Insulin Resistance-Associated Genetic Variants in Type 1 Diabetes

Abstract

Aims:

To examine candidate insulin resistance single nucleotide polymorphisms (SNPs) for associations with glycemic control, insulin resistance, BMI, and complications in an observational type 1 diabetes (T1D) cohort: the Pittsburgh Epidemiology of Diabetes Complications (EDC) study.

Methods:

In 422 European-ancestry participants, we assessed associations using additive models between 15 candidate SNPs and 25-year mortality, cardiovascular disease, microalbuminuria, overt nephropathy and proliferative retinopathy, and 25-year mean HbA1c, estimated glucose disposal rate (eGDR, inverse measure of insulin resistance), and BMI.

Results:

The A allele of rs12970134 was associated with higher mean HbA1c (β=+0.34±0.09, p=0.00009) and nominally associated with worse eGDR (p=0.02). Further analyses suggest the HbA1c association may be modified by diabetes therapy regimen: rs12970134 AA genotype was associated with higher HbA1c under non-intensive therapy conditions (<3 insulin injections/day or monitoring blood glucose<3 times/day [p=0.004]), but not under intensive therapy (≥3 injections/day or insulin pump and monitoring glucose≥3 times/day [p=0.71]). There were no significant associations between any SNPs and BMI or complications.

Conclusions:

rs12970134, near MC4R, is strongly associated with HbA1c in this cohort. Further exploration of this genomic region is warranted, as it may hold promise for discovering new therapeutic targets to improve glycemic control in T1D.

1. INTRODUCTION

Insulin resistance is known to increase with weight gain and physical inactivity.¹ In addition, many single nucleotide polymorphisms (SNPs) in multiple loci have been shown to be associated with insulin resistance in the general population.² As in the general population, insulin resistance is often associated with weight gain and obesity in type 1 diabetes (T1D), but also occurs in their absence.³ While the role of insulin resistance in T1D has yet to be fully elucidated,⁴ increased insulin resistance is associated with poor glycemic control, greater insulin dose requirements, and increased risk of complications.^5–7 Thus, insulin resistance itself has been proposed as a potential target for adjunctive therapies to insulin treatment in T1D,^8,9 raising the possibility that early identification of individuals at increased risk for insulin resistance may help to guide diabetes management. To our knowledge, the association between insulin resistance genetic variants and outcomes in T1D has not previously been explored. Thus, our objective was to examine whether 15 SNPs, which have previously been associated with insulin resistance in the general population (Supplementary Table 1), are associated with glycemic control, estimated insulin resistance, body mass index (BMI), mortality, and complications in a US T1D cohort, the Pittsburgh Epidemiology of Diabetes Complications (EDC) study.

2. SUBJECTS, MATERIALS AND METHODS

2.1. Study Population

The Pittsburgh EDC Study is an historical-prospective cohort study of childhood-onset (<17 years old) T1D. All participants (n=658) were diagnosed with T1D, or seen within one year of their diagnosis, at Children’s Hospital of Pittsburgh between 1950 and 1980. The cohort, which has been described in detail elsewhere,^10,11 has been followed since 1986–1988, with biennial examinations and questionnaires for the first ten years and thereafter with biennial questionnaires and further examinations at 18- and 25-years post-baseline. Research protocols were approved by the University of Pittsburgh institutional review board (approval #19040065) and all participants provided written informed consent.

Genotyping and Candidate SNPs

Genotyping was performed using the Infinium HumanCore Exome-24 BeadChip (Illumina, San Diego, CA, USA), following the manufacturer’s protocol. A flow chart describing how the analytic sample was determined is provided in Supplementary Figure 1. A total of n=496 participants gave consent for DNA extraction from whole blood between 1988–1998. Of those, n=43 participants’ DNA specimens were depleted or degraded by the time genotyping was completed in 2015, leaving n=453 in the genotyped sample. After quality control, those with extreme heterozygosity, genotyping rate <95%, sample exclusions (discordance between reported and genotype-inferred sex), or genetic outliers were removed, leaving a sample of n=441 participants with quality genetic data. Due to a low proportion of non-European ancestry participants in the EDC study (<2%), these analyses were restricted to participants of European ancestry, leaving n=432. Finally, there were ten first-degree relative pairs, thus one individual from each pair was randomly selected for exclusion, resulting in the final analytic sample size of n=422. SNPs deviating from Hardy Weinberg equilibrium were excluded from the genotyping data during quality control. Minimac3/Minimac3-omp version 1.0.14 was used for imputation, with the 1000 Genomes (1KG) Phase 3 version 5 reference panel (updated Oct 20, 2015). The 15 candidate SNPs (Supplementary Table 1) examined here were chosen because they have been reported to be significantly associated with measures of insulin resistance in the general population, as recently reviewed.²

2.2. Phenotype Definitions

HbA1 was assessed at baseline and repeated at 2-, 4-, 6-, 8-, and 10-years, and HbA1c was assessed at 18- and 25-years of follow-up. For the first 18 months of the study, HbA1 was measured in fasting blood samples using microcolumn cation exchange (Isolab, Akron, OH, USA). For the remainder of the first 10 years of follow-up, HbA1 was measured using automated high-performance liquid chromatography (Diamat; Biorad, Hercules, CA, USA). The two assays had high agreement (r=0.95; Diamat HbA1=−0.18+1.00[Isolab HbA1]). HbA1 values were converted to DCCT-aligned HbA1c values using a regression equation derived from duplicate assays (DCCT HbA1c=0.14 + 0.83[EDC HbA1]).¹² At the 18- and 25-year examinations, HbA1c was measured using the DCA 2000 analyzer (Bayer Healthcare LLC. Elkhart, IN, USA) and converted to DCCT-aligned HbA1c with the equation: DCCT HbA1c=(EDC HbA1c-1.13)/0.81.

Height and weight were measured using standard methods to calculate BMI. Waist and hip circumference were measured at least twice; the average of each was used to calculate the waist-hip ratio (WHR). Blood pressure was measured according to the Hypertension Detection and Follow-up Program protocol,¹³ with a random-zero sphygmomanometer from baseline through the 10-year exam, replaced by an aneroid sphygmomanometer beginning with the 18-year exam. Estimated glucose disposal rate (eGDR), a validated, inverse measure of insulin resistance derived from hyperinsulinemic-euglycemic clamp studies,¹⁴ was calculated using the equation:

$$\text{eGDR}(\text{mg}/\text{kg}/\text{min})=24.395-\left(12.971\ast \text{Waist}-\text{Hip}\text{Ratio}\right)-(3.388\ast \text{Hypertension})-(0.601\ast \text{HbA}1\text{c})$$

where Hypertension is blood pressure ≥140/90 mmHg or use of blood-pressure lowering medication (0=no, 1=yes).

Insulin regimen, medication use, smoking, and alcohol consumption were self-reported. Leisure physical activity (kcal/week) was estimated using the Paffenbarger questionnaire.¹⁵ Total caloric intake, sodium intake, percent calories from carbohydrate, protein, total fat, and saturated fat intake were estimated from the Harvard-Willet Semi-Quantitative Food Frequency Questionnaire, a 1-year, 115-item food frequency questionnaire.¹⁶ The questionnaire was administered at baseline (1986–88), visit two (1988–90), and visit six (1996–98).

Cause-specific mortality was determined over 25 years using medical records, death certificates, autopsy reports, and/or interview with next of kin. Causes of death were classified using all available information according to the Diabetes Epidemiology Research International (DERI) system¹⁷ by a committee of physicians. Cardiovascular disease (CVD) mortality was defined as fatal CVD, myocardial infarction, or stroke as either the primary or a contributing cause of death. Renal disease mortality was defined as death due to renal failure as either a primary or contributing cause of death. The two aforementioned cause-specific mortality classifications were not mutually exclusive.

Total CVD was defined as CVD mortality, fatal or nonfatal myocardial infarction (MI, including clinical events and subclinical myocardial infarction on ECG, i.e. Minnesota code 1.1 or 1.2), coronary revascularization procedure, blockage ≥50%, ischemic EGC at an EDC study visit (Minnesota codes 1.3, 4.1–4.3, 5.1–5.3, 7.1), EDC physician-diagnosed angina, or fatal or nonfatal stroke. Medical records were used to confirm self-reported CVD events.

The assessment of proliferative diabetic retinopathy (PDR) has been described in detail elsewhere.¹⁸ Briefly, stereoscopic color fundus photographs of three standard fields (1, 2, and 4) were taken with a Zeiss camera (Carl Zeiss, Oberkochen, Germany) after pupil dilation. Photos were graded using a modification of the Early Treatment Diabetic Retinopathy Study (ETDRS) adaptation of the modified Airlie House classification of PDR.¹⁹ Grading of stereoscopic fundus photographs was performed at the Fundus Photograph Reading Center at the University of Wisconsin-Madison. PDR was defined as ETDRS score ≥60 or laser photocoagulation for PDR.

At the baseline, 2-, 4-, 6-, 8, 10-, and 18 year EDC examinations, urinary albumin was measured by immunonephelometry. Albumin excretion rate (AER) was calculated for each of three timed urine samples (24-hour, overnight, and 4-hour collections obtained over a two-week period). Microalbuminuria (MA) was defined as ever having an albumin excretion rate (AER) ≥20 μg/min in 2 of 3 timed urine samples prior to the 25-year exam, Urinary Albumin/Creatinine Ratio (UCAR) ≥30 mg/g at the 25-year examination in 2 of 3 samples, or a history of dialysis or renal transplantation. Overt nephropathy (ON) was defined as ever having AER ≥200 μg/min in 2 of 3 timed urine samples prior to the 25-year examination, UACR ≥300 mg/g at the 25-year examination in 2 of 3 samples, or a history of dialysis or renal transplantation.

2.3. Statistical Analyses

For each candidate SNP, additive linear models were fit to estimate the increase in 25-year mean HbA1c, eGDR, and BMI associated with each additional risk allele. Similarly, for each SNP, additive logistic regression models were used to assess the association between each additional risk allele and odds of being positive for each complication. All models were adjusted for sex and principal components of ancestry. To account for the number of comparisons being made, we used a Bonferroni-corrected significance level of p=0.003 (i.e., 0.05 corrected for 15 SNPs).

Separate general additive mixed models were fit to assess the association between rs12970134 genotype and 25-year longitudinal HbA1c, eGDR, and BMI trajectories, with genotype modeled as a fixed effect and intercept and time as random effects. All available repeated measures data were incorporated into these mixed models, thus no participants were excluded due to missing data. In secondary, exploratory analyses, similar general additive mixed models were fit with total caloric intake and percent calories from carbohydrate, protein, total fat, and saturated fat intake as outcomes, as these data were only available at three time points (baseline, visit two, and visit six). All models were adjusted for sex, principal components of ancestry, T1D duration at baseline, and insulin dose and current smoking status at each time point. The models for HbA1c and eGDR were also adjusted for BMI at each time point. As EDC is an exclusively childhood-onset T1D cohort, age and T1D duration are highly correlated (r=0.86, p<.0001). Thus, the models presented here were adjusted for T1D duration only, however, alternative models were fit adjusting for age instead of T1D duration and results remained the same. Waist-hip ratio and hypertension, the other individual components of eGDR (besides HbA1c), were also examined in separate models.

2.4. Replication

Replication analyses were performed in the Scottish Diabetes Research Network Type 1 Bioresource (SDRNT1BIO).²⁰ The analytic sample comprised 5,259 participants with available DNA and HbA1c measured at baseline. The cross-sectional association between rs12970134 and HbA1c was examined at baseline (2010–2012) using general additive models adjusted for BMI, sex, diabetes duration, age, and principal components of ancestry.

3. RESULTS

3.1. Associations Between Candidate SNPs and Insulin Resistance Traits and Complications

The baseline characteristics of the genotyped EDC cohort are shown in Table 1. Associations between each SNP and insulin resistance-related traits are shown in Table 2. Only the association between rs12970134 (located in a region containing the candidate gene, MC4R) and mean HbA1c was statistically significant. Each additional A allele of rs12970134 was associated with a 0.34% increase in HbA1c (p=0.00009). While not reaching statistical significance using the adjusted significance level, rs12970134 was also associated with lower insulin sensitivity, as measured by eGDR (β=−0.32, p=0.02). Additional associations approaching statistical significance were observed between rs17046216 in MSMO1 and eGDR (β=−0.31, p=0.01), rs7077836 in TCERG1L and eGDR (β=+0.36, p=0.03), and rs8050136 in FTO and BMI (β=+0.58, p=0.02). There were no significant associations between the candidate SNPs and mortality (Supplementary Table 2) or complication outcomes (Supplementary Table 3), although the association between rs17046216 (an intronic SNP in MSMO1) and microalbuminuria approached the Bonferroni-corrected significance threshold (OR=1.62, 95% CI: 1.17, 2.24, p=0.004).

Table 1.

Baseline characteristics of the subgroup of the EDC study cohort with genotyping data

	n=422
Age (years)	27.2 (7.9)
Age at Type 1 Diabetes Onset (years)	8.3 (4.0)
Type 1 Diabetes Duration (Years)	18.8 (7.5)
% Female (n)	47% (199)
HbA1c
%	8.7 (1.5)
mmol/mol	72 (16.1)
Estimated Glucose Disposal Rate (mg/kg/min)	7.9 (1.8)
Body Mass Index (kg/m2)	23.7 (3.3)
Waist-Hip Ratio (cm/cm)	0.82 (0.07)
Insulin Dose (units/kg body weight)a	0.80 (0.25)
Systolic BP (mmHg)	112.8 (15.3)
Diastolic BP (mmHg)	72.2 (10.7)
Total Cholesterol (mg/dl)	187.1 (40.6)
HDL Cholesterol (mg/dl)	53.9 (12.1)
LDL Cholesterol (mg/dl)b	112.5 (32.4)
Non-HDL Cholesterol (mg/dl)	133.2 (40.0)
Triglycerides (mg/dl), median (IQR)c	81 (59–117)
Albumin Excretion Rate (μg/min), median (IQR)	13 (8–122)
Estimated Glomerular Filtration Rate (mL/min/1.73m2)	103.3 (31.8)
White Blood Cell Count (x109 cells/L)	6.6 (1.9)
% Ever Smoker (n)	35% (149)

Values are mean (standard deviation) unless otherwise noted,

^a14 missing/not reported,

^b27 missing due to non-fasting blood draw or could not be calculated due to high triglycerides,

^c20 missing due to non-fasting blood draw

Table 2.

Candidate SNP associations with insulin resistance-related traits in the EDC cohort

SNP	Chr.a	Nearest Gene	Allelesb	Risk Allele Frequency	eGDRc,d	HbA1cc	BMIc
rs2943641	2	IRS1	T/C	0.62	0.06±0.12, p=0.63	0.03±0.08, p=0.72	−0.32±0.28, p=0.18
rs780094	2	GCKR	T/C	0.61	−0.11±0.12, p=0.34	0.08±0.08, p=0.29	−0.12±0.23, p=0.60
rs6723108	2	TMEM163	G/T	0.62	0.07±0.13, p=0.57	−0.16±0.09, p=0.07	0.42±0.26, p=0.10
rs998451	2	TMEM163	G/A	0.37	−0.03±0.13, p=0.82	0.14±0.09, p=0.10	−0.51±0.26, p=0.05
rs13081389	3	PPARG	A/G	0.06	0.07±0.24, p=0.79	−0.21±0.17, p=0.20	0.62±0.21, p=0.21
rs17046216	4	MSMO1	T/A	0.32	−0.31±0.12, p=0.01	0.08±0.08, p=0.33	−0.05±0.24, p=0.84
rs702634	5	ARL15	G/A	0.70	0.24±0.13, p=0.06	−0.19±0.09, p=0.03	0.02±0.26, p=0.93
rs4311394	5	ARL15	A/G	0.26	0.22±0.13, p=0.10	0.02±0.09, p=0.86	0.33±0.27, p=0.22
rs972283	7	KLF14	A/G	0.52	−0.19±0.11, p=0.09	−0.007±0.08, p=0.93	−0.004±0.23, p=0.99
rs1208	8	NAT2	G/A	0.56	−0.08±0.11, p=0.46	−0.13±0.07, p=0.08	0.10±0.22, p=0.64
rs7903146	10	TCF7L2	C/T	0.28	−0.03±0.13, p=0.79	0.09±0.09, p=0.29	−0.32±0.25, p=0.21
rs7077836	10	TCERG1L	G/A	0.15	0.36±0.16, p=0.03	0.01±0.11, p=0.91	−0.10±0.32, p=0.75
rs35767	12	IGF1	A/G	0.83	0.08±0.16, p=0.64	0.10±0.11, p=0.35	−0.17±0.32, p=0.58
rs8050136	16	FTO	C/A	0.41	0.07±0.12, p=0.57	−0.11±0.08, p=0.19	0.58±0.24, p=0.02
rs12970134	18	MC4R	G/A	0.26	−0.30±0.13, p=0.02	0.34±0.09, p=0.00009	0.33±0.26, p=0.20

Bolded text indicates a significant association p<0.003

^aChromosome

^bReference allele/risk allele

^cβ±standard error, p-value. All models adjusted for sex and first three principal components of ancestry

^dEstimated glucose disposal rate

3.2. Examination of Characteristics by rs12970134 Genotype

Given the strong association between rs12970134 and HbA1c, we performed additional analyses to further explore this relationship. Baseline characteristics by rs12970134 genotype are presented in Table 3. Baseline HbA1c increased with each additional rs12970134 A allele (p=0.004). There were also associations between a greater number of A alleles and lower baseline eGDR (p=0.06), greater proportion who were current smokers (p=0.04), and higher baseline saturated fat intake as a percent of total calories (p=0.04). None of the other factors examined differed by genotype. Longitudinal trajectories of mean HbA1c, eGDR, and BMI by rs12970134 genotype (Figure 1) demonstrated that each additional rs12970134 A allele was associated with significantly higher HbA1c (Figure 1a) (β=0.35, SE=0.08, p=0.00003) and lower eGDR (Figure 1b, β=−0.30, SE=0.11, p=0.01) over follow-up, adjusting for principal components of ancestry, diabetes duration, sex, current smoking status, insulin dose, and BMI. Adjusting for waist-hip ratio instead of BMI in HbA1c models did not alter the results. There was no significant association between rs12970134 and longitudinal increase in BMI (Figure 1c, β=0.33, SE=0.23, p=0.16). We additionally examined the two remaining individual components of eGDR, waist-hip ratio (Supplementary Figure 2) and hypertension (Supplementary Figure 3), and observed no significant association between rs12970134 genotype and either component.

Table 3.

Baseline characteristics of the EDC study cohort by rs12970134 genotype

	GG (n=236)	GA (n=153)	AA (n=33)	β ±SEa	p-value
Age (years)	27.0 (7.6)	27.6 (8.0)	26.7 (8.1)	0.04 ±0.60	0.95
Age at Type 1 Diabetes Onset (years)	8.1 (4.1)	8.8 (3.9)	8.1 (3.8)	0.23 ±0.31	0.45
Type 1 Diabetes Duration (Years)	18.9 (7.5)	18.9 (7.6)	18.6 (8.0)	−0.19 ±0.58	0.74
% Female (n)	47.9% (113)	46.4% (71)	45.5% (15)	−0.06 ±0.15	0.71
HbA1c (%)	8.5 (1.5)	8.8 (1.5)	9.3 (1.4)	0.33 ±0.11	0.004
eGDR (mg/kg/min)	8.0 (1.8)	7.8 (1.8)	7.3 (2.0)	−0.25 ±0.13	0.06
Body Mass Index (kg/m2)	23.6 (3.3)	23.6 (3.4)	24.3 (2.9)	0.30 ±0.25	0.23
Waist-Hip Ratio (cm/cm)	0.82 (0.07)	0.82 (0.07)	0.84 (0.07)	0.005 ±0.004	0.17
Insulin Dose (units/kg body weight)	0.80 (0.26)	0.79 (0.24)	0.86 (0.27)	0.008 ±0.02	0.70
Systolic BP (mmHg)	112.3 (14.1)	112.8 (17.1)	115.7 (14.8)	1.03 ±1.14	0.37
Diastolic BP (mmHg)	72.1 (10.6)	72.0 (10.7)	73.5 (11.3)	0.33 ±0.80	0.68
% ACE Inhibitor Use, (n)	3.5% (8)	4.0% (6)	0% (0)	−0.23 ±0.24	0.62
% Diuretic Use, (n)	9.3% (21)	10.6% (16)	0% (0)	−0.29 ±0.29	0.32
Total Cholesterol (mg/dl)	187.7 (38.9)	182.9 (42.1)	202.9 (41.9)	2.7 ±3.1	0.39
HDL Cholesterol (mg/dl)	54.4 (12.2)	53.3 (12.1)	53.9 (11.0)	−0.45 ±0.86	0.60
LDL Cholesterol (mg/dl)	112.8 (30.8)	110.1 (36.0)	121.3 (24.9)	1.7 ±2.6	0.52
Non-HDL Cholesterol (mg/dl)	133.3 (38.3)	129.6 (41.6)	148.9 (41.8)	3.2 ±3.1	0.31
Triglycerides (mg/dl), median (IQR)	82.0 (59–117)	79.5 (60.5–106.5)	85.0 (61–165)	0.04 ±0.04b	0.38
Albumin Excretion Rate (μg/min), median (IQR)	33.1 (7.8–134)	14.0 (7.9–73.0)	25.0 (11.00–284)	0.15 ±0.15b	0.31
eGFR (mL/min/1.73m2)	102.4 (33.8)	103.6 (29.4)	108.3 (27.8)	2.65 ±2.45	0.28
White Blood Cell Count (x109 cells/L)	6.58 (1.98)	6.47 (1.77)	6.82 (1.89)	0.02 ±0.15	0.89
Leisure physical activity (kcal/week), median (IQR)	1602 (672–3138)	1725 (802–2668)	2548 (952–3084)	0.11 ±0.10b	0.26
% Ever Smoker (n)	34.3% (81)	34.6% (53)	45.5% (15)	0.12 ±0.16	0.46
% Current Smoker (n)	18.6% (44)	24.8% (38)	33.3% (11)	0.37 ±0.18	0.04
Alcoholic beverages/week, median (IQR)	0 (0–2)	0 (0–1)	0 (0–0)	−0.03 ±0.08	0.72
Total Caloric Intake (cal/day)	2127 (779)	2104 (713)	1870 (472)	−89.8 ±51.1	0.08
Total Sodium Intake (mg/day)	2490 (933)	2418 (844)	2476 (1371)	−37.9 ±68.6	0.58
Percent Carbohydrate Intake (% of total calories)	48.1 (7.0)	48.3 (6.8)	45.6 (6.6)	−0.63 ±0.54	0.24
Percent Protein Intake (% of total calories)	17.9 (3.1)	17.6 (3.2)	17.8 (2.4)	−0.13 ±0.24	0.57
Percent Total Fat Intake (% of total calories)	34.1 (5.3)	33.4 (5.8)	36.8 (6.1)	0.40 ±0.44	0.36
Percent Saturated Fat Intake (% of total calories)	12.0 (2.6)	11.8 (2.3)	13.6 (3.0)	0.41 ±0.19	0.04

Values are mean (SD), unless noted.

^aAdjusted for sex and principal components of ancestry,

^bNatural log transformed

Figure 1.

Longitudinal trajectories of HbA1c (panel A), estimated glucose disposal rate (eGDR, panel B), and body mass index (BMI, panel C) by rs12970134 genotype in the EDC cohort. Solid lines represent the first 10 years of the study with regular biennial clinical examinations (1986–88 to 1996–98); dashed lines represent the longer interim periods which occurred after that time.

3.3. Association between rs12970134 and HbA1c by Intensive Insulin Therapy Status

There was no significant interaction between rs12970134 and time with respect to HbA1c (p=0.38), eGDR (p=0.26), or BMI (p=0.83); however, the differential between genotypes in HbA1c is notably reduced beginning at 18 years of follow-up (Figure 1a). At that time, 79% of participants were using ≥3 insulin injections/day or an insulin pump, and 28% of participants were self-monitoring blood glucose (SMBG) ≥3 times/day. Thus, a genotype × treatment regimen interaction with respect to HbA1c was examined in secondary analyses using cross-sectional data from the 10-year (1996–1998) EDC follow-up visit (n=331). This time point was chosen as it occurred shortly after the results of the Diabetes Control and Complications Trial (DCCT), a time period during which intensive diabetes therapy was being widely adopted into clinical practice, and it was the only EDC clinical follow-up visit that had adequate numbers of participants both using (23.4%, n=78) and not using (76.6%, n=256) intensive therapy (defined as ≥3 insulin injections/day or use of an insulin pump and SMBG ≥3 times/day, Supplementary Table 4).

Participants using intensive therapy had lower HbA1c and systolic blood pressure, higher eGDR, and saw their physicians more often than those not using intensive therapy (Supplementary Table 4). As shown in Figure 2, the association between each additional rs12970134 A allele and higher HbA1c was observed only in those who were using non-intensive therapy (Figure 2a, additive model p=0.13, recessive model p=0.006), while no difference was observed in those using intensive therapy (Figure 2b, additive model p=0.99, recessive model p=0.68). We also examined the non-intensive therapy group according to three therapeutic regimens: 1) <3 insulin injections/day and SMBG< 3 times/day, 2) <3 insulin injections/day, but SMBG≥3 times/day, and 3) ≥3 insulin injections/day or use of an insulin pump, but SMBG<3 times/day. For the AA genotype, HbA1c was similarly high in the two treatment groups using <3 insulin insulin/day regardless of SMBG status, however numbers were sparse (Supplementary Table 5).

Figure 2.

Smoothed density plots of HbA1c by rs12970134 genotype and intensive therapy status at the EDC 1996–98 (year 10) study visit. Non-intensive therapy is defined as <3 injections of insulin/day or self-monitoring blood glucose <3 times/day. Intensive therapy is defined as ≥3 injections of insulin per day or use of insulin pump and self-monitoring blood glucose ≥3 times/day.

3.4. Replication Analysis of the rs12970134 and HbA1c Association in SDRNT1BIO

There was no association between rs12970134 and HbA1c in SDRNT1BIO overall (Supplementary Figure 4). To increase comparability with the EDC cohort, the analyses were repeated in SDRNT1BIO participants with type 1 diabetes onset <17 years old; again no association was observed (Supplementary Figure 5). When examined by intensive insulin therapy status, as in EDC, there was no association in those using intensive therapy (Supplementary Figure 6). In participants not using intensive therapy, the rs12970134 AA genotype had slightly lower mean HbA1c in the SDRNT1BIO cohort (p=0.009), in contrast to EDC, where the AA genotype was associated with higher mean HbA1c.

4. DISCUSSION

In this examination of candidate IR SNPs, we found a strong association between rs12970134 and glycemic control (i.e., HbA1c) and a weaker association with eGDR, an inverse measure of insulin resistance, despite no association with BMI or waist-hip ratio, in this T1D cohort. To our knowledge, the current study is the first to report an association between rs12970134 and HbA1c, as well as the first to examine any phenotypic associations with rs12970134 in a T1D cohort. The rs12970134 SNP is located in a region containing the melanocortin-4 receptor (MC4R) gene (18q21.32), which is involved in regulating appetite and energy homeostasis. The role of MC4R in human body weight was first described over two decades ago.²¹ Since then, mutations in MC4R have been determined to be a common cause of monogenic obesity in European-ancestry populations and are also associated with hyperinsulinemia in the general population.²² The SNP examined here, rs12970134, is located 154 kb downstream from the 3’ end of MC4R and has been associated with obesity^23,24 and insulin resistance²³ in genome-wide association studies. In subsequent meta-analyses of studies in the general population, the A allele of rs12970134 was confirmed to be associated with higher BMI, as well as with greater insulin resistance and type 2 diabetes, independent of BMI.^25,26 The lack of a strong association between rs12970134 and BMI in the EDC cohort could be due to a differences in the distribution of BMI in this T1D cohort, which was lower on average and had less variability compared to BMI in the aforementioned general population studies where mean BMI was in the overweight-obese range.^23–26 Similar to our findings, the rs12970134 A allele has recently been associated with having a 12-year fasting glucose trajectory > 75^th percentile in non-diabetic children.²⁷

It is important to note that, to date, the function of rs12970134 has not been determined. It is hypothesized to be related to leptin-melantocortin pathway signaling, due to its proximity to and common phenotypic associations with the MC4R gene, but its specific function has yet to be elucidated. Functional characterization of variants in the MC4R proximal 3’ linkage disequilibrium non-coding region, where rs12970134 is located, has been examined in detail, as it may contain a possible insulator element, potentially modulating interactions between MC4R promotor and enhancers.²⁸ The insulator in this region could explain why non-coding SNPs, such as rs12970134, have been consistently associated with MC4R-related phenotypes but not RNA expression in tissues examined by the GTEx consortium. Given the consistent phenotypic associations with SNPs proximal to MC4R, more exploration of this genomic region is warranted.

The potential effect modification of the association between rs12970134 and HbA1c by insulin regimen observed in this in T1D cohort is intriguing. We observed that while individuals with the high risk AA genotype have higher absolute HbA1c, they have similar relative improvement in HbA1c compared to the GA and GG genotypes beginning in 1994–96 (Figure 1). This improvement occurred after publication of the DCCT results in 1993, which led to widespread adoption of intensive diabetes therapy. Our analyses stratifying by insulin regimen at the 1996–98 study visit further suggest that under conditions of intensive diabetes therapy, individuals with the AA genotype are able to attain the same level of glycemic control as those with the lower risk genotypes. It has been shown in prior studies that insulin therapy leads to increased leptin levels in T1D, which is a leptin deficient state at onset due to loss of adipose tissue stores prior to insulin therapy initiation.³ If rs12970134 is indeed involved in the leptin-melanocortin pathway, then a possible explanation for the difference in its association with HbA1c by intensive therapy status is that A allele carriers may have greater impairment of leptin expression or signaling, compared to the G allele, when insulin levels are less physiologic. However, under conditions of intensive therapy, leptin function in A-allele carriers may be restored to similar levels as the G allele carriers, leading to comparable glycemic control. Exogenous administration of insulin has previously been shown to improve leptin levels in both new-onset T1D^29,30 and poorly controlled T2D.³¹ Additionally, in a small study of T1D patients, those who were using intensive insulin therapy had higher leptin levels than those on conventional insulin therapy, independent of body weight.³² Our findings here support further investigation, in T1D specifically, of the leptin-melanocortin pathway and MC4R signaling, which has previously been postulated to hold promise for identifying new adjunctive treatments to improve glycemic control and reduce adverse consequences of insulin therapy (i.e., hypoglycemia and weight gain).³³ In a recent pilot study, after 20 weeks of recombinant methionyl human leptin (metreleptin) therapy in T1D, there were decreases in insulin dose and adiposity, though glycemic control did not improve significantly.³⁴ However, the finding that the same level of glycemic control was maintained at a lower insulin dose supports further exploration of this pathway in T1D.

4.1. Strengths and Limitations

The EDC study is a well-characterized, prospective cohort study which allows the simultaneous assessment of multiple mortality and complication outcomes, as well as clinical risk factors, over 25 years of follow-up. For our longitudinal models, we have used mixed models with maximum likelihood estimation, which utilize all available information over follow-up. Limitations of the study include the possibility of survivor bias at the later time points. As shown in Supplementary Table 2, rs12970134 genotype was not associated with mortality. However, the AA genotype did have somewhat higher mortality (33%) by the end of follow-up than the other genotypes (22% in GG and 24% in GA), which may at least partially account for the lower HbA1c observed in the remaining survivors in the AA genotype at the 2011–13 study visit. An additional limitation is that, due to small numbers, we were unable to use the DCCT definition of intensive diabetes therapy, which includes SMBG≥4 times/day, as only 51 participants (GG n=31, GA n=18, AA n=2) in the current analyses met this definition. Another limitation is that while a replication analysis was performed in the SDRNT1BIO cohort, the finding of an association between the A allele of rs12970134 and higher HbA1c observed in EDC was not validated. Given the different calendar years of follow-up between the two cohorts, there are potentially important differences that could not be explored here due to sample size limitations (e.g., differences in insulin regimen not fully captured by the intensive insulin status definition, dietary intake, and BMI distribution). Thus, further replication studies are needed to validate the EDC findings and determine whether they generalize to other populations. Finally, this epidemiologic study does not include validation at the cellular or molecular level, but the phenotypic associations we observed are consistent with current knowledge regarding the MC4R region and insulin resistance.^23–26

4.2. Conclusions

We have shown that rs12970134, near MC4R, is strongly associated with HbA1c in this T1D cohort. Our findings support the potential for using genetic markers to identify patients susceptible to poor glycemic control soon after T1D onset to help guide treatment regimens. While the association between rs12970134 and HbA1c was not replicated in the SDRNT1BIO cohort, the noncoding region near MC4R, as well as the leptin-melanocortin pathway in general, warrant further study in other independent cohorts to potentially identify targets for adjunctive therapies that may help to reduce adverse effects of insulin therapy, including hypoglycemia and weight gain in T1D.

Highlights.

• Insulin resistance genes may help identify new therapy targets in T1D

• The candidate polymorphism rs12970134, near MC4R, was strongly associated with HbA1c

• The association weakened with intensive insulin therapy, implying gene-treatment interaction

• Therapies targeting the MC4R region could hold promise for improving glycemic control in T1D