CVB-5-DS  and molecular clones derived from echovirus 9 strain DM (E-9-DM) [9, 17] were used in the study. The infective clones were constructed as described previously . The E-9 clones have been characterised at the molecular and biological levels (A. Paananen, P. Ylipaasto, T. Smura, O. Korsgren, J. Galama and M. Roivainen, unpublished data).
Infection and viral growth
The pancreatic islets were infected as described previously . For viral growth determination, islets were harvested at different intervals and freeze–thawed four times. The infectivity of each sample was determined by endpoint titration in 96-well plates of green monkey kidney (GMK) cells .
The islet samples were stained using antivirus and anti-insulin antibodies (The Binding Site, Birmingham, UK), as described previously . Slides were analysed using a confocal microscope (Leica TCS SPE; Wetzlar, Germany)
The viability of islets was monitored using Live/Dead assay (Invitrogen, Carlsbad, CA, USA) as described previously .
Total RNA of the islets was extracted using Trizol reagent (Invitrogen) according to manufacturer’s instructions. The RNA was cleaned using RNeasy (Qiagen, Hilden, Germany) cleanup protocol with DNA digestion treatment (Qiagen).
The quality and quantity of RNA was controlled with electrophoresis (Biorad Laboratories, Hercules, CA, USA) and spectrophotometer (Nanodrop ND-100; Thermo Fisher Scientific, Waltham, MA, USA), respectively. Affymetrix HG U133A chips or Illumina Sentrix HumanHT-12 Expression BeadChips were hybridised with the samples according to manufacturers’ protocols.
Gene expression analysis
The mRNA expression levels of infected islets and mock-infected control islets were compared. For Illumina probes the gene expression levels were considered to be changed if the mean fold change in expression signal between infected islets and mock-infected control islets was equal to or more than twofold or statistically significant. For Affymetrix probes, manufacturer’s recommendations were used, as described previously .
Gene set enrichment analysis
Gene set enrichment analysis using GSEA 2.06 and Molecular Signature Database (MSigDB) was performed to detect statistically significant and concordant changes of genes belonging to known gene sets [26, 27]. Data from Affymetrix gene chip were analysed using a pre-ranked list based on the signal log ratio. The data from Illumina arrays were analysed using signal to noise ratio as the metric for ranking genes. Gene set permutations were used to determine the statistical significance of the enrichment score of pathways.
Multiplex bead-based assays based on xMAP technology (Bio-Plex; Biorad Laboratories) were used to assay cytokine/chemokine concentrations in the culture-medium supernatant fractions of the islets. ESM Table 2 contains the limits of detection and the coefficients of variability of these assays. Statistical significance was evaluated using Student’s t test.
The statistical significance of difference between means in the quantification of mRNA was calculated with paired two-tailed Student’s t test. A p value less than 0.05 was considered statistically significant.
Pancreatic islet gene expression profile induced by beta cell tropic CBV-5-DS strain
In search of HEV-induced gene expression patterns that result in islet destruction, the dynamics of islet mRNA expression and cytokine secretion were measured after infection with the beta cell tropic and highly cytolytic CVB-5-DS strain.
Global mRNA expression and cytokine secretion were measured 1 and 3 days p.i, representing the time of maximum virus concentration and the time before apparent cellular damage. GSEA analysis indicated significant upregulation in the pathways related to inflammation, cytokines, innate immune response, antigen processing and presentation, cell death (both apoptosis and necrosis) and Janus kinase (JAK)/signal transducer and activator of transcription (STAT), mitogen-activated protein kinase (MAPK) and nuclear factor of kappa light polypeptide gene enhancer in B cells (NFκB) signalling cascades, whereas the metabolic pathways (including glycolysis and oxidative phosphorylation) were downregulated (ESM Fig. 1).
Immunoregulatory gene expression pattern after CVB-5-DS infection
Corresponding to increased IFN production, the expression of several genes coding for IFN-inducible antiviral defence molecules were upregulated (ESM Table 3). Likewise, upregulation of intracellular (viral) dsRNA and single-stranded RNA (ssRNA) receptors DDX58 (RIG-I), IFIH1 (MDA-5), DHX58 (LGP2) toll-like receptor 3 (TLR3) and nucleotide-binding oligomerisation domain containing (NOD)-like receptor 2 (NOD2) was detected (Fig. 2a).
The effects of PRR activation are mediated mainly via three signal transduction cascades–JAK/STAT-, MAPK- and NFκB-mediated pathways, all of which showed upregulation in GSEA analysis (ESM Fig. 1). For NFκB and MAPK, upregulation of inhibitors were also detected (Fig. 2a). The expression of genes for three MAPKs, extracellular-signal-regulated kinase (ERK) (MAPK3), c-Jun N-terminal kinase (JNK) (MAPK10) and p38 (MAPK11), were downregulated. MAPKs regulate the activity of the activator protein-1 (AP-1) transcription factor complex. AP-1 transcription factor members ATF3, ATF4 and ELK4 as well as AP-1 complex suppressors BATF2 (SARI) and BATF3 (JDP1) were upregulated after CBV-5-DS infection.
The expression of multiple cytokine and chemokine genes were upregulated by CVB-5-DS infection (Fig. 2b). Furthermore, most of the observed upregulations increased from 1 to 3 days p.i. and were also detected in the tissue culture medium after infection (ESM Table 2).
The expression of several genes involved in antigen processing and presentation were upregulated after CVB-5-DS infection. These included regulatory members of immunoproteasome complex (PSMB8, PSMB9, PSMB10, PSME2), antigen peptide transporters (TAP1 and TAP2), as well as several MHC class I molecules (Fig. 2a).
Beta cell function-related gene expression after CVB-5-DS infection
In addition to its immunomodulatory effects, CVB-5-DS infection changed the expression of several genes important for beta cell function and metabolism (ESM Fig. 1), including genes associated with glycolysis and insulin secretion.
The signalling cascade for insulin secretion in pancreatic beta cells begins with the closure of ATP-dependent K+ channels due to an increased intracellular ATP/ADP ratio. This leads to cell membrane depolarisation, opening of voltage-sensitive Ca2+ channels and release of insulin stored in secretory vesicles due to increased cytosolic Ca2+ concentration. GSEA analysis suggested downregulation of glycolysis and oxidative phosphorylation after CVB-5-DS infection (Fig. 2c, ESM Fig. 1). Likewise, genes for a modulator of ATP-dependent K+ channels (ABCC8 [SUR1]) and voltage-dependent Ca2+ channel subunits (CACNB2 and CACNA2D2) as well as ATP2A3 (SERCA), an endoplasmic reticulum (ER) Ca2+ pump that translocates calcium from the cytosol to the ER lumen were downregulated (Fig. 2c, ESM Table 3). The expression of several genes coding for other (K+, Na+ and Cl−) ion channel subunits and modulators were also downregulated after CVB-5-DS infection (ESM Table 3).
In addition to apparently impaired intracellular Ca2+ homeostasis and membrane potential regulation, the expression of several genes involved in insulin production and secretion were changed after CVB-5-DS infection (Fig. 2c). These included genes associated with insulin secretory granule assembly, hydrolases responsible for synthesis of insulin from proinsulin, enzymes required for acidification of the granule lumen and molecules associated with granule traffic and release, as well as granule cargo proteins co-secreted with insulin. These results are in line with previous studies that indicate impaired beta cell insulin secretion after CBV-5 infection [14, 15].
Pancreatic islet gene expression profiles induced by lytic vs non-lytic virus strains
Several genes were similarly upregulated after infection with any of the enterovirus strains (i.e. CVB-5-DS vs both E-9 strains). These included genes encoding PRRs (DDX58, IFIH1 and NOD2), IFNs (IFNB1, IL28 and IL29), antiviral molecules, STAT1, the regulatory members of immunoproteasome complex (PSMB8, PSMB9, PSMB10 and PSME2) and antigen peptide transporters (TAP1 and TAP2).
Also differences between lytic and non-lytic strains were detected (Fig. 4c, ESM Fig. 2, ESM Table 4). Upregulation of chemokine gene expression was stronger after lytic E-9-M1-12 infection than after benign E-9-M2-2 infection. The expression of genes encoding pro-inflammatory cytokines and beta cell apoptosis inducers IL-1α, IL-1β, TNF-α and tumour necrosis factor superfamily 10 (TNFSF10; TRAIL) were upregulated by cytolytic virus strains CVB-5-DS and E-9-M1-12, but not by the benign strain E-9-M2-2 (Fig. 2b, Fig. 4c, ESM Fig. 2). Furthermore, TRAIL receptor gene TNFRSF10D was downregulated after infection with the benign E-9-M2-2 strain.
Both apoptotic and anti-apoptotic genes were upregulated after infection with the cytolytic virus strains. These included genes associated with IAP (inhibitor of apoptosis) protein family members (BIRC3), X-linked inhibitor of apoptosis (XIAP, BIRC4) and XIAP antagoniser (XAF1, XIAP associated factor 1) as well as genes for B cell lymphoma-2 protein family members, such as PMAIP1 (NOXA) and MCL1 (Fig. 4c). Benign E-9-M2-2 infection induced downregulation of genes encoding a cell-cycle and apoptosis regulator, tumour protein p53 (TP53), and p53 activator ataxia telangiectasia mutated (ATM) as well as upregulation of p53 inhibitor gene MDM4 (Fig. 4c).
Out of the genes involved in cellular responses to reactive oxygen species (ROS), all three virus strains induced upregulation of mitochondrial superoxide dismutase 2 (SOD2). After CVB-5-DS infection the expression of genes coding for glutathione peroxidase 3 (GPX3), peroxiredoxin 2 (PRDX2) and sestrin 1 (SESN1) (that regenerates overoxidised peroxiredoxins), were downregulated, whereas E-9-M1-12 infection induced downregulation of the gene for catalase (CAT) (Fig. 4c, ESM Table 3).
ROS-induced DNA damage is counteracted by DNA repair enzymes such as poly-(ADP-ribose) polymerase (PARP). The expression of genes encoding several members of the PARP family was upregulated by all of the HEV strains, although PARP1 was downregulated after CVB-5-DS infection (Fig. 4c, ESM Table 3).
The ESM text discusses changes observed in metabolomics studies after infection of islets with CVB-5-DS.
Innate immunity and inflammatory mediators have a major role in the development of type 1 diabetes . The potential pathogenetic link between HEV infection and type 1 diabetes can therefore be the linkage of virus infection in beta cells with an inflammatory response. Here we studied the connection between HEV-induced islet destruction, viral genetics and host-cell response to infection and detected a highly pro-inflammatory islet response to the infection. Furthermore, the expression of pro-inflammatory cytokine genes (IL1A, IL1B, TNF and TRAIL) correlated with the lytic potential of a virus.
The outcome of islet infection (i.e. destructive vs benign) and inflammatory response are probably interconnected, since the major mechanism of islet destruction after HEV infection is necrosis . This is likely to have a profound effect on the local inflammatory milieu at the site of infection, since necrosis represents a severe form of death where the cell content leaks into the surrounding tissue causing damage to bystander cells and inducing strong inflammatory response . Thus, in addition to impaired beta cell function, islet infection with a lytic HEV strain may lead to autoantigen presentation in an inflammatory milieu, which may induce autoimmune reactions.
Some HEV strains are highly cytolytic in pancreatic islets whereas others replicate with no apparent islet destruction [15–17]. This indicates that viral strain-specific genetic factors partly determine the outcome of infection. Since there is a significant delay between viral growth and islet destruction [14, 15], we hypothesised that lytic HEV strains induce a distinct host response profile leading to islet destruction, a profile that differs from that caused by other strains having a more benign effect. This hypothesis was tested by comparing the gene expression patterns after lytic and ‘benign’ HEV infections.
Complete islet preparations were used in this study. Thus, the gene expression patterns monitored reflect the combined effects of virus and virus-induced factors on all of the cell types present. This is likely to mimic the situation in vivo, where the fate of a beta cell is affected by responses of neighbouring non-beta cells. Since the islet response to HEV infection involves multiple interconnected networks of genes, the experiments were designed to reveal large-scale patterns of changes in gene expression over a prolonged period rather than the impact of a single or few pathways. Therefore, the conclusions deduced from this study need to be further elucidated with more detailed experiments.
In a previous study, a complex network of genes participating in islet responses to a viral infection was identified . In the present study, the changes in gene expression were measured at two time points to gain information about temporal changes in gene expression patterns. The number of regulated genes increased from the first to the second time point for the lytic strains, but remained constant for the non-lytic strain (Fig. 4a, b, ESM Table 3). This suggests that the dynamics of the changes in gene expression is associated with islet destruction. Furthermore, the changes induced by CVB-5 were compared with those of lytic and non-lytic E-9 infections to study which of the genes are regulated by HEV infection in general and which of the gene expression changes are specific for lytic infections.
The benign HEV strain induced fewer changes in gene expression than the lytic HEV strains. All of the virus strains upregulated the expression of genes associated with key mediators of the innate immune response, including type I and type III IFNs, dsRNA receptors, antiviral molecules and corresponding signal transduction cascades; this may have implications on the outcome of infection. IFNs upregulate the expression of genes encoding PRRs in human pancreatic islets [19, 32], thus forming a positive feedback loop that has to be tightly regulated. The lack of accurate negative regulation of PRRs would lead to overt innate immune reactions, whereas lack of accurate positive regulation would lead to an impaired antimicrobial response. The temporally increasing upregulation of genes involved in the innate immune response suggests that a lytic HEV infection may drive this positive feedback loop towards overt innate immune response, whereas for the benign strains the innate immune response is weaker, potentially providing a possibility for establishment of chronic infection.
Many of the PRR- and IFN-elicited responses in beta cells are mediated by NFκB , STAT-1  and MAPK  signal transduction pathways, all of which showed upregulation after infection. However, NFκB and MAPK pathways also showed increased inhibition (both pathways) and/or downregulation of the signal cascade molecules (MAPK) after CVB-5-DS infection. This may reflect negative feedback mechanisms aimed at inhibiting overt activation of these pathways.
PRR-mediated induction of IFNs and transcription factors leads to the expression of pro-inflammatory cytokines, chemokines and adhesion molecules, which collectively regulate the recruitment of immune cells to sites of infection and modulate the magnitude and duration of the immune response. The expression of genes associated with pro-inflammatory cytokines IL-1α, IL-1β and TNF-α were upregulated by lytic strains (CBV-5-DS and E-9-M1-12) only. Furthermore, consistent with previous studies [19, 32, 35–37], the cytolytic enterovirus strains induced strong stimulation of genes involved in chemokine secretion, whereas after infection with the benign (E-9-M2-2) strain, the chemokine response was weaker. These results suggest that lytic HEV strains are more potent inducers of a pro-inflammatory micro-environment and probably recruit immune cells to the site of infection more efficiently.
IL-1α, IL-1β and TNF-α induce cellular death in pancreatic islets [38, 39]. However, cytokine production induced by HEV infection cannot be the sole factor leading to destruction of pancreatic islets. Exposure of human islets to these cytokines induces cell death by apoptosis [21, 40, 41], whereas HEV infection induces secondary necrosis after initial nuclear pyknosis . Only a small fraction of beta cells show signs of apoptosis after infection [14, 42]. Accordingly, the GSEA analysis indicated upregulation of apoptosis and necrosis pathways together with shutdown of metabolic pathways. The reduction of phosphatidylcholines, the major constituents of biological membranes, revealed by the lipidomics, may be a result of cell lysis resulting from necrosis.
The switch between apoptosis and necrosis may be determined by the availability of ATP [43, 44]. Downregulation of glycolysis and oxidative phosphorylation by CVB-5 infection might lead to intracellular depletion of ATP and NADH and, thereby, combined with IL-1, TNF-α and TRAIL expression, induce secondary necrosis. In this respect HEV-infected human islets may resemble rodent islets, which, when exposed to IL-1β, IFN-γ and TNF-α, die predominantly by necrosis , possibly due to depletion of cellular ATP and NAD+ [22, 46]. However, in murine beta cells the ATP and NAD+ depletion is dependent on nitric oxide produced by the inducible form of nitric oxide synthase (iNOS) while in human beta cell death the role of iNOS is controversial [19, 22].
The ATP levels may be further lowered by PARP family members [46, 47] that are involved in ROS-induced DNA strand-break repair. Excessive poly-(ADP-ribose) formation by PARP may deplete cellular NAD+ pools and therefore induce necrosis in several cell types, including pancreatic beta cells . Consistently, several members of the PARP family were upregulated by HEV infection.
These results suggest that (analogous to a hypothesis by Eizirik and Mandrup-Poulsen ) HEV-induced islet cell necrosis might result from a combined effect of cytotoxic cytokines and ATP depletion due to downregulated glycolysis and oxidative phosphorylation and possibly via PARP overactivation.
In conclusion, the results support the hypothesis that a distinct gene expression profile correlates with HEV-induced human pancreatic islet necrosis. This profile includes temporally increasing gene expression levels of pro-inflammatory cytokines (IL-1α, IL-1β, TNF-α and TRAIL), which are considered to be key mediators of cytokine-induced beta cell dysfunction. This virally induced destructive gene expression profile may be linked to generation or maintenance of autoimmune reactions since it combines pro-inflammatory cytokine/chemokine expression with beta cell destruction.
Human pancreatic islets were provided by Islet Cell Resource (ICR) Center Basic Science Islet Distribution Program (City of Hope National Medical Center, Duarte, CA, USA). The authors thank M. Eskelinen (National Institute for Health and Welfare [THL], Helsinki, Finland), M. Kyläniemi (Turku Centre for Biotechnology, University of Turku and Åbo Akademi University, Turku, Finland) and S. Heinonen (Turku Centre for Biotechnology, University of Turku and Åbo Akademi University, Turku, Finland), for expert technical assistance.
This study was supported by The Juvenile Diabetes Research Foundation (USA), The European Union (EP7-HEALTH-2007, DIAPREPP N202013), The Academy of Finland, Päivikki and Sakari Sohlberg Foundation, The Finnish Diabetes Foundation and The Maud Kuistila Memorial Foundation.
Duality of interest
The authors declare that there is no duality of interest associated with this manuscript.
PY, PG, AP, SK, HA analysed and interpreted data and edited the manuscript. TS analysed and interpreted data and drafted the manuscript. OK, JRTL, RL, TSL, MO, LP and JG designed experiments and revised the manuscript. MR is the principal investigator, conceived and designed experiments, revised the manuscript and critically reviewed the manuscript for intellectual content. The final version of the manuscript was approved by all the authors.