Longitudinal Pattern of First-Phase Insulin Response Is Associated With Genetic Variants Outside the Class II HLA Region in Children With Multiple Autoantibodies


Abstract

A declining first-phase insulin response (FPIR) is associated with positivity for multiple islet autoantibodies, irrespective of class II HLA DR-DQ genotype. We examined the associations of FPIR with genetic variants outside the HLA DR-DQ region in the Finnish Type 1 Diabetes Prediction and Prevention (DIPP) study in children with and without multiple autoantibodies. Association between FPIR and class I alleles A*24 and B*39 and eight single nucleotide polymorphisms outside the HLA region were analyzed in 438 children who had one or more FPIR results available after seroconversion. Hierarchical linear mixed models were used to analyze repeated measurements of FPIR. In children with multiple autoantibodies, the change in FPIR over time was significantly different between those with various PTPN2 (rs45450798), FUT2 (rs601338), CTSH (rs3825932), and IKZF4 (rs1701704) genotypes in at least one of the models. In general, children carrying susceptibility alleles for type 1 diabetes experienced a more rapid decline in insulin secretion compared with children without susceptibility alleles. The presence of the class I HLA A*24 allele was also associated with a steeper decline of FPIR over time in children with multiple autoantibodies. Certain genetic variants outside the class II HLA region may have a significant impact on the longitudinal pattern of FPIR.

Introduction

The first-phase insulin response (FPIR), a marker reflecting functional capacity of the β-cells in the pancreas, increases physiologically over time in children and adolescents (1). As a sign of deteriorating β-cell function, a decline in FPIR can, however, be observed several years before clinical type 1 diabetes (T1D) (1).

The class II HLA DR-DQ region has been shown to affect the appearance of islet-specific autoantibodies. Children with multiple autoantibodies have a high risk of progressing to clinical disease, and the presence of multiple autoantibodies seems to represent a point of no return (2). However, class II HLA does not have any effect on the progression rate from advanced islet autoimmunity to clinical diabetes (3), which in turn is influenced by some class I HLA alleles (4). Genetic variants outside the HLA region also affect the development of islet autoimmunity and/or progression to clinical diabetes (57).

We recently observed that the association between FPIR and class II HLA DR-DQ is secondary to the presence of multiple autoantibodies (8). The declining pattern of FPIR was associated with multiple autoantibodies irrespective of HLA class II risk group. However, it is possible that other genetic polymorphisms are specifically associated with the evolution of FPIR during progression from autoimmunity to clinical disease.

Here, we studied the role of two class I HLA alleles and eight selected non-HLA gene polymorphisms in the development of insulin secretory capacity as measured by FPIR in children participating in the Finnish Type 1 Diabetes Prediction and Prevention (DIPP) study. Because the presence of multiple islet autoantibodies is strongly associated with β-cell failure, we analyzed separately children with and without multiple biochemical autoantibodies. The selected HLA class I alleles and non-HLA markers have previously been shown to associate with the progression rate from islet autoimmunity to clinical diabetes (4,5,9,10). However, it is not known how or whether these markers are associated with insulin response. The genetic variants of INS and CTSH genes were selected because of their known role in β-cell function (11,12).

Research Design and Methods

The population-based DIPP study was launched in 1994 to screen for diabetes-associated risk by genotyping the major HLA DR-DQ haplotypes at birth (3). The study participants were followed regularly for the appearance of islet autoantibodies at 3–12-month intervals. Children who developed islet autoantibodies (islet cell antibodies and biochemical autoantibodies to insulin, GAD 65, and IA2) underwent an intravenous glucose tolerance test (IVGTT) (1), whereas autoantibodies to zinc transporter 8 were analyzed after IVGTT. β-Cell function was estimated by FPIR and change in FPIR (ΔFPIR) as described previously (8).

Genotyping Methods

HLA typing of major DR-DQ haplotypes was performed with a PCR-based lanthanide-labeled hybridization method using time-resolved fluorometry for detection (3). Genotyping using the Sequenom platform (San Diego, CA) of eight single nucleotide polymorphisms (SNPs), including PTPN22 (rs2476601), IFIH1 (rs1990760), INS (rs689), IKZF4 (rs1701704), ERBB3 (rs2292239), CTSH (rs3825932), PTPN2 (rs45450798), and FUT2 (rs6013380), was performed at the University of Eastern Finland (Kuopio, Finland) (5); CTSH (rs3825932) genotyping was performed using the Taqman SNP Genotyping Assay (Thermo Fisher Scientific, Waltham, MA). The assays of class I HLA alleles (B*39, A*24, and B*39:06) were analyzed on the DELFIA platform (4). SNPs in ERBB3 and IKZF4 polymorphisms were highly correlated (Fisher exact test P < 0.0001).

Autoantibody Analyses

Autoantibodies to insulin, GAD 65, IA2, and zinc transporter 8 were measured in serum samples by a radiobinding assay (13,14). Islet cell antibodies were measured by classical immunofluorescence method applied to sections of human pancreas, blood group O (15).

Study Participants

The 438 study children (268 [61.2%] males) with one or more FPIR results (133 [30.4%] who had progressed to T1D, 35 with a single biochemical, 65 with multiple biochemical autoantibodies who did not progress to T1D during the study period) had been categorized according to the biochemical autoantibody status (none/one or multiple [at least two] biochemical islet autoantibodies) at the time of the first IVGTT. The median age at the first IVGTT, which was performed at least 2 years before diagnosis in progressors, was 4.6 years. Diabetes was diagnosed according to World Health Organization criteria (16).

Statistical Analyses

ΔFPIR was calculated in children with and without multiple biochemical autoantibodies. Before data analysis, the response variable FPIR was log-transformed. Age-adjusted hierarchical linear models (8) applied to analyze the repeated measurements of FPIR included autoantibody status (0 or 1 autoantibody) in children without multiple autoantibodies, genotypes (three groups except for class I HLA genotypes, which were categorized into two groups), and interaction terms genotype by time and autoantibody group by time. The period of 0–5 years from the first IVGTT was examined.

Three types of models (additive, recessive, and dominant) were investigated for the SNP genotypes. In the additive model, all three groups were compared. In the recessive model, children homozygous for the risk allele were compared against those who were not homozygous for the risk allele (two groups). In the dominant model, children carrying the risk allele were compared with those who did not have a risk allele (two groups).

Statistical analyses were performed with JMP Pro version 11.2 and SAS 9.4 for Windows (SAS Institute, Cary, NC) software. P < 0.05 (two-tailed) was considered statistically significant.

Ethical Considerations

This study was conducted according to the guidelines of the Declaration of Helsinki II and was approved by local ethics committees. Written informed consent was obtained from all participants and/or their primary caregivers.

Data and Resource Availability

The data sets generated and analyzed during the current study are not publicly available because of privacy regulations. No applicable resources were generated or analyzed during the current study.

Results

The median FPIR levels and ΔFPIR over the observation period are shown in children with and without multiple biochemical autoantibodies (Tables 1 and 2). FPIR increased over time in children without multiple autoantibodies (Table 2), whereas it declined in those with multiple autoantibodies (Table 1). When the hierarchical linear mixed models were used in children with multiple autoantibodies, modest associations were observed between the evolution of FPIR and three of the gene regions studied (PTPN2 [rs45450798], FUT2 [rs601338], and CTSH [rs3825932]) in the additive model (P = 0.013, P = 0.020, and P = 0.0042, respectively) (Table 3).

Table 1

The median of the first FPIR and ΔFPIR over time according to different genotypes in 195 children with multiple (at least two) biochemical autoantibodies during follow-up

Table 2

The median of the first FPIR and ΔFPIR over time according to different genotypes from 243 children with zero or one biochemical autoantibody at the time of the first IVGTT

Table 3

FPIR as analyzed by a hierarchical linear mixed model adjusted for age from 195 children with multiple autoantibodies at the time of the first IVGTT

In general, children carrying susceptibility alleles had a more rapid decline in insulin secretion compared with those who did not carry a susceptibility allele. Children homozygous for the diabetes-associated risk allele in IKZF4 and PTPN2 genes had a steeper decline of FPIR than those who were not homozygous for the risk allele in these genes (recessive model P = 0.026 and P = 0.0035, respectively) (Table 3). Children carrying the T1D-associated risk allele in FUT2 and CTSH genes experienced also a steeper decline of FPIR than those without the risk allele in these genes (dominant model P = 0.0098 and P = 0.0011, respectively) (Table 3). In an analysis where risk scores were calculated on the basis of T1D risk in four SNPs that were significant in the model, there were no clearly additive effects (data not shown).

The class I HLA A*24 allele was also associated with the evolution of FPIR in children with multiple autoantibodies (P = 0.037) (Table 3) so that the presence of the A*24 allele was associated with a steeper coefficient estimate of FPIR (−0.00037, SE 0.000098, P = 0.0002) (Table 3). In children without multiple autoantibodies, the FPIR increased over time independent on A*24 allele status (Table 2). Furthermore, in children without multiple autoantibodies, ERBB3 (rs2292239) showed a significant association with FPIR in the recessive model (P = 0.0075) (Table 4).

Table 4

FPIR as analyzed by a hierarchical linear mixed model adjusted for age and the number of autoantibodies from 243 children without multiple autoantibodies

Discussion

In this study, we identified novel associations between FPIR and genetic variants known to affect T1D. In children with multiple autoantibodies, the change in FPIR over time was different between those categorized by their PTPN2 (rs45450798), FUT2 (rs601338), CTSH (rs3825932), and IKZF4 (rs1701704) genotypes. Children carrying disease susceptibility alleles had a more rapid decline in insulin secretion over time compared with those who did not carry the allele associated with susceptibility for T1D.

Homozygosity for the risk alleles in the IKZF4 and PTPN2 genes was associated with a steeper decline of FPIR compared with nonhomozygosity. IKZF4 encodes for Eos, which is known to play an important role in lymphoid development (17). A decreased tyrosine phosphatase expression associated with the PTPN2 variant has been shown to sensitize β-cells to cytokine-induced apoptosis (18).

Children with multiple autoantibodies carrying at least one risk allele in the CTSH and FUT2 genes were characterized by a steeper decline of FPIR compared with those who did not carry a risk allele. In recently diagnosed children, however, it was, the CT genotype of CTSH that was associated with the lowest dose of insulin, and the children with the CT genotype were most often in remission 12 months after onset compared with those with other genotypes (11). Interestingly, in healthy adults, the CTSH genotype affected β-cell function in the oral glucose tolerance test but showed no effect on FPIR (11).

Fructosyltransferase 2 enzyme in the Golgi apparatus is involved in the creation of a precursor of the H antigen, which is needed in the synthesis of A and B antigens found in secretions. Individuals carrying the major allele G are called secretors, and they have a functional FUT2 gene (19). In the current study, we observed a difference between children carrying the AA or AG genotype versus the GG genotype. The mechanisms underlying the association between FUT2 and FPIR are not known but could be related by the observation that the secretor status has been associated with composition of the human microbiome (20), although this is controversial (21).

IFIH1, PTPN22, and INS did not show any association with FPIR in this study, which could partly be explained by the observation that they all have been found in the DIPP study to have their main effect on the development of islet autoimmunity (5). It is not known whether associations between insulin secretion and various genotypes would be different in children without or before the appearance of islet autoantibodies. In autoantibody-positive children carrying both INS risk alleles but without class II HLA risk, the increase of FPIR was slower than in children who carried one or no INS risk alleles (12). Some effect of these genes could potentially be seen in subgroups; for example, the association of caesarean section with the development of T1D was reported to be affected by the IFIH1 genotype (22).

Hyperexpression of class I HLA antigens is often seen in pancreatic islets from patients with T1D (23). In this study, the presence of the class I HLA A*24 allele was associated with a steeper decline of FPIR in children with multiple autoantibodies. The presence of the A*24 allele has previously been reported to predict rapid progression to clinical disease in autoantibody-positive relatives of patients with T1D (24).

The unique possibility to analyze young, genetically predisposed children followed intensively over a relatively long period is a strength of this study. A weakness is the low number of observations within some genotypes, which reduces the statistical power. We did not analyze FPIR and its changes over time in relation to the initiating autoantibody (5,9). Although the overall effect of the genetic markers studied on FPIR is modest, it is conceivable that quite a variation in the β-cell mass exists. A wide range of the estimated β-cell mass was observed in adults, even in subjects with low FPIR and multiple autoantibodies (25).

In conclusion, our results show that certain genetic variants outside the class II HLA region can have a significant impact on the longitudinal pattern of FPIR. In children with multiple autoantibodies, the diabetes risk alleles were associated with more rapid loss in β-cell secretory capacity. The underlying mechanisms are still unknown.

Article Information

Acknowledgments. The authors thank all DIPP personnel and families who have participated in this important study.

Funding. This study was funded by JDRF; Special Research Funds for Turku, Oulu and Tampere University Hospitals in Finland; Academy of Finland; Turku University Foundation; University of Turku Graduate School Doctoral Programme; Diabetes Research Foundation in Finland; Sigrid Juselius Foundation; Päivikki and Sakari Sohlberg Foundation; Emil Aaltonen Foundation; Alma and KA Snellman Foundation; Kyllikki and Uolevi Lehikoinen Foundation; and Turku Centre of Lifespan Research.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. M.K.K. wrote the first draft. M.K.K., M.-L.M., A.-P.L., J.L., E.L., P.V., A.H., T.H., M.Ki., O.S., M.Kn., R.V., J.I., and J.T. reviewed the manuscript and approved the final version. M.K.K., M.-L.M., A.-P.L., P.V., A.H., T.H., M.Ki., O.S., M.Kn., R.V., J.I., and J.T. acquired the data. M.K.K., A.-P.L., J.L., R.V., J.I., and J.T. interpreted the data. M.K.K. and E.L. analyzed data. M.K.K., O.S., R.V., J.I., and J.T. designed the study. M.K.K. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

  • Received March 29, 2019.
  • Accepted October 2, 2019.



Source link