, Volume 57, Issue 9, pp 1859–1868 | Cite as

Type 2 diabetes susceptibility gene variants predispose to adult-onset autoimmune diabetes

  • Mette K. AndersenEmail author
  • Maria Sterner
  • Tom Forsén
  • Annemari Käräjämäki
  • Olov Rolandsson
  • Carol Forsblom
  • Per-Henrik Groop
  • Kaj Lahti
  • Peter M. Nilsson
  • Leif Groop
  • Tiinamaija Tuomi



Latent autoimmune diabetes in adults (LADA) is phenotypically a hybrid of type 1 and type 2 diabetes. Genetically LADA is poorly characterised but does share genetic predisposition with type 1 diabetes. We aimed to improve the genetic characterisation of LADA and hypothesised that type 2 diabetes-associated gene variants also predispose to LADA, and that the associations would be strongest in LADA patients with low levels of GAD autoantibodies (GADA).


We assessed 41 type 2 diabetes-associated gene variants in Finnish (phase I) and Swedish (phase II) patients with LADA (n = 911) or type 1 diabetes (n = 406), all diagnosed after the age of 35 years, as well as in non-diabetic control individuals 40 years or older (n = 4,002).


Variants in the ZMIZ1 (rs12571751, p = 4.1 × 10−5) and TCF7L2 (rs7903146, p = 5.8 × 10−4) loci were strongly associated with LADA. Variants in the KCNQ1 (rs2237895, p = 0.0012), HHEX (rs1111875, p = 0.0024 in Finns) and MTNR1B (rs10830963, p = 0.0039) loci showed the strongest association in patients with low GADA, supporting the hypothesis that the disease in these patients is more like type 2 diabetes. In contrast, variants in the KLHDC5 (rs10842994, p = 9.5 × 10−4 in Finns), TP53INP1 (rs896854, p = 0.005), CDKAL1 (rs7756992, p = 7.0 × 10−4; rs7754840, p = 8.8 × 10−4) and PROX1 (rs340874, p = 0.003) loci showed the strongest association in patients with high GADA. For type 1 diabetes, a strong association was seen for MTNR1B (rs10830963, p = 3.2 × 10−6) and HNF1A (rs2650000, p = 0.0012).


LADA and adult-onset type 1 diabetes share genetic risk variants with type 2 diabetes, supporting the idea of a hybrid form of diabetes and distinguishing them from patients with classical young-onset type 1 diabetes.


Adult-onset type 1 diabetes GADA GAD autoantibodies Genetics HNF1A LADA Latent autoimmune diabetes in adults MTNR1B TCF7L2 ZMIZ1 



All New Diabetics in Scania


Diabetes Registry in Vasa


GAD autoantibodies


Latent autoimmune diabetes in adults


Malmö Diet and Cancer Study


Single-nucleotide polymorphism


Diabetes is a complex disease with two main subtypes: type 1 diabetes, characterised by autoimmune destruction of the insulin-producing beta cells, and type 2 diabetes, characterised by a combination of insulin resistance and reduced insulin secretion. The subdivision into type 1 and type 2 diabetes is especially intricate in adults, in whom the disease is more like a spectrum of phenotypes. This spectrum also includes the subtype termed latent autoimmune diabetes in adults (LADA).

LADA accounts for about 10% of European patients initially diagnosed with type 2 diabetes in whom type 1 diabetes-associated autoantibodies are present [1, 2, 3]. The most prevalent autoantibodies among these adults are GAD autoantibodies (GADA), present in 90%. Other autoantibodies, such as those to protein tyrosine phosphatase IA-2 or zinc transporter 8 are reported in only 1–2% of GADA-negative patients [3, 4]. The level of autoantibodies, particularly GADA, has a clear influence on the phenotype of LADA. A high GADA level is associated with a phenotype more like type 1 diabetes, including reduced beta cell function, increased need for insulin treatment, lower BMI and lower prevalence of dyslipidaemia (reviewed by [5, 6]).

The genetic component of diabetes has been intensively studied in patients with type 2 diabetes and in patients with type 1 diabetes diagnosed in childhood, which has resulted in identification of several susceptibility gene variants [7]. Genetic studies in LADA comprising >100 patients are rare but indicate that LADA shares a genetic predisposition with childhood-onset type 1 diabetes. Association with HLA-DQB1 and PTPN22 risk alleles has consistently been shown, while association with INS is disputed. Furthermore, the frequency of these type 1 diabetes risk alleles has consistently been shown to correlate with GADA level among LADA patients (reviewed by [6]). For type 1 diabetes the genetic predisposition seems to be similar in children and adults diagnosed predominantly at >15 years of age; only for HLA-DQB1 and IL2RA have effects related to age at onset been reported [8, 9, 10].

Whether type 2 diabetes-associated variants predispose to adult-onset type 1 diabetes or LADA is largely unknown. To date, consistent association has only been reported for TCF7L2 and LADA [11, 12]. We hypothesised that LADA, and possibly adult-onset type 1 diabetes, is a genetic hybrid of type 1 and type 2 diabetes with increased frequency of type 2 diabetes risk genotypes. Further, we expected the type 2 diabetes-associated gene variants to show strongest association in LADA patients with low GADA levels, whose disease is phenotypically more like type 2 diabetes. In this study, we assessed the association of type 2 diabetes susceptibility gene variants in LADA and adult-onset type 1 diabetes diagnosed at age >35 years, to explore the degree of genetic overlap in adult-onset forms of diabetes.


Study participants

We applied a two-stage study design with Finnish (phase I) and Swedish (phase II) data sets. The Finnish data set comprised individuals from the Diabetes Registry in Vasa (DIREVA), the Botnia Study [13, 14] and the FinnDiane Study [15]. The Swedish data set comprised individuals from the Scania Diabetes Registry [16], the diabetes registry (DiabNorth) in the population-based Västerbotten Study [17] and the Malmö Diet and Cancer Study (MDC) [18] (Fig. 1). The studies were approved by the local ethics committees, and informed consent was obtained from all study participants. The 25 patients included in both the Botnia Study and DIREVA were analysed as part of the Botnia Study.
Fig. 1

The two-staged study design, with indication of the contribution of patients and control individuals from each of the cohorts. T1D, adult-onset type 1 diabetes

We included patients with LADA or type 1 diabetes diagnosed at >35 years of age and non-diabetic control individuals. The control individuals were aged ≥ 40 years at the time of investigation (Finnish/Swedish, mean ± SD, 56.6 ± 10.3/56.8 ± 6.2 years), had a BMI of 26.5 ± 4.0/25.4 ± 3.7 kg/m2 and included 46.1/48.8% men.

Diagnosis of LADA was based on presence of GADA and sufficient beta cell function at the time of diagnosis, indicated by no insulin treatment and/or C-peptide level >0.2 nmol/l. The majority of the LADA patients had initially been diagnosed with type 2 diabetes. Diagnosis of type 1 diabetes was based on initial diagnosis of type 1 diabetes made by the treating physician, a fasting serum C-peptide concentration< 0.2 nmol/l at the time of investigation and initiation of permanent insulin treatment within 6 months from diagnosis.

Measurements and assays

C-peptide concentrations were determined by a radioimmunoassay (Human C-peptide RIA; Linco, St Charles, MO, USA; or Peninsula Laboratories, Belmont, CA, USA), or with Cobas e411 (Roche Diagnostics, Mannheim, Germany). GADA were measured with an ELISA assay (RSR, Cardiff, UK) in the DIREVA patients (positive cut-off: 10 IU/ml) and with radiobinding assays using 35S-labelled protein [2] in the rest of the patients (positive cut-off: 5 RU or 32 IU/ml). The radiobinding assay showed 62–88% sensitivity and 91–99% specificity in workshops (Combinatorial Autoantibody or Diabetes/Islet Autoantibody Standardization Programs) from 1998 to 2013, and the ELISA assay showed 72% sensitivity and 99% specificity in the 2013 Islet Autoantibody Standardization Program workshop.

To assess the relationship between gene variants and GADA level, the LADA patients were stratified into LADAhigh and LADAlow according to the median GADA level among Finnish LADA patients. The same cut-off values (Botnia, 20.4 RU or 73.0 IU/ml; DIREVA, 59.0 IU/ml) were applied to the Swedish patients (65.9% were LADAhigh) as they were typed using the same methods.


All cases (n = 1,317) and a subset of the controls (n = 987) were genotyped with a single-nucleotide polymorphism (SNP) panel, custom-designed for the All New Diabetics in Scania (ANDIS) study in Sweden (; accessed 7 April 2014), including SNPs reported to be associated with type 2 diabetes or related phenotypes. Genotyping was performed by matrix-assisted laser desorption–ionisation time-of-flight mass spectrometry on a MassARRAY platform (Sequenom, San Diego, CA, USA). We excluded SNPs with a minor allele frequency (MAF) < 5%, with genotyping success rate < 90% or disagreeing with Hardy–Weinberg equilibrium (p < 0.00029, corresponding to 0.05/No. of SNPs). Individuals with missing genotypes for >5% of the SNPs were also excluded. Altogether, 143 SNPs (genotyping success rate > 99%), 1,272 cases and 905 control individuals passed the quality control.

The rest of the controls (Botnia, n = 873; MDC, n = 2,142) had been genotyped with the Metabochip [19], where genotypes were called with GenomeStudio (Illumina, San Diego, CA, USA). SNPs included both on the Metabochip and in the ANDIS panel numbered 105; of these SNPs, 46 were in or near loci previously associated with type 2 diabetes in Europeans [20] and 40 passed the quality control (electronic supplementary material [ESM] Table 1). In addition, we included a SNP in GLIS3, as this is one of the few loci that have shown an association with both type 1 and type 2 diabetes (type 2 diabetes association only in Asians). Hence, 41 SNPs were analysed.

Statistical analysis

Statistical Package for Social Science software (version 19.0; SPSS, Chicago, IL, USA) was applied to compare clinical characteristics between patient groups by the non-parametric Mann–Whitney U test or the χ 2 test. Genetic association analyses were performed with PLINK v1.07 ( [21]). Differences in allele frequencies between cases and controls were assessed with logistical regression adjusted for the effect of age, sex and BMI. ORs are presented for the reported type 2 diabetes-associated alleles. For the association tests in the Finnish phase I data set a p value of < 0.0012 was considered statistically significant, corresponding to a Bonferroni correction for the number of SNPs tested (n = 41). RevMan v5.2.6 ( was applied to conduct fixed-effect meta-analyses estimating the combined effect sizes for the Finnish and Swedish data sets. Cochran’s Q test was used to assess the heterogeneity between data sets in the meta-analyses.

The possible effect of HLA-DQB1 risk genotypes (*0302 or *0302/02) on association results was tested in part of the Finnish data set with available HLA-DQB1 genotype data (LADA, n = 264; type 1 diabetes, n = 245; control individuals, n = 503). HLA-DQB1 genotype was included as covariate in logistical regression analyses; the main results were similar with this adjustment (data not shown).


Clinical characteristics

Compared with the Swedish LADA patients, the Finnish LADA patients were older at the time of investigation and had longer disease duration as well as lower levels of fasting serum C-peptide and fasting plasma glucose (Table 1). Moreover, the Finnish patients included a smaller proportion of patients with high GADA levels. The two groups of LADA patients were similar with respect to age at diagnosis, glycaemic control and BMI. Among patients with type 1 diabetes, the Finnish patients were younger at diagnosis, had longer disease duration and higher BMI than the Swedish patients (Table 1). C-peptide, lipid levels and age at investigation were similar for Swedish and Finnish patients with type 1 diabetes.
Table 1

Clinical characteristics



Type 1 diabetes



p value



p value

Men, n (%)

472 (48.1)

439 (57.8)


306 (49.7)

100 (45.0)


LADAhigh, %

244 (51.7)

335 (65.9)





Age at diagnosis, years

55.9 (11.6)

56.8 (10.9)


42.5 (7.2)

48.4 (9.9)


Age at investigation, years

63.5 (11.2)

59.4 (11.3)


54.3 (9.4)

55.6 (11.4)


Duration, years

7.8 (7.6)

3.7 (5.6)


11.9 (8.9)

7.2 (5.9)


fP-glucose, mmol/l

9.2 (3.8)

10.3 (4.1)



12.8 (4.8)


HbA1c, %

7.3 (1.6)

7.6 (2.4)


8.4 (1.4)

8.5 (2.5)


HbA1c, mmol/mol

56.7 (17.6)

59.6 (26.7)


68.6 (15.0)

69.9 (27.2)


BMI, kg/m2

28.5 (5.2)

28.1 (5.3)


25.3 (3.7)

23.3 (3.6)


fS C-peptide, nmol/l

0.63 (0.56)

0.84 (0.60)


0.03 (0.05)

0.04 (0.07)


Total cholesterol, mmol/l

5.24 (1.13)

5.55 (1.17)


5.05 (0.93)

5.20 (1.02)


Triacylglycerol, mmol/l

1.61 (1.13)

1.79 (1.44)


1.13 (0.68)

1.10 (0.48)


HDL-cholesterol, mmol/l

1.34 (0.38)

1.22 (0.38)


1.49 (0.47)

1.49 (0.40)


Data are number (%) or mean (SD)

Only significant p values are shown

fP, fasting plasma; fS, fasting serum; LADAhigh, LADA patients with GADA above the median, defined based on the patients from the Botnia Study; NA, not applicable

Phase I: Finnish patients


Nominal association was observed for 11 variants (Table 2 and ESM Table 1). Applying the conservative Bonferroni-corrected cut-off, association remained for two loci, namely ZMIZ1, which is a novel finding in LADA (rs12571751, OR [95% CI]: 1.35 [1.15, 1.60], p = 3.2 × 10−4), and the previously reported TCF7L2 (rs7903146, 1.44 [1.18, 1.77], p = 3.8 × 10−4).
Table 2

Association results: LADA

SNP – Candidate gene

RAF: controls/cases

OR (95% CI)

p value

rs12571751 – ZMIZ1

A: 0.49/0.56

1.35 (1.15, 1.60)

3.2 × 10−4a

rs7903146 – TCF7L2

T: 0.17/0.22

1.44 (1.18, 1.77)

3.8 × 10−4a

rs10842994 – KLHDC5

C: 0.81/0.85

1.44 (1.15, 1.80)


rs896854 – TP53INP1

T: 0.46/0.52

1.28 (1.09, 1.50)


rs108116661 – CDKN2A/B

T: 0.82/0.87

1.42 (1.11, 1.80)


rs10830963 – MTNR1B

G: 0.29/0.34

1.27 (1.07, 1.51)


rs1111875 – HHEX

C: 0.52/0.56

1.23 (1.05, 1.46)


rs2237895 – KCNQ1

C: 0.48/0.52

1.20 (1.02, 1.41)


rs7578597 – THADA

T: 0.94/0.96

1.49 (1.03, 2.16)


rs7756992 – CDKAL1

G: 0.27/0.31

1.20 (1.01, 1.43)


rs7754840 – CDKAL1

C: 0.30/0.34

1.19 (1.01, 1.40)


Loci obtaining a p value <0.05 in the Finnish data set, with logistical regression adjusted for age, sex and BMI

aSignificant below the Bonferroni cut-off (p < 0.0012)

RAF, risk allele (reported for type 2 diabetes) frequency

When the LADA patients were stratified according to GADA level, nominal association was seen for ten variants (Fig. 2) with either LADAlow (GADA level below the median) or LADAhigh (GADA above the median). The strongest evidence for association in the LADAlow patients was observed for HHEX (rs1111875: 1.43 [1.13, 1.80], p = 0.0024) and CDKN2A/B (rs10811661: 1.65 [1.16, 2.35], p = 0.0057). For KLHDC5 (rs10842994: 1.65 [1.23, 2.22], p = 9.5 × 10−4), TP53INP1 (rs896854: 1.36 [1.12, 1.66], p = 0.0023), CDKAL1 (rs7756992: 1.39 [1.13, 1.72], p = 0.0020; rs7754840: 1.32 [1.08, 1.62], p = 0.0075) and ANKRD55 (rs459193: 1.28 [1.02, 1.61], p = 0.030) association was restricted to LADAhigh patients. Although statistically significant association for MTNR1B was limited to LADAhigh patients (rs10830963: 1.27 [1.03, 1.56], p = 0.028), the risk-allele frequency was similar in LADAhigh and LADAlow patients. The ZMIZ1 (LADAlow, 1.47 [1.17, 1.85], p = 0.0011; LADAhigh, 1.30 [1.06, 1.59], p = 0.011) and TCF7L2 (LADAlow 1.46 [1.11, 1.93], p = 0.0073; LADAhigh, 1.42 [1.11, 1.82], p = 0.0052) associations were independent of GADA level (Fig. 2 and ESM Table 1).
Fig. 2

Risk-allele frequencies (%) for variants that show association with LADAlow (a), LADAhigh (b) or both (c). *p < 0.05, p < 0.005 and p < 0.0012, risk-allele frequency for patients with LADA (grey bars), LADAlow (white bars) or LADAhigh (black bars) vs control individuals (hatched bars)

Adult-onset type 1 diabetes

Nominal association was observed for nine variants (Table 3). Applying the Bonferroni-corrected cut-off, association remained for three loci: MTNR1B (rs10830963: 1.59 [1.32, 1.93], p = 1.7 × 10−6), PROX1 (rs340874; 0.68 [0.56, 0.82], p = 6.1 × 10−5) and HNF1A (rs2650000; 1.43 [1.19, 1.72], p = 1.7 × 10−4).
Table 3

Association results: adult-onset type 1 diabetes

SNP – Candidate gene

RAF: controls/cases

OR (95% CI)

p value

rs10830963 – MTNR1B

G: 0.29/0.39

1.59 (1.32, 1.93)

1.7 × 10−6a

rs340874 – PROX1

C: 0.49/0.40

0.68 (0.56, 0.82)

6.1 × 10−5a

rs2650000 – HNF1A

A: 0.35/0.43

1.43 (1.19, 1.72)

1.7 × 10−4a

rs7754840 – CDKAL1

C: 0.30/0.36

1.26 (1.05, 1.52)


rs7756992 – CDKAL1

G: 0.27/0.33

1.26 (1.04, 1.53)


rs10811661 – CDKN2A/B

T: 0.82/0.86

1.38 (1.05, 1.80)


rs864745 – JAZF1

T: 0.48/0.42

0.81 (0.67, 0.97)


rs10401969 – CILP2

C: 0.08/0.05

0.64 (0.43, 0.96)


rs7957197 – HNF1A

T: 0.74/0.79

1.27 (1.02, 1.57)


Loci obtaining a p value <0.05 in the Finnish data set, with logistic regression adjusted for age, sex and BMI

aSignificant below the Bonferroni cut-off (p < 0.0012)

RAF, risk allele (reported for type 2 diabetes) frequency

Phase II: Swedish patients and meta-analyses

SNPs obtaining a p value  <0.05 (adjusted for age, sex and BMI) in phase I were tested in the Swedish phase II data set.


The ZMIZ1 association was replicated in the Swedish LADA patients (1.27 [1.08, 1.50], p = 0.0049), but driven by LADAhigh (1.40 [1.15, 1.70], p = 7.2 × 10−4; LADAlow: 1.02 [0.76, 1.37], p = NS). The TCF7L2 risk allele was significantly increased in the Swedish LADAlow patients (1.55 [1.13, 2.13], p = 0.0066), but failed to reach statistical significance in the whole group. None of the other associations observed in Finnish LADA patients were replicated in the Swedish patients (ESM Table 1).

Combining the Finnish and the Swedish data sets in a meta-analysis, both ZMIZ1 (p = 4.1 × 10−5) and TCF7L2 (p = 5.8 × 10−4) remained significant below the Bonferroni-corrected threshold in the whole LADA group, without heterogeneity between the data sets (Fig. 3). Variants in KCNQ1 (rs2237895, 1.17 [1.06, 1.31]), CDKAL1 (rs7756992, 1.19 [1.06, 1.34]; rs7754840, 1.19 [1.06, 1.32]), KLHDC5 (rs10842994, 1.23 [1.07, 1.42]) and CILP2 (rs10401969, 1.27 [1.07, 1.50]) obtained p values < 0.005 and heterogeneity was only observed with respect to CILP2 (p = 0.02).
Fig. 3

Forest plots for loci obtaining p ≤ 0.0012 in the meta-analyses of LADA, LADAhigh or LADAlow. The OR (95% CI) is indicated for the Finnish and Swedish data sets (squares), as well as for the meta-analysis (diamonds). Associations are shown for: (a) ZMIZ1 rs12571751 with LADA and LADAhigh with no heterogeneity between data sets (p = 0.350 and p = 0.946, respectively); (b) TCF7L2 rs7903146 with LADA and LADAlow with no heterogeneity between data sets (p = 0.251 and p = 0.693, respectively); (c) CDKAL1 rs7756992 with LADA and LADAhigh with no heterogeneity between data sets (p = 0.677 and p = 0.364, respectively); (d) KCNQ1 rs2237895 with LADA and LADAlow with no heterogeneity between data sets (p = 0.950 and p = 0.328, respectively)

In meta-analyses where LADA patients were stratified based on GADA level, association with LADAlow was accentuated for TCF7L2 (p = 2.4 × 10−4) and KCNQ1 (p = 0.0012) (Fig. 3) and suggestive association was observed for MTNR1B (1.27 [1.08, 1.50], p = 0.0039). With respect to LADAhigh, ZMIZ1 (p = 1.1 × 10−4) and CDKAL1 (rs7756992, p = 7.0 × 10−4; rs7754840, p = 8.8 × 10−4) showed significant association below the Bonferroni-corrected threshold (Fig. 3) and suggestive association was seen for PROX1 (rs340874, 1.22 [1.07, 1.39], p = 0.003) and TP53INP1 (rs896854, 1.21 [1.06, 1.37], p = 0.005). Only the TP53INP1 SNP showed signs of heterogeneity between the data sets (p = 0.09).

Adult-onset type 1 diabetes

None of the loci associated with adult-onset type 1 diabetes in the Finnish patients were replicated in the Swedish data set (ESM Table 1) but in the meta-analysis MTNR1B (p = 3.2 × 10−6) and HNF1A (p = 0.0012) remained significant, although with heterogeneity between data sets (Fig. 4). Suggestive association was observed for CDKAL1 (rs7756992, 1.27 [1.08, 1.50], p = 0.0040) and CENDT2 (rs1552224, 0.77 [0.65, 0.92], p = 0.0042).
Fig. 4

Forest plots for the variants obtaining p ≤ 0.0012 in the meta-analyses of adult-onset type 1 diabetes. The OR (95% CI) is indicated for the Finnish and Swedish data sets (squares), as well as for the meta-analysis (diamonds). (a) MTNR1B rs10830963; (b) HNF1A rs2650000. Both loci showed heterogeneity between data sets (p = 0.064 and p = 0.032, respectively). T1D, adult-onset type 1 diabetes


To improve the genetic characterisation of LADA and to address the question of genetic overlap in the adult-onset forms of diabetes, we studied type 2 diabetes-associated gene variants in patients with LADA or adult-onset type 1 diabetes diagnosed at > 35 years of age. We show that both subgroups of adult-onset autoimmune diabetes share genetic risk variants with type 2 diabetes. The reported associations show effect sizes and direction of effect consistent with those reported for type 2 diabetes, which indicate that the associations are true. With respect to LADA, the level of GADA had a distinct effect on the genetic associations. Previously, we have assessed type 1 diabetes-associated gene variants in the LADA patients included in the current study and reported association with HLA-DQB1 and PTPN22, although with lower risk-allele frequencies in the LADA patients than in patients with young and adult-onset type 1 diabetes [9, 11]. LADA, thus, seems to be a genetic hybrid of type 1 and type 2 diabetes distinct from classical young-onset type 1 diabetes.

A great strength of our study is the inclusion of a unique collection of patients with adult-onset type 1 diabetes diagnosed after the age of 35 years, and the largest published collection of genetically characterised LADA patients, numbering over 900. Further, we analysed LADA and adult-onset type 1 diabetes as separate phenotypes since type 1 diabetes risk variants show different frequencies in the two diabetes subtypes. Even though this stratification reduces the sample size in the groups analysed, the more precise phenotyping may pinpoint associations that would be missed in the combined group of unspecified adult-onset autoimmune diabetes.

The sample size is a limitation of our study, especially when considering the reported effect sizes of variants associated with type 2 diabetes ranging from OR of 1.05 to 1.38 [20]. Thus, we can detect the variants showing strongest association with LADA or adult-onset type 1 diabetes but lack statistical power to exclude association with other variants. This problem was accentuated in the analyses of GADA-level subgroups. Another limitation is the fact that our study is based on current knowledge of the genetic predisposition to type 2 diabetes. For the great majority of associated loci the causal variant is still unknown and therefore most of the examined variants represent statistical association and not necessarily biological association. The lack of knowledge about the causal variant within each locus is reflected in the difficulties of pinpointing the mechanistic links between genetic variants and disease.

Variants in TCF7L2 and ZMIZ1 showed strong evidence of association with LADA. In Europeans, the intronic TCF7L2 rs7903146 variant is the strongest identified genetic risk factor for type 2 diabetes [22]. Association has also been shown between TCF7L2 and LADA [12] in work encompassing a subset of the patients included in the current study [11]. TCF7L2 encodes a transcription factor, which is involved in maintaining the secretory function of beta cells and has been suggested to be a key determinant of beta cell mass (reviewed in [23]). The rs7903146 variant is associated with impaired insulin secretion, although the exact mechanistic link is unclear [24]. The variant most likely imposes a non-autoimmune reduction in insulin secretion, which in LADA patients exacerbates the presumably already reduced beta cell function caused by autoimmune destruction.

In agreement with the Italian Non-Insulin Requiring Autoimmune Diabetes (NIRAD) study [25], the meta-analysis showed stronger TCF7L2 association in patients with low GADA level. This GADA-level effect was, however, driven by the Swedish data set, while the association in the Finnish data set was independent of GADA level. This discrepancy may reflect differences in the differentiation between LADA and type 1 diabetes in Finnish and Swedish patients. The frequency of the TCF7L2 risk T allele among the adult-onset type 1 diabetes patients and control individuals was similar, in line with previous reports in a mixture of patients with adult-onset autoimmune diabetes [8] and in childhood-onset type 1 diabetes [26]. Thus, TCF7L2 represents a distinguishing factor between LADA and type 1 diabetes, irrespective of age at diagnosis, and the T allele seems to protect against type 1 diabetes possibly by sufficient impairment of insulin synthesis, resulting in reduced antigenic burden suggested to be operative in type 1 diabetes.

Importantly, the association of ZMIZ1 rs12571751 with LADA is novel. An association was seen both in Finnish and Swedish LADA patients, but in the Swedish patients it was strongest in those with a high GADA level. Other intronic ZMIZ1 variants have been associated with a variety of autoimmune diseases, including childhood-onset type 1 diabetes [27] and coeliac disease [28]. Even though we observed no association between the variant and adult-onset type 1 diabetes, the ZMIZ1 loci seems to play a key role in the development of autoimmune diseases.

ZMIZ1 is an interesting candidate gene for LADA as the encoded protein is a transcriptional co-activator interacting with Smad3 and p53 [29, 30]. Both these transcription factors are involved in pathways possibly linked to diabetes susceptibility. Smad3 modulates the TGF-β signalling pathway, which is linked to immune tolerance, inhibition of autoimmune diabetes and induction of anti-inflammatory responses [31, 32]. On the other hand, p53 is linked to apoptosis, which may be a determining factor for beta cell mass but also in the autoimmune process where apoptosis may induce the formation of autoantibodies by increasing the exposure to autoantigens [33]. The p53 link is also interesting in light of the suggestive association between TP53INP1 and LADAhigh, as TP53INP1 is a direct target gene for p53 and has been suggested to regulate p53-dependent apoptosis [34]. Moreover, reduced activity of transcription factor 7-like 2 has been linked to increased expression of TP53INP1 and beta cell apoptosis [23]. Thus, p53 and apoptosis could provide a molecular connection between three LADA-associated genes.

We hypothesised that among LADA patients, those with low GADA levels would have a genotype more similar to type 2 diabetes (i.e. they show higher frequencies of the type 2 diabetes risk variants than those with high GADA levels). However, only a few type 2 diabetes variants supported this hypothesis and the results were inconsistent in the Finnish and Swedish data sets. This heterogeneity in GADA-level association may, of course, be explained by the small sample size. An alternative explanation is that the associations, which seem to be GADA-level dependent, are mediated by factors that correlate with the GADA level, such as BMI, and that the type 2 diabetes variants are not directly linked to the autoimmune process in LADA. Considering LADA, it is, however, very interesting that some of the type 2 diabetes variants are associated only with LADAhigh, as this supports LADA being a true hybrid of type 1 diabetes and type 2 diabetes and not just a mixed group of misclassified patients with type 1 diabetes or type 2 diabetes.

With respect to type 1 diabetes, we hypothesised that non-autoimmune mechanisms may contribute to disease development, especially in adult-onset type 1 diabetes where the autoimmune disease process seems less pronounced. A genetic overlap between type 1 and type 2 diabetes is supported by the reported associations for variants in ZMIZ1 and GLIS3 with both subtypes of diabetes [20, 27, 35] and possibly also by associations for variants in HHEX and PPARG, for which inconsistent results have been published [36, 37]. Our data showed no support for these previous findings, as none of the four loci were associated with adult-onset type 1 diabetes and only ZMIZ1 was associated with LADA. Of note, with the current sample size we were unable to rule out association for these loci. However, in support of the hypothesis of a genetic overlap between type 1 and type 2 diabetes, we identified novel associations between adult-onset type 1 diabetes and variants in two type 2 diabetes-associated loci, MTNR1B and HNF1A. Both associations were driven by the Finnish patients, but remained significant in the meta-analysis. The discrepancy between the data sets may be due to the small sample size or the significant difference between Finnish and Swedish patients with respect to age at diagnosis.

The intronic MTNR1B rs10830963 variant has been associated with type 2 diabetes, elevated fasting glucose level and impaired insulin response to oral and intravenous glucose. MTNR1B encodes the melatonin receptor 1B (MT2), which is expressed in human pancreatic islets and beta cells and plays a key role in the regulation of circadian rhythms. Disruption of the circadian rhythms has been linked to metabolic disorders such as type 2 diabetes (reviewed by [38]). Furthermore, melatonin has been shown to be an immune stimulator (reviewed by [39]) and seems to be directly involved in the modulation of insulin secretion. However the effect of melatonin on insulin secretion is debatable, as melatonin has been shown to inhibit insulin secretion in vivo and in vitro in rodent models, whereas in one study it enhanced insulin secretion in human islets [40, 41].

The functional link between MTNR1B and type 2 diabetes is unclear. The mechanism has been suggested to be gain-of-function, as the rs10830963 risk allele is associated with increased mRNA expression of the MT2 receptor in islets from non-diabetic individuals [40]. In contrast, studies of rare coding MTNR1B variants point to loss of receptor function as the causal mechanism linked to type 2 diabetes [42].

A role for MTNR1B in the pathogenesis of autoimmune diabetes is further supported by a possible association between the MTNR1B variant and LADA, especially in the Finnish patients, in whom the risk-allele frequency was similar in LADAhigh and LADAlow patients. However, as no association has been observed between the variant and childhood-onset type 1 diabetes [43], these data may indicate that MTNR1B is a candidate gene only for the adult-onset forms of diabetes. This hypothesis is supported by the observed age-related reduction in endogenous melatonin levels. Reduced melatonin levels have been linked to immunosenescence, a process that is thought to contribute to autoimmune diseases in elderly people (reviewed by [39]).

Rare mutations in HNF1A cause MODY [44], and the common variants have been linked to LDL-cholesterol (rs2650000) and type 2 diabetes (rs7957197) [45, 46]. HNF1A encodes a transcription factor essential for pancreatic beta cell development and function [47] and is thus a good candidate gene for diabetes in general. This HNF1A–type 1 diabetes association, together with the majority of the LADA associations, point to compromised beta cell mass as a key mechanism contributing to insulin deficiency also in adult-onset autoimmune diabetes.

Our study adds to the genetic knowledge about the much-debated diabetes subtype LADA. However, so far all genetic studies of LADA have been candidate-gene based and have focused on type 1 or type 2 diabetes-associated variants, thus it is still unknown whether LADA harbours its own unique risk variants. A genome-wide association study in a larger sample of LADA patients will be needed to answer this question. A larger sample size would also permit extended stratification of the LADA patients according to HLA genotype, which could reveal interesting genetic interactions, and according to the presence of additional autoantibodies, which could reveal genetic differences as carriers of multiple autoantibodies show a more rapid disease progression.



The Botnia Research Group, the FinnDiane Research group, the Västerbotten Study group and the DIREVA study group are acknowledged for recruiting and clinically studying the participants.


The Botnia Study was supported by grants from the Academy of Finland (263401 and 267882), the Sigrid Juselius Foundation, the Finnish Diabetes Research Foundation, the Folkhälsan Research Foundation, the Finska Läkaresällskapet, the Ollqvist Foundation, Korsholm, Malax, Närpes, and Vasa Health Care Centers and The Helsinki University Central Hospital. The FinnDiane Study was supported by grants from the Folkhälsan Research Foundation, the Wilhelm and Else Stockmann Foundation and the Academy of Finland. The Västerbotten Study was funded by the Västerbotten County Council. The MDC study was supported by a project grant from the Swedish Research Council (K2011-65X-20752-04-6). The ANDIS Study was supported by infrastructure grants from the Swedish Research Council (2010-5983 and 2012-5538). LG was supported by an ERC Advanced Research Grant GENETARGET-T2D (GA 269045).

Duality of interest

P-HG has received research grants from Eli Lilly and Roche, as well as lecture honoraria from Boehringer Ingelheim, Eli Lilly, Genzyme, Medscape, MSD, Novartis and Novo Nordisk. P-HG is an advisory board member of Abbott, AbbVie, Boehringer Ingelheim, Cebix, Eli Lilly and Novartis. All other authors declare that there is no duality of interest associated with their contribution to this manuscript.

Contribution statement

The study was planned by MKA, OR, P-HG, LG and TT. MS, TF, AK, CF, KL and PMN contributed to phenotype and genotype data acquisition. MKA and TT analysed and interpreted the data and drafted the initial version of the manuscript. All co-authors reviewed the manuscript and approved the submitted version. MKA is the guarantor of this work.

Supplementary material

125_2014_3287_MOESM1_ESM.xls (46 kb)
ESM Table 1 (XLS 45 kb)


  1. 1.
    Turner R, Stratton I, Horton V et al (1997) UKPDS 25: autoantibodies to islet-cell cytoplasm and glutamic acid decarboxylase for prediction of insulin requirement in type 2 diabetes. UK Prospective Diabetes Study Group. Lancet 350:1288–1293PubMedCrossRefGoogle Scholar
  2. 2.
    Tuomi T, Carlsson A, Li H et al (1999) Clinical and genetic characteristics of type 2 diabetes with and without GAD antibodies. Diabetes 48:150–157PubMedCrossRefGoogle Scholar
  3. 3.
    Hawa MI, Kolb H, Schloot N (2013) Adult-onset autoimmune diabetes in Europe is prevalent with a broad clinical phenotype: action LADA 7. Diabetes Care 36:908–913PubMedCentralPubMedCrossRefGoogle Scholar
  4. 4.
    Davis TM, Wright AD, Mehta ZM et al (2005) Islet autoantibodies in clinically diagnosed type 2 diabetes: prevalence and relationship with metabolic control (UKPDS 70). Diabetologia 48:695–702PubMedCrossRefGoogle Scholar
  5. 5.
    van Deutekom AW, Heine RJ, Simsek S (2008) The islet autoantibody titres: their clinical relevance in latent autoimmune diabetes in adults (LADA) and the classification of diabetes mellitus. Diabet Med 25:117–125PubMedCrossRefGoogle Scholar
  6. 6.
    Tuomi T, Santoro N, Caprio S, Cai M, Weng J, Groop L (2013) The many faces of diabetes: a disease with increasing heterogeneity. Lancet 383:1084–1094PubMedCrossRefGoogle Scholar
  7. 7.
    Groop L, Pociot F (2014) Genetics of diabetes - are we missing the genes or the disease? Mol Cell Endocrinol 382:726–739PubMedCrossRefGoogle Scholar
  8. 8.
    Howson JM, Rosinger S, Smyth DJ et al (2011) Genetic analysis of adult-onset autoimmune diabetes. Diabetes 60:2645–2653PubMedCentralPubMedCrossRefGoogle Scholar
  9. 9.
    Andersen MK, Lundgren V, Turunen JA et al (2010) Latent autoimmune diabetes in adults differs genetically from classical type 1 diabetes diagnosed after the age of 35 years. Diabetes Care 33:2062–2064PubMedCentralPubMedCrossRefGoogle Scholar
  10. 10.
    Klinker MW, Schiller JJ, Magnuson VL et al (2010) Single-nucleotide polymorphisms in the IL2RA gene are associated with age at diagnosis in late-onset Finnish type 1 diabetes subjects. Immunogenetics 62:101–107PubMedCrossRefGoogle Scholar
  11. 11.
    Cervin C, Lyssenko V, Bakhtadze E et al (2008) Genetic similarities between latent autoimmune diabetes in adults, type 1 diabetes, and type 2 diabetes. Diabetes 57:1433–1437PubMedCrossRefGoogle Scholar
  12. 12.
    Lukacs K, Hosszufalusi N, Dinya E, Bakacs M, Madacsy L, Panczel P (2012) The type 2 diabetes-associated variant in TCF7L2 is associated with latent autoimmune diabetes in adult Europeans and the gene effect is modified by obesity: a meta-analysis and an individual study. Diabetologia 55:689–693PubMedCrossRefGoogle Scholar
  13. 13.
    Groop L, Forsblom C, Lehtovirta M et al (1996) Metabolic consequences of a family history of NIDDM (the Botnia study): evidence for sex-specific parental effects. Diabetes 45:1585–1593PubMedCrossRefGoogle Scholar
  14. 14.
    Isomaa B, Forsen B, Lahti K et al (2010) A family history of diabetes is associated with reduced physical fitness in the Prevalence, Prediction and Prevention of Diabetes (PPP)-Botnia study. Diabetologia 53:1709–1713PubMedCrossRefGoogle Scholar
  15. 15.
    Thorn LM, Forsblom C, Fagerudd J et al (2005) Metabolic syndrome in type 1 diabetes: association with diabetic nephropathy and glycemic control (the FinnDiane study). Diabetes Care 28:2019–2024PubMedCrossRefGoogle Scholar
  16. 16.
    Lindholm E, Agardh E, Tuomi T, Groop L, Agardh CD (2001) Classifying diabetes according to the new WHO clinical stages. Eur J Epidemiol 17:983–989PubMedCrossRefGoogle Scholar
  17. 17.
    Rolandsson O, Norberg M, Nystrom L et al (2012) How to diagnose and classify diabetes in primary health care: lessons learned from the Diabetes Register in Northern Sweden (DiabNorth). Scand J Prim Health Care 30:81–87PubMedCentralPubMedCrossRefGoogle Scholar
  18. 18.
    Berglund G, Elmstahl S, Janzon L, Larsson SA (1993) The Malmo Diet and Cancer Study. Design and feasibility. J Intern Med 233:45–51PubMedCrossRefGoogle Scholar
  19. 19.
    Voight BF, Kang HM, Ding J et al (2012) The metabochip, a custom genotyping array for genetic studies of metabolic, cardiovascular, and anthropometric traits. PLoS Genet 8:e1002793PubMedCentralPubMedCrossRefGoogle Scholar
  20. 20.
    Morris AP, Voight BF, Teslovich TM et al (2012) Large-scale association analysis provides insights into the genetic architecture and pathophysiology of type 2 diabetes. Nat Genet 44:981–990PubMedCentralPubMedCrossRefGoogle Scholar
  21. 21.
    Purcell S, Neale B, Todd-Brown K et al (2007) PLINK: a tool set for whole-genome association and population-based linkage analyses. Am J Hum Genet 81:559–575PubMedCentralPubMedCrossRefGoogle Scholar
  22. 22.
    Grant SF, Thorleifsson G, Reynisdottir I et al (2006) Variant of transcription factor 7-like 2 (TCF7L2) gene confers risk of type 2 diabetes. Nat Genet 38:320–323PubMedCrossRefGoogle Scholar
  23. 23.
    Zhou Y, Zhang E, Berggreen C et al (2012) Survival of pancreatic beta cells is partly controlled by a TCF7L2-p53-p53INP1-dependent pathway. Hum Mol Genet 21:196–207PubMedCrossRefGoogle Scholar
  24. 24.
    Saxena R, Gianniny L, Burtt NP et al (2006) Common single nucleotide polymorphisms in TCF7L2 are reproducibly associated with type 2 diabetes and reduce the insulin response to glucose in nondiabetic individuals. Diabetes 55:2890–2895PubMedCrossRefGoogle Scholar
  25. 25.
    Zampetti S, Spoletini M, Petrone A et al (2010) Association of TCF7L2 gene variants with low GAD autoantibody titre in LADA subjects (NIRAD Study 5). Diabet Med 27:701–704PubMedCrossRefGoogle Scholar
  26. 26.
    Field SF, Howson JM, Smyth DJ, Walker NM, Dunger DB, Todd JA (2007) Analysis of the type 2 diabetes gene, TCF7L2, in 13,795 type 1 diabetes cases and control subjects. Diabetologia 50:212–213PubMedCentralPubMedCrossRefGoogle Scholar
  27. 27.
    Barrett JC, Clayton DG, Concannon P et al (2009) Genome-wide association study and meta-analysis find that over 40 loci affect risk of type 1 diabetes. Nat Genet 41:703–707PubMedCentralPubMedCrossRefGoogle Scholar
  28. 28.
    Dubois PC, Trynka G, Franke L et al (2010) Multiple common variants for celiac disease influencing immune gene expression. Nat Genet 42:295–302PubMedCentralPubMedCrossRefGoogle Scholar
  29. 29.
    Li X, Thyssen G, Beliakoff J, Sun Z (2006) The novel PIAS-like protein hZimp10 enhances Smad transcriptional activity. J Biol Chem 281:23748–23756PubMedCrossRefGoogle Scholar
  30. 30.
    Lee J, Beliakoff J, Sun Z (2007) The novel PIAS-like protein hZimp10 is a transcriptional co-activator of the p53 tumor suppressor. Nucleic Acids Res 35:4523–4534PubMedCentralPubMedCrossRefGoogle Scholar
  31. 31.
    Luo X, Tarbell KV, Yang H et al (2007) Dendritic cells with TGF-beta1 differentiate naive CD4+CD25- T cells into islet-protective Foxp3+ regulatory T cells. Proc Natl Acad Sci U S A 104:2821–2826PubMedCentralPubMedCrossRefGoogle Scholar
  32. 32.
    Yoshimura A, Wakabayashi Y, Mori T (2010) Cellular and molecular basis for the regulation of inflammation by TGF-beta. J Biochem 147:781–792PubMedCentralPubMedCrossRefGoogle Scholar
  33. 33.
    Trudeau JD, Dutz JP, Arany E, Hill DJ, Fieldus WE, Finegood DT (2000) Neonatal beta-cell apoptosis: a trigger for autoimmune diabetes? Diabetes 49:1–7PubMedCrossRefGoogle Scholar
  34. 34.
    Okamura S, Arakawa H, Tanaka T et al (2001) p53DINP1, a p53-inducible gene, regulates p53-dependent apoptosis. Mol Cell 8:85–94PubMedCrossRefGoogle Scholar
  35. 35.
    Cho YS, Chen CH, Hu C et al (2011) Meta-analysis of genome-wide association studies identifies eight new loci for type 2 diabetes in east Asians. Nat Genet 44:67–72PubMedCentralPubMedCrossRefGoogle Scholar
  36. 36.
    Qu HQ, Grant SF, Bradfield JP et al (2008) Association analysis of type 2 diabetes Loci in type 1 diabetes. Diabetes 57:1983–1986PubMedCentralPubMedCrossRefGoogle Scholar
  37. 37.
    Raj SM, Howson JM, Walker NM et al (2009) No association of multiple type 2 diabetes loci with type 1 diabetes. Diabetologia 52:2109–2116PubMedCentralPubMedCrossRefGoogle Scholar
  38. 38.
    Karamitri A, Renault N, Clement N, Guillaume JL, Jockers R (2013) Minireview: toward the establishment of a link between melatonin and glucose homeostasis: association of melatonin MT2 receptor variants with type 2 diabetes. Mol Endocrinol 27:1217–1233PubMedCrossRefGoogle Scholar
  39. 39.
    Hardeland R (2013) Melatonin and the theories of aging: a critical appraisal of melatonin's role in antiaging mechanisms. J Pineal Res 55:325–356PubMedGoogle Scholar
  40. 40.
    Lyssenko V, Nagorny CL, Erdos MR et al (2009) Common variant in MTNR1B associated with increased risk of type 2 diabetes and impaired early insulin secretion. Nat Genet 41:82–88PubMedCentralPubMedCrossRefGoogle Scholar
  41. 41.
    Ramracheya RD, Muller DS, Squires PE et al (2008) Function and expression of melatonin receptors on human pancreatic islets. J Pineal Res 44:273–279PubMedCrossRefGoogle Scholar
  42. 42.
    Bonnefond A, Clement N, Fawcett K et al (2012) Rare MTNR1B variants impairing melatonin receptor 1B function contribute to type 2 diabetes. Nat Genet 44:297–301PubMedCentralPubMedCrossRefGoogle Scholar
  43. 43.
    Bradfield JP, Qu HQ, Wang K et al (2011) A genome-wide meta-analysis of six type 1 diabetes cohorts identifies multiple associated loci. PLoS Genet 7:e1002293PubMedCentralPubMedCrossRefGoogle Scholar
  44. 44.
    Vaxillaire M, Boccio V, Philippi A et al (1995) A gene for maturity onset diabetes of the young (MODY) maps to chromosome 12q. Nat Genet 9:418–423PubMedCrossRefGoogle Scholar
  45. 45.
    Kathiresan S, Willer CJ, Peloso GM et al (2009) Common variants at 30 loci contribute to polygenic dyslipidemia. Nat Genet 41:56–65PubMedCentralPubMedCrossRefGoogle Scholar
  46. 46.
    Voight BF, Scott LJ, Steinthorsdottir V et al (2010) Twelve type 2 diabetes susceptibility loci identified through large-scale association analysis. Nat Genet 42:579–589PubMedCentralPubMedCrossRefGoogle Scholar
  47. 47.
    Odom DT, Zizlsperger N, Gordon DB et al (2004) Control of pancreas and liver gene expression by HNF transcription factors. Science 303:1378–1381PubMedCentralPubMedCrossRefGoogle Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 2014

Authors and Affiliations

  • Mette K. Andersen
    • 1
    • 2
    • 3
    Email author
  • Maria Sterner
    • 4
  • Tom Forsén
    • 5
  • Annemari Käräjämäki
    • 6
    • 7
  • Olov Rolandsson
    • 8
  • Carol Forsblom
    • 1
    • 2
    • 9
  • Per-Henrik Groop
    • 1
    • 2
    • 9
    • 10
  • Kaj Lahti
    • 6
  • Peter M. Nilsson
    • 11
  • Leif Groop
    • 3
    • 4
  • Tiinamaija Tuomi
    • 1
    • 2
    • 12
  1. 1.Research Programs Unit, Diabetes and ObesityUniversity of HelsinkiHelsinkiFinland
  2. 2.Folkhälsan Research CenterHUSHelsinkiFinland
  3. 3.Institute for Molecular Medicine Finland (FIMM)University of HelsinkiHelsinkiFinland
  4. 4.Department of Clinical Sciences, Diabetes and EndocrinologyLund UniversityMalmöSweden
  5. 5.Department of General Practice and Primary Health CareUniversity of HelsinkiHelsinkiFinland
  6. 6.Vasa Health Care CenterVasaFinland
  7. 7.Department of Primary Health CareVasa Central HospitalVasaFinland
  8. 8.Department of Public Health and Clinical Medicine, Family MedicineUmeå UniversityUmeåSweden
  9. 9.Division of NephrologyHelsinki University Central HospitalHelsinkiFinland
  10. 10.Baker IDI Heart and Diabetes InstituteMelbourneAustralia
  11. 11.Department of Clinical Sciences, Division of MedicineLund UniversityMalmöSweden
  12. 12.Department of MedicineHelsinki University HospitalHelsinkiFinland

Personalised recommendations