CDKN2A/B T2D Genome-Wide Association Study RiskSNPs Impact Locus Gene Expression and Proliferationin Human IsletsYahui Kong,1 Rohit B. Sharma,1 Socheata Ly,1 Rachel E. Stamateris,1 William M. Jesdale,2 andLaura C. Alonso1

Diabetes 2018;67:872–884 | https://doi.org/10.2337/db17-1055

Genome-wide association studies link the CDKN2A/B lo-cus with type 2 diabetes (T2D) risk, but mechanisms increas-ing risk remain unknown. The CDKN2A/B locus encodescell cycle inhibitors p14, p15, and p16; MTAP; and ANRIL,a long noncoding RNA. The goal of this study was to de-termine whether CDKN2A/B T2D risk SNPs impact locusgene expression, insulin secretion, or b-cell proliferationin human islets. Islets from donors without diabetes (n =95) were tested for SNP genotype (rs10811661, rs2383208,rs564398, and rs10757283), gene expression (p14, p15, p16,MTAP, ANRIL, PCNA, KI67, and CCND2), insulin secretion(n = 61), and b-cell proliferation (n = 47). Intriguingly, locusgenes were coregulated in islets in two physically over-lapping cassettes: p14-p16-ANRIL, which increased withage, and MTAP-p15, which did not. Risk alleles atrs10811661 and rs2383208 were differentially associatedwith expression of ANRIL, but not p14, p15, p16, or MTAP,in age-dependent fashion, such that younger homozygousrisk donors had higher ANRIL expression, equivalent toolder donor levels. We identified several risk SNP combi-nations that may impact locus gene expression, suggestingpossible mechanisms by which SNPs impact locus biology.Risk allele carriers at ANRIL coding SNP rs564398 had re-duced b-cell proliferation index. In conclusion,CDKN2A/Blocus SNPs may impact T2D risk by modulating islet geneexpression and b-cell proliferation.

Type 2 diabetes (T2D) risk has a strong genetic component.Significant research investment has identified .100 geno-mic regions that influence T2D risk in human populations(1–3). Most T2D risk single nucleotide polymorphisms

(SNPs) are noncoding, and the mechanism by which theyimpact local genome biology remains unclear for many loci(3). Risk alleles may act in multiple ways, interacting withother genes and polymorphism effects in a tissue-specificmanner. Genome-wide expression quantitative trait locistudies seek to identify how polymorphisms impact biologyat any given locus (1,4–7); however, depth of information atindividual loci is limited in genome-wide studies. Most T2DSNPs influence risk by impacting islet biology (8), but thecost and inaccessibility of human islets, and poor utility ofnonhuman models to study the human genome, haveslowed progress in clarifying mechanisms.

SNPs at the CDKN2A/B genomic locus impact risk ofT2D and related diseases, such as gestational diabetes mel-litus, cystic fibrosis–related diabetes, and posttransplant di-abetes, across ethnicities and cultures, suggesting a centraldiabetogenic mechanism (9). Multiple SNPs in different link-age blocks at the CDKN2A/B locus confer T2D risk (9); mech-anisms impacting risk remain unknown. The CDKN2A/Blocus encodes four genes (Fig. 1):MTAP, CDKN2A, CDKN2B,and a long noncoding RNA (lncRNA) named ANRIL. CDKN2Aand CDKN2B are well studied, encoding cell cycle inhib-itors (p14 and p16 are splice variants of CDKN2A, andp15 is encoded at CDKN2B) that impact aging, senescence,and tumorigenesis via regulation of retinoblastoma andp53 (10,11). Three T2D SNPs at this locus, rs10811661,rs2383208, and rs10757283, are noncoding, located down-stream of known genes; rs2383208 and rs10811661 are inone linkage block, and rs10757283 is in a separate linkageblock immediately downstream. A fourth SNP, rs564398,;100,000 base pairs upstream of these, falls within exon

1Diabetes Center of Excellence, Department of Medicine, University of Massa-chusetts Medical School, Worcester, MA2Department of Quantitative Health Sciences, University of Massachusetts Med-ical School, Worcester, MA

Corresponding author: Laura C. Alonso, laura.alonso@umassmed.edu.

Received 1 September 2017 and accepted 29 January 2018.

This article contains Supplementary Data online at http://diabetes.diabetesjournals.org/lookup/suppl/doi:10.2337/db17-1055/-/DC1.

© 2018 by the American Diabetes Association. Readers may use this article aslong as the work is properly cited, the use is educational and not for profit, and thework is not altered. More information is available at http://www.diabetesjournals.org/content/license.

872 Diabetes Volume 67, May 2018

ISLETSTUDIES

2 of ANRIL. These SNPs were identified in large populationstudies seeking to identify genomic regions associated withT2D risk (12–14) (more details have previously been pub-lished [9]). The three downstream SNPs are mostly associ-ated with T2D risk and not other diseases; the rs564398SNP is also associated with coronary heart disease andglaucoma (15). The absolute magnitude of T2D risk is lowwith all identified SNPs (at this and other loci); for ex-ample, reported odds ratio for the linkage region containingrs10811661 and rs2383208 ranges from 1.18 to 1.46 (4,12,14,16,17). Weaker odds ratios were seen for rs564398(1.12–1.26) (example previously published in 13); intrigu-ingly, multiple studies show that rs564398 is associatedwith T2D risk in Caucasian but not Asian populations(18). A risk-risk combination of rs10811661/rs2383208and rs10757283 conferred an odds ratio of 1.24, with stron-ger association than individual risk alleles (13). Althougheach T2D SNP is in linkage disequilibrium with multiple

other SNPs, fine mapping has not identified linked SNPswith greater disease association than these genome-wideassociation study–identified SNPs (1,4). The causal SNP inany of these linkage blocks is not yet known.

In human populations, the rs10811661 risk allele is as-sociated with reduced insulin secretory capacity after oral orintravenous glucose challenge (16,19–22). Insulin secretorycapacity is a composite end point influenced by b-cell mass,insulin production, glucose sensing, and stimulus-secretioncoupling (23), factors that cannot currently be effectivelyseparated in living human subjects. Intriguingly, given theaging and senescence roles played by CDKN2A/B genes, theimpact of rs10811661 on T2D risk was influenced by subjectage (18). SNPs at this locus also influence insulin sensitivityand biology of other metabolic tissues, demonstrating thecomplexity of even a single genomic locus on T2D biology (9).

Since human studies suggest that CDKN2A/B locusSNPs impact T2D risk, at least in part, by reducing insulin

Figure 1—CDKN2A/B locus genes were expressed coordinately in human islets. A: Diagram of the CDKN2A/B locus at 9p21, adapted from theUniversity of California, Santa Cruz, Genome Browser GRCh38/hg38 assembly. Vertical arrows show the locations of T2D SNPs tested in thisstudy, by linkage block: green (rs564398 [leftmost]), blue (rs2383208 and rs10811661 [middle two]), and red (rs10757283 [rightmost]). B–D:Abundances of p14, p16, and ANRIL were highly correlated in human islet samples. E–G: p15 abundance did not correlate with p16 and onlymarginally correlated with p14 and ANRIL. MTAP expression was marginally correlated with p14, p16, and ANRIL (H and not shown) but highlycorrelated with p15 expression (I). mRNA abundance is expressed as DCt, normalized to ACTB. Dashed line, P values and R2 values werecalculated by linear regression; n = 95 for all panels. Red lines highlight correlations with higher R2 values.

diabetes.diabetesjournals.org Kong and Associates 873

secretory capacity, we hypothesized that locus SNPs influ-ence pancreatic islet biology. Here, we present a detailedanalysis of CDKN2A/B biology in nondiabetic humanislets. We identified two overlapping coregulated genesets: p14-p16-ANRIL and p15-MTAP. p14-p16-ANRIL ex-pression, but not p15-MTAP expression, increased withdonor age. Of the four T2D risk SNPs tested, rs2383208and rs10811661 risk alleles were associated with inappro-priate high expression of the ANRIL lncRNA in samplesfrom younger donors. No other SNP-gene interaction wasidentified, but our data suggest certain SNP pairs that mayimpact locus gene expression in combinatorial fashion. Fi-nally, risk alleles at rs564398 were associated with reducedb-cell proliferation index, suggesting a functional implicationfor this SNP, and perhaps the ANRIL lncRNA, in accrual ormaintenance of human b-cell mass.

RESEARCH DESIGN AND METHODS

Human IsletsHuman islets were obtained from the Integrated IsletDistribution Program (IIDP) at the City of Hope, supportedby the National Institute of Diabetes and Digestive andKidney Diseases (NIDDK), National Institutes of Health, orfrom a collaborative group headed at Vanderbilt University(24). Human islet studies were determined by the Univer-sity of Massachusetts Institutional Review Board to notqualify for institutional review board review or exemptionbecause they do not involve the use of human subjects.De-identified islet samples from 95 subjects without diabe-tes were live shipped in Prodo islet transport media. Donors(Supplementary Table 1) included 42 females, 48 males, and5 without sex reported, with mean6 SD age 406 16 yearsand ethnicity as follows: 1 Asian, 8 black or AfricanAmerican, 14 Hispanic/Latino, 66 white, and 6 unknown.Upon receipt, islets were plated in islet culture medium(RPMI, 10% FBS, 5 mmol/L glucose, and penicillin/streptomycin) and incubated at 37°, 5% CO2, overnightfor recovery from isolation and shipment. After recovery,800 islet equivalents (IEQs) were handpicked, washed inPBS containing 100 nmol/L Na3VO4, and flash frozen at280°C in 200-IEQ aliquots for future DNA and RNA anal-ysis. Additional islets from a subset of donors were culturedas described below for glucose-stimulated proliferation.

GenotypingDNA and RNA were extracted from flash-frozen 200-IEQaliquots using the Norgen RNA/DNA/Protein PurificationKit (Norgen Biotek Corp. Thorold, Ontario, Canada) ac-cording to the manufacturer’s protocol. Genotyping for fourCDKN2A/B SNPs (rs564398 [C/T], rs10811661 [C/T],rs2383208 [G/A], and rs10757283 [C/T]) was performed induplicate with commercial (C_2618017_10, C_31288917_10,C_15789011_10, and C_31288916_10) TaqMan SNP geno-typing assays (Thermo Fisher Scientific, Waltham, MA) onthe C1000 Touch Thermal Cycler (Bio-Rad) or the RealPlexCycler (Eppendorf) real-time PCR platforms, using 20 ng DNAin a 10-mL reaction volume under conditions recommended

by the manufacturer. SNP determination was confirmed byboth allelic discrimination and by manual cycle threshold(Ct) value assessment for all samples and all SNPs. Minorallele frequencies (MAFs) in our cohort were in agreementwith expected MAF based on the 1000 Genomes Project(25) (Supplementary Table 2), and the observed frequencyof SNP combinations predicted linkage disequilibrium sim-ilar to 1000 Genomes–reported values for these SNPs (Sup-plementary Table 3) (26).

Gene Expression AssaysTotal RNA was reverse transcribed using a SuperScript IV VILOMasterMix kit (Thermo Fisher Scientific). The expression levelsof target genes in human islets were quantitatively assessedin duplicate using Taqman-validated human gene expressionassays (Thermo Fisher Scientific). Primers/probes used wereas follows: ANRIL, Hs04259476_m1; p15, Hs00793225_m1; p14, Hs99999189_m1; p16, Hs02902543_mH;MTAP, Hs00559618_m1; KI67, Hs01032443_m1; PCNA,Hs00696862_m1; CCND2, Hs00153380_m1; ACTB,Hs01060665_g1; and GAPDH, Hs02758991_g1. ACTBand GAPDH were used as endogenous reference to normal-ize gene expression. Reproducibility of duplicate measure-ments was high, as assessed by the R2 of the correlationbetween duplicates and by the absolute value of the relativepercentage difference between the duplicates (Supplemen-tary Fig. 1). Transcript expression levels were presented aslog2-transformed expression (DCt).

Human Islet Culture ExperimentsHuman islets cultured overnight in islet culture mediumwere dispersed to single cells using single-use–apportioned0.05% trypsin and plated on uncoated glass coverslips(Fisherbrand) as previously described (27–29). Dispersedcells were cultured in islet culture medium containing either5 mmol/L or 15 mmol/L glucose for 96 h, with 20 mg/mLBrdU included for the entire time. After culture, the isletcells were fixed for 10 min in 4% paraformaldehyde (Sigma-Aldrich). Immunofluorescence staining was performed afterunmasking in 1 N HCl for 25 min at 37°C for insulin(ab7842, Abcam, or A056401-2; Dako), BrdU (ab6326;Abcam), and DAPI as previously described (27–29). b-Cellproliferation, defined as the percent of insulin-stainingcells that were also BrdU labeled, was quantified onblinded images (30). Data were expressed as the prolifer-ation index, calculated as the ratio of %BrdU+ b-cells in15 mmol/L glucose divided by the %BrdU+ b-cells in5 mmol/L glucose.

StatisticsUnivariate analyses were performed using GraphPad Prism,and data are expressed as mean 6 SD. P values were de-termined by two-tailed Student t test for comparison oftwo conditions, with F test to compare variances, one-wayANOVA with Tukey posttest for correction for multiplecomparisons for comparison of more than two condi-tions, or by linear regression for assessment of the re-lationship between two continuous variables. Multivariable

874 CDKN2A/B T2D SNPs Impact Human Islet Biology Diabetes Volume 67, May 2018

linear models were performed to examine gene expressions(p14, p15, p16, ANRIL, and MTAP) simultaneously adjustedfor donor sex, race/ethnicity, age (continuous), and BMI(continuous); additional models further adjusted for expres-sion of the other gene products. Missing values were mod-eled with a missing indicator, replacing unknown valueswith sample means for linear variables. Infrequent or un-known race/ethnicity was grouped in a residual category.RNA expression associated with SNPs was estimated inlinear multivariable models adjusting for demographics(as described above) in two fashions: first, by settingthe population with no risk alleles as the common ref-erence group, and second, by estimating the linear effecton a per-allele basis (treating the number of risk alleles asadditive). Insulin secretion index was estimated as a func-tion of demographic variables as described above and byincluding each SNP as a predictor of insulin secretion index.Interpretation of these models can be found in the Supple-mentary Data. P , 0.05 was considered significant, al-though this may be too generous for the exploratoryanalyses with multiple comparisons. For the SNP combina-tion hypothesis-generating analyses, the false discovery rate(FDR), calculated by the original method of Benjaminiand Hochberg, was set at 10%, based on our estimationthat a hypothesis with 90% likelihood of being correctwarranted experimental follow-up.

RESULTS

CDKN2A/B Locus Gene Expression Is CoordinatelyRegulated in Human IsletsTo understand the context of how T2D risk SNPs mightimpact biology at this locus in human islets, we firstquantified expression of all locus genes (Fig. 1A). ValidatedTaqman probes were chosen that could independentlyquantify transcripts including MTAP, p14 (CDKN2A:ARF),p15 (CDKN2B:INK4B), p16 (CDKN2A:INK4A), and theANRIL (CDKN2B-AS1) lncRNA. p14 and p16 are splicevariants of CDKN2A, sharing exons 2 and 3 but withdifferent first exons; exons 2 and 3 are in different read-ing frames, and thus p14 and p16 encode entirely differ-ent proteins with different functions (31). The ANRILprobe spans exons 5 and 6, thus detecting all known iso-forms. In this cohort of islet samples from 95 uniquedonors without diabetes (Supplementary Table 1), RNAabundances of p14, p16, and ANRIL were highly corre-lated with each other (Fig. 1, normalized to ACTB, andSupplementary Fig. 2, normalized to GAPDH). In contrast,abundance of p15, despite being physically located withinthe first intron of ANRIL, was poorly (ACTB normaliza-tion) or not (GAPDH normalization) correlated with p14,p16, or ANRIL. On the other hand, p15, but not p14, p16,or ANRIL, was highly correlated with MTAP expression.When the data were examined in multivariable linearmodels, with integration of donor characteristics such asage into the model, again p14-ANRIL and p14-p16 werehighly correlated, as were p15-MTAP (Supplementary Table4). These results suggest two independent but overlapping

coregulatory cassettes at the CDKN2A/B locus in humanislets, with p14-p16-ANRIL in one and p15-MTAP in theother.

Age-Dependent Gene Expression Increase of p14, p16,and ANRIL but Not p15 or MTAPIn many tissues, including islets, some CDKN2A/B locusgenes increase with advancing age (9,10,32). In this cohortof human islets, expression of p14, p16, and ANRIL showeda modest positive correlation with donor age, whereas p15andMTAP did not (Fig. 2A–E). Donor BMI could potentiallyconfound the impact of age on gene expression; however,BMI was similar across donor ages (Fig. 2F). Furthermore,we observed no correlation between donor BMI, sex, orethnicity and expression of any CDKN2A/B locus gene inunivariate analysis (Supplementary Figs. 3–5). Multivariablelinear models integrating age, sex, race, and BMI confirmeda positive correlation of p14, p16, and ANRIL with donorage and confirmed a lack of impact of sex, race, or BMI onlocus gene expression (Supplementary Table 4). Scatterplotsof gene expression versus age showed that some genes wereexpressed in very low abundance in islets from juvenile(age ,10 years) donors, with points falling well below thelinear regression curve (Fig. 2A, C, and D). Focused analysisof juvenile (,10 years) versus adolescent/adult (.10 years)islets (Fig. 2G–K) revealed that expression of p14, p16 , andANRIL was markedly lower in juvenile islets, but expressionof MTAP and p15 was not, again suggesting altered regula-tory characteristics of these two genes relative to otherlocus genes. Interestingly, an F test showed that the var-iances were reduced in juvenile islets (see SD bars in Fig.2G–K) for p14 (P , 0.0001), p16 (P , 0.001), and ANRIL(P , 0.0001) but not for p15 or MTAP (P = NS for both),despite the much smaller sample size, again suggestingfundamentally different biology of the juvenile samples.In sum, older age increased expression of p14, p16, andANRIL, but not p15 or MTAP, and the youth-associatedsuppression was exaggerated in islets from very youngdonors.

T2D Risk SNPs at rs10811661 and rs2383208 IncreasedANRIL Expression in an Age-Defined Subset of IsletSamplesWe next tested whether T2D-related SNPs at CDKN2A/Bimpact locus gene expression. Validated Taqman geno-typing procedures ascertained and confirmed the geno-type of all n = 95 preps for four T2D SNPs: rs564398(hg38 chr9:22029548), rs2383208 (hg38 chr9:22132077),rs10811661 (hg38 chr9:22134095), and rs10757283 (hg38chr9:22134173). Measured MAFs (Supplementary Table 2)were similar to reported MAFs for ethnicity-matched pop-ulations, supporting genotyping accuracy. SNPs rs2383208and rs10811661 were tightly linked, with only 2 of 95samples differing in our cohort, consistent with the link-age disequilibrium reported in LDpair and HaploReg (26,33,34) (Supplementary Tables 3 and 5). In raw analysisacross the entire cohort, no SNP genotype correlated withabundance of any locus transcript by univariate (Fig. 3)

diabetes.diabetesjournals.org Kong and Associates 875

or multivariable (Supplementary Table 6) analysis. Sincedonor age impacted expression of p14, p16, and ANRIL,we assessed whether age interfered with the assessment ofSNP effect on gene expression. Mean age was not sig-nificantly different between genotypes for any SNP (datanot shown). However, expression of transcript abundanceas a function of age revealed that for ANRIL, but not for

p14 or p16, the age-dependent increase was genotype de-pendent, evident only in protective allele–carrying samplesat rs2383208 (Fig. 4) and rs10811661 (SupplementaryFig. 6). Homozygous risk samples had high levels of ANRILacross all ages .10 years (Fig. 4B [flat slope of AA regres-sion line even despite the influence of the juvenile sam-ples]). Age-genotype interaction was not observed for any

Figure 2—Abundance of p14, p16, and ANRIL, but not p15 orMTAP, was correlated with donor age and strongly reduced in juvenile islets. A–E:Consistent with prior observations, p16 mRNA abundance was positively correlated with donor age (in years). p14 and ANRIL were alsocorrelated with age, but p15 and MTAP were not. Age accounted for only a small proportion of the variance in gene expression, even forp16. F: BMI partitioned equally across donor age in this cohort (dotted lines demarcate BMI 18–25 kg/m2 [normal weight], 25–30 kg/m2

[overweight], and .30 kg/m2 [obese]). G–K: Islets from juvenile donors (age ,10 years) contained markedly less p14, p16, and ANRIL, butnot p15 or MTAP, than older islets. mRNA abundance is expressed as DCt, normalized to ACTB. Data are mean 6 SD. P values and R2 valueswere calculated by linear regression (A–F) or by Student t test (G–K). For all panels, n = 92; missing values are the result of lack of donorinformation for age and BMI (three samples).

876 CDKN2A/B T2D SNPs Impact Human Islet Biology Diabetes Volume 67, May 2018

locus gene for rs10757283 or rs564398 (not shown). Whenthe samples were reanalyzed using a different methodology,binning by quartiles, it was again evident that samples witha protective allele at rs2383208 or rs10811661 showed an

age-dependent increase in ANRIL, but homozygous risksamples did not. In contrast, for p16 the slope of the age-dependent–gene expression regression lines (Fig. 4A) andbinning analysis (Fig. 4C) were similar across genotypes. For

Figure 3—In crude analysis, individual SNP identity did not impact expression of locus genes in human islets. Risk allele for each SNP, therightmost genotype in each case, is in red. All comparisons were nonsignificant by ANOVA with correction for multiple comparisons. ANRILshowed a trend toward higher abundance in islets with homozygous risk for rs10811661 (P = 0.08) and rs2383208 (P = 0.07) compared withprotective allele–carrying samples. mRNA abundance is expressed as DCt, normalized to ACTB. Data are mean 6 SD. n = 95 for all subpanels.rs108, rs10811661; rs238, rs2383208; rs107, rs10757283; rs564, rs564398.

diabetes.diabetesjournals.org Kong and Associates 877

testing of whether ANRIL abundance was inappropriatelyincreased by homozygous risk at rs2383208 or rs10811661in samples from young donors, samples of those betweenthe ages of 10 years (to exclude juveniles, which were allsuppressed independent of genotype) and 50 years (definedby the intersection of the linear regression curves in Fig.4B), was stratified by genotype (Fig. 4E and F). ANRIL, butnot p16, abundance was significantly increased in youngerhomozygous risk samples compared with protective-allelecarriers. Taken together, T2D homozygous risk geno-type at rs2383208 or rs10811661 prematurely increasedANRIL expression in islets of younger donors to older-donorlevels.

SNPs May Interact With Each Other to CombinatoriallyInfluence Gene ExpressionThis cohort was not powered for us to perform subgroupanalyses to definitively detect gene expression impact ofSNP combinations. For example, a sample size analysis

using our ANRIL expression mean and SD for rs2383208genotypes reveals that we would require n = 54 per sub-group to achieve a power of 80% and type 1 error of 0.05.Given the diminishing number of samples as we partitionby SNP combination, we do not approach this sample sizefor subanalyses. Instead, we analyzed our data set using anFDR approach to prioritize hypotheses to test in futurestudies such as ex vivo promoter-enhancer experiments.We estimated that a risk tolerance of 90% likelihoodthat a hypothesis was correct would support future ex-perimental investment. We then stratified our gene ex-pression data by all SNP combinations and analyzedeach comparison for likelihood of difference, defined byan FDR of 10% (Fig. 5 and Supplementary Fig. 7). By thesecriteria, we determined that genotype at rs564398 andrs10757283 may influence the impact of rs2383208 and/orrs10811661 on gene expression. For rs564398, in homozy-gous protective rs564398, but not risk allele–containingsamples, protective alleles at rs2383208 and rs10811661

Figure 4—Age interacted with genotype at rs2383208 to determine ANRIL abundance; young donors with protective alleles had lower ANRILexpression. A and B: Expression of p16 (A) or ANRIL (B) abundance as a function of donor age, stratified by genotype, showed that unlike forp16, age dependence of ANRIL was driven by samples of rs2383208 GG+GA genotype and was absent in samples with AA (homozygous risk)genotype. C and D: Binning analysis of the cohort (nonjuvenile samples separated by quartiles) illustrated the age-dependent ANRIL increase inGG+GA samples but not in homozygous risk AA samples. Juveniles ,10 years of age showed markedly lower abundance, independent ofgenotype. E and F: In younger donors (ages 10–50 years [lower threshold defined by juvenile cutoff and upper threshold defined by theintersection of the regression curves in B, which is 50.8]) homozygous risk increased ANRIL abundance. mRNA abundance is expressed asDCt, normalized to ACTB. Statistics by linear regression (A and B) and ANOVA (D–F) with overall ANOVA significance in upper-left corner andsignificant pairwise comparisons after correction for multiple comparisons labeled. Sample size: (A–D) n = 92 (three samples missing age) and(E and F) n = 57 samples between the ages of 10–50 years.

878 CDKN2A/B T2D SNPs Impact Human Islet Biology Diabetes Volume 67, May 2018

may decrease abundance of p16 compared with homozy-gous risk carriers. For rs10757283, in homozygous riskrs10757283 samples, but not protective allele samples, pro-tective allele at rs2383208 or rs10811661 may decreaseabundance of p15. These observations suggest that in-dividual SNPs may contribute risk by impacting locusbiology only in the presence of other locus SNP genotypes,support investment in future experiments to test spe-cific combinations, and help narrow which combinationsto target.

CDKN2A/B T2D SNPs Did Not Impact Glucose-Stimulated Insulin SecretionA subset of islet samples (n = 61) was tested for insulinsecretion stimulation index at their respective islet isolationcenters. Insulin secretion index was positively correlatedwith BMI but showed no relationship with donor age, sex,or islet isolation center by univariate (Supplementary Fig. 8)or multivariable (Supplementary Table 7) analysis. Wheninsulin secretion index was stratified by SNP genotype, con-trary to the hypotheses that CDKN2A/B T2D risk SNPs

Figure 5—SNP combinations may influence locus gene expression. A: Schematic showing approximate locations of the T2D SNPs analyzed inthis study, relative to the ANRIL gene. SNP colors, as in Fig. 1A, indicate linkage disequilibrium. B and C: Protective alleles of rs10811661(shown) and rs2383208 (Supplementary Fig. 4) may decrease expression of p16 in homozygous protective rs564398 CC samples. The samecomparison for p14 did not meet FDR,10% (q value 17%); for ANRIL, FDR.20% (data not shown). D and E: Homozygous risk alleles at bothneighboring SNPs rs10757283 and rs10811661 may collaboratively increase p15 expression. The same comparison for MTAP showedFDR .20%. *FDR ,10%, our predetermined risk tolerance for future experiments exploring haplotype hypotheses. mRNA abundance isexpressed as DCt, normalized to ACTB. n = 95 for all panels. All other inter-SNP comparisons, both shown and not shown (aside from thosein Supplementary Fig. 4), resulted in FDR .10% or had insufficient data points to evaluate (defined as n = 2 or fewer).

diabetes.diabetesjournals.org Kong and Associates 879

impair glucose sensing, insulin production, or stimulus se-cretion coupling, samples with T2D risk alleles did not showevidence for impairment in ex vivo insulin secretion in thiscohort (Fig. 6 and Supplementary Table 8).

Risk Alleles at rs564398 Reduced Glucose-Inducedb-Cell ProliferationSince CDKN2A/B locus genes are best known for inhibitingthe cell cycle, we assessed transcript markers of prolifera-tion (KI67, PCNA, and CCND2) in this cohort, as well as theactual rate of S-phase entry in growth-stimulatory cultureconditions, by BrdU labeling, in a subset of samples. Sur-prisingly, although PCNA and CCND2 showed a high degreeof correlation with each other, KI67 did not correlate witheither PCNA or CCND2 (Fig. 7A–C). No SNP genotype wascorrelated with abundance of PCNA, CCND2, or KI67 (notshown). Transcript level is only a surrogate for proliferationand lacks sensitivity in a tissue with a very low frequency ofproliferation events and a mixture of cell types. To measureactual cell cycle entry in b-cells, we cultured a subset (n =47) of islet preparations in 5 mmol/L or 15 mmol/L glucoseand quantified nuclear BrdU incorporation in insulin-positivecells (Supplementary Fig. 9). BrdU incorporation rate in5 mmol/L glucose was nominally correlated with basalPCNA abundance but not with KI67 or CCND2 (Supplemen-tary Fig. 10). As previously observed (27,28,35), glucoseincreased human b-cell proliferation (P , 0.0001 [notshown]). To test whether any T2D SNP genotype impactedb-cell proliferation, the proliferation index (ratio of BrdU+b-cells in 15 mmol/L compared with 5 mmol/L glucose)was stratified by SNP identity. Genotype at rs2383208,rs10811661, and rs10757283 did not influence the prolif-eration index (Fig. 7D–E). However, genotype at rs564398

was strongly associated with the human b-cell proliferationindex, with homozygous protective alleles showing approx-imately doubled stimulation of proliferation by 15 mmol/Lglucose compared with islet samples harboring risk alleles atthis SNP (Fig. 7F).

DISCUSSION

We have performed a comprehensive analysis of how oneT2D genome-wide association study locus associated withinsulin secretory capacity in human populations influenceshuman islet biology. In n = 95 islet samples, we quantifiedlocus gene expression, SNP genotype, donor characteristics,b-cell function (insulin secretion), and b-cell proliferation.We have made several important observations. First, thelocus contains two distinct gene cassettes that are physicallyoverlapping but have different regulatory characteristics,with one age dependent and the other not. Juvenile isletshave markedly suppressed expression of p14-p16-ANRILbut not p15-MTAP. Second, individual T2D SNPs at thelocus do not substantially alter expression of locus genes,but subtle age-dependent influences are detected, which,contrary to expectations, impact the ANRIL lncRNA ratherthan the most prominent locus product, p16. SNPs mayinteract with each other to influence gene expression, in-creasing the complexity of genotype interpretation and rais-ing intriguing mechanistic hypotheses. Third, genotype didnot impact insulin secretion index. Finally, risk allele atrs564398, which is located within a transcribed exon ofANRIL, decreased the b-cell proliferation index. This workimproves understanding of how CDKN2A/B T2D SNPs im-pact human islet biology and suggests that the influence ofthe locus on human insulin secretory capacity may beeffected via b-cell mass rather than function.

How SNPs influence T2D risk is an important questionin genetics today (3). Although the CDKN2A/B locus isassociated with a number of different disease syndromes(36), the SNPs that we selected to study are most stronglyrelated to T2D risk, with the exception of rs564398, whichis also associated with coronary disease and glaucoma (15).CDKN2A/B SNPs are associated with a range of diabetes-related syndromes beyond T2D, such as gestational diabe-tes mellitus and transplant-related diabetes (9). Since thislocus is active in many different cell types, and the coronarydisease risk region is mostly nonoverlapping with theT2D risk region (36), it is likely that CDKN2A/B impacton diseases other than T2D is mediated by effects outsidethe islet. Given the association of rs10811661/rs2383208with impaired insulin production capacity, to identify pos-sible T2D risk mechanisms we performed our study inislets. Our data highlight the complexity of genetic inputsto human metabolism. Even starting with a genomic locusrepeatedly associated with disease risk across ethnicitiesand T2D-related syndromes (9), with in vivo evidencethat the islet is the risk-mediating tissue (16,19–21), abun-dant preclinical locus knowledge in model systems (9,32),and a fairly large sample size of the relevant tissue, wefound the impact of risk SNPs to be subtle. As with many

Figure 6—Insulin secretion was similar across SNP genotypes. Sixty-one of the islet preparations were tested for glucose-stimulated insulinrelease “stimulation index” at islet isolation centers; the IIDP-derivedinsulin secretion index is plotted against donor genotype. No relation-ship is evident between T2D SNP genotype and IIDP-reported insulinsecretory index. n = 61 for all SNPs. rs108, rs10811661; rs238,rs2383208; rs107, rs10757283; rs564, rs564398.

880 CDKN2A/B T2D SNPs Impact Human Islet Biology Diabetes Volume 67, May 2018

studies in human islets, our data illustrate the marked var-iability from one donor to the next, which reflects the var-iability of outbred human populations. We incorporateddonor parameters such as age and BMI in our analyses butcould not measure many potential other confounding pre-mortem influences such as insulin sensitivity (liver, muscle,brain, and fat), coexisting diseases and medications, environ-mental effects (diet, stresses, and toxins), exercise history,prenatal events, and others. Islet stress related to donordemise, isolation, shipping, and culture may also introducevariability (37). In this context, subtle effects observed inthis challenging system may reflect large effects in certainsubpopulations or small effects present uniformly acrossvariable conditions.

The presence of two gene expression cassettes at thelocus, only one of which is age dependent, suggests inter-esting biology. p16 is well-known to increase with age in

many tissues and organisms (10). Our observation that p15abundance did not increase with age in human islets con-flicts with published results in other tissues (38,39). Therelationship betweenMTAP and CDKN2A/B locus genes hasnot been studied extensively, but these are contained in thesame Hi-C–defined topologically associating domain in thehuman genome (Supplementary Fig. 11) (40,41). Whethercoregulation of p15 andMTAP has functional importance inthe islet remains unknown.

rs2383208/rs10811661 impacted ANRIL abundance inage-dependent fashion. Islets containing homozygous riskalleles showed a “premature aging” phenotype, with youngrisk allele islets having ANRIL levels similar to those of olderprotective-allele islets. The low MAF for these SNPs pre-cluded comparison between homozygous protective and ho-mozygous risk, which might have revealed a larger effectsize. We do not currently know whether one of these, or

Figure 7—Risk allele at rs564398 suppressed glucose induction of b-cell proliferation. A–C: RNA abundance of proliferation-related genesPCNA, CCND2, and KI67 in flash-frozen islets showed a strong correlation with PCNA and CCND2 (A) but not with KI67 (B and C). D–F:rs564398, but not rs2383208, rs10811661, or rs10757283, was associated with b-cell proliferation. Islets containing one or two T2D risk allelesat rs564398 had lower proliferation index than islets containing homozygous protective alleles at rs564398. Dispersed islets were cultured onglass coverslips for 96 h in either 5 mmol/L glucose (unstimulated) or 15 mmol/L (stimulated) glucose with BrdU present for the whole 96 h.Cultures were fixed, immunostained, imaged, and blinded, and the percent insulin(+) cells that were also BrdU(+) were quantified by manualcounting. Plotted is the proliferation index, which is the ratio of 15 to 5 mmol/L. n = 95 (A–C) and n = 43 (D–F) (47 preps tested, but 4 preps had0% BrdU+ in 5 mmol/L glucose, and thus an index could not be calculated). mRNA abundance is expressed as DCt, normalized to ACTB. Dataare mean 6 SD. P values are by linear regression (A–C) and ANOVA with correction for multiple comparisons (D–F).

diabetes.diabetesjournals.org Kong and Associates 881

another SNP in linkage with them, is causative. Fine map-ping of this region has not revealed SNPs with greaterimpact on T2D risk (1,4), but islet ANRIL abundance wasnot the end point in those studies. These SNPs fall neara known regulatory region downstream of the 39 end ofANRIL, which may regulate ANRIL transcription. Mechanis-tically, how higher ANRIL abundance in islets might in-crease T2D risk is unknown. In other cell types, ANRIL isproproliferative (42), an effect mediated by ANRIL-depen-dent suppression of locus cell cycle inhibitors p14, p15, andp16 (43). Our RNA analyses did not detect any hint ofnegative correlation between ANRIL and p14/p16 in islets;in fact, the strong positive correlation between these tran-scripts calls into question whether ANRIL negatively regu-lates other locus genes in human islets.

Beyond SNP regulation of ANRIL abundance, a secondobservation also points to a role for ANRIL in human islets:rs564398, which influences the b-cell proliferation index,is a transcribed polymorphism in this lncRNA. Althoughthe causative SNP remains unknown, it is possible thatrs564398 itself impacts lncRNA activity. ANRIL is a compli-cated gene, with 20 exons and at least 14 reported isoforms(42). Some ANRIL variants are circular (44). Exon 2 is notcontained in all isoforms but is generally associated withlinear variants (44). Whether rs564398 identity impactsANRIL isoform production, splicing, stability, or interac-tion with other cellular DNA, RNA, or protein, to regulatehuman b-cell proliferation, remains unknown. Genotypeat rs564398 did not correlate with expression of cell cy-cle genes, or BrdU incorporation, under basal conditions;rs564398 may be a marker for b-cell responsiveness toproliferation-inducing conditions rather than increased pro-liferation in unstimulated conditions. Importantly, thereare many SNPs in linkage with rs564398, and any of these,or a combination of these, could be influencing biology in-stead of rs564398 itself. Also important, while rs564398has been repeatedly confirmed to be associated with T2D inCaucasian populations, it has little to no relationship withT2D risk in Asian populations (45–47). Taken together withthe “premature aging” influence of rs10811661/rs2383208on ANRIL expression, CDKN2A/B SNPs may impact T2Drisk by adversely impacting accrual of b-cell mass duringearly adulthood.

Combinatorial SNP regulation of gene expression in-creases the complexity of how genomic variationmay impactcellular function. Whole-genome studies restricting locusanalyses to single “lead” SNPs cannot detect biology relatedto two or more local SNPs interacting with each other. Themechanism by which SNPs interact to regulate CDKN2A/Blocus gene expression in human islets is unknown. Interac-tion between rs10757283 and rs2383208 or rs10811661linked SNPs to regulate p15 abundance could be via modu-lation of transcription factor occupancy or epigenetic regu-lation of the enhancer region in which they are located (48).Intriguingly, our observed interaction between these SNPsis supported by a complementary linkage disequilibriumblock analysis, which revealed that a haplotype consisting

of rs2383208/rs10811661 and rs10757283 was associatedwith T2D risk (13). There are multiple SNPs in linkage dis-equilibrium with rs10811661 and also with rs10757283;actual causal polymorphisms are unknown. A physical orfunctional interaction between the ANRIL lncRNA and thisenhancer may mediate cooperation between rs564398 andrs2383208, rs10811661, or linked SNPs to regulate p16expression. Our study cannot distinguish between in-cisand in-trans interaction. Focused, cell type–specific studiesare needed to determine how SNP combinations influencelocus gene expression.

This study adds to the body of knowledge debating therelative influence of b-cell mass versus function on T2Drisk. Our study found that CDKN2A/B SNPs did not influ-ence glucose-stimulated insulin secretion either in univari-ate analysis or in multivariable models incorporating donorage, sex, race, and BMI and islet isolation center. This is per-haps surprising given the in vivo data linking risk allele atrs10811661 with impaired insulin secretion (16,22). Sincein vivo insulin secretion is a composite outcome of bothmass and function, CDKN2A/B SNPs may impact b-cell massbut not b-cell function. This concept is in agreement withthe widespread assumption that CDKN2A/B SNPs influ-ence b-cell proliferation because of the known cell cycle–inhibitory effects of locus genes p14, p15, and p16 (49). Ourrs564398 data are the first demonstration, to our knowledge,of a CDKN2A/B locus SNP impacting b-cell proliferation.

Our study has caveats. Multiple cell types are found inhuman islets. Other than the proliferation measurements,which were assessed in insulin-positive cells, all other stud-ies were performed on whole islets. We did not assess therelative proportion of islet cells that were b-cells, and tothe extent that gene expression may be cell type–specific,variable cellular makeup may have influenced results. Inaddition, there is considerable heterogeneity even amongb-cells (50,51). Repetition of our current analyses on sortedb-cells, or on single cells, exceeds our current resources. Theinsulin secretion data have caveats; use of the IIDP-reportedinsulin secretion index introduces technical variability butbenefits from freshly isolated, preshipment tissue. Our con-firmation that insulin secretion correlated with BMI, butnot with isolation center, is reassuring in this regard.

In sum, this work provides new information about howCDKN2A/B T2D SNPs impact islet biology, suggests thatthe ANRIL lncRNA may play a role in human islets, anduncovers a link between a T2D SNP and b-cell proliferation.Further studies into the CDKN2A/B locus to developa mechanistic understanding of how these SNPs impactislet biology to influence T2D risk could one day open thedoor for using personalized genomic information to informT2D subtype definitions and therapeutic choice.

Acknowledgments. Human pancreatic islets were provided by the NIDDK-funded IIDP at City of Hope and by Sambra Redick and David Harlan from theUniversity of Massachusetts Medical School and Alvin C. Powers from VanderbiltUniversity. The authors are grateful to Ahmet Rasim Barutcu, from the Broad Institute

882 CDKN2A/B T2D SNPs Impact Human Islet Biology Diabetes Volume 67, May 2018

and Harvard University, for helpful guidance with the topologically associatingdomain analysis. The authors thank the Beta Cell Biology Group at the Universityof Massachusetts Medical School for many helpful discussions.Funding. This work was supported by grants from NIDDK, National Institutes ofHealth (R01-DK-095140 to L.C.A.; DK-104211 and DK-106755 to VanderbiltUniversity group, headed by Al Powers; and 2UC4-DK-098085 to the IIDP). L.C.A.received support from the American Diabetes Association grant 1-15-BS-003 incollaboration with the Order of the Amaranth.Duality of Interest. No potential conflicts of interest relevant to this articlewere reported.Author Contributions. Y.K. performed the majority of the experiments.R.B.S., S.L., and R.E.S. also performed experiments. Y.K. and L.C.A. analyzed data.W.M.J. performed the multivariable linear modeling. L.C.A. devised and planned theexperiments and wrote the manuscript. All authors viewed and had the opportunityto edit and approve the manuscript. L.C.A. is the guarantor of this work and, assuch, had full access to all the data in the study and takes responsibility for theintegrity of the data and the accuracy of the data analysis.Prior Presentation. Parts of this study were presented in abstract form at the77th Scientific Sessions of the American Diabetes Association, San Diego, CA, 9–13June 2017.

References1. Morris AP. Fine mapping of type 2 diabetes susceptibility loci. Curr Diab Rep2014;14:5492. Morris AP, Voight BF, Teslovich TM, et al.; Wellcome Trust Case ControlConsortium; Meta-Analyses of Glucose and Insulin-related traits Consortium(MAGIC) Investigators; Genetic Investigation of ANthropometric Traits (GIANT) Con-sortium; Asian Genetic Epidemiology Network–Type 2 Diabetes (AGEN-T2D) Con-sortium; South Asian Type 2 Diabetes (SAT2D) Consortium; DIAbetes GeneticsReplication And Meta-analysis (DIAGRAM) Consortium. Large-scale associationanalysis provides insights into the genetic architecture and pathophysiology oftype 2 diabetes. Nat Genet 2012;44:981–9903. Fuchsberger C, Flannick J, Teslovich TM, et al. The genetic architecture oftype 2 diabetes. Nature 2016;536:41–474. Gaulton KJ, Ferreira T, Lee Y, et al.; DIAbetes Genetics Replication And Meta-analysis (DIAGRAM) Consortium. Genetic fine mapping and genomic annotation definescausal mechanisms at type 2 diabetes susceptibility loci. Nat Genet 2015;47:1415–14255. van de Bunt M, Manning Fox JE, Dai X, et al. Transcript expression data fromhuman islets links regulatory signals from genome-wide association studies fortype 2 diabetes and glycemic traits to their downstream effectors. PLoS Genet2015;11:e10056946. Fadista J, Vikman P, Laakso EO, et al. Global genomic and transcriptomicanalysis of human pancreatic islets reveals novel genes influencing glucose me-tabolism. Proc Natl Acad Sci U S A 2014;111:13924–139297. Varshney A, Scott LJ, Welch RP, et al.; NISC Comparative Sequencing Program.Genetic regulatory signatures underlying islet gene expression and type 2 diabetes.Proc Natl Acad Sci U S A 2017;114:2301–23068. Gaulton KJ. Mechanisms of type 2 diabetes risk loci. Curr Diab Rep 2017;17:729. Kong Y, Sharma RB, Nwosu BU, Alonso LC. Islet biology, the CDKN2A/B locusand type 2 diabetes risk. Diabetologia 2016;59:1579–159310. Kim WY, Sharpless NE. The regulation of INK4/ARF in cancer and aging. Cell2006;127:265–27511. Sharpless NE, Sherr CJ. Forging a signature of in vivo senescence. Nat RevCancer 2015;15:397–40812. Scott LJ, Mohlke KL, Bonnycastle LL, et al. A genome-wide association study oftype 2 diabetes in Finns detects multiple susceptibility variants. Science 2007;316:1341–134513. Zeggini E, Weedon MN, Lindgren CM, et al.; Wellcome Trust Case ControlConsortium (WTCCC). Replication of genome-wide association signals in UK samplesreveals risk loci for type 2 diabetes. Science 2007;316:1336–134114. Saxena R, Voight BF, Lyssenko V, et al.; Diabetes Genetics Initiative of BroadInstitute of Harvard and MIT, Lund University, and Novartis Institutes of BioMedical

Research. Genome-wide association analysis identifies loci for type 2 diabetes andtriglyceride levels. Science 2007;316:1331–133615. Cunnington MS, Santibanez Koref M, Mayosi BM, Burn J, Keavney B. Chro-mosome 9p21 SNPs associated with multiple disease phenotypes correlate withANRIL expression. PLoS Genet 2010;6:e100089916. Grarup N, Rose CS, Andersson EA, et al. Studies of association of variants nearthe HHEX, CDKN2A/B, and IGF2BP2 genes with type 2 diabetes and impaired insulinrelease in 10,705 Danish subjects: validation and extension of genome-wide as-sociation studies. Diabetes 2007;56:3105–311117. Xu M, Bi Y, Xu Y, et al. Combined effects of 19 common variations on type 2diabetes in Chinese: results from two community-based studies. PLoS One 2010;5:e1402218. Peng F, Hu D, Gu C, et al. The relationship between five widely-evaluatedvariants in CDKN2A/B and CDKAL1 genes and the risk of type 2 diabetes: a meta-analysis. Gene 2013;531:435–44319. Hribal ML, Presta I, Procopio T, et al.; EUGENE2 Consortium. Glucose tolerance,insulin sensitivity and insulin release in European non-diabetic carriers of a poly-morphism upstream of CDKN2A and CDKN2B. Diabetologia 2011;54:795–80220. ’t Hart LM, Simonis-Bik AM, Nijpels G, et al. Combined risk allele score of eighttype 2 diabetes genes is associated with reduced first-phase glucose-stimulatedinsulin secretion during hyperglycemic clamps. Diabetes 2010;59:287–29221. Dimas AS, Lagou V, Barker A, et al.; MAGIC Investigators. Impact of type2 diabetes susceptibility variants on quantitative glycemic traits reveals mechanisticheterogeneity. Diabetes 2014;63:2158–217122. Jonsson A, Ladenvall C, Ahluwalia TS, et al. Effects of common genetic variantsassociated with type 2 diabetes and glycemic traits on a- and b-cell function andinsulin action in humans. Diabetes 2013;62:2978–298323. Brereton MF, Rohm M, Ashcroft FM. b-Cell dysfunction in diabetes: a crisis ofidentity? Diabetes Obes Metab 2016;18(Suppl. 1):102–10924. Dai C, Hant Y, Shostak A, et al. Age-dependent human b cell proliferationinduced by glucagon-like peptide 1 and calcineurin signaling. J Clin Invest 2017;127:3835–384425. 1000 Genomes Project Consortium, Auton A, Brooks LD, Durbin RM, et al. Aglobal reference for human genetic variation. Nature 2015;526:68–7426. Machiela MJ, Chanock SJ. LDlink: a web-based application for exploringpopulation-specific haplotype structure and linking correlated alleles of possiblefunctional variants. Bioinformatics 2015;31:3555–355727. Sharma RB, O’Donnell AC, Stamateris RE, et al. Insulin demand regulates b cellnumber via the unfolded protein response. J Clin Invest 2015;125:3831–384628. Stamateris RE, Sharma RB, Kong Y, et al. Glucose induces mouse b-cellproliferation via IRS2, mTOR, and cyclin D2 but not the insulin receptor. Diabetes2016;65:981–99529. Pascoe J, Hollern D, Stamateris R, et al. Free fatty acids block glucose-inducedb-cell proliferation in mice by inducing cell cycle inhibitors p16 and p18. Diabetes2012;61:632–64130. Alonso LC, Yokoe T, Zhang P, et al. Glucose infusion in mice: a new model toinduce beta-cell replication. Diabetes 2007;56:1792–180131. Robertson KD, Jones PA. Tissue-specific alternative splicing in the humanINK4a/ARF cell cycle regulatory locus. Oncogene 1999;18:3810–382032. Krishnamurthy J, Ramsey MR, Ligon KL, et al. p16INK4a induces an age-dependent decline in islet regenerative potential. Nature 2006;443:453–45733. Ward LD, Kellis M. HaploReg: a resource for exploring chromatin states,conservation, and regulatory motif alterations within sets of genetically linked var-iants. Nucleic Acids Res 2012;40:D930–D93434. Gaunt TR, Rodríguez S, Day IN. Cubic exact solutions for the estimation ofpairwise haplotype frequencies: implications for linkage disequilibrium analyses anda web tool ‘CubeX’. BMC Bioinformatics 2007;8:42835. Levitt HE, Cyphert TJ, Pascoe JL, et al. Glucose stimulates human beta cellreplication in vivo in islets transplanted into NOD-severe combined immunodeficiency(SCID) mice. Diabetologia 2011;54:572–58236. Jeck WR, Siebold AP, Sharpless NE. Review: a meta-analysis of GWAS andage-associated diseases. Aging Cell 2012;11:727–731

diabetes.diabetesjournals.org Kong and Associates 883

37. Stiegler P, Stadlbauer V, Hackl F, et al. Prevention of oxidative stress in porcineislet isolation. J Artif Organs 2010;13:38–4738. Menthena A, Koehler CI, Sandhu JS, et al. Activin A, p15INK4b signaling, andcell competition promote stem/progenitor cell repopulation of livers in aging rats.Gastroenterology 2011;140:1009–102039. Wang F, Zhang L. p15(INK4b) regulates cell cycle signaling in hippocampalastrocytes of aged rats. Aging Clin Exp Res 2016;28:813–82140. Zhou X, Maricque B, Xie M, et al. The Human Epigenome Browser at Wash-ington University. Nat Methods 2011;8:989–99041. Wang Y, Zhang B, Zhang L, et al. The 3D Genome Browser: a web-basedbrowser for visualizing 3D genome organization and long-range chromatin inter-actions.[Internet], 2017. Available from https://www.biorxiv.org/content/early/2017/02/27/112268. Accessed 27 February 201742. Congrains A, Kamide K, Ohishi M, Rakugi H. ANRIL: molecular mechanisms andimplications in human health. Int J Mol Sci 2013;14:1278–129243. Aguilo F, Zhou M-M, Walsh MJ. Long noncoding RNA, polycomb, and theghosts haunting INK4b-ARF-INK4a expression. Cancer Res 2011;71:5365–536944. Burd CE, Jeck WR, Liu Y, Sanoff HK, Wang Z, Sharpless NE. Expression oflinear and novel circular forms of an INK4/ARF-associated non-coding RNA correlateswith atherosclerosis risk. PLoS Genet 2010;6:e1001233

45. Wu Y, Li H, Loos RJF, et al. Common variants in CDKAL1, CDKN2A/B, IGF2BP2,SLC30A8, and HHEX/IDE genes are associated with type 2 diabetes and im-paired fasting glucose in a Chinese Han population. Diabetes 2008;57:2834–284246. Ng MCY, Park KS, Oh B, et al. Implication of genetic variants near TCF7L2,SLC30A8, HHEX, CDKAL1, CDKN2A/B, IGF2BP2, and FTO in type 2 diabetes andobesity in 6,719 Asians. Diabetes 2008;57:2226–223347. Horikawa Y, Miyake K, Yasuda K, et al. Replication of genome-wide associationstudies of type 2 diabetes susceptibility in Japan. J Clin Endocrinol Metab 2008;93:3136–314148. Pasquali L, Gaulton KJ, Rodríguez-Seguí SA, et al. Pancreatic islet enhancerclusters enriched in type 2 diabetes risk-associated variants. Nat Genet 2014;46:136–14349. Hannou SA, Wouters K, Paumelle R, Staels B. Functional genomics of theCDKN2A/B locus in cardiovascular and metabolic disease: what have we learnedfrom GWASs? Trends Endocrinol Metab 2015;26:176–18450. Gutierrez GD, Gromada J, Sussel L. Heterogeneity of the pancreatic beta cell.Front Genet 2017;8:2251. Roscioni SS, Migliorini A, Gegg M, Lickert H. Impact of islet architecture onb-cell heterogeneity, plasticity and function. Nat Rev Endocrinol 2016;12:695–709

884 CDKN2A/B T2D SNPs Impact Human Islet Biology Diabetes Volume 67, May 2018