Assignment of Functional Relevance to Genes at Type 2 Diabetes-Associated Loci Through Investigation of β-Cell Mass Deficits

Abstract

Type 2 diabetes (T2D) has been associated with a large number of genomic loci, many of which encompass multiple genes without a definitive causal gene. This complexity has hindered efforts to clearly identify functional candidate genes and interpret their role in mediating susceptibility to disease. Here we examined the relevance of individual genes found at T2D-associated loci by assessing their potential contribution to a phenotype relevant to the disease state: production and maintenance of β-cell mass. Using transgenic zebrafish in which β-cell mass could be rapidly visualized in vivo, we systematically suppressed the expression of orthologs of genes found at T2D-associated genomic loci. Overall, we tested 67 orthologs, many of which had no known relevance to β-cell mass, at 62 human T2D-associated loci, including eight loci with multiple candidate genes. In total we identified 25 genes that were necessary for proper β-cell mass, providing functional evidence for their role in a physiological phenotype directly related to T2D. Of these, 16 had not previously been implicated in the regulation of β-cell mass. Strikingly, we identified single functional candidate genes at the majority of the loci for which multiple genes were analyzed. Further investigation into the contribution of the 25 genes to the adaptive capacity of β-cells suggested that the majority of genes were not required for glucose-induced expansion of β-cell mass but were significantly necessary for the regeneration of β-cells. These findings suggest that genetically programmed deficiencies in β-cell mass may be related to impaired maintenance. Finally, we investigated the relevance of our findings to human T2D onset in diabetic individuals from the Old Order Amish and found that risk alleles in β-cell mass genes were associated with significantly younger age of onset and lower body mass index. Taken together, our study offers a functional approach to assign relevance to genes at T2D-associated loci and offers experimental evidence for the defining role of β-cell mass maintenance in genetic susceptibility to T2D onset.

Genome-wide association (GWA) studies have uncovered more than 75 genomic loci associated with type 2 diabetes (T2D), potentially implicating at least as many genes (reviewed in reference 1). Despite this finding, many of the causative gene(s) at those loci and the discrete mechanistic roles each play in the disease have not been extensively elucidated. This is due in part to the difficulty of large-scale functional interpretation for a physiologically complex phenotype. Identification of causal genes at associated loci and elucidation of their relevance to T2D requires an approach to systematically assess the contribution of candidate genes to specific phenotypes that are relevant to the disease state and therefore indicative of a potential functional link.

One discrete phenotype directly relevant to T2D etiology is the loss of β-cell functional capacity (reviewed in reference 2). Insulin resistance drives increased insulin secretion that over time impairs β-cell function. Mounting evidence suggests that depletion of β-cell mass also occurs in T2D patients as a result of increased apoptosis (3). This is consistent with evidence indicating a tight correlation between β-cell area and function in diabetes (2). Loss of β-cell mass may therefore be indicative of individual susceptibility to β-cell dysfunction and T2D onset, and this predisposition may be mediated by deficiencies in specific genes. The nature of β-cell mass, however, makes it difficult to study in vivo necessitating the use of model organisms in which it can be easily observed. The zebrafish is particularly amenable to such study because β-cell mass can be visualized in vivo in transparent embryos and larvae (4). Specification of β-cells and other endocrine cell types are conserved between zebrafish and mammals (5). Moreover, zebrafish β-cell mass exhibits adaptive capacity similar to mammalian systems and the processes important to β-cell specification during embryonic stages are also recapitulated during adaptation (5), suggesting that the deficiencies in β-cell neogenesis might not only impact β-cell reserve but may also impair the maintenance of β-cell mass.

Here we carried out a large functional screen with the goal of identifying genes at T2D-associated loci that may have functional relevance to the etiology of the disease. We tested the hypothesis that genes that are functionally relevant to T2D could impact β-cell capacity. We reasoned that β-cell deficits observed as a result of loss of candidate gene function would provide evidence of its relevance to T2D. Such observations also would potentially support a β-cell mediated effect. Using the zebrafish system to systematically suppress expression of individual orthologs for genes found at T2D-associated loci, we identified a subset necessary for the production and maintenance of pancreatic β-cells and included several genes previously identified for their role in β-cell maintenance. Identification of these genes included the identification of individual functional genes at multigene GWAS loci. Finally, investigation of diabetic carriers of risk alleles in this subset of β-cell mass genes revealed phenotypes consistent with increased susceptibility to β-cell deficits including younger age of diabetes onset and lower body mass index (BMI). Taken together, these findings demonstrate the utility and feasibility of our in vivo approach for assigning functional relevance to genes at T2D-associated loci by examining their contribution to a discrete phenotype relevant to disease etiology. Our findings also suggest that distinct risk genotypes may drive specific aspects of dysfunction contributing to the disease state.

Materials and Methods

Zebrafish husbandry, orthology, and embryonic gene expression

Adult transgenic fish were maintained and bred at 28–30°C. Embryos were raised at 28.5°C until harvesting for experimental studies. Genes at T2D susceptibility loci were identified based on published reports (6–30). Zebrafish orthologs were then identified for each T2D-associated gene using reciprocal Basic Local Alignment Search Tool search. All zebrafish orthologs are listed in Table 1. Expression of 67 T2D-associated genes was verified in wild-type embryos (Tubingen) by RT-PCR analysis at 1, 3, and 5 days post fertilization (dpf). Staging was carried out according to published guidelines (31). RNA was isolated using Isol-RNA (5 PRIME, Inc.) and cDNA was generated using RevertAid first-strand cDNA synthesis kit (Thermo Scientific). HotMaster Taq DNA polymerase (5 PRIME, Inc.) was used for PCR; conditions varied by primer set (melting temperature range 49°C–63°C) as did agarose gel electrophoresis conditions (range 1%–2.5% Tris-acetate-EDTA buffer). Primer sequences are available upon request.

Table 1.

T2D-Associated Gene Orthologs Implicated in β-Cell Mass in Zebrafish

Gene	β-Cell Number	Significance (P Value)	β-Cell Mass Area, μm2	Significance (P Value)
Control	33.41 ± 3.32		23.59
pdx1a	9.20 ± 3.92	1.31E-86	14.09	1.92E-06
wfs1a	14.60 ± 3.68	1.08E-40	15.24	3.06E-09
camk1d	14.72 ± 4.35	4.02E-76	9.19	2.60E-10
gckr	14.80 ± 4.56	8.17E-32	9.62	4.06E-07
hnf4aa	16.04 ± 4.16	1.18E-37	15.46	4.51E-04
c2cd4ab	16.24 ± 4.07	2.36E-55	10.89	9.49E-06
kcnj11	16.50 ± 4.05	3.90E-21	16.16	4.46E-04
ide	17.78 ± 3.71	1.05E-52	12.96	1.18E-06
hnf1aa	17.79 ± 4.91	2.62E-55	13.55	2.08E-04
cdkal1	18.18 ± 3.19	1.51E-60	7.61	2.96E-08
klhdc5	18.34 ± 3.19	1.05E-54	13.29	8.30E-06
srebf1a	18.44 ± 3.46	1.69E-48	11.11	1.89E-06
pax4a	18.95 ± 3.45	7.40E-50	8.41	1.11E-08
notch2	19.03 ± 4.22	1.12E-43	11.75	1.15E-05
fitm2	19.15 ± 3.01	1.80E-51	11.18	9.37E-05
thadaa	19.23 ± 4.54	2.89E-42	11.06	2.91E-07
adamts9	19.36 ± 4.06	2.22E-35	9.44	2.10E-07
gck	19.81 ± 4.22	1.79E-39	14.65	2.95E-04
pepd	19.92 ± 3.03	5.42E-55	6.67	3.71E-07
klf14	20.10 ± 3.70	7.55E-49	10.3	4.25E-06
hnf1ba	20.73 ± 3.53	2.43E-47	4.3	4.10E-04
zfand3	21.25 ± 2.81	1.71E-45	8.76	4.40E-07
slc30a8	21.30 ± 3.62	3.25E-46	12.64	1.84E-06
ankrd55	22.33 ± 3.11	3.08E-43	14.09	5.29E-05
glis3	29.30 ± 1.88	1.63E-14	17.5	1.40E-03

Of 67 orthologs targeted by morpholino injection, the 25 listed were significantly necessary for proper β-cell mass area and β-cell number at 5 dpf relative to control morpholino-injected larvae.

^aNine genes had been previously implicated in β-cell mass in mouse models.

Morpholino design, injection, and validation

Morpholino antisense oligonucleotides (Gene Tools, LLC) were designed to target each ortholog. Embryos were injected at the one- to two-cell stage at three concentrations to verify dose response and to determine optimal morpholino concentration for subsequent studies. Subsequent experiments were carried out using the highest morpholino concentration eliciting β-cell phenotypes in the absence of overt morphological defects or developmental delay. A control morpholino (4 ng/nL) was used for all experiments. A total of 200–400 embryos were injected for each morpholino. Morpholino sequences are available upon request. For validation of morpholino efficacy, injected embryos were collected at 3, 5, and 7 dpf. RNAs and cDNAs were isolated and generated, respectively, and RT-PCR was carried out using primers flanking targeted exons. To quantify gene-specific down-regulation, specifically to assess quantity of wild-type transcript, quantitative RT-PCR (qRT-PCR) was performed at 5 dpf. To measure wild-type transcript levels in morphant relative to control, the reverse primer for each gene was designed to sit within the identified region targeted for splicing as identified via the RT-PCR outlined above. Therefore, gene expression results represent a decrease in wild-type transcript. Samples and controls/standards were run in triplicates on the Roche LightCylcer 480. Gene expression levels were normalized relative to β-actin. Primer sequences are available upon request. Morpholino-induced off-target effects were assessed by quantification of δ113 p53 expression, a diagnostic marker of such effects, via qRT-PCR (32). Samples were run in triplicate and gene expression levels were normalized relative to β-actin.

CRISPR/Cas9 disruption of genomic loci

To confirm the results obtained by morpholino injection, CRISPR-Cas9 was used to generate mutant zebrafish lines using guide RNA (gRNA) generated by plasmid-free in vitro transcription (protocol provided by the laboratory of S. Burgess, National Institutes of Health/National Human Genome Research Institute). Briefly, guide RNAs composed of a 22-bp target sequence flanked by a 5′ T7 promoter sequence (5′-TAATACGACTCACTATA-3′) and a 3′ overlap sequence (5′-GTTTTAGAGCTAGAAATAG-3′) were annealed to a generic oligo (5′-AAAAGCACCGACTCGGTGCCACTTTTTCAAGTTGATAACGGACTAGCCTTATTTTAACTTGCTATTTCTAGCTCTAAAAC-3′). The assembled oligos were transcribed in vitro using the MaxiScript kit (Ambion; AM1314M). Cas9 transcript was generated using the mMESSAGE mMACHINE T3 kit (Life Technologies; AM1348). Approximately 150 single-celled Tg(ptf1a:eGFP;insa:mCherry) embryos were coinjected with each T2D gene-specific, 25-pg guide RNA and 300 pg Cas9 RNA. For each gRNA, 75 embryos were imaged at 5 dpf for their individual β-cell mass. A subset (∼35) of the imaged embryos was isolated for assessment of β-cell number. To verify CRISPR-Cas9-induced mutagenesis, genomic DNA was extracted and T7 endonuclease-based confirmation of mismatch (33) after the PCR amplification of the target region was performed.

β-Cell mass and number quantification

Double-transgenic zebrafish from the Tg(ptf1a:eGFP;insa:mCherry) line (S. Leach, Johns Hopkins University) were used to assay pancreatic β-cells relative to exocrine pancreas mass. Area and intensity of mCherry expression was quantified in 40–90 larvae per injection concentration using ImageJ software (National Institutes of Health, Bethesda, Maryland) at 24 hours post fertilization (hpf) and 5 dpf, to identify β-cell mass in the principal islet and mature β-cells, respectively. mCherry-expressing β-cells were quantified at the effective concentration. Similarly, the area of green fluorescent protein (GFP) expression was quantified using ImageJ software (National Institutes of Health) to assess the possibility of an exocrine defect. The mean ratio of the area of insulin expression to the area of exocrine pancreas was calculated using these values to identify potential false-positive results. For each gene, specificity of the observed reduction in β-cell mass was assessed using the Tg(neurod:eGFP) reporter line. At 72 hpf, embryos were fixed in 4% paraformaldehyde (PFA) overnight and deyolked to obtain a nonobstructed view of the developing endocrine progenitor mass. The area of GFP (ie, neurod1) expression in the developing pancreas (n ≥ 50) was quantified using ImageJ software (National Institutes of Health) to determine the possibility of β-cell mass decline as a result of general endocrine progenitor cell deficiencies. For the quantification of β-cell number, injected embryos were fixed in 4% PFA overnight at 5 dpf and mounted individually on a glass slide in one drop of ProLong Gold antifade reagent (Life Technologies; P36930). Embryos were covered with a glass coverslip such that the islet was dispersed to allow for single-cell resolution of mCherry-expressing cells. Values were compared with β-cell counts in control morpholino- or Cas9-only-injected controls. For all analyses, a two-tailed t test was used to evaluate statistical significance of observed changes relative to controls and across the various morpholino concentrations.

Insulin- and histone H3-β-cell immunostaining

Embryos from the Tg(insa:mCherry) transgenic line were injected with morpholino at the one- to two-cell stage. Embryos were cultured in embryo medium until 5 dpf, at which point they were fixed in 4% PFA overnight. Embryos were dehydrated in sequential concentrations of methanol-PBS for 10 minutes each, followed by incubation in acetone for 30 minutes at −20ºC, and washed in PBS with Tween 20 for 5 minutes. Larvae were blocked in blocking buffer for 1 hour and incubated (1:50 or 1:100) with the primary antibody (ab7842 or ab5176; Abcam) overnight at 4ºC followed by a wash in immunofluorescence buffer and incubation with the secondary antibody (A11073 or A11008; Life Technologies) for 2 hours at room temperature. For the insulin-β-cell assay, colocalization of mCherry and GFP signal in 5 dpf larvae were analyzed using both a Zeiss Lumar.V12 stereoscope and a Zeiss LSM 510 Meta confocal microscope (×63/1.4 oil differential interference contrast). For the stereoscope analyses, larvae (n ≥ 20 per morpholino) were flat mounted, dorsal side up, on a glass slide and visualized for both mCherry and GFP fluorescence. Signals were quantified using ImageJ software (National Institutes of Health). For confocal microscopy (n = 4 per morpholino), larvae were flat mounted, left side down on a glass slide. A coverslip was placed on top so that the β-cell mass was slightly dispersed. Localization of GFP (insulin) and mCherry (preproinsulin/β-cells) signals was assessed. Images were stacked using ImageJ software (National Institutes of Health). To assess the proliferating cells, embryos were mounted individually on a glass slide in one drop of ProLong Gold antifade reagent (Life Technologies; P36930) and covered with a glass coverslip to disperse the islet and allow for single-cell resolution of mCherry-expressing cells. Colocalization events were quantified and recorded to identify proliferating β-cells.

β-Cell expansion and regeneration

Double-transgenic Tg(ptf1a:eGFP;insa:mCherry) embryos were cultured until 5 dpf in embryo medium (34) with or without 40 mM glucose medium, starting at 24 hpf and assessed for β-cell mass and β-cell number (n = 20–30 per T2D gene ortholog). For regeneration studies, Tg(insa:nsfB-mCherry) ([4]; M. Parsons, Johns Hopkins University) β-cells were ablated by 36-hour treatment with 10 mM metronidazole, starting at 56 hpf. Embryos were subsequently transferred to normal embryo medium for 36 hours of recovery, after which β-cell mass and number of 20–30 embryos (per T2D gene ortholog) were recorded.

Quantification of glucose

Beginning on 5 dpf, control or T2D morphant larvae were fed one of two diets: AP100 food only (CD) or AP100 food soaked in 40 mM glucose solution (HGD). Larvae were fed twice daily until 7 dpf. Larvae (n = 50) were then collected and used in the glucose assay (Biovision, Inc) as outlined in the manufacturer's protocol. Samples and controls/standards were run in triplicates. Absolute glucose levels were calculated as described in the manufacturer's procedure.

Genotyping and T2D risk allele analyses in the Amish

The Old Order Amish (OOA) (Lancaster County, Pennsylvania) are a founder population of Europe with extensive genealogical records (35), allowing for linkage of virtually all participants into a single, 14-generation pedigree. More than 3500 samples with both DNA and extensive phenotype information on various aspects of health including glucose homeostasis and diabetes make up the Amish Complex Disease Research Program from which samples for this study were ascertained.

Genome-wide genotyping was carried out using either the Affymetrix 500K or 6.0 genotyping chip according to the manufacturer's recommendations. Genotype calling was done using BirdSeed version 2 (36). High-quality genotypes were imputed to the HapMap 2 CEPH (Utah residents with ancestry from northern and western Europe) reference population using MACH (37). Previously reported (6–15, 17–23, 25–30, 38, 39) GWA variants were identified and clustered into two groups, non-β-cell mass or β-cell mass, based on linked gene and effect in zebrafish. Diabetic individuals from the Amish Complex Disease Research Program with GWA study data and information on age of diabetes onset were identified and assigned a count of risk variants, regressed against age of diabetes using a variance components framework that accounts for the relationship structure of the OOA population (http://edn.som.umaryland.edu/mmap/index.php). In young, lean or old, obese persons with diabetes, we assessed the average number of risk alleles/single nucleotide polymorphism (SNP) normalized to the total number of SNPs associated with each gene group. Significance was determined using a t test with Bonferroni correction.

Results

Identification and targeting of T2D orthologs in zebrafish

To disrupt T2D-associated genes in zebrafish, we first identified orthologs for genes associated with the disease in human. We screened the zebrafish genome for genes orthologous to each of 76 human genes found across 64 genomic loci identified in GWA and linkage studies (6, 8–15, 17, 19–23, 25–30, 38–40). Single-candidate genes were selected at 54 loci based on proximity of significantly associated SNPs and known biological evidence (6–17, 19–21, 25–29, 39). For 10 additional loci, multiple genes were selected for identification of orthologs (18, 22, 25, 39). We were unable to identify zebrafish orthologs at two of those loci, one at chr9q21.31 (TLE4 and CHCHD9 [6]) and one at chr12q21.1 (LGR5 and TSPAN8 [25]). In total, we identified 67 zebrafish orthologs for human T2D genes (Supplemental Table 1). We assessed expression at embryonic stages and found that transcripts for all 67 genes were detected starting at 1 dpf with the exception of two genes, wfs1 and gckr, which were detected starting at 3 dpf (Supplemental Figure 1). Embryonic expression supported the relevance of transient knockdown during embryonic stages at which initial production of β-cells occurs. We therefore designed morpholino antisense oligonucleotides targeted against each gene.

Depletion of T2D-associated gene orthologs reduced β-cell mass

To investigate the role of each T2D ortholog in initial production of β-cells, we injected individual morpholinos into double-transgenic embryos expressing preproinsuin promoter-driven mCherry in β-cells (insa:mCherry) and ptf1a promoter-driven GFP in exocrine pancreas (4). Injected embryos were imaged at 5 dpf (n ≥ 50 per experiment, repeated three or more times per morpholino; Figure 1), and changes in β-cell area were assessed by quantification of the area of mCherry expression (Figure 1, B–H). The specificity of significant changes in the average β-cell area was determined by also confirming the absence of gross developmental defects or delay, which was verified by general morphology (Figure 1A and Supplemental Figure 2A), by validating the dose-responsive nature of observed changes in β-cell area (Supplemental Figure S2B), or by assessing exocrine pancreas morphology via GFP expression (Figure 1, I–O). Relative to control morpholino-injected larvae, most morpholinos tested (40 of 67) did not result in a significant decrease in β-cell area (Figure 1P and Supplemental Figure 3; P ≤ .001). Not surprisingly, this included genes that have not previously been implicated in β-cell function, such as jazf1 and r3hdml (Figure 1, C and D). In contrast, a subset of 25 morpholinos resulted in a specific reduction in the area of β-cells at a level that was significant (P ≤ .001; Figure 1, E–H and P, Supplemental Figure 3, and Table 1). These included genes for which previous studies in mouse had revealed necessary functions in β-cell production, such as wfs1 and srebf1 (Figure 1, E and F) as well as genes with no previously described role, such as pepd and zfand3 (Figure 1, G and H). Two morpholinos targeted against insa and pparg resulted in a significant increase in β-cell area, consistent with reports in mouse in which either whole animal or conditional knockout, respectively, resulted in significant islet hyperplasia driven primarily by increased β-cell mass (41, 42) (Figure 1P and Supplemental Figure 3; P ≤ .0001).

Figure 1.

Effects of reduced T2D gene expression in vivo on β-cell area. Zebrafish embryos were injected with individual morpholinos targeting each of 67 T2D-associated gene orthologs. A, Live control morpholino-injected Tg(ptf1a:eGFP;insa:mCherry) zebrafish larva at 5 dpf with inset (dashed line, A′) showing the pancreas. B–O, Pancreata of larvae injected with control morpholino or morpholinos against T2D-associated gene orthologs imaged under fluorescence for either insa:mCherry in β-cells (B–H) or ptf1a:eGFP expression in exocrine pancreas (I–O). Examples of genes for which knockdown produced no effect included jazf1 (C) and r3hdml (D), whereas a reduced β-cell area was observed with the depletion of genes previously implicated in β-cell mass, such as wfs1 (E) and srebf1 (F) and for genes not previously implicated, such as pepd (G) and zfand3 (H). P, Quantification of β-cell area by ImageJ quantification of the area of mCherry expression, shown for all 67 T2D-associated genes. Forty genes did not significantly alter β-cell area. Twenty-five genes, however, displayed significantly reduced β-cell mass relative to control morpholino (student's t test, P ≤ 1 × 10⁻⁴), whereas a significant increase was observed for two genes (students t test, P ≤ 1 × 10⁻⁴).

To verify that decreased expression of mCherry indeed represented reduced β-cell mass, we validated the presence of Insulin protein expression in cells expressing mCherry by whole-mount immunofluorescence. In control morpholino-injected larvae, colocalization of insulin and mCherry expression could be observed, and the area of insulin expression was not significantly differently from that of mCherry (Supplemental Figure 4, A–E). Localization of insulin to the β-cells was confirmed via confocal microscopy (Supplemental Figure 4, F–H). To assess the concordance of this colocalization in morphants that exhibited a decrease in β-cell mass, we examined larvae depleted of wfs1 expression (Supplemental Figure 4, I–L and N–P) and also observed colocalization of Insulin suggesting concordance between transgene expression and insulin protein in β-cells.

To determine whether the reduced area of mCherry expression indeed represented changes in β-cell mass, we calculated for each gene targeted the ratio of the average β-cell area relative to the average area of exocrine pancreas, determined by GFP expression. Down-regulation of genes that were not identified as contributors to β-cell area resulted in similar proportional β-cell area to control morpholino (∼4%; Supplemental Figure 5). In contrast, genes for which morpholino knockdown depleted the β-cell mass also exhibited significant decreases in the ratio of β-cell area to exocrine pancreas area (0.6%–2.5%; P ≤ .0001; Supplemental Figure 5). We also observed a significant increase (9.9%–5.9%; P ≤ .0001; Supplemental Figure 5) in the ratio of β-cell area to exocrine pancreas area for three genes (pparg, ins, and pdx1), consistent with the increase in β-cell mass observed by the depletion of pparg and ins and the disruption of exocrine pancreatic development with the reduction of pdx1 (43, 44).

To verify that these effects on β-cells were a result of reduced gene expression and not other off-target effects conferred by morpholino injection, we pursued multiple avenues to validate the sensitivity and specificity of our approach. We first assessed the efficacy of morpholinos to disrupt targeted genes. For splice-blocking morpholinos, we verified the disruption of splicing at 5 dpf (Supplemental Figure 6A). We also assessed the presence of each targeted transcript by RT-PCR in 5 dpf larvae injected with translation-blocking morpholino and observed reduced abundance (Supplemental Figure 6B). Reduction of wild-type T2D transcripts in morphants, relative to control, was more accurately confirmed for all targeted genes via qRT-PCR (Supplemental Figure 6C). To preclude the possibility that morpholino-induced phenotypes were not simply the result of off-target toxicity and widespread cell death imparted by the morpholinos, we examined δ113 p53 expression levels, which has been reported as a diagnostic signature for off-target morpholino-induced toxicity (32). Five of the 67 morpholinos resulted in significant increases in δ113 p53 expression, potentially suggesting the induction of morpholino off-target effects (Supplemental Figure 7, P ≤ .0001). However, these included three genes (th, tmem195, and tp53inp1) that were not implicated in β-cell mass phenotypes and two genes (camk1d and cdkal1) that were (Supplemental Figure 7). These observations indicate that the observed β-cell deficits did not correlate with off-target toxicity induced by morpholinos and indicate that such effects are unlikely to explain the decreased production of β-cells.

Finally, we used CRISPR/Cas9-mediated disruption (33) for a subset of 10 genes for which morpholinos either did or did not produce significant changes in β-cell mass. We assessed the area of mCherry expression in insa:mCherry larvae coinjected with Cas9 endonuclease and gRNA targeted at each gene and validated disruption of the genomic locus by T7 endonuclease assay (Supplemental Figure 8 [33]). Genes that did not impact β-cell mass by morpholino injection did not significantly contribute to the β-cell area or number when targeted by CRISPR-Cas9. These CRISPR-Cas9 mutants exhibited similar mCherry expression and β-cell number as Cas9-only-injected controls (Supplemental Figure 8, A–F′ and L). However, those genes for which morpholino knockdown resulted in a significant loss of β-cell mass also produced a reduction when targeted by CRISPR-Cas9, which appeared to be even more severe upon quantification of β-cell number (Supplemental Figure 8, G–K′ and L). The high level of concordance of phenotypes in morpholino- and CRISPR-treated larvae potentially indicates the importance of these genes in pancreatic or β-cell specification, but these observations are also consistent with a recent report of phenotypes present in morpholino- and gRNA/Cas9-injected animals but not germline mutants (45).

Quantification of β-cell number

To more accurately assess changes in the production of β-cells, we quantified the number of β-cells present in the principal islet at 5 dpf by fixing larvae and dispersing the islet to observe individual mCherry-positive cells (Figure 2A). Consistent with previous reports, we found that control larval islets comprised an average of 33 β-cells (Figure 2, A and H [5]). Larvae injected with morpholinos contributing to the depletion of β-cell mass, however, had significantly lower numbers of β-cells (P ≤ 1 × 10⁻¹⁴; Figure 2, A and D–H, Table 1, and Supplemental Figure 9), a finding that was not observed for any morpholinos that did not significantly deplete the area of β-cell mass (Figure 2, B and C, and Supplemental Figures 3 and 9). Quantification of β-cell number also revealed that genes with marginally significant roles in β-cell area (0.001 < P < .01) were not necessary for proper β-cell number. These included 17 genes. Upon quantification of β-cell number, however, none of these genes were significantly necessary for the production of the proper number of β-cells (P > .01), suggesting that the 25 genes that contributed significantly to both the area of β-cell mass and the β-cell number were indeed those necessary for the proper production of β-cells.

Figure 2.

Reduction in β-cell number confirms β-cell area phenotype. To more accurately assay changes in the production of β-cells, the number of β-cells present in the principal islet at 5 dpf was quantified. Larvae were fixed and individual mCherry-positive cells, dispersed from the principal islet, were counted. A–G, Representative larvae mounted for single-cell resolution of mCherry-expressing cells (arrowhead). H, β-Cell mass reduction was consistent with the significant reduction in β-cell number observed per larva, whereas those genes previously identified to induce an increase in β-cell mass via morpholino knockdown, ins and pparg, remained significantly different from control. *, P ≤ 1 × 10⁻⁵; **, P ≤ 1 × 10⁻¹⁴; ***, P ≤ 1 × 10⁻²⁰ (two tailed t test). ns, not significant. I, In total, 25 T2D-associated gene orthologs were identified as being necessary for β-cell mass. J, Identified genes included both T2D-genes that also contribute to NDM (green) and MODY (blue) as well as genes associated with only common T2D (red). Only two genes associated with both T2D and either form of young-onset diabetes were not necessary for β-cell mass in zebrafish, although the loss of ins expression increased β-cell mass and loss of abcc8 reduced β-cell mass at marginal significance.

In total, we found that 25 of the 67 genes tested were highly significant for proper production of β-cells during the embryonic stages (Figure 2I), only nine of which (cdkn2a, hnf1a, hnf1b, hnf4a, notch2, pax4, pdx1, srebf1, and wfs1) had been previously implicated in β-cell mass (26, 27, 42, 46–51), supporting the relevance of zebrafish to mammalian β-cell specification but also indicating novel roles in production of β-cells for 16 genes. These included genes such as klhdc5 and ankrd55, for which little or no biological relevance to T2D etiology have been assigned. Despite the abundance of genes lacking prior evidence of pancreatic function, all genes, for which expression data had been reported in at least one of three separate databases (Pancreatic Expression Database, EMBL-EBI, and RefExA), were expressed in the pancreas (52–55). Reassuringly, we found that nearly all of the T2D genes that are also associated with younger-onset forms of diabetes, mature-onset diabetes of the young (MODY), and neonatal diabetes mellitus (NDM) were identified as necessary for β-cell mass production, suggesting that our observations were consistent with their known roles (Figure 2, I and J). Two exceptions to this were INS and ABCC8, both of which are associated with MODY and NDM (56, 57). Our findings do not preclude a possible involvement for these genes in β-cell mass, however, given the increased β-cell mass with suppression of ins and the fact that abcc8 reduced it at a level that was marginally significant (P < .03). These findings might suggest their involvement in β-cell mass via opposing mechanisms or to a lesser extent, respectively.

Assessment of the specificity and origins of β-cell mass depletion

To examine the specificity of the observed defects to β-cells, rather than a global deficit of endocrine cells, we examined the endocrine progenitor population prior to cell type differentiation using the Tg(neurod:eGFP) transgenic line (58). At 72 hpf, a majority (n = 15) of the 25 β-cell mass genes were not necessary for endocrine progenitor populations because their disruption resulted in normal or a slightly increased neuroD:eGFP expression (Supplemental Figure 10), suggesting that β-cells are specifically perturbed in those animals. In contrast, disruption of the 10 remaining genes resulted in a significant reduction in endocrine progenitors relative to controls (P ≤ 1 × 10⁻³; Supplemental Figure 10), suggesting that depletion of β-cell mass may be the result of broader disruption of endocrine cell type specification. Interestingly, loss of ins or pparg expression resulted in either normal or expanded neurod:eGFP expression, respectively, suggesting that the increased β-cell mass in those animals is either specific to β-cells or reflects a general expansion of endocrine cell specification, respectively (Supplemental Figure 11, A–D).

To investigate the potential developmental origins of β-cell mass deficits, we next set out to assess the contribution of the two major early populations of cells to the pancreas, those in the dorsal bud, which originate at 15 hpf and establish the principal islet by 24 hpf, and those in the ventral bud, which develop at 32 hpf and migrate to the principal islet between 48 and 54 hpf (59). We assessed dorsal bud development in embryos depleted of β-cell mass genes by quantifying the number of mCherry expressing cells at 24 hpf. We then assessed potential defects in ventral bud populations by calculating the difference in cell number between the dorsal bud mass at 24 hpf and the mature principal islet at 5 dpf. Strikingly, we found that all genes contributing to the mature principal islet deficit at 5 dpf were also necessary for proper production of dorsal bud cells (Supplemental Figure 12; P < 1 × 10⁻¹⁴), suggesting that all 25 T2D genes necessary for mature β-cell mass contributed to this initial step. In addition, the disruption of most genes resulted in either normal or significantly smaller ventral bud populations (Supplemental Figure 12, L–AD, P < 1 × 10⁻⁶), indicating either an inability to make up for the loss of dorsal bud cells or compounding that initial deficit. Three exceptions were glis3, klf14, and ankrd55, which, when disrupted, resulted in an increase in the ventral bud cells, albeit not enough to offset the initial loss of dorsal bud cells (Supplemental Figure 12, AC–AD). Importantly, the two genes whose depletion increased mature β-cell mass, ins and pparg, contributed differentially to these two early populations. Loss of pparg did not disrupt dorsal bud populations but resulted in a significant increase in ventral bud populations (Supplemental Figure 11, E–I) offering an explanation for the overall increase in β-cells by 5 dpf. In contrast, loss of ins resulted in a significant decrease in dorsal bud cells counterbalanced by a significant increase in ventral bud cells (Supplemental Figure 11, E–I).

Given that the mature size of the principal islet increases by the addition and proliferation of ventral bud cells (60), genes necessary for ventral bud expansion potentially contribute to that proliferation process during mature β-cell mass formation. To examine this possibility, we assessed a subset of the genes for their contribution to proliferation of ventral bud-derived β-cells. To do so, we ablated only the dorsally derived β-cells using a transgenic line, Tg(insa:nsfB-mCherry), allowing for the ablation of β-cells by treatment with a prodrug, metronidazole (4). Removal of the prodrug allows for the recovery and regeneration of β-cells. We ablated dorsal bud cells by metronidazole treatment from 20 hpf until 32 hpf and then allowed a period of recovery until dorsal-ventral bud fusion at 54 hpf (59). We then quantified the number of proliferative mCherry-expressing ventral bud-derived cells by whole-mount immunofluorescent staining using an antibody against the proliferation marker phosphohistone-H3. Control embryos exhibited, on average, 1.4 proliferating ventral bud β-cells (Supplemental Figure 13, A–A″ and E). We examined three genes whose depletion resulted in either an increase of ventral bud cells (glis3), no change (slc30a8), or a decrease (notch2). Consistent with their effects on ventral bud cells, we observed a slight proliferation increase, albeit not significant, in glis3 morphants (Supplemental Figure 13, B–B″ and E), no significant change in proliferative capacity for slc30a8 morphants (Supplemental Figure 13, C–C″ and E), and a significant decrease in ventral bud-derived β-cell proliferative capacity in notch2 morphants (Supplemental Figure 13, D and E). The proliferative cells identified could be derived from a number of sources in addition to the ventral-derived β-cells. These β-cells might also be derived by de novo neogenesis from established precursors in the ductal epithelium or could be the result of transdifferentiation from other endocrine or exocrine populations or extrapancreatic tissues (reviewed in reference 61). These alternative scenarios could be plausible because studies in pdx1 mophant zebrafish suggest that the late endocrine cells emerge predominantly from a postmitotic cell type (60).

Functional mapping of multigene loci associated with T2D

The corroboration of known β-cell mass genes supported the relevance of our system to human functional genetics. We therefore reasoned that this functional characterization approach might offer a novel strategy to examine the putative contribution of multiple candidate genes at multigene GWA study loci. To examine this possibility, we selected eight loci at which multiple genes of unknown causality were found surrounding associated variants and that harbored multiple genes for which we could identify zebrafish orthologs (6, 8–10, 18, 21, 22, 25). For each locus we depleted the expression of orthologs for at least two genes. At one locus, found at chr7p21.2, neither of the two orthologs tested (dgkb and tmem195) were involved in production of β-cell mass (Figure 1P and Supplemental Figure 3). At another locus, found at chr11p15.5, two genes, th and ins, were targeted. Suppression of th did not have any impact on β-cell mass, but reduced ins resulted in an increase (Figure 1 and Supplemental Figures 3 and 9). Single genes at five of the other six loci were necessary for β-cell mass. These included genes known to play a role in β-cell function, pax4 and kcnj11 (Figure 3, A and B) as well as three novel genes, ide, camk1d, and c2cd4a (Figure 3, C–E). C2CD4B at chr15p22.2 did not have an obvious zebrafish ortholog and therefore could not be tested. Two genes, fitm2 and hnf4a at 20p13.12 locus tagged by rs6017317 (9), played a significant role in β-cell mass, potentially suggesting that the β-cell mass effects of that locus are driven by both genes. However, the third gene, r3hdml, played no such role (Figures 1D, 2C, and 3F). These findings not only implicate seven genes in the production of β-cells but also likely exclude roles in β-cell mass for at least six other genes at the loci tested.

Figure 3.

Functional fine-mapping at multigene loci identifies and excludes candidate genes involved in β-cell mass. Functional testing of multiple genes at individual loci was used to assign or exclude roles in β-cell mass. Loci were based on published associations of known SNPs at each locus (green). A–E, Individual gene orthologs (red) were identified as being necessary for β-cell mass area (graphs shown for genes at each locus), and other genes (black) were excluded based on no significant difference from control. *, c2cd4b was not tested at the chr15q22.2 locus because no ortholog was identified in zebrafish. F, Two gene orthologs, fitm2 and hnf4a, were implicated at the chr20q13.12 locus, although r3hdml was excluded.

T2D genes necessary for production of β-cells contribute to maintenance

T2D is an adult-onset disease, so it is unlikely that deficiencies in embryonic production of β-cells result in a dysfunctional β-cell mass early in life. Given this and the fact that many processes necessary for the developmental specification of β-cells also contribute to the adaptive capacity of β-cell mass, we reasoned that the genes identified in our screen may also play a role in adaptive capacity. To test this possibility, we first examined the ability of β-cells in morpholino-injected insa:mCherry larvae to expand in response to chronic high-glucose conditions, a treatment that results in increased β-cell number in zebrafish (5). Transgenic embryos depleted for each of the 25 β-cell mass genes were cultured in medium supplemented with 40 mM glucose. Consistent with previous reports (5), this produced an average of 39.3 β-cells by 5 dpf in control larvae, an expansion of 17.8% over those cultured in embryo medium (Figure 4, A, B, and I). By comparison, a majority of genes (n = 17) that we identified as playing a role in β-cell specification were not necessary for glucose-induced expansion of β-cell mass because their loss resulted in similar or increased expansion relative to control medium (Figure 4, A–I). A subset of 9 genes, however, were required for this aspect of β-cell adaptive capacity. These included several genes involved in β-cell neogenesis, such as pax4 and glis3 (62, 63), potentially confirming the importance of neogenesis in the β-cell mass glucose response (5). As expected, depletion of gck, the primary glucose sensor in β-cells, also resulted in decreased glucose-induced expansion.

Figure 4.

Assessment of glucose induced β-cell mass expansion in larvae depleted of β-cell mass genes. A–I, Morpholino-injected insa:mCherry larvae cultured in either control embryo medium or medium supplemented with 40 mM glucose until 5 dpf. Quantification of β-cells (arrowhead) was assessed and calculated for each of the β-cell mass genes as a rate of expansion based on the percentage change in β-cell number relative to control (I). *, P ≤ 1 × 10⁻⁵, two tailed t test. ns, not significant.

These observations suggest that, although the baseline reserve of β-cells is impaired in animals deficient for β-cell mass genes, their ability to sense glucose and produce more β-cells in response is likely generally intact. In light of this, we reasoned that an alternate mechanism of eventual decline in β-cell function might occur as a result of inability to recuperate exhausted β-cell mass. To test this possibility, we again turned to the transgenic line, allowing for specific ablation of β-cells by the addition of metronidazole to culture media and regeneration upon its removal (4). After complete ablation of β-cells and recovery, we found that all but seven genes (ankrd55, hnf1a, kcnj11, klf14, pdx1, slc30a8, and zfand3) were necessary for the regeneration of β-cells at a rate comparable with controls (Figure 5, A–I), suggesting that most genes (n = 18) were significantly necessary for the ability to generate β-cells after loss. We did not observe any morphants in which regeneration was enhanced relative to controls (Figure 5I). The exception to this was in larvae depleted of ins expression, which exhibited a significantly higher rate of regeneration, relative to controls (Supplemental Figure 11, J–M and P; P ≤ 1 × 10⁻⁶), although they expanded in response to glucose at a significantly lower rate (Supplemental Figure 11, Q–T and W). In contrast, larvae depleted of pparg expression exhibited normal regeneration of β-cells and glucose-induced expansion (Supplemental Figure 11, N, O, and U–W), suggesting no deficiencies in this aspect of β-cell adaptive capacity. In total, 14 of the 25 genes necessary for proper β-cell mass were also necessary for either glucose-induced expansion or regeneration (Supplemental Figure 14), suggesting that deficits in β-cell mass production are in large part also associated with deficits in the responsiveness of β-cell mass to adaptive requirements.

Figure 5.

Deficiencies in β-cell regeneration suggest an inability to recuperate lost β-cell mass. A–I, β-Cell regeneration was observed at 128 hpf after metronidazole-induced ablation and recovery in Tg(insa:nsfB-mCherry) embryos, which produce β-cells normally in the absence of metronidazole (untreated). The number of β-cells regenerated after ablation was quantified and a regeneration rate was calculated for each morpholino by the number of β-cells produced per day of recovery. I, Regeneration rates shown as a percentage change relative to control. *, P = .01, **, P = .002, ***, P ≤ 1 × 10⁻⁵, two tailed t test. ns, not significant.

Maintenance of glucose homeostasis in larvae deficient in β-cell mass genes

We next set out to test whether the loss of β-cell mass as a result of the loss of gene expression would also result in a reduced ability to maintain glucose levels. To do so, we treated 5 dpf larvae to either a CD or HGD (40 mM glucose infused diet) for 48 hours followed by quantification of absolute glucose levels in homogenates of pooled zebrafish larvae. Under the CD conditions, we identified nine genes that were also necessary for the maintenance of glucose levels similar to controls (Supplemental Figure 15A, P ≤ .0001). The addition of the HGD expanded this to 11 genes that were necessary for the proper control of glucose levels (Supplemental Figure 15B). These include genes previously associated with elevated glucose levels in other mutant models such as gckr, hnf1a, pdx1, wfs1, and pax4 (8, 47, 48, 64, 65) as well as the identification of novel genes for their role in this functional mechanism, such as klhdc5 and fitm2. These findings suggest that although many genes contributing to proper production of β-cell mass are also necessary for the control of glucose homeostasis, there is not absolute concordance between β-cell mass and the regulation of total systemic glucose either in control conditions or under exposure to high-glucose conditions. The increased number of genes that were necessary for the regulation of glucose homeostasis upon the addition of high glucose to the larval diet, however, suggests that increased demand may reveal glucose regulation deficits that would not be evident otherwise.

Diabetic carriers of risk alleles in β-cell mass genes develop disease with fewer additional risk factors

The inability to maintain β-cell mass over time has been reported in T2D, and evidence suggests a correlation between the loss of β-cell mass and loss of β-cell function (2, 66). We therefore hypothesized that diabetic individuals who are more genetically susceptible to β-cell dysfunction may be more susceptible to the development of disease, even with fewer additional risk factors, such as advanced age or obesity. To examine this possibility and the relevance of our findings in zebrafish to human T2D, we examined diabetic individuals in the OOA (67) for the presence of risk alleles in T2D-associated genes. In diabetic individuals for whom genotype, age of onset, and BMI data were available (n = 45), we identified 60 previously reported risk variants (Table 2). Based on our functional assays we classified each variant as either necessary for β-cell mass (23 variants; Table 2) or not (37 variants; Table 2), allowing us to assign a count of risk alleles in β-cell mass genes and in non-β-cell mass genes. We observed a significant association between a higher number of risk alleles in β-cell mass genes and a lower age of diabetes onset (P = .03; Figure 6A). In contrast, no significant association was observed between age of diabetes onset and the number of risk alleles in non-β-cell mass genes (P = .24; Figure 6B). Moreover, T2D patients who were young (age of onset ≤ 45 y) and lean (BMI ≤ 25 kg/m²) exhibited a 1.4-fold higher average risk allele count in β-cell mass genes, representing a significant enrichment after normalization of the total number of risk alleles per SNP (P ≤ .01; Figure 6C). Persons with diabetes who were older (age of onset ≥ 60 y) and obese (BMI ≥ 30 kg/m²) did not exhibit a similar enrichment (Figure 6D). To preclude the possible contribution of undiagnosed forms of young onset diabetes to these data, we excluded variants in known MODY genes from the analyzed variants. This did not diminish the significance of the association between increased number of risk alleles in β-cell mass genes and younger age of onset (P = .05; Supplemental Figure 16). These findings suggest a more significant role for impaired function of β-cell mass genes in individuals who develop T2D with fewer additional risk factors.

Table 2.

T2D Risk Alleles Analyzed in Amish Diabetic Individuals

Gene	SNP(s)	Gene	SNP(s)
ADAMTS9	rs4607103a	IGF2BP2	rs4402960; rs1470579
ADCY5	rs11708067; rs2191349	INS	rs10770141
ANK1	rs516946	IRS1	rs7578326
ANKRD55	rs459193a	JAZF1	rs864745
BCAR1	rs7202877	KCNQ1	rs231362
BCL11A	rs243021	KLF14	rs972283a
BCL2	rs12454712	KLHDC5	rs10842994a
C2CD4A	rs7172432a	MC4R	rs12970134
CAMK1D	rs12779790a	MTNR1B	rs1387153
CDKAL1	rs7754840a	NOTCH2	rs10923931a
CDKN2A	rs564398	PAX4	rs6467136a
CENTD2	rs1552224	PEPD	rs3786897a
CILP2	rs10401969	PPARG	rs1801282
FITM2	rs6017317a	PRC1	rs8042680
FTO	rs8050136	PROX1	rs340874
GATAD2A	rs3794991	PSMD6	rs831571
GCK	rs4607517a	SLC30A8	rs13266634a
GCKR	rs780094a	SREBF1	rs4925115a
GIPR	rs10423928	TCF7L2	rs7903146
GLIS3	rs7041847a	TLE1	rs2796441
GRB14	rs3923113	THADA	rs7578597a
HMG20A	rs7178572	TP53INP1	rs896854
HMGA2	rs1531343	WFS1	rs1801214a
HNF1A	rs7957197a	VPS13C	rs17271305
HNF1B	rs4430796a	ZFAND3	rs9470794a
HNF4A	rs4812829a; rs6017317a	ZFAND6	rs11634397
IDE	rs1111875a	ZMIZ1	rs12571751

To examine the relevance of our functional findings in zebrafish to human T2D, we investigated the presence of T2D-associated risk alleles identified by GWA study. Diabetic carriers of risk alleles were identified in the OOA and used for analyses of additional risk factors (age and obesity).

^aVariant in β-cell mass gene.

Figure 6.

Analyses of diabetic carriers of risk alleles in β-cell mass genes indicate onset of T2D with less contribution from age and obesity. A, Diabetic individuals from the OOA cohort were assessed for the number of risk alleles in β-cell mass genes (x-axis) relative to age of T2D onset (y-axis). A higher number of risk alleles was associated with a lower age of onset (P = .03). B, Assessment of risk alleles in non-β-cell mass genes relative to age of T2D onset indicated no significant association (P = .24). C and D, Persons with diabetes were classified as young (<45 y) and lean (BMI < 25 kg/m²) or old (>60 y), and obese (BMI > 30 kg/m²); the average number of risk alleles in either β-cell mass genes or non-β-cell mass genes was calculated, revealing a significant enrichment of alleles in β-cell mass genes in young, lean persons with diabetes.

Discussion

The identification of many genomic loci associated with risk for T2D has shed light onto the genetic complexity of the disease. Less insight has been provided, however, into the mechanisms by which specific genes mediate T2D susceptibility. This has been partly due to the difficulty of interpreting the functional relevance of genes at T2D-associated loci with respect to disease biology. In this study, we carried out a systematic in vivo functional screen of genes at T2D-associated loci with the goal of identifying candidate genes that may play a role in disease etiology. Our observations suggest that we have identified a subset of genes that may be functionally relevant to the disease at least in part by the capacity to produce and maintain β-cell mass (Supplemental Table 2). We capitalized on the unique utility of the zebrafish system for rapid functional investigation owing to the short development time of transparent larvae, visibility of β-cells in vivo via transgene expression (4), and ease of large-scale suppression of gene expression. These features allow for the experimental characterization of T2D gene function in a manner that is more feasible than in mammalian systems and for a phenotype that is difficult to assess in humans. This system allowed for a robust approach, which entailed physiologically relevant in vivo interpretation of human GWA study data using a vertebrate system followed by validation of findings in human. The relevance of our approach was not only supported by data from diabetic patients suggesting susceptibility to β-cell loss with fewer additional risk factor, but also by two additional lines of evidence. The first is our ability to corroborate previous findings in mammalian systems for several genes. Second, T2D-associated MODY and NDM genes having known roles in β-cell production were identified in our screen. Taken together, our findings provide evidence for the utility of systematic large-scale investigation of gene function in zebrafish and the relevance of this model to human T2D phenotypes.

A surprising finding from multiple GWA studies for T2D susceptibility has been the identification of genes important for β-cell function in association with the disease. However, beyond those genes already known to play a role in β-cell function, the extent to which genetic modulation of β-cells is related to T2D susceptibility is unclear. Our findings suggest that the regulation of the ability to produce and maintain β-cells may be important for certain genotypes. This, however, raises an important distinction between β-cell mass and β-cell function. This study assessed only the former. Recent studies from other groups have provided support for the contribution of genotype-specific regulation of insulin secretion by certain T2D-associated variants (68, 69), but the relationship to β-cell mass is unclear. This is also consistent with our observations that ins or abcc8 were not significantly necessary for proper production of β-cells. Given the known role of these genes specifically in the production and release of insulin from β-cells, these findings perhaps suggest distinct differences between regulation of β-cell number and function. In contrast, reduced kcnj11 resulted in a significant reduction in β-cell mass, potentially suggesting a more central role of kcnj11 in β-cell survival. This is supported by reports in mice that complete loss of Kcnj11 resulted in reduced numbers of β-cells with age, potentially due to apoptosis (70), a finding not observed in Abcc8 knockout mice (71).

It is also important to note the possibility that changes in insulin secretion contributes to regulation of systemic glucose homeostasis, which indirectly affects β-cell mass. Moreover, although many of our findings corroborate previous reports in zebrafish and other models, our observations for one gene, pax4, contrasted with prior studies. In the mouse, Pax4 is essential for β- and δ-cell lineage differentiation and repression of the α-cell determining transcription factor Arx (47, 72). Similarly, our observations in zebrafish indicated reduction of β-cells at 24 hpf in pax4 morphants (Figures 1 and 2). However, prior morpholino studies targeting pax4 in zebrafish suggested no effect on early β-cells via an in situ hybridization analysis of insulin expression (73). This discrepancy could be due to differences in morpholino potency, the developmental stage assessed, the zebrafish stocks used, or the differences in analysis via in situ hybridization vs quantification of mCherry-expressing cells.

T2D is a complex disease state that is mediated by complex signaling in many different cell types. Therefore, we cannot rule out the involvement of β-cell mass genes in other tissues or even β-cell function. We also cannot make statements about the involvement of non-β-cell mass genes in other aspects of disease etiology. This is particularly poignant with respect to the use of our functional investigation to rule out candidates at multigene loci. Our agnostic functional mapping approach offers a novel strategy to address a major challenge in post-GWA study genetics. Our findings represent a proof of principle to indicate the possibility of assigning physiologically relevant function to genes of ambiguous causal nature at GWA study-identified loci. These observations do not, however, preclude possible roles for these genes in other aspects of T2D etiology or even pancreatic function. For example, the chr10q23.33 locus encompasses two genes, HHEX and IDE, with roles that are potentially relevant to pancreatic function (Figure 3C [6, 22]). Our assay, however, implicated only ide in β-cell mass (Figure 1P and Supplemental Figure 3). This was not the case for hhex, although it was indeed necessary for exocrine pancreas specification, consistent with reports in mice (Figure 1P and Supplemental Figure 3 [74]). Although this strategy provides evidence implicating the identified genes in β-cell mass specifically, we did not carry out a comprehensive analysis of all loci, and this present study does not rule out the potential contribution of genes to other features of T2D, including those in other aspects of β-cell function. To fully clarify this latter distinction, it will be necessary to rule out the involvement of non-β-cell mass genes in aspects of β-cell function beyond β-cell number.

To initially identify genes involved in the production of β-cells, our investigation focused on the developmental stages. Although T2D is an adult-onset disease and impaired β-cell function is not evident early in life, reduced or delayed capacity to produce β-cells initially may lead to a reduced β-cell reserve. Evidence from various studies support the possibility that deficits in islet development may predispose to T2D later in life (75). It is not clear, however, whether individuals who may be susceptible to T2D are so because of a limited β-cell mass with which to respond to increased demand for insulin. Although β-cell number was significantly impaired, we observed an intact adaptive expansion for 17 of the 25 genes. The exceptions, such as pax4, gck, and pdx1, have been implicated in β-cell neogenesis, glucose sensing, or both, respectively (47, 64). The normal or enhanced production of β-cells that we observed under chronic hyperglycemic condition may be consistent with islet hyperplasia and hyperinsulinemia that initially accompanies insulin resistance prior to β-cell failure. In contrast, our results indicated that 18 of the 25 β-cell mass genes were necessary for the regeneration of lost β-cells. This might suggest that initial embryonic production is an indicator of the capacity to generate more β-cells later in life. This is consistent with reports that have implicated β-cell neogenesis in regeneration (4). Importantly, there were no genes for which their reduced expression enhanced regeneration relative to controls (Figure 5I). This suggests that it is unlikely that individuals who are genetically susceptible to reduced β-cell capacity are better able to recover β-cells once they are lost. These observations potentially offer an explanation for the inability to maintain β-cell function in individuals with compromised function of these genes.

The approach described in this study offers a strategy to examine the contribution of T2D genes to other aspects of disease, including insulin sensitivity, inflammation, obesity, and other factors. This raises the possibility of using systematic in vivo experimental analyses to decipher the genetic architecture of T2D. T2D is a heterogeneous disease, including vastly different susceptibility, severity, and prognosis across different individuals. Understanding the discrete role of individual genes and variants in each aspect of T2D onset and prognosis will allow for the delineation of each individual's disease and indicate the relationship of genotype to differential susceptibility and prognosis. Evidence suggests that the number of risk alleles in an individual may not be an effective predictor of the likelihood of disease onset (76). However, our findings support the possibility that the nature of the risk alleles carried by an individual may dictate the nature of their susceptibility by differential modulation of distinct aspects of glucose regulation.