Genetic variation at glucose and insulin trait loci and response to glucose-insulin-potassium (GIK) therapy: the IMMEDIATE trial

Abstract

The mechanistic effects of intravenous glucose, insulin and potassium (GIK) in cardiac ischemia are not well understood. We conducted a genetic sub-study of the Immediate Myocardial Metabolic Enhancement During Initial Assessment and Treatment in Emergency care (IMMEDIATE) Trial to explore effects of common and rare glucose and insulin-related genetic loci on initial to 6-h and 6- to 12-h change in plasma glucose and potassium. We identified 27 NOTCH2/ADAM30 and 8 C2CD4B variants conferring a 40–57% increase in glucose during the first 6 h of infusion (P < 5.96 × 10⁻⁶). Significant associations were also found for ABCB11 and SLC30A8 single-nucleotide polymorphisms (SNPs) and glucose responses, and an SEC61A2 SNP with a potassium response to GIK. These studies identify genetic factors that may impact the metabolic response to GIK, which could influence treatment benefits in the setting of acute coronary syndromes (ACS).

INTRODUCTION

Among treatments that have shown promise in acute coronary syndromes (ACS) is acute myocardial metabolic support. Cellular, animal and human studies have suggested that if instituted at the onset of ischemia, metabolic therapy in the form of intravenous glucose-insulin-potassium (GIK), which promotes glycolysis and lowers circulating free fatty acids, helps maintain the viability of an at-risk myocardium and improves ACS outcomes.¹ On the basis of animal results, administration of GIK very early in ACS is considered most effective in protecting ischemic myocardium.² The Immediate Myocardial Metabolic Enhancement During Initial Assessment and Treatment in Emergency care (IMMEDIATE) Trial was designed to deliver this metabolic support immediately upon the patient’s presentation with ACS,³ recapitulating the timing of myocardial support shown effective in experimental studies. The trial demonstrated that among patients with suspected ACS, out-of-hospital administration of GIK, compared with placebo, was associated with lower rates of cardiac arrest and in-hospital mortality, smaller infarct size and lower free fatty acid levels.⁴

Recent advances in DNA technology through genome-wide association studies (GWAS) have allowed the identification of hundreds of common gene variants associated with metabolic diseases and related traits, including glucose sensitivity and insulin resistance.≥ Moreover, there is emerging evidence pointing toward the role of low frequency variants in metabolic traits.⁶ The goal of this study was to examine whether genetic variants previously associated with glucose and insulin-related traits by GWAS, or located within susceptibility loci, appear to contribute to responses to GIK among IMMEDIATE Trial patients. Identifying genetic variation that modifies the glucose, potassium and insulin response to GIK may provide greater insight into the metabolic factors that influence the overall course of patients with ACS.

MATERIALS AND METHODS

Study population

The IMMEDIATE Trial was a double-blind randomized controlled clinical trial conducted in 13 cities across the United States that assessed the effectiveness of intravenous GIK administered early in the course of ACS for 12 h.³ Out-of-hospital intravenous administration of GIK or placebo commenced immediately after a patient was considered to likely be having ACS. Inclusion and exclusion criteria for the IMMEDIATE Trial have been previously described,⁴ and can be seen in more detail in Supplementary Methods.

The IMMEDIATE Genetic Ancillary Study aimed to identify genetic modifiers of response to GIK therapy. A total of 321 participants were recruited at 9 participating sites during the parent trial enrollment. All IMMEDIATE Trial participants were eligible for inclusion, provided that consent for genetic analysis was granted. This study was approved by the Icahn School of Medicine at Mount Sinai’s IRB (Institutional Review Board) in addition to the IRB at each site.

Metabolite measurements

Glucose and potassium levels were measured by the study site hospital laboratory upon arrival at the emergency department and after 6 and 12 h of study drug infusion, or alternatively at the point the drug was stopped prematurely. A subset of Genetic Ancillary Study participants (N= 117) was concurrently enrolled in the Biological Mechanisms Cohort³ and had glucose and insulin levels measured initially and at 6 and 12 h at a core laboratory. Glucose levels generated by the hospital laboratories were correlated with glucose levels in the Biological Mechanisms Cohort core laboratory at all time points (r² = 0.643–0.941, P=1.4×10⁻¹³ to 1.50×10⁻⁵²). Study site hospital measurements were used for these analyses as substantially more data were available. Metabolite levels were generated by the core laboratory at Tufts Clinical and Translational Science Institute.

Genotyping and candidate variant selection

Genomic DNA was extracted from whole blood using the Gentra Puregene Blood Kit (Qiagen, Valencia, CA, USA) and from saliva using the Oragene*•DNA Saliva Collection Kit (DNA Genotek, Ontario, Canada). Genotyping was carried out using the lllumina Metabochip⁷ and HumanExome Beadchip⁸ at the Children’s Hospital of Philadelphia. Single-nucleotide polymorphisms (SNPs) were clustered into genotypes using the lllumina BeadStudio software, and zCall for rare variants.⁹

We downloaded the GWAS catalog (accessed 30 July 2013)⁵ and selected all SNPs associated with fasting glucose, type-2 diabetes, glycated hemoglobin levels, diabetes-related insulin, diabetes (incident, gestational), proinsulin levels, 2-h glucose challenge, insulin resistance and/or response, and fasting insulin (Supplementary Table S1). For intergenic SNPs, we also selected variants ±250 kb from the index GWAS SNP. Annotation documentation from Metabochip was used to further select variants that had been included for type-2 diabetes, fasting glucose, 2-h glucose tests, fasting insulin and hemoglobin A1c traits. SNPs were removed if they had call rates < 95% and deviated from Hardy-Weinberg equilibrium (P< 0.0001), resulting in the selection of 15 159 SNPs with a mean genotyping success rate of 99.8%. Duplicate concordance was assessed for three quality control samples assayed on Metabochip and five samples assayed on Exome Chip. Genotype concordance for each duplicate was ≥98.3%. Samples were filtered for individual call rates < 95% and relatedness (identical by descent). After excluding 2 samples that failed identical by descent analysis (same individuals recruited twice), and samples with a low call rate, the Genetic Ancillary Study cohort comprised 318 individuals.

Statistical analysis

Phenotype derivation.

Associations between gene variants and changes in non-fasting plasma glucose and potassium during 12-h infusion were investigated. Due to non-linear trends over 12 h, the change in each trait was determined between the initial and 6-h measurement (calculated as 6-h level- initial level), and between the 6-h and 12-h measurements (calculated as 12-h level‒6-h level). Glucose data exhibited skewed distributions and were log-transformed before analysis. Therefore, their 6-h log-changes are interpreted as ratios. For potassium, 6-h changes can be interpreted as absolute differences between the two measurements. Correlations between the metabolites were generated using Pearson’s correlation.

Population stratification.

To control for population admixture, we computed principal components to be used as covariates in regression analyses. We calculated principal components in EIGENSTRAT¹⁰ on a combined set of 130 539 independent SNPs (linkage disequilibrium, LD, r² < 0.3) from Metabochip and Exome chip that passed quality control, had minor allele frequency (MAF) ≥5%, a call rate of >0.98 and a Hardy-Weinberg equilibrium 0050-value of >0.001.

Association analysis.

For common variant analysis (MAF ≥ 5%), we used a two degree of freedom (d.f.) test to detect joint significance for the main effect of the SNP and the SNP × intervention interaction in the same model.¹¹ All models adjusted for age, sex, treatment arm, diabetes status and the first two principal components. For each trait, associations with initial to 6-h change also adjusted for the time from infusion initiation to first blood draw. For gene-based association testing, all non-synonymous variants (non-synonymous, splice variants, 3’ UTR and 5’ UTR), regardless of MAF, were collapsed using the ‘adjusted optimal’ method (SKAT-O) within the SNP-set (Sequence) Kernel Association Test¹² shown to have the greatest statistical power to detect gene-level associations.¹³ Statistical analyses were carried out using the R software package (www.r-project.org).

Selection of significance threshold.

For association analyses of 14 389 common variants (MAF ≥ 5%), we determined the number of uncorrelated markers to be 8387, after accounting for LD using the Li and Ji approach.¹⁴ Therefore, after adjustment for multiple hypothesis testing, a P-value threshold for statistical significance was set at 5.96× 10⁻⁶. For gene-based analysis, including 208 genes (Supplementary Table S2) and represented by 1310 SNPs, a P-value threshold was set at 2.40 × 10⁻⁴ (0.05/208).

RESULTS

Patient characteristics

The characteristics of the participants in the IMMEDIATE Genetic Ancillary Study (Tables 1 and 2) show no significant differences in basic demographics, including age, gender and medical history of myocardial infarction, diabetes or heart failure between randomization groups (P>0.05). The average age at time of recruitment was 63.4 years (31–96 years), and 74.8% of participants were male. Participants were predominantly of White ethnicity (87.7%), 7.6% were African-Americans, 1.6% were Asian-Americans, 3.2% were classified as ‘other’ and 12.7% identified as Hispanic race (Table 1).

Table 1.

Demographics, pre-hospital clinical characteristics and medical history by treatment group for the IMMEDIATE Genetic Ancillary Study

	n	GIK	n	Placebo	P-valuea
Demographics/clinical characteristics
Age (years)	157	63.8 ±12.5	160	63.0 ±13.6	0.60
Gender, n = Male (%)	157	117 (74.5)	160	120 (75.0)	0.92
Ethnicity, n (%)	157		160		0.07
White		132 (84.1)		146 (91.3)
Black		17 (10.8)		7 (4.4)
Asian		4 (2.5)		1 (0.6)
Other		4 (2.5)		6 (3.8)
Hispanic, n (%)	157	15 (9.5)	159	25 (15.7)	0.01
Initial out of hospital blood pressure (mm Hg)
Systolic	157	141.9 ±35.4	159	144.3 ±30.3	0.52
Diastolic	155	84.9 ±24.6	157	86.8 ±21.9	0.47
BMI (kg m−2)	150	28.6 ±5.5	152	28.8 ± 6.6	0.77
Time for symptom onset until infusion initiation (hours)	128	2.7 ±3.9	141	2.3 ±2.7	0.33
Medical history, n (%)
Previous myocardial infarction (yes)	157	55 (35.0)	160	57 (35.6)	0.91
History of diabetes (yes)	157	49 (31.2)	160	38 (23.8)	0.14
Previous heart failure (yes)	157	27 (17.2)	160	24 (15.0)	0.38

Abbreviations: BMI, body mass index; GIK, glucose-insulin-potassium; IMMEDIATE, Immediate Myocardial Metabolic Enhancement During Initial Assessment and Treatment in Emergency care. Data are presented as mean±s.d. and percentages.

^aStatistical significance P< 0.05.

Table 2.

In-hospital clinical characteristics and analyte measurements by treatment group for the IMMEDIATE Genetic Ancillary Study

	n	GIK	n	Placebo	P-valueb
Clinical characteristics
Duration of treatment (hours)	157	8.4 ± 4.8	132	9.1 ±4.5	0.17
Time from treatment initiation until first blood draw—Standard of Care	156	0.64 ± 0.76	159	0.55 ±0.61	0.28
(hours)
Time from treatment initiation until first blood draw—Biological	50	2.21 ±1.20	55	2.63 ± 1.27	0.09
Mechanisms Cohorts (hours)
Non-fasting plasma glucose levels (mg dl−1)a
Initial blood glucose	156	188.7 (177.7–202.4)	159	154.5 (145.5–165.7)	3.10× 10−5
6-h blood glucose	92	188.6 (167.3–214.9)	114	148.4 (138.4–159.2)	3.70×10−4
12-h blood glucose	137	146.9 (131.6–162.4)	142	142.6 (134.3–151.4)	0.64
Initial to 6-h fold change in glucose	91	1.01 (0.91–1.13)	113	0.97 (0.92–1.02)	0.47
6- to 12-h fold change in glucose	90	0.74 (0.66–0.83)	113	0.95 (0.91–1.00)	1.64×10−5
Non-fasting plasma potassium levels (mmol 1−1)
Initial potassium	155	3.99 ± 0.64	159	3.93 ± 0.55	0.37
6-h potassium	91	4.35 ±0.52	112	4.08 ± 0.46	9.30×10−5
12-h potassium	146	4.43 ± 0.58	152	3.97 ± 0.46	3.95 ×10−13
Initial to 6-h change in potassium	89	0.42 ±0.53	111	0.17 ±0.53	0.001
6- to 12-h change in potassium	84	0.17 ±0.52	107	−0.13 ±0.38	5.62×10−6
Non-fasting plasma Insulin Levels (μU ml −1)a
Initial insulin	50	107.8 (83.9–137.0)	55	18.3 (13.5–25.0)	5.30× 10−14
6-h insulin	49	154.5 (115.6–206.4)	58	22.4 (16.6–30.3)	6.02 × 10−15
12-h insulin	51	100.5 (70.8–141.2)	61	21.1 (15.5–29.4)	1.43×10−9
Initial to 6-h fold change in insulin	48	1.45 (1.10–1.91)	52	1.16 (0.93–1.43)	0.20
6- to 12-h fold change in insulin	49	0.68 (0.53–0.86)	58	0.89 (0.72–1.10)	0.09

Abbreviations: GIK, glucose–insulin–potassium; IMMEDIATE, Immediate Myocardial Metabolic Enhancement During Initial Assessment and Treatment in Emergency care. Data are presented as unadjusted means ± s.d. or percentages unless otherwise stated.

^aData were log-transformed before analysis. Data are presented as geometric means and 95% confidence intervals. Ratios are presented for initial to 6-h, and 6- to 12-h changes for logged data.

^bStatistical significance P < 0.05.

Initial measurements were carried out on average 35 min after the initiation of infusion for glucose and potassium (Table 2). As expected, glucose, potassium and insulin levels were higher in the GIK group than in placebo (Table 2). Plasma glucose was increased with GIK initially and at 6h (P≤ 3.70 × 10⁻⁴); however, at 12 h glucose levels were similar to those observed in the placebo group (P = 0.64). Insulin levels reflecting endogenous and exogenous insulin from GIK were greater in the GIK group at all three measurements (P≤ 1.43× 10⁻⁹). No differences in potassium levels were observed with intervention at the initial measurement, but modestly higher levels were detected with GIK at both 6 and 12 h (P≤ 9.30 × 10⁻⁵). Treatment-stratified and treatment-combined correlations of all study metabolites at different time points are presented in Supplementary Table S3. Glucose and potassium levels, as well as initial to 6-h and 6- to 12-h changes, were representative of the overall IMMEDIATE Trial. There were no significant differences in glucose or potassium levels when comparing Genetic Ancillary Study participants with all other individuals recruited into the IMMEDIATE Trial (P>0.05, data not shown).

Summary of genetic data

The SNP selection criteria resulted in the inclusion of 15 159 SNPs for analysis. Of those, 14 389 were common variants (MAF ≥0.05), including 165 non-synonymous, 110 synonymous, 9262 intronic, 4181 intergenic, 206 3’ UTR, 39 5’ UTR and 2 splice variants. There were 770 rare variants (MAF < 0.05) selected for gene- based analysis. Of these, 585 were non-synonymous, 6 were splice variants, 164 were located in the 3’ UTR and 15 in the 5’ UTR.

Association analysis of glucose traits

There were 37 SNPs exceeding the corrected threshold for statistical significance. Specifically, 27 common variants (MAF 10–12%) in strong LD (r²≥ 0.85) at the NOTCH2/ADAM30 locus on chromosome 1 modified the glucose response to GIK (Figure 1a; Supplementary Table S4). In GIK-treated individuals, each copy of ‘C’ allele for the lead SNP, rs7534586, was associated with a 57% increase in plasma glucose between the initial measurement and 6h, compared with a 9% increase in the placebo group (2 d.f. test’s P, P_2d.f. = 2.33 × 10⁻⁶; Table 3, Figure 1b). Among the associated variants was rs10923931 (P_2d.f. = 3.80× 10⁻⁶) identified by GWAS as associated with type-2 diabetes.^{1 b} Variants with the smallest P-value at each locus are shown in Table 3.

Figure 1.

Association between NOTCH2/ADAM30 locus and initial to 6-h change in glucose: (a) Regional plots of the top association signals ±250 kb. The X axis shows the chromosome and physical distance (kb), the left Y axis shows the negative base ten logarithm of the P-value and the right Y axis shows recombination activity (cM/Mb) as a blue line. The linkage disequilibrium of surrounding single-nucleotide polymorphisms (SNPs) with the leading SNP is indicated by a scale of intensity of red color filling as shown in the legend at the upper right- hand corner of each plot. The genome-wide association studies (GWAS) SNP P-value is shown in blue, and is annotated with the SNP identifier. Positions, recombination rates and gene annotations are according to the NCBI’s build 36 (hg18). (b) Fold change by treatment arm and genotype. Data are presented as the geometric mean±s.e.m. after the adjustment for age, sex, treatment arm, diabetes, PCI, PC2 and time from infusion initiation until initial glucose measurement. The P-value is from the 2 degree of freedom (d.f.) test. GIK, glucose-insulin-potassium.

Table 3.

Significant associations between the top candidate SNPs and metabolite traits^a

SNP	Chr	Position	Minor allele	MAF	Function	Gene (distance from closest gene, bp)	Placebo change per allele	GIK change per allele	P (SNP)	P (SNP × treatment interaction)	P (2 d.f.)
Log initial to log 6-h glucose
rs7534586	1	120496127	C	0.11	Intron	NOTCH2	1.09b	1.57b	0.28	0.002	2.33×10−06
rs1406227	15	62459614	A	0.23	Intergenic	C2CD4B (2132) MGC15885 (469757)	1.04b	1.41b	0.52	0.001	2.63×10 −06
Log 6-h to log 12-h glucose
rs34453330	2	169781530	A	0.05	Intron	ABCB11	1.01b	0.52b	0.94	6.5 ×10−5	5.65 ×10−07
rs10105491	8	118217271	G	0.25	Intergenic	SLC30A8 (28318) MED30 (315694)	0.98b	1.37b	0.74	9.0×10 −5	5.55×10 −06
6 to 12-h potassium
rs74474517	10	12182121	A	0.06	Intron	SEC61A2	− 0.04c	−0.68c	0.69	0.0005	2.28 × 10−06

Abbreviations: d.f., degree of freedom; GIK, glucose–insulin–potassium; MAF, minor allele frequency; SNPs, single-nucleotide polymorphisms.

^aOne variant with the smallest P-value (2 d.f.) at each locus is presented. Included in the model are age, sex, randomization group, PCI, PC2 and history of diabetes. For initial to 6-h change in glucose, time from infusion initiation to initial blood measurement was also adjusted for. FC = fold-change between the initial to 6-h, or 6-h to 12-h measurements.

^bFold change.

^cAbsolute change.

The C2CD4B region on chromosome 1 also conferred a larger change in glucose in the first 6 h of infusion in GIK patients compared with placebo (Table 3) with eight variants (MAF ≥0.22) in strong LD with the lead SNP rs1406227 (LD>0.95), demonstrating associations with initial response to GIK (Figure 2a; Supplementary Table S4). For each copy of the minor ‘A’ allele of rs1406227, a 41% increase in glucose was observed with GIK treatment compared with 4% in placebo-treated patients (≥2d.f. = 2.63×1 O’⁶; Figure 2b).

Figure 2.

Association between the C2CD4B locus and initial to 6-h change in glucose: (a) Regional plots of the top association signals and (b) fold change by treatment arm and genotype. See Figure 1 for plot description. GIK_; glucose-insulin-potassium.

There were two variants that modified response to GIK between 6 and 12 h. The intronic variant, rs34453330 (MAF = 0.05) in the ABCB11 gene, was associated with a 48% decrease in glucose per ‘A’ allele in GIK-treated individuals, however no difference was observed in the placebo group (P_2d.f. = 5.65 × 10⁻⁷; Figure 3a). The minor ‘G’ allele of SLC30A8 rs10105491 (MAF = 0.25) interacted with treatment to increase glucose by 37% between 6 and 12 h with GIK but not with placebo (P_2d.f. = 5.55 × 10⁻⁶; Figure 3b). Other suggestive genetic associations (P_2d.f. < 0.001) for initial to 6-h, and 6- to 12-h change in glucose are presented in Supplementary Tables S4 and S5.

Figure 3.

Genetic association for 6- to 12-h change in glucose by treatment arm and genotype: (a) For ABCB17 rs34453330 and (b) for SLC30A8 rs10105491. Data are presented as the geometric mean±s.e.m. and adjustment for age, sex, treatment arm, diabetes, PC1 and PC2. The P-value is from the 2 degree of freedom (d.f.) test. GIK, glucose-insulin-potassium.

When analysis was carried out on Whites only, the largest population sub-group, similar associations between each of these variants and glucose response to GIK were observed (data not shown).

Association analysis of potassium traits

An association between rs74474517 and change in potassium between 6 and 12 h was identified (Table 3). This intronic variant located within the SEC61A2 gene (MAF = 0.06) modified response to intervention, resulting in a decrease in potassium between 6 and 12 h in the GIK arm but not in the placebo (P_2d.f. = 2.28 × 10⁻⁶;β_gik = −0.68 mmol I⁻¹,β_Placebo = −0.05mmol1⁻¹ Figure 4). A similar association was observed when Whites were analyzed separately (P = 4.12 × 10⁻⁶). No significant genetic associations with initial to 6-h change in potassium were observed; however, a number of suggestive associations within or near the KCNQ 7, CDKN2B, SLC2A2 and DGKB genes were observed (P_2d.f.⁼ 4.07 × 10⁻⁵ to 0.001). Suggestive associations are shown in Supplementary Tables S6 and S7.

Figure 4.

Genetic association for 6- to 12-h change in potassium by treatment arm and genotype for SEC61A2. Data are presented as the arithmetic mean±s.e.m. Data have been adjusted for age, sex, treatment arm, diabetes, PC1 and PC2. The P-value is from the 2 degree of freedom (d.f.) test. GIK, glucose-insulin-potassium.

Gene-based analysis of glucose and potassium

Gene-based analysis of common and rare functional variants in 208 genes identified a study-wise significant association (P < 2.40 × 10⁻⁴) between ZNF750 (P = 6.99× 10⁻⁵), FAM180B (P= 1.29× 10⁻⁴), CELF1 (P= 1.35× 10⁻⁴) and initial to 6-h change in glucose (Table 4). Suggestive associations were identified including NOTCH2 (P = 0.002) and ADAM30 (P=0.017, data not shown), the top region identified in the single-point common variant analysis (Table 3; Figures 1 and 2). A nominal association between NOTCH2 was also observed for 6- to 12-h change in glucose (P = 0.03) (data not shown). For potassium, gene-based testing did not provide substantial evidence for novel association signals beyond those already implicated in the single-point common variant analysis. Top gene-based associations for each trait are shown in Table 4.

Table 4.

Top genes (P < 0.01) from gene-based analysis of non-coding variants

Gene	Chr.	Start position	End position	No of SNPs	P-alue	Phenotype
NOTCH2	1	120455095	120506308	14	0.002	Log initial to log 6-h change in glucose
SPC25	2	169732622	169732622	1	0.001	Log initial to log 6-h change in glucose
LRP2	2	169985338	170177325	33	0.006	6 to 12-h change in potassium
ZFAND3	6	38084405	38084405	1	0.008	Log 6 to log 12-h change in glucose
SLC30A8	8	118184783	118188536	13	4.56 × 10−4	Log initial to log 6-h change in glucose
IDE	10	94211444	94333827	7	0.007	Log initial to log 6-h change in glucose
ABCC8	11	17414570	17450177	3	0.006	Initial to 6-h change in potassium
CELF1	11	47487740	47574716	9	1.35×10−4	Log initial to log 6-h change in glucose
FAM 180B	11	47609867	47610271	2	1.29 × 10 − 4	Log initial to log 6-h change in glucose
MTNR1B	11	92714749	92715916	7	0.007	6- to 12-h change in potassium
NCKAP1L	12	54902264	54928894	3	0.006	Log 6-h to log 12-h change in glucose
SPPL3	12	121201156	121202664	4	0.005	Log initial to log 6-h change in glucose
FN3KRP	17	80676855	80685655	10	0.004	Log initial to log 6-h change in glucose
ZNF750	17	80788899	80790442	10	6.99 × 10−5	Log initial to log 6-h change in glucose
QPCTL	19	46201812	46206425	3	0.002	6- to 12-h change in potassium

Abbreviation: SNPs, single-nucleotide polymorphisms.

In several cases, different genes from the same region showed signals for association with the same or related traits. That is, ZNF750, the strongest association with the first 6-h change in glucose, is located near to FN3KRP, which was also suggestively associated with the same phenotype. Similarly, both FAM180B and CELF1 were associated with the first 6-h change in glucose. The LRP2 and SPC25 genes located ~250 kb apart were suggestively associated with 6- to 12-h change in potassium and initial to 6-h change in glucose, respectively.

DISCUSSION

This study was the first to investigate genetic modifiers of response to GIK infusion in individuals presenting with likely ACS. We investigated the association between genetic variation at glucose and insulin-related loci and short-term response to GIK infusion, assessed as initial to 6-h, and 6- to 12-h change in plasma glucose and potassium.

We observed significant associations for 27 common variants at the NOTCH2-ADAM30 locus, all in strong LD, and the initial to 6-h glucose response to therapy. Participants treated with GIK, who carried a minor allele at the NOTCH2-ADAM30 region, exhibited a 40–57% greater increase in glucose in the first 6h of infusion compared with those without a variant. In the placebo group, these same alleles conferred a 4–9% larger change in plasma glucose between the initial and 6 h measurement. In addition, a gene-based approach that combined the effects of 14 rare and common functional variants further suggested that NOTCH2 has a role in modifying response to GIK.

The Diabetes Genetics Replication and Meta-Analysis consortium identified multiple signals within the NOTCH2-ADAM30 locus that were strongly associated with type-2 diabetes.¹⁵ Within this region, the ‘A’ allele of the NOTCH2 variant, rsl0923931 was associated with increased risk of diabetes, and in our study this same allele was associated with a higher increase in glucose levels in response to GIK. The role of NOTCH2 in the development of type-2 diabetes has been proposed to occur through effects on pancreatic β-cell function.¹⁶

A greater increase in glucose between the initial measurement and 6h was also observed for eight variants near the C2CD4B gene. Expression of the C2CD4B gene, which encodes the nuclear localized factor 2 in the pancreas, is stimulated by pro-inflammatory cytokines.^17,18 A meta-analysis of 21 GWA studies, and follow-up investigation in 76 558 subjects, associated C2CD4B with fasting glucose homeostasis.¹⁹ The region has also been associated with an increase in type-2 diabetes risk,¹⁸ plasma glucose 2 h after an oral glucose load,²⁰ diminished insulin release²¹ and diminished glucose-stimulated insulin response.²²

On average, glucose levels decreased in the final 6 h of infusion in both treatment arms. In this study, we identified two genes that modified the glucose response to GIK during this time. The minor ‘A’ allele of rs34453330, located in intron 26 of the ATP-binding cassette sub-family B, member 11 (ABCB11) gene, was associated with a 48% larger decrease in glucose between 6 and 12 h than T allele carriers. Interestingly, the effect of this variant was limited to GIK-treated subjects. In a meta-analysis of two GWA studies, ABCB1 /, which encodes the ABC transporter BSEP (bile slat export pump), and neighboring G6PC2 (glucose-6-phosphatase) genes were associated with elevated plasma glucose. Variation at this locus has also been associated with increased basal hepatic glucose production, increased insulin release²³ and a reduced risk of type-2 diabetes.²⁴ This protective effect plausibly occurs via a hyper-response to postprandial elevation in circulating glucose levels.²³

The second region associated with 6- to 12-h change in glucose was near SLC30A8, which encodes an islet-specific zinc transporter connected to insulin granule function.²⁵ In our, study, the ‘G’ allele of a variant near SLC30A8, rs10105491, was associated with a smaller decrease in glucose between 6 and 12 h in GIK-treated individuals, but no real difference with placebo. SLC30A8 was first linked to type-2 diabetes by GWAS in 2007, with the index variant, rs13266634, conferring an 18% increased risk for the ‘A’ allele.²⁰ In a prospective study, genetic variation within SLC30A8 was also shown to result in impaired beta-cell function over time,²⁷ and in a study of non-diabetics was associated with either abnormal insulin processing or secretion.²⁸ In our study, the rs13266634 ‘A’ allele was associated with a larger 6- to 12-h decrease in the GIK group (P = 0.03), but not in placebo.

We identified one intronic variant (rs74474517) in SEC61A2 that was associated with 6- to 12-h change in potassium levels. The ‘A’ allele was associated with a larger decrease in potassium levels (−0.64 mmol 1⁻¹ per allele) in GIK-treated patients than in placebo (−0.05 mmol 1⁻¹ per allele). SEC61A2 belongs to a family of proteins that mediate the translocation and insertion of membrane proteins, including potassium channels, in the endoplasmic reticulum.²⁹ The observed association may indicate that SEC61- mediated differences in potassium channel number or structure modify the response to GIK therapy. Moreover, SEC61A2 is located <150 kb away from the CDC123/CAMK1D locus. A study investigating expression quantitative trait loci and co-expression networks of top type-2 diabetes associated gene variants detected that the expression of SEC61A2 was highly correlated with CDC123/CAMK1D rs12779790 genotype,³⁰ previously associated with type-2 diabetes¹⁵ and insulin-related traits.¹⁰

Strengths of this study include the random assignment of GIK to individuals at high risk of ACS. Moreover, availability of repeated measurements during the infusion provided additional information about the dynamics of the response. The dense genomic coverage of the candidate regions provided by the Metabochip and Exome chip has also allowed us to examine the role of the functional variants potentially responsible for the GWAS signals in gene-based analysis. Importantly, most of the top associations detected in our study did not involve previously reported GWAS SNPs, even when the analyses were limited to Whites only. Instead, other variants within the susceptibility loci were identified, suggesting that a better coverage used in our analyses could improve the accuracy of the detection or the involvement of different variants in GIK response. Moreover, gene-based analysis indicated that different genes in close proximity to each other were associated with correlated traits, further emphasizing the importance of fine mapping and functional studies to determine causal genes/variants that alter phenotypic traits.

We acknowledge several limitations of our study, including the lack of pre-treatment and fasting metabolite measurements, and the sample size of this ancillary study, which limited the statistical power to detect modest effects. Also of note, in spite of the large glucose load in the GIK infusion, endogenous insulin secretion did not vary significantly over the 12-h infusion as indicated by mostly unchanged C-peptide levels (data not shown). This suggests that the relative levels of glucose and insulin during GIK infusions were balanced in regard to the impact on blood glucose levels. However, due to the lack of fasting glucose and insulin (or C-peptide) levels, a formal estimation of insulin resistance, as assessed by the homeostasis model assessment index, HOMA2-IR, was not possible. Novel regions not queried in the present analysis may also contribute to GIK response, leaving the possibility that genome-wide studies may prove fruitful. Finally, given the interventional nature of this study and unavailability of DNA samples from other GIK trials, replication of our findings was not possible.

In summary, this is the first study to examine genetic modifiers of GIK therapy in patients with likely ACS using 15 149 genetic variants previously linked to glucose and insulin-related traits. We identified significant associations between variants within or near NOTCH2-ADAM30, C2CD4B, ABCB17, SLC30A8, SEC61A2 and glucose and potassium traits during 12-h infusion with GIK. Findings from this study suggest that genetic factors have an important role in determining the response to GIK treatment and may provide new insights into GIK mechanisms.