Genetics of New-Onset Diabetes after Transplantation

Abstract

New-onset diabetes after transplantation is a common complication that reduces recipient survival. Research in renal transplant recipients has suggested that pancreatic β-cell dysfunction, as opposed to insulin resistance, may be the key pathologic process. In this study, clinical and genetic factors associated with new-onset diabetes after transplantation were identified in a white population. A joint analysis approach, with an initial genome-wide association study in a subset of cases followed by de novo genotyping in the complete case cohort, was implemented to identify single-nucleotide polymorphisms (SNPs) associated with the development of new-onset diabetes after transplantation. Clinical variables associated with the development of diabetes after renal transplantation included older recipient age, female sex, and percentage weight gain within 12 months of transplantation. The genome-wide association study identified 26 SNPs associated with new-onset diabetes after transplantation; this association was validated for eight SNPs (rs10484821, rs7533125, rs2861484, rs11580170, rs2020902, rs1836882, rs198372, and rs4394754) by de novo genotyping. These associations remained significant after multivariate adjustment for clinical variables. Seven of these SNPs are associated with genes implicated in β-cell apoptosis. These results corroborate recent clinical evidence implicating β-cell dysfunction in the pathophysiology of new-onset diabetes after transplantation and support the pursuit of therapeutic strategies to protect β cells in the post-transplant period.

One-year graft survival after renal transplantation is now excellent, exceeding 93% for organs donated after brain death and 96% for those from living donors.^1–3 Technical advancements in surgery, improved understanding of immunology, and innovative developments in pharmacology have altered the landscape of renal transplantation. The goal of preventing early graft loss has largely been achieved and arguably the greatest challenge now is the avoidance of late graft failure. Although there has been a considerable improvement in 1-year renal transplant survival, the rate of graft attrition after the first year remains frustratingly constant.^2,4

New-onset diabetes after transplantation (NODAT) is a common and serious disorder that curtails recipient survival.^5–7 NODAT is associated with cardiovascular complications^8–11 and develops in 2%–50%¹² of renal transplant recipients. Approximately 50% of recipients with NODAT require insulin therapy.^{6–8,13–15} A number of clinical variables have been associated with NODAT, including black ethnicity, older recipient age, female sex, obesity, immunosuppression, and viral infections.^{5,6,8,13,16,17}

Until recently, the pathophysiology of NODAT was considered to be analogous to type 2 diabetes mellitus. Renal transplant recipients have increased insulin resistance compared with transplant-naïve persons with normal renal function.¹⁸ In a nondiabetic renal transplant population, the main determinants of insulin resistance are obesity and corticosteroid therapy.¹⁹ Insulin resistance improves in renal transplant recipients after successful transplantation^20,21 and recipients have enhanced insulin sensitivity compared with dialysis patients.²² At 1 year, there is no significant difference in insulin resistance between renal transplant recipients with NODAT and those with normal glucose tolerance.^18,23 Furthermore, insulin resistance indices before transplantation and in the early post-transplant period do not predict NODAT development.¹¹

Pancreatic β-cell dysfunction may prove to be the main pathologic contributor to NODAT. In glucose clamp studies, a deficit in insulin secretion was common to renal transplant recipients with NODAT.^18,20,21,24 There are a number of possible mechanisms for β-cell toxicity in renal transplantation, including hyperglycemia,²⁵ elevated free fatty acids,²⁶ and the effect of immunosuppressive medication.²⁷ A recent proof-of-concept clinical trial demonstrated that aggressive management of post-transplant hyperglycemia with insulin significantly reduced the 1-year incidence of NODAT.²⁸ This provides further evidence that post-transplant hyperglycemia plays a key role in NODAT development.

This study investigates clinical and genetic factors associated with NODAT in a relatively large, white renal transplant population. Clinical variables were identified in a carefully phenotyped, ethnically homogeneous cohort. Initial exploratory analysis was conducted via a genome-wide association study (GWAS) in a subgroup of NODAT cases patients and controls to identify genetic variants associated with NODAT. De novo genotyping was then performed in a larger cohort of NODAT patients and controls to validate the findings.

Results

Patient Cohort

There were 707 first, deceased donor kidney transplants performed at Belfast City Hospital (Belfast, UK) between May 1986 and May 2005. Over 99% of both recipients and donors were white; genetic analysis was restricted to those of recorded white ancestry. The average age of recipients was 37 years (range, 2–77 years) and the average age of donors was 42 years (range, 1–75 years). There were 439 male recipients (62.1%) and 428 male donors (59.1%). All recipients had their primary renal diagnosis classified according to the European Dialysis and Transplantation Association coding system. Diagnoses were categorized as glomerular disease (21%), pyelonephritis/interstitial nephritis (20%), autosomal dominant polycystic kidney disease (15%), diabetic nephropathy (9%), other specified miscellaneous etiologies (22%), and CKD not defined (13%). The median follow-up time was 12.2 years (range, 0–26.0 years).

There were changes to the routine post-transplant immunosuppression during the study period. Before 1989, all recipients received dual therapy with prednisolone and azathioprine. Subsequently, calcineurin inhibitor (CNI)–based maintenance therapy was introduced. Mycophenolate mofetil became available in 1998 and from this time, approximately 25% of patients had CNI-free maintenance regimens. All patients received prednisolone for at least 1 year after transplantation.

In our study, the NODAT clinical phenotype was strictly defined as a new requirement for oral hypoglycemic agents or insulin for management of hyperglycemia after transplantation. NODAT status was available for 605 recipients; 58 of 605 recipients (9.6%) developed NODAT during the follow-up period.

Clinical Analyses

At 12 months, 529 adult renal transplant recipients had a functioning graft; 57 of these patients developed NODAT during the follow-up period.

The median graft survival was 10.4 years. During the follow-up period, there were 162 cases of death-censored graft failure. A further 159 recipients died with a functioning graft. Biopsy-proven acute rejection (P<0.001), donor age (P=0.001), and CNI usage (P=0.004) were associated with death-censored graft failure after 1 year. In this cohort, HLA mismatching did not significantly influence death-censored graft survival (P=0.85). This is likely to reflect the policy of favorable HLA matching at our center.^29,30

NODAT

Comparison of clinical variables between NODAT cases and controls is shown in Table 1. The median time to NODAT was 100 months (interquartile range, 113 months). Clinical variables associated with NODAT were age, female sex, and percentage weight change in the first year after transplantation. There was no association between CNI therapy and NODAT. In multivariate analysis, the associations with recipient age, female sex, and percentage weight change in the first year remained significant. The association between weight at transplantation and NODAT was also significant after adjustment for confounding (Table 2).

Table 1.

Comparison of clinical variables between NODAT cases and controls

Variable	NODAT (n=57)	Control (n=370)	P Value
Recipient age (yr)	49.1±13.2	42.4±14.0	0.001
Recipient sex
Male	30 (53)	249 (67)
Female	27 (47)	121 (33)	0.04
Donor age (yr)	37.1±14.8	37.6±16.1	0.80
Donor sex
Male	33 (58)	213 (58)
Female	24 (42)	157 (42)	1.0
Primary renal disease
GN	7 (12)	89 (24)	0.07
Interstitial/pyelonephritis	15 (26)	79 (21)	0.50
ADPKD	16 (28)	60 (16)	0.05
Other	10 (18)	84 (23)	0.40
Unknown	9 (16)	58 (16)	0.70
Decade of transplantation
1986–1995	24 (42)	174 (47)
1996–2005	33 (58)	196 (53)	0.60
HLA mismatch (n)	2.1±1.1	2.2±1.1	0.60
Acute rejection	9 (16)	48 (13)	0.90
Immunosuppression
Calcineurin inhibitors	41 (72)	299 (81)	0.40
Azathioprine	35 (61)	224 (61)	1.0
Mycophenolate mofetil	28 (49)	185 (50)	0.90
mTOR inhibitor	7 (12)	34 (9)	0.60
Weight at transplant (kg)	72.2±13.3	68.5±14.3	0.07
Weight change at 1 yra
Weight loss	9 (17)	71 (21)
Weight gain<10%	11 (20)	122 (36)
Weight gain>10%	34 (63)	149 (43)	0.04b

Data are expressed as n (%) or mean±SD. ADPKD, autosomal dominant polycystic kidney disease; mTOR, mammalian target of rapamycin.

^aValues for weight change at 1 year were available for 396 of 427 recipients.

^bLinear by linear.

Table 2.

Multivariate logistic regression analysis for NODAT

Variable	OR (95% Confidence Interval)	P Value
Recipient agea	1.4 (1.1 to 1.8)	0.003
Recipient sex	2.2 (1.2 to 4.3)	0.02
Weight at transplantation	1.03 (1.01 to 1.05)	0.01
Weight change in first year	2.0 (1.3 to 3.1)	0.002

^aPer decade.

Genetic Analyses

GWAS

There were 561,233 single-nucleotide polymorphisms (SNPs) genotyped in 256 individuals consisting of 26 NODAT patients and 230 controls. After quality control, the average genotyping rate was >99%. Twenty-six SNPs were provisionally associated with NODAT (P<10⁻⁵); these SNPs were taken forward for second-stage genotyping (Supplemental Table 1).

Reported NODAT SNPs

A literature review identified 27 SNPs previously reported to be associated with NODAT with an additional 10 SNPs investigated for association (Table 3).^31–45 Nine of these SNPs were genotyped in the GWAS and proxy SNPs were identified for 17 SNPs that were not directly genotyped (r²>0.80). No association was identified between these previously reported SNPs and NODAT in our population (Supplemental Table 2). Proxy SNPs were unavailable for the remaining 11 reported SNPs. Only one of these SNPs (rs1050450) had previously been associated with NODAT in a white population.

Table 3.

Reported SNPs investigated for association with NODAT

Gene	SNP	Population	Reference
ATF6	rs2340721a	White	Fougeray et al.31
CAPN10	rs5030952	White	Kurzawski et al.32
CCL5	rs2107538	Korean	Jeong et al.33
	rs2280789	Korean	Jeong et al.33
	rs3817655	Korean	Jeong et al.33
ENPP1	rs1044498a	Hispanic	Yang et al34
GPX1	rs1050450	White	Dutkiewicz et al.35
HNF1A	rs1169288a	Hispanic	Yang et al.34
	rs1800574a	Hispanic	Yang et al.34
HNF4A	rs2144908	Hispanic	Yang et al.34
	rs1884614	Hispanic	Yang et al.34
	rs1800961a	Hispanic	Yang et al.34
IFNγ	rs2430561	White	Babel et al.36
IL-17E	rs1124053	Korean	Kim et al.37
IL-17RA	rs2229151	Korean	Kim et al.37
	rs4819554	Korean	Kim et al.37
IL-17RB	rs1043261	Korean	Kim et al.37
	rs1025689	Korean	Kim et al.37
IL-1B	rs3136558	Korean	Kim et al.37
IL-2	rs2069762	Korean	Kim et al.37
	rs2069763	Korean	Kim et al.37
IL-4	rs2243250	Korean	Kim et al.37
	rs2070874	Korean	Kim et al.37
IL-6	rs1800795	White	Bamoulid et al.38
IL-7R	rs1494558	Korean	Kim et al.37
	rs2172749	Korean	Kim et al.37
IRS1	rs1801278	Hispanic	Yang et al.34
KCNJ11	rs5219	Spanish	Tavira et al.39
		Hispanic	Yang et al.34
KCNQ1	rs2237895	Spanish	Tavira et al.40
NFATC4	rs10141896	Hispanic	Chen et al.41
PPARG	rs1801282a	Hispanic	Yang et al.34
PPARGC1	rs8192678a	Hispanic	Yang et al.34
SUR1	rs1799854a	Hispanic	Yang et al.34
	rs1801261a	Hispanic	Yang et al.34
SLC30A8	rs13266634	Korean	Kang et al.42
TCF7L2	rs7903146	White	Kurzawski et al.43
	rs12255372a	White	Ghisdal et al.44
		Korean	Kang et al.45
		Hispanica	Yang et al.34
		Hispanic	Yang et al.34

^aNonsignificant association.

Linkage disequilibrium was assessed for all SNPs (Supplemental Figures 1–50) and is displayed for the top-ranked SNPs on chromosomes 1 and 6 (Figures 1 and 2).

Figure 1.

Regional association plots illustrating linkage disequilibrium between SNPs associated with NODAT. (A) On chromosome 6. (B) On chromosome 1.

Figure 2.

Haplotype block on chromosome 6.

Second-Stage Genotyping

De novo genotyping was undertaken in all NODAT patients and 383 controls. The top-ranked SNPs associated with NODAT in the GWAS (n=26), previously published NODAT SNPs investigated in the GWAS (n=26), and the SNP with a reported association in a white population for which a proxy was unavailable (n=1) were genotyped. Sequenom iPLEX (Hamburg, Germany) technology was used to genotype 44 SNPs; TaqMan (Applied Biosystems, Warrington, UK) was used to genotype six SNPs (Supplemental Figures 51–100, Supplemental Table 3). It was not possible to genotype three SNPs due to problems with primer design (Supplemental Table 4). The average genotyping success rate was >99% and no SNP demonstrated deviation from the Hardy–Weinberg equilibrium (HWE).

In logistic regression analysis, eight SNPs were associated with NODAT (P≤0.001). After adjustment for the clinical variables known to be associated with NODAT (Table 2), adjusted P values, adjusted odds ratios (ORs), and 95% confidence intervals for ORs were calculated (Table 4). Using Haploview,⁴⁶ a single haplotype block was identified between adjacent SNPs on chromosome 6 (Figure 2). Haplotype analysis did not increase the strength of the association between these SNPs and NODAT (P=0.01).

Table 4.

Association results for de novo genotyped SNPs with NODAT

SNP	Gene	Location	MAF	P Value	Padj	ORadj (95% Confidence Interval)
rs10484821	ATP5F1P6	6:139868910	0.14	1.5×10−7	0.000002	3.5 (2.1 to 5.8)
rs7533125	DNAJC16	1:15883744	0.23	0.001	0.001	2.4 (1.5 to 3.6)
rs2861484	CELA2B	1:15812665	0.20	0.001	0.0002	2.4 (1.5 to 3.7)
rs11580170	AGMAT	1:15909744	0.26	<0.00	0.0002	2.2 (1.4 to 3.4)
rs2020902	CASP9	1:15834360	0.18	<0.001	0.0003	2.3 (1.5 to 3.6)
rs1836882	NOX4	11:89232161	0.12	0.001	0.001	2.7 (1.5 to 4.8)
rs198372	NPPA	1:11909514	0.13	<0.001	0.001	2.5 (1.5 to 4.2)
rs4394754	INPP5A	10:134343062	0.28	0.001	0.001	2.1 (1.4 to 3.2)
rs7145618	PPP2R5C	14:102329098	0.08	0.01	0.002	3.1 (1.5 to 6.4)
rs17722392	KIDINS220	2:8940154	0.06	0.02	0.003	3.2 (1.5 to 6.9)
rs2265919	SHPRH	6:146221753	0.42	0.002	0.003	2.0 (1.3 to 3.1)
rs1783606	SHANK2	11:70576651	0.12	0.02	0.003	2.4 (1.3 to 4.1)
rs2265477	SHPRH	6:146212338	0.44	0.01	0.01	1.8 (1.2 to 2.8)
rs2069763a	IL-2	4:123377482	0.28	0.01	0.01	0.48 (0.3 to 0.8)
rs6903252	Intergenic	6:145922777	0.47	0.01	0.01	1.8 (1.2 to 2.8)
rs2265917	SHPRH	6:146212285	0.42	0.01	0.01	1.8 (1.2 to 2.8)
rs10899444	SHANK2	11:70606500	0.12	0.05	0.01	2.2 (1.2 to 3.9)
rs16936667	PRDM14	8:70975726	0.14	0.02	0.01	2.2 (1.2 to 3.8)
rs341497	DIAPH3	13:60429001	0.06	0.004	0.01	2.6 (1.3 to 5.2)
rs1016429	GRIN3A	9:104402364	0.07	0.01	0.01	2.5 (1.2 to 5.0)
rs10117679	GRIN3A	9:104378479	0.04	0.01	0.02	2.8 (1.2 to 6.4)
rs1871184	ITGA1	5:52234323	0.15	0.01	0.02	1.9 (1.1 to 3.1)
rs3212574	ITGA2	5:52366779	0.22	0.01	0.02	1.7 (1.1 to 2.7)
rs2240747	ZNRF4	19:5456930	0.18	0.01	0.03	1.8 (1.1 to 3.1)
rs6793265	ITPR1	3:4735533	0.12	0.08	0.07	1.7 (0.9 to 3.0)
rs2070874a	IL-4	5:132009710	0.13	0.10	0.10	0.56 (0.3 to 1.1)
rs17657199	NDST1	5:149950246	0.05	0.10	0.10	2.2 (0.8 to 5.6)
rs2069762a	IL-2	4:123377980	0.33	0.10	0.10	1.4 (0.9 to 2.1)
rs7903146a	TCF7L2	10:114758349	0.31	0.20	0.20	1.4 (0.9 to 2.1)
rs1043261a	IL-17RB	3:53899276	0.07	0.20	0.20	0.54 (0.2 to 1.4)
rs2280789a	CCL5	17:34207003	0.11	0.30	0.20	0.61 (0.3 to 1.3)
rs2172749a	IL-7R	5:35855264	0.32	0.30	0.30	0.77 (0.5 to 1.2)
rs1494558a	IL-7R	5:35861068	0.32	0.40	0.30	0.77 (0.5 to 1.2)
rs1800795a	IL-6	7:22766645	0.39	0.60	0.30	1.3 (0.8 to 2.0)
rs12255372a	TCF7L2	10:114808902	0.30	0.30	0.30	1.3 (0.8 to 1.9)
rs5219a	KCNJ11	11:17409572	0.35	0.70	0.30	0.80 (0.5 to 1.3)
rs2107538a	CCL5	17:34207780	0.17	0.40	0.40	0.77 (0.4 to 1.4)
rs1801282a	PPARG	3:12393125	0.13	0.50	0.40	0.75 (0.4 to 1.5)
rs3817655a	CCL5	17:34199641	0.17	0.40	0.40	0.79 (0.4 to 1.4)
rs1799854	ABCC8	11:17448704	0.43	0.30	0.50	1.2 (0.8 to 1.8)
rs4819554a	IL-17RA	22:17565035	0.22	0.50	0.50	0.83 (0.5 to 1.4)
rs1169288a	HNF1A	12:121416650	0.32	0.60	0.60	1.1 (0.7 to 1.8)
rs2340721a	ATF6	1:161849385	0.32	0.40	0.60	1.1 (0.7 to 1.7)
rs1025689a	IL-17RB	3:53883722	0.36	0.70	0.60	1.1 (0.7 to 1.8)
rs1124053a	IL-17E	14:22914819	0.26	0.50	0.70	1.1 (0.7 to 1.7)
rs2144908a	HNF4A	20:42985717	0.14	0.70	0.70	0.88 (0.4 to 1.8)
rs8192678a	PPARGC1A	4:23815662	0.34	0.40	0.70	0.93 (0.6 to 1.5)
rs13266634a	SLC30A8	8:118184783	0.31	0.60	0.70	1.1 (0.7 to 1.7)
rs1044498a	ENPP1	6:132172368	0.15	0.90	0.80	0.93 (0.5 to 1.7)
rs1800961a	HNF4A	20:43042364	0.03	0.99	NA	NA (NA)

NA, no rs1800961 minor alleles present in the NODAT cases.

^aSNPs previously reported to be associated with NODAT.

Gene Enrichment and Pathway Analyses

Gene enrichment and pathway analyses were undertaken for the top-ranked SNPs from the GWAS using Partek pathway within Genomics Suite 6.6. The top hit was the “response to stimulus” gene set (P=0.01) (Supplemental Figure 101, Supplemental Table 5). In pathway analysis, the P13K-AKT signaling pathway had the highest enrichment score (P<0.001) (Table 6, Supplemental Figure 102).

Table 6.

Top-ranked (P<0.05) gene pathways from GWAS results

Pathway Name	Enrichment Score	Enrichment P Value
PI3K-Akt signaling pathway	8.0	<0.001
Glutamatergic synapse	6.8	0.001
Regulation of actin cytoskeleton	5.0	0.01
Phosphatidylinositol signaling system	4.8	0.01
Extracellular matrix–receptor interaction	4.7	0.01
Arrhythmogenic right ventricular cardiomyopathy	4.6	0.01
Hypertrophic cardiomyopathy	4.2	0.02
Small cell lung cancer	4.1	0.02
Dilated cardiomyopathy	4.1	0.02
Oocyte meiosis	4.0	0.02
Hematopoietic cell lineage	4.0	0.02
Dopaminergic synapse	3.7	0.03

P13K-Akt, phosphatidylinositol 3-kinase and protein kinase B.

Discussion

NODAT is a common and serious complication of solid organ transplantation.^5,6 The growing prevalence of obesity among renal transplant recipients combined with improved access to transplantation for older patients is likely to increase the incidence of NODAT.^59–61 Therapeutic strategies for NODAT have focused on altering immunosuppressive regimens with limited success.^13,17,62 It is unlikely that NODAT can be avoided by tailoring immunosuppression without a detrimental effect on graft outcomes.

This is the first exploratory GWAS to be undertaken for NODAT. The secondary validation phase confirmed associations between the top-ranked loci; candidate genes, the top-ranked gene enrichment set, and the most significantly associated pathway are implicated in β-cell apoptosis (Tables 5 and 6, Supplemental Table 5).^47–58 β-Cell dysfunction may be the primary pathologic process in NODAT^18,20,21 and is discussed further in view of the genetic associations identified. Insulin resistance may also contribute to NODAT pathogenesis but this has not been conclusively proven.^{11,18,20,24,28}

Table 5.

Candidate genes and association with diabetes for top-ranked SNPs

SNP	Candidate Genesa	Gene Function	Other SNP Associations	Gene Association with Diabetes Mellitus
rs10484821	ATP5F1P6b	Mitochondrial pseudogene	None	None known
rs7533125	DNAJC16, CASP9, AGMAT, CELA2B, DDI2	Component of heat shock protein 40	None	Differential expression in pancreatic islets of type 2 patients with diabetes47
rs2861484	CELA2B, CASP9, CELA2A, DNAJC16, CTRC, EFHD2, AGMAT	Pancreatic elastase	None	None known
rs11580170	AGMAT, DNAJC16, CASP9, DDI2, CELA2B	Agmatine	None	Reduces apoptosis,48 increases insulin secretion,49 and insulin sensitivity50 in rats
rs2020902	CASP9, DNAJC16, CELA2B, CELA2A, AGMAT, CTRC, EFHD2	Caspase 9	Non-Hodgkin’s lymphoma	Component of intrinsic apoptotic pathway in pancreatic β cells51
rs1836882	NOX4	Subunit of NADPH oxidase complex	None	Increases ROS generation, reduces insulin gene expression and increases β-cell apoptosis in mice and in vitro,52 blockade replenishes islet insulin stores in animal models53
rs198372	NPPA, NPPB, CLCN6, METHFR, KIAA2013, PLOD1	Natriuretic peptide	None	Inhibits adipokine and cytokine production,54 reduced natriuretic peptide levels are associated with the development of diabetes55
rs4394754	INPP5A	Membrane associated type 1 inositol 1,4,5-triphosphate 5-phosphatase	None	Reduces β-cell proliferation in vitro,56 stimulates calcium induced insulin signaling and exocytosis in mice,57 regulates hepatic gluconeogenesis in mice58

Where there is more than one candidate, the gene in closest proximity to the SNP was chosen (underlined).

^aCandidate genes within 100 kB of SNP recorded in order of proximity.

^bPseudogene.

Glucose-Stimulated Insulin Secretion

Hyperglycemia upregulates Fas receptor expression on the β cell surface (Figure 3); initially, this promotes cell proliferation and insulin secretion.^63,64 The SNP rs4394754 is upstream from the 5′-untranslated region of INPP5A. The inositol polyphosphate 5-phosphatases are implicated in insulin signaling and exocytosis and inhibit the proliferation of insulin-producing cells in vitro.^56,57,65

Figure 3.

Glucose-stimulated insulin secretion. (1) Hyperglycemia induces Fas receptor expression on the β-cell surface. (2) Fas receptor expression and IL-1β production increase transcription of insulin genes and β-cell proliferation. (3) Mitochondrial generation of ATP is stimulated by hyperglycemia. (4) Insulin production in the ER is enhanced and increased insulin secretion restores normoglycemia.

Hyperglycemia stimulates ATP generation in mitochondria, causing membrane depolarization, calcium ion influx, and increased insulin exocytosis.^63,66

Glucotoxicity

Hyperglycemia initiates a train of events that begins with enhanced insulin secretion but, unchecked, will result in β-cell apoptosis. After gene enrichment, the top-ranked gene set associated with NODAT was “response to stimulus.” β Cells demonstrate increased apoptosis and reduced proliferation after 2 days of exposure to moderately elevated glucose concentrations in vitro.⁶⁷ Within 4 days, human β cells have almost completely lost their secretory function.⁶³ The period of exposure to elevated glucose concentrations is more important than the degree of hyperglycemia.^68,69 In the immediate post-transplantation period, hyperglycemia is virtually universal among renal transplant recipients; 87% of nondiabetic recipients at the Mayo Clinic recorded a blood glucose≥200 mg/dl during this period.^28,70 In a recent clinical trial, all renal transplant recipients recorded a blood glucose ≥140 mg/dl during hospitalization.²⁸ Factors contributing to post-transplant hyperglycemia include the catecholamine stress response to surgery,^71,72 corticosteroid treatment,⁷³ and restoration of normal renal degradation and excretion of insulin.^74,75

Hyperglycemia in the immediate post-transplant period is associated with a 4-fold increase in NODAT.^10,76 A recent proof-of-concept trial suggests that aggressive management of hyperglycemia in the postoperative period can reduce the incidence of NODAT. Hecking et al. compared insulin therapy when blood glucose exceeded 140 mg/dl to standard diabetic treatment in a cohort of kidney transplant recipients. At 1 year, there were no cases of NODAT in the treatment group compared with an incidence of 28% in the control group.²⁸ It is plausible that early management of hyperglycemia with insulin prevents β-cell glucotoxicity and apoptosis.

Hyperglycemia-induced β-cell apoptosis occurs via multiple pathways (Figure 4). Prolonged hyperglycemia attenuates the expression of Fas-associated protein with death domain–like IL-1β–converting enzyme inhibitory protein.^63,64,77 In the absence of Fas-associated protein with death domain–like IL-1β–converting enzyme inhibitory protein, Fas receptors induce apoptosis via both caspase-dependent and independent pathways.^69,77–79 Protracted stimulation of IL-1β promotes apoptosis by downregulating other protective proteins within the β cell.^64,80 In addition, both Fas activation and IL-1β induce mitochondrial dysfunction⁷⁷ and in vitro studies of β cells in NODAT have demonstrated a role for mitochondrial dysfunction.⁸¹ During prolonged hyperglycemia, generation of reactive oxygen species (ROS) by both mitochondria and the NADPH pathway exceed the oxidative stress threshold of the β cell.^66,82,83 ATP5F1P6 (rs10484821) is an inferred mitochondrial ATP synthase pseudogene.⁸⁴ NOX4 (rs1836882) encodes an enzyme of the NADPH oxidase complex.⁸⁵ In animal models of type 2 diabetes, NADPH oxidase blockade reduces oxidative stress and replenishes insulin stores.⁵³ Agmatine, a protein encoded by AGMAT (rs11580170), acts as a ROS scavenger to protect the mitochondrial membrane.⁴⁸ In animal models, agmatine also enhances insulin secretion.⁴⁹ Damaged mitochondria release cytochrome C and induce apoptosis via the activation of caspase 9⁷⁷; rs2020902 is located in the CASP9 gene.⁴⁹

Figure 4.

β-Cell glucotoxicity. (1) Prolonged hyperglycemia results in large numbers of Fas receptors being expressed on the β-cell membrane. (2) Fas activation induces β-cell apoptosis via both caspase-dependent and independent pathways. (3) Fas activation and elevated IL-1β concentrations induce mitochondrial dysfunction. (4) Excessive mitochondrial stimulation generates ROS that cause oxidative damage to the β cell. (5) Damaged mitochondria release cytochrome C, which activates the caspase cascade. (6) Reduced mitochondrial generation of ATP limits the production of insulin by the ER. (7) Prolonged insulin demands result in ER stress and intracytoplasmic accumulation of calcium ions, which stimulates apoptosis. (8) There is inadequate insulin exocytosis to restore normoglycemia.

Sustained demands on the endoplasmic reticulum (ER) during hyperglycemia result in cytoplasmic accumulation of calcium,^86,87 causing activation of the caspase cascade⁸⁶ and mitochondrial damage.⁴⁸ In β cells, ER stress upregulates heat shock protein 40 expression; this correlates with increased apoptosis.⁴⁷ The transcription product of the DNAJC16 (rs7533125) gene forms part of the heat shock protein 40 complex. NODAT is associated with elevated serum proinsulin concentrations^15,20 and the exocytosis of this inactive precursor reflects ER stress.⁸⁸

Lipotoxicity

Body weight and, in this study, percentage weight gain in the first year after transplantation are associated with NODAT.^5,6 Leptin is secreted by adipose tissues and is associated with both pretransplantation body fat and post-transplantation weight gain.^89,90 Leptin enhances IL-1β production in pancreatic islets promoting apoptosis.⁹¹ Elevated leptin concentrations correlate with increased serum free fatty acids. These directly induce β-cell apoptosis by caspase activation,⁷⁷ induction of ER stress,⁸⁷ and ROS generation.⁶⁶ The natriuretic peptides, encoded by NPPA (rs198372), inhibit leptin secretion and low natriuretic peptide levels are associated with the development of diabetes mellitus.^54,55

Immunosuppression

The contribution of corticosteroids and CNIs to the development of NODAT is well recognized.^17,92 Corticosteroids increase insulin resistance and promote β-cell apoptosis by exacerbating mitochondrial and ER stress.^13,73,93,94 The calcineurin/nuclear factor of activated T cell pathway modulates insulin exocytosis and β-cell proliferation. In animal models and in vitro, calcineurin/nuclear factor of activated T-cell blockade reduces insulin exocytosis and β-cell mass.^27,95,96 However, the evidence for an association between CNIs and NODAT is mixed; a recent meta-analysis found only a weak association between CNI therapy and NODAT.¹⁷ In this study, a statistically significant association was not demonstrated, which may reflect the low incidence of NODAT in our population. Mycophenolate also induces β-cell death by upregulation of the caspase cascade and induction of ER stress.^27,77 The contribution of each of these immunosuppressants explains the disappointing results of limiting one group in reducing NODAT incidence.

Strengths and Limitations

This study identified clinical and genetic variables associated with NODAT in an ethnically homogeneous population. It is the first to utilize an exploratory GWAS with confirmation by de novo genotyping; all previous studies have focused on candidate genes. The majority of SNPs previously reported to be associated with NODAT have been identified in a single population with failure of replication in other ethnic groups.^33,34,37 The TCF7L2 SNP rs7903146 is an exception to this; rs7903146 variants have been associated with NODAT in Korean⁴⁵ (P=0.02), white European⁴⁴ (P=0.002), and Polish Caucasian⁴³ (P=0.02) populations, although not all reports demonstrate an association.^34,97,98 This study had a comparable cohort size and rs7903146 minor allele frequency to the other studies but did not find an association between rs7903146 and NODAT. This may reflect the use of an extreme NODAT phenotype, which could prevent identification of genetic variants associated with milder disease. It is also possible that there may be an interaction between rs7903146 and CNIs in the risk for NODAT. Recipients were receiving CNIs in all three studies reporting an association, whereas 25% of this cohort had CNI-free immunosuppression.

Strengths of this study include the availability of high-quality clinical follow-up data over 25 years, the identification and correction for clinical variables in the same population, the use of multiple genotyping technologies, and second-stage typing of additional DNA samples to validate discovery findings. Stringent diagnostic criteria were used to militate against phenotypic heterogeneity and to enrich the population for risk alleles. The reported prevalence of NODAT in renal transplant recipients varies between 2% and 50%, reflecting differences in diagnostic criteria and reporting.¹² Although the diagnostic criteria used in this study will have potentially underestimated the NODAT incidence, we are confident that all recipients who developed a “severe diabetic phenotype” were identified. The small number of NODAT cases is a weakness of this study, resulting in the possibility that SNPs associated with a milder phenotype or that are less strongly associated with NODAT may not have been identified. However, GWAS have been successfully utilized to identify loci associated with extreme phenotypes in a number of complex diseases.^99–102

As a consequence of the small number of NODAT cases, these results did not reach the conventional threshold for genome-wide significance. As described, the stringent diagnostic criteria were used to generate phenotypic homogeneity with reduced case numbers as a consequence. A joint analysis approach was implemented to maximize power in this study; SNPs of interest were identified in the exploratory GWAS and de novo genotyping was used in the complete cohort of NODAT patients to validate these findings. This study has <30% power to identify a risk allele, with a minor allele frequency (MAF) of 10% and an OR of 1.5 with 95% confidence; however, the strong effects observed confirm that the smaller sample size utilized was sufficient to identify risk alleles for NODAT. Larger studies are required to identify genetic variants that confer a smaller effect on the risk of developing NODAT. The individuals genotyped for this study provided at least 60%, 80%, and 90% power to identify a risk allele at MAFs of 10%, 20%, or 30% respectively and OR of 2.0 with 95% confidence. The top-ranked SNPs associated with NODAT are in biologically plausible pathways that influence -cell apoptosis and corroborate current clinical evidence regarding NODAT pathogenesis. The associated genes or their transcription products have been associated with diabetes in vitro as well as in animal models and in longitudinal follow-up studies.

In conclusion, NODAT is a serious complication of solid organ transplantation that is detrimental to recipient survival.^5–7 Recent clinical studies suggest that β-cell toxicity, predominantly mediated by hyperglycemia, may be the primary pathologic process in NODAT^28,76 and that aggressive management of elevated serum glucose in the post-transplant period may prevent NODAT.²⁸ This study is the first exploratory GWAS with secondary validation for NODAT undertaken in a population of solid organ transplant recipients. SNPs associated with NODAT were associated with β-cell apoptotic pathways. This provides further evidence that β-cell dysfunction and death are key components of NODAT pathogenesis.

Concise Methods

Patient Cohort

All kidney transplant procedures in Northern Ireland are performed at the Belfast City Hospital. Consecutive renal transplant recipients who received first, deceased donor kidney transplants between May 1986 and May 2005 inclusive were included in this study. Donor and recipient clinical variables, graft outcomes, and survival are prospectively recorded in the Northern Ireland Kidney Transplant Database. Available data include donor age and sex, recipient age and sex, primary renal disease, HLA match, immunosuppression, acute rejection episodes, weight, incidence of NODAT, allograft survival, recipient survival, and cause of death. There were significant advances in transplantation during the study period. To allow for this, the variable “decade of transplantation” was considered. The first decade terminated on December 31, 1995, with the second decade commencing on January 1, 1996. Follow-up was continued until August 2012.

NODAT

NODAT was defined as a new requirement for oral hypoglycemic therapy or insulin after kidney transplantation. Controls were renal transplant recipients who did not have diabetes mellitus at transplantation and did not develop a requirement for oral hypoglycemic therapy or insulin during the follow-up period.

Clinical Analyses

All adults (aged≥16 years) who received a first, deceased donor transplant between May 1986 and May 2005 inclusive and who had a functioning renal transplant at 1 year were included. One-year graft survival was a prerequisite to investigating the effect of weight gain in the first year on NODAT incidence. Clinical variables investigated for association with NODAT were recipient age, recipient sex, primary renal disease, donor age, donor sex, HLA mismatch, acute rejection, immunosuppressive regimen, weight at transplantation, percent weight gain in the first year after transplantation, and decade of transplantation. SPSS for Windows (version 21.0; SPSS, Inc., Chicago, IL) was utilized for all analyses, with P values<0.05 considered significant.

Genetic Analyses

GWAS

DNA from individuals in the patient cohort was genotyped using the Illumina 660K array as part of the Wellcome Trust Case Control Consortium 3 Study into Renal Transplant Dysfunction (http://www.wtccc.org.uk/ccc3/). No sample was excluded for poor DNA quality, quantity, or poor genotype concordance with previous genotypes during a fingerprint evaluation stage. There were 26 NODAT patients and 230 controls successfully genotyped. Standard quality control was utilized with exclusion if the genotyping call rate was <90%, if MAF was <1%, if there was deviation from HWE (P<10⁻⁶), or if the sample call rate was <95%. Extreme heterozygosity and cryptic relatedness were not observed in this cohort. Known copy number variation and mitochondrial SNPs were excluded from analyses. These quality control steps and association analysis was performed using PLINK.¹⁰³ The level of statistical significance was set at P<10⁻⁵ considering the relatively small number of cases available; second-stage genotyping was conducted for validation and supporting data for all SNPs of interest. Linkage disequilibrium was calculated and results were plotted using Haploview⁴⁶ with regional association plots generated in LocusZoom.¹⁰⁴

In addition, there are 37 SNPs previously published as associated with NODAT. The association between these SNPs and NODAT was investigated in our GWAS cohort. Where reported SNPs were not genotyped in the GWAS, proxy SNPs were identified (r²>0.80) and association with NODAT was investigated.

Second-Stage Genotyping

All SNPs associated with NODAT with a P value<10⁻⁵ from the GWAS were selected for further analysis. We included previously reported NODAT SNPs as well as a single SNP (rs1050450) that was not genotyped in the GWAS and for which a proxy was unavailable. De novo genotyping for the above SNPs was undertaken in 58 NODAT patients and 383 controls from the patient cohort using Sequenom iPLEX and TaqMan technologies. DNA samples were randomly arranged in a 384-well format with four father-mother-proband trios and four negative controls per plate. Assays were performed according to the manufacturers’ instructions. The genotyping success rate and HWE were calculated for all SNPs using PLINK. Haploview (version 4.2; http://www.broadinstitute.org/haploview)⁴⁶ was used to identify linkage disequilibrium between SNPs and visualize haplotype blocks.

Statistical analysis was performed using PLINK.¹⁰³ A logistic regression multivariate model was utilized to investigate the confounding effect of clinical variables significantly associated with NODAT and the SNP associations. We included those clinical covariates that remained significantly associated with NODAT in multivariate logistic regression analysis (Table 2).

Gene Set Enrichment and Pathway Analyses

Gene set enrichment and pathway analysis were performed on the top hits from the GWAS using Partek software (version 6.6; Partek, Inc., St. Louis, MO). For gene set enrichment and pathway analysis, Fisher’s exact test was utilized, gene sets were restricted to those with two or more genes included, and genes were analyzed according to the hg19 genome build.

Ethics Statement

Ethical approval for this study was granted by the Office for Research Ethics Committees, Northern Ireland (http://www.hscbusiness.hscni.net/orecni.htm, 08/NIR03/79, 12/NI/0178). This study adhered to the Declaration of Istanbul.

Disclosures

None.