In Utero Exposure to Maternal Hyperglycemia and Offspring Type 2 Diabetes Genetic Risk Score Are Independently Associated With Risk of Impaired Glucose Tolerance in Youth

Abstract

OBJECTIVE

We tested associations of type 2 diabetes genetic risk score (T2D-GRS) and exposure to maternal hyperglycemia with childhood impaired glucose tolerance (IGT) and T2D and glycemic outcomes in youth from the Hyperglycemia and Adverse Pregnancy Outcome Follow-Up Study.

RESEARCH DESIGN AND METHODS

We calculated T2D-GRS using 1,150 known genetic variants associated with T2D in adults. In utero exposures included gestational diabetes mellitus (GDM) and sum-of-glucose z scores during oral glucose tolerance test at ∼28 weeks’ gestation. IGT + T2D and continuous glycemic outcomes were measured when children were 10–14 years old.

RESULTS

In 3,444 children (mean age, 11.4 years), higher maternal sum-of-glucose z scores and child T2D-GRS were both associated with higher glucose levels. In children exposed to GDM and with T2D-GRS >75th percentile, 15.9% had IGT + T2D, compared with 5.6% in nonexposed children.

CONCLUSIONS

High genetic risk for diabetes and in utero exposure to maternal hyperglycemia are additively associated with IGT + T2D and glycemic outcomes in youth.

Graphical Abstract

Introduction

Childhood type 2 diabetes (T2D) is a growing health crisis with faster progression and worse outcomes compared with that of adult counterparts and type 1 diabetes (1–3). Early identification of youth at high risk of T2D is key to prevention. Genetic risk scores (GRSs) improved T2D prediction risk in adults, with now >1,000 loci (4). Most loci discovered in youth overlap adult findings (5–9), plus T2D-GRS have shown better prediction in younger versus older adults (10), arguing for using GRS in youth. In parallel, previous work in the Hyperglycemia and Adverse Pregnancy Outcome Follow-Up Study (HAPO-FUS) showed that maternal hyperglycemia is associated with impaired glucose tolerance (IGT) and higher glucose levels during an oral glucose tolerance test (OGTT) in children 10–14 years later (11,12). In this study, we hypothesized that in utero exposure to maternal hyperglycemia and T2D-GRS are independently associated with IGT + T2D and other glycemic indices at ages 10–14 years.

Research Design and Methods

Participants

This study used data from the parent HAPO study (11) and HAPO-FUS (12). The protocols were approved by each center’s institutional review board and described previously (11,12). The parent HAPO study was conducted between July 2000 and April 2006 across 15 international field centers in which a total of 25,505 pregnant individuals underwent a 75-g OGTT at between 24 and 32 weeks of gestation (11). Women were classified as having gestational diabetes mellitus (GDM) post hoc using International Association of the Diabetes and Pregnancy Study Groups and World Health Organization criteria (12).

The HAPO-FUS enrolled 4,160 children between ages 10 and 14 years (12). Children underwent a 2-h OGTT after an overnight fast. Collection, processing, and quality control measures of samples for glucose and C-peptide are described elsewhere (12). Type 1 diabetes was excluded with detection of serum diabetes autoantibodies.

DNA Processing

Child DNA was extracted using Gentra Puregene Blood kit (Qiagen). We included genotyped data from 3,444 children’s DNA. Processing and imputation of the genotyped data set were previously described (13). Imputation was performed on the TOPMed Imputation Server using Minimac4 (version 1.5.7) and TOPMed reference panel (14,15).

Childhood T2D-GRS

We used the 1,289 single nucleotide polymorphisms (SNPs) identified (P < 5 × 10⁻⁸) from the most recent largest multiancestry genome-wide association study (GWAS) meta-analysis of T2D in adults (4) to construct a T2D-GRS that included 1,150 SNPs available within our data set (Supplementary Table 1). We generated the T2D-GRS by summing the number of risk alleles at each SNP, using a weighted genetic additive model [using β estimates from the T2D-GGI multiancestry GWAS (4)]. We standardized the T2D-GRS into a z score (centered at 0, SD = 1) reflecting genetic burden for T2D. We categorized T2D-GRSs ≥75th percentile as “higher genetic risk.”

In Utero Exposure to Maternal Hyperglycemia

We calculated maternal sum of glucose z scores during the pregnancy OGTT by transforming fasting, 1-h, and 2-h glucose measurements into a z score and summing the individual z scores. We used GDM as a dichotomous variable.

Childhood Glycemic Outcomes: Ages 10–14 Years

Continuous glycemic outcomes included fasting, 30-min, 1-h, and 2-h glucose measurements from OGTTs, plus sum of glucose z scores. Using C-peptide and glucose values from the OGTTs, we calculated a modified Matsuda Index [reflecting insulin sensitivity (16)], insulinogenic index [estimates insulin secretion (17)], and disposition index (derived by multiplying Matsuda and insulinogenic indices) (18). IGT was defined as 2-h glucose concentration of 140–199 mg/dL. T2D was defined as fasting glucose ≥126 mg/dL and/or 2-h glucose ≥200 mg/dL or self-reported diabetes.

Statistical Analyses

We used logistic and linear regression analyses for dichotomous and continuous glycemic outcomes, respectively. All analyses included both maternal sum of glucose z scores (or GDM) and child T2D-GRSs (continuous or dichotomized) to determine if they were independently associated with child glycemic outcomes. Model 1 included the first three genetic principal components, field center, child sex, and age; model 2 included model 1 variables plus maternal BMI, age, and parity. We explored a potential interaction between maternal hyperglycemia (GDM yes/no) and child T2D-GRS (≥75th percentile vs. <75th percentile) on children’s glycemic outcomes, and between the two determinants as continuous variables.

We used two-sided P < 0.05 for evaluating statistical significance for the predictor-outcome associations. We considered interaction P < 0.10 suggestive of potential interactions. We conducted analyses using R, version 4.2.2.

Data and Resource Availability

Data from HAPO-FUS are available in the National Institute of Diabetes and Digestive and Kidney Diseases Central Repository (https://repository.niddk.nih.gov/).

Results

We included 3,444 mother-child pairs in this analysis (Table 1 and Supplementary Fig. 2). Mean (SD) maternal age was 30.1 (5.6) years. GDM was present in 14.9%. The mean (SD) age of children was 11.4 (1.2) years. Seven children (0.2%) had T2D and 214 (6.2%) had IGT.

Table 1.

Maternal and childhood demographic and metabolic characteristics (N = 3,444)

Characteristic	Data
Maternal
Age at OGTT, mean (SD), years	30.1 (5.6)
Gestational age at OGTT, mean (SD), years	27.8 (1.6)
BMI, mean (SD), kg/m2	27.5 (4.9)
Sum of glucose z scores, mean (SD)	0.1 (2.3)
Parity (any previous delivery at ≥20 weeks), n (%)	1,763 (51.2)
GDM (defined by IADPSG criteria), n (%)	514 (14.9)
Child
Age, mean (SD), years	11.4 (1.2)
BMI, mean (SD), kg/m2	19.3 (4.4)
BMI z score, mean (SD)	0.5 (1.3)
Sex (female), n (%)	1,688 (49.0)
Field center, n (%)
Bangkok	240 (7)
Barbados	559 (16.2)
Belfast	387 (11.2)
Bellflower	436 (12.7)
Chicago	242 (7)
Cleveland	184 (5.3)
Hong Kong	690 (20)
Manchester	377 (10.9)
Petah Tiqva	1 (0)
Toronto	328 (9.5)
Child glycemic traits, mean (SD)
Fasting plasma glucose (mg/dL)	90.5 (6.9)
30-min glucose (mg/dL)	141.8 (23.2)
1-h glucose (mg/dL)	124.8 (30.1)
2-h glucose (mg/dL)	108.8 (20.0)
Fasting plasma glucose (mmol/L)	5.0 (0.4)
30-min glucose (mmol/L)	7.9 (1.3)
1-h glucose (mmol/L)	6.9 (1.7)
2-h glucose (mmol/L)	6.0 (1.1)
Matsuda index	32.3 (13.4)
Insulinogenic index	0.1 (0.1)
Sum of glucose z scores	0.3 (2.7)
Disposition index	4.1 (3.8)
IGT, n (%)*	214 (6.2)
T2D, n (%)†	7 (0.20)

IADPSG, International Association of the Diabetes and Pregnancy Study Groups.

*IGT: 2-h PG 140–199 mg/dL.

†T2D: FPG ≥126 mg/dL and/or 2-h PG ≥200 mg/dL or self-reported diabetes on treatment at HAPO-FUS visit.

Higher maternal sum of glucose z scores and child T2D-GRSs were independently associated with greater risk of IGT + T2D at ages 11–14 years and with higher childhood glucose levels at all OGTT time points (Table 2). Odds ratios (ORs) and β values across adjusted models remained consistent for all glucose outcomes. Higher maternal sum of glucose z scores and higher child T2D-GRSs were both associated with lower Matsuda and disposition indices; only higher child T2D-GRS was associated with lower insulinogenic index. In the adjusted models, we did not detect significant interactions between the two continuous determinants for any childhood glycemic outcomes (Supplementary Table 3). Dichotomous determinants were additive, with an overall 15.9% of children with IGT + T2D among children with GDM exposure and with ≥75th percentile of T2D-GRS (P = 0.06 for interaction; Fig. 1).

Table 2.

Outcomes for childhood glycemic traits with continuous determinants^*

		Determinants
Child outcome	Model†	Maternal glycemia,‡ OR or β (95% CI); P value	T2D child GRS (z score); P value
IGT + T2D	Model 1	1.11§ (1.05–1.17); 4.27 × 10−4	1.36§ (1.13–1.63); 9.56 × 10−4
	Model 2	1.12§ (1.05–1.19); 4.04 × 10−4	1.35§ (1.13–1.63); 1.07 × 10−3
Fasting glucose (mg/dL)	Model 1	0.25 (0.16–0.34); 8.08 × 10−8	0.27 (0.01–0.54); 4.43 × 10−2
	Model 2	0.26 (0.16–0.36); 1.95 × 10−7	0.27 (0.00–0.53); 4.68 × 10−2
30-min glucose (mg/dL)	Model 1	1.08 (0.75–1.41); 1.30 × 10−10	2.00 (1.04–2.96); 4.73 × 10−5
	Model 2	1.12 (0.77–1.47); 5.92 × 10−10	2.00 (1.04–2.97); 4.65 × 10−5
1-h glucose (mg/dL)	Model 1	1.25 (0.80–1.69); 4.67 × 10−8	3.50 (2.20–4.79); 1.35 × 10−7
	Model 2	1.28 (0.81–1.76); 1.50 × 10−7	3.51 (2.21–4.80); 1.31 × 10−7
2-h glucose (mg/dL)	Model 1	0.70 (0.41–0.99); 2.86 × 10−6	2.55 (1.70–3.40); 4.57 × 10−9
	Model 2	0.69 (0.38–1.01); 1.41 × 10−5	2.54 (1.69–3.39); 5.07 × 10−9
Sum of glucose z scores	Model 1	0.15 (0.11–0.19); 1.01 × 10−14	0.34 (0.22–0.45); 5.13 × 10−9
	Model 2	0.16 (0.12–0.20); 7.27 × 10−14	0.34 (0.22–0.45); 6.07 × 10−9
Insulinogenic index	Model 1	0.00 (0.00–0.00); 5.33 × 10−1	−0.01 (−0.01 to 0.00); 2.82 × 10−2
	Model 2	0.00 (0.00–0.00); 7.52 × 10−2	−0.01 (−0.01 to 0.00); 2.66 × 10−2
Matsuda index	Model 1	−0.68 (−0.87 to 0.50); 1.38 × 10−12	−1.18 (−1.73 to 0.63); 2.80 × 10−5
	Model 2	−0.52 (−0.72 to 0.32); 3.50 × 10−7	−1.19 (−1.74 to 0.64); 2.11 × 10−5
Disposition index (log)	Model 1	−0.02 (−0.03 to 0.01); 1.18 × 10−7	−0.07 (−0.09 to 0.04); 1.12 × 10−7
	Model 2	−0.02 (−0.03 to 0.01); 2.32 × 10−7	−0.07 (−0.09 to 0.04); 9.60 × 10−8

*Interaction P values were >0.05 for all models.

†Maternal glycemia was defined by maternal sum of glucose z scores.

‡Model 1: three genetic principal components, field center, child age, and sex. Model 2: Model 1 plus maternal BMI, maternal age, and parity.

§Odds ratio. Values not marked with § are β values.

Figure 1.

Frequency of child IGT and T2D by maternal GDM status and childhood genetic risk. Childhood genetic risk was broken down by top quartile vs. lower three quartiles. †P = 0.06 for interaction. GDM absent + lower quartiles T2D-GRS, n = 120 of 2,130. GDM absent + top quartile T2D-GRS, n = 47 of 762. GDM present + lower quartiles T2D-GRS, n = 29 of 347. GDM present + top quartile T2D-GRS, n = 25 of 157.

Conclusions

This study demonstrates the independent and additive associations of in utero exposure to maternal hyperglycemia and high individual T2D genetic burden on childhood risk of IGT + T2D, as well as measures of glycemia and insulin metabolism at ages 10–14 years.

We extended prior literature by using a T2D-GRS derived from the most recent GWAS (4) and evaluating its association with IGT + T2D in addition to glycemic and insulin metabolism measures in a large and diverse pediatric cohort. Previous reports investigating T2D-GRSs in pediatric cohorts have been limited by small sample size, selection bias, and/or relatively homogeneous race/ethnicity (7,8,19). For example, a higher T2D-GRS (composed of 62 SNPs) was associated with lower insulinogenic and disposition indices, along with risk of impaired fasting glucose (OR = 1.08 per risk allele; P = 0.008) but not with IGT, acknowledging limited power due to low frequency (8). In 356 adolescents from the Exploring Perinatal Outcomes Among Children cohort, a T2D-GRS based on 272 known T2D-SNPs showed that a higher T2D-GRS was associated with higher 2-h glucose level and with lower insulinogenic and disposition indices (7). The authors of that study did not report analyses investigating IGT or T2D (7). Notably, they also found that the association between higher T2D-GRS and 2-h glucose measurement was stronger in individuals exposed to GDM in utero (7), in line with our observations in this study (Fig. 1). The association between maternal hyperglycemia and offspring risk of IGT above the genetic risk could be due to shared pre- and/or postnatal environments between mother and child, and/or to epigenetic changes, as suggested by DNA methylation findings at birth (20), that may reflect fetal programming of β-cell dysfunction and insulin resistance.

We observed positive associations of both determinants with all child OGTT glucose measures and IGT + T2D, as well as negative associations with insulin sensitivity. The addition of maternal covariables did not appreciably affect GRS estimates but did modestly affect maternal glycemia estimates on child insulin sensitivity. Remarkably, maternal glycemia was not associated with child’s insulinogenic index. Similarly, child T2D-GRS was only weakly associated with child’s insulinogenic index (P = 0.03) compared with other traits (Table 2). This is in contrast with previous reports that found clear associations between prior T2D-GRS and insulinogenic index, and not with HOMA for insulin resistance (7,8). This discrepancy may be due to the younger mean age of our participants. It is also possible that the now-expanded knowledge of the genetics of T2D is not driven as much by variants that affect β-cell function, in contrast to earlier genetics discovery (10).

This study had several strengths. We used a large and diverse data set, representing multiple countries. We also used the most recent GWAS to generate our T2D-GRS and tested its association with clinically defined IGT + T2D and continuous glucose and insulin metabolic measures in childhood. Among our limitations, despite having a larger sample size than prior reports, our power was still limited due to the number of IGT + T2D cases, especially to detect interactions. Additionally, we based our T2D-GRS on loci identified in adult populations; additional GWAS are needed in large and diverse pediatric cohorts. The β estimates used in T2D-GRS were derived from fixed-effect meta-analysis of ancestry-specific GWAS β values (4); we applied these β values uniformly across our multiancestry population. This may have resulted in misclassification of the genetic risk for T2D in our HAPO participants; future GWAS should use newer methods to account for genetic heterogeneity (21).

Because genotyping is becoming more accessible, we have the unique opportunity to identify those at risk for youth-onset dysglycemia earlier, before clinical risk factors arise. Given the additive and independent genetic risks, implementation of genetic risk screening, beginning as early as the perinatal period, could be undertaken in at-risk children with known in utero exposure to maternal hyperglycemia. Perinatal identification could help target lifestyle interventions in early and mid-childhood and more strategically allocate resources to prevent IGT. Future studies should investigate if this type of early-life identification strategy is applicable in real-world settings to implement precision prevention interventions to reduce T2D in youth.

This article contains supplementary material online at https://doi.org/10.2337/figshare.28851254.

Funding Statement

This study was funded by the National Institutes of Health (grants DK095963, DK117491, HD34242, HD34243, HG-004415 and R03CA211318). A.C.D. was supported by the Ruth L. Kirschstein National Research Service Award T32 DK007169 from the National Institute of Diabetes and Digestive and Kidney Diseases.