Excess of rare noncoding variants in several type 2 diabetes candidate genes among Asian Indian families

Abstract

Background: Type 2 diabetes (T2D) etiology is highly complex due to its multiple roots of origin. Polygenic risk scores (PRS) based on genome-wide association studies (GWAS) can partially explain T2D risk. Asian Indian people have up to six times higher risk of developing T2D than European people, and underlying causes of this disparity are unknown.

Methods: We have performed targeted sequencing of ten T2D GWAS/candidate regions using endogamous Punjabi Sikh families and replication studies using unrelated Sikh people and families from three other Indian endogamous ethnic groups (EEGs).

Results: We detect rare and ultra-rare variants (RVs) in KCNJ11-ABCC8 and HNF4A (MODY genes) cosegregated with late-onset T2D. We also identify RV enrichment in two new genes, SLC38A11 and ANPEP, associated with T2D. Gene-burden analysis reveals the highest RV burden contributed by HNF4A (p = 0.0003), followed by KCNJ11/ABCC8 (p = 0.0061) and SLC38A11 (p = 0.03). Some RVs detected in Sikh people are also found in Agarwals from Jaipur, both from Northern India, but were monomorphic in other two EEGs from South Indian people. Despite carrying a high burden of T2D and RVs, most families have a significantly lower burden of PRS. Functional studies show that an intronic regulatory variant (RV) in ABCC8 affects the binding of Pax4 and NF-kB transcription factors, influencing downstream gene regulation.

Conclusions: The high burden of T2D in these families may stem from the enrichment of noncoding RVs in a small number of major known genes (including MODY genes) with oligogenic inheritance alongside RVs from genes associated with polygenic susceptibility. These findings highlight the need to conduct deeper evaluations of families from non-European ancestries to identify potential novel therapeutics and implement preventative strategies.

Fig. 1. The burden of rare variants of multiple MODY genes (ABCC8, KCNJ11, and HNF4A) in Ped 36 of Sikh families.

Highlighted fonts depict polygenic risk scores using Asian Indian (PRS_AI) and European (PRS_EU) models. Details of the rare variants from ABCC8, KCNJ11, HNF4A, SLC38A11, and ANPEP detected in this family are listed on the right side. Detection of heterozygous variant alleles in each family member is indicated by the numerical code for the listed SNPs under the individual’s ID.

Fig. 2. Segregation and in silico functional characterization of a missense RV in a MODY gene (KCNJ11) in four Sikh families.

A Pedigrees of Sikh families show the representation of a new rare damaging variant rs41282930 (Ser385Cys) in KCNJ11 gene segregating with T2D; mutant carriers of this variant are shown in red squares; B Sequence alignment reveals absolute conservation of rs41282930 at position Ser385Cys of the KCNJ11 gene across species; C Allele frequency of Ser385Cys among different populations; D The wildtype residue (blue) forms hydrogen bonds with Serine at position 383. However, the mutant residue (red) Cysteine forms hydrogen bonds not only with Serine 383 but also with Isoleucine at position 384.

Fig. 3. Bar graph representing the number of rare variant carriers in each family.

The correlation coefficient (r) shows the correlation of rare variants with T2D. *(r) and p-value could not be determined due to the absence of controls in these pedigrees. ^Rare variant burden calculated excluding variants rs5216 and rs376706487 in PED36 and PED99.

Fig. 4. A Sikh family (PED332) showing rare variants burden in ANPEP, DDX50, and HK1 genes.

Despite having a high load of T2D, no rare variant was detected from MODY genes in this family. Highlighted fonts depict polygenic risk scores using Asian Indian (PRS_AI) and European (PRS_EU) models.

Fig. 5. Detection and characterization of two missense RVs in SLC38A11 gene segregating with T2D.

A Pedigree of a Sikh family (PED180) shows the representation of new rare damaging variants rs139693718 (Phe47Ser) and rs140708593 (Arg116Gln) in SLC38A11 gene; B Sequence alignment reveals absolute conservation of rs139693718 at position Phe47Ser of the SLC38A11 gene across species; allele frequency of Phe47Ser among different populations and the wildtype residue (blue) forms hydrogen bonds with Asparagine at position 44. However, the mutant residue (red) Serine forms hydrogen bonds not only with Asparagine 44 but also with Lysine at position 328. C Sequence alignment reveals absolute conservation of rs140708593 at position Arg116Gln of the SLC38A11 gene across species; allele frequency of Arg116Gln among different populations and the wildtype residue (blue) forms hydrogen bonds with Alanine at position 119. However, the mutant residue (red) Glutamine forms hydrogen bonds with Alanine119 and an additional hydrogen bond with Phenylalanine at position 261. Only Ped#180 was carrying these variants.

Fig. 6. The rare intronic variant rs117727754 detected in the ABCC8 gene was present in the transcription factor binding site.

A Rare variant represses the activity of transcription factors Pax4, Sp1, and NF-kβ with scores of −10.52, −4.90, and −2.91, respectively, and enhances the activity of Ap1 (Score = 4.46). B The ABCC8 variant also forms a chromatin loop with NCR3LG1 gene. C The Pax4-ABCC8 and NF-kβ-ABCC8 pull-down assay using biotinylated probes: DNA-transcription factor (RNA) binding in Western blots showing the binding of NF-kβ - ABCC8 and Pax4-ABCC8 among two mutant (CG) human carriers vs. wildtype (CC) carriers. D Bar graphs representing the decreased intensity (2.6-fold) for Mut1 and (4.2-fold) Mut2 for Nf-kβ, and (2.3-fold) for Mut1 and (2.1-fold) for Mut2 compared to the WT band for the Pax4 transcription factor. The bands were quantified and analyzed using ImageJ.

Fig. 7. Bar charts representing the distribution of polygenic risk scores_.

PRS_AI (A) and PRS_EU (B) showing significant differences between family cohort (Discovery) (N = 288), families with high rare variant burden from the selected GWAS candidate genes (N = 82) vs. the unrelated individuals of AIDHS/SDS (N = 4602) among T2D cases. Errors bars represent the standard error and individual data points are added.