Genetic Linkage of Serum Homocysteine in Dominican families: The Family Study of Stroke Risk and Carotid Atherosclerosis

Abstract

Background and Purpose

Homocysteine levels are determined by genetic and environmental factors. Several studies have linked high plasma levels of total Hcy (tHcy) to the increased risk of cardiovascular disease (CVD), stroke, and many other conditions. However the exact mechanism of documented and novel tHcy quantitative trait loci (QTL) to that risk is unknown.

Methods

We have performed linkage analysis in 100 high-risk Dominican families with 1362 members. Probands were selected from the population-based Northern Manhattan Study. A set of 405 microsatellite markers was used to screen the whole genome. Variance components analysis was used to detect evidence for linkage, after adjusting for stroke risk factors. Ordered-Subset Analysis (OSA) based on Dominican Republic (DR) enrollment was conducted.

Results

tHcy levels had a heritability of 0.44 (p<0.0001). The most significant evidence for linkage was found at chromosome (Ch) 17q24 (MLOD=2.66, p=0.0005) with a peak at D17S2193 and was significantly increased in a subset of families with a high proportion of DR enrollment (MLOD=3.92, p=0.0022). Additionally, modest evidence for linkage was found at Ch 2p21 (MLOD=1.77, p=0.0033) with a peak at D2S1356 and was significantly increased in a subset of families with a low proportion of DR enrollment (MLOD=2.82, p=0.0097).

Conclusions

We found a strong evidence for novel QTLs on Ch 2 and 17 for tHcy plasma levels in Dominican Families. Our Family Study provides essential data for a better understanding of the genetic mechanisms associated with elevated tHcy levels leading to CVD after accounting for environmental risk factors.

Cardiovascular disease (CVD) and stroke are the most common causes of death in Western countries.¹ Several studies have demonstrated that increased plasma levels of total homocysteine (tHcy)² are associated with premature onset of CVD³ and stroke.⁴ Homocysteine (Hcy)² is formed from methionine as a result of cellular methylation reactions.⁵ The exact mechanisms by which Hcy promotes CVD are not yet fully understood although it has been proposed that Hcy may have a role in endothelial injury, high-density lipoprotein (HDL) inhibition, thrombogenesis and autoimmune response.⁵ However, clinical trials using vitamin B12 and folic acid to decrease the levels of tHcy failed to demonstrate a clinical benefit in secondary prevention against stroke⁶ or myocardial infarction.⁷ In contrast, other trials have shown benefits from B-vitamin supplementation in high risk stroke patients⁸ but not in patients with myocardial infarction,⁹ suggesting that tHcy may play a pivotal role in stroke.

For these reasons, a great effort has been made to identify the genetic determinants of plasma tHcy. Polymorphisms in genes encoding for Methylenetetrahydrofolate Reductase (MTHFR) have been associated with variations in plasma levels of tHcy. Specifically, the MTHFR 677 C→T polymorphism was the most important known genetic determinant of folate and tHcy status.¹⁰ We have previously reported that vascular risk associated with elevated tHcy levels is greatest among whites and Hispanics compared to blacks.⁴ Few studies have documented differences in heritability for tHcy by race-ethnicity,^{11, 12} but the data is still limited. The aim of the present study was to detect novel quantitative trait loci (QTL), a region on a chromosome which influences the trait, for tHcy among high-risk Dominican families.

Materials and Methods

Subjects

Details of the Family Study of Stroke Risk and Carotid Atherosclerosis have been described in full elsewhere.¹³ Briefly, high-risk probands were selected from the population-based Northern Manhattan Study (NOMAS) according to the following criteria: (1) report of a sibling with a history of myocardial infarction or stroke; or (2) having 2 of 3 quantitative risk phenotypes (maximal carotid plaque thickness, left ventricular mass, or tHcy level above the 75^th percentile in the NOMAS cohort). Most probands (80%) were recruited based on the first criterion. Families were enrolled if the proband was able to provide a family history, obtain consent from family members, and had at least 3 first-degree relatives able to participate. No probands were excluded by disabling or fatal vascular events prohibiting consent of family members. Although probands were identified in Northern Manhattan, we enrolled family members in New York (Columbia University) and in the Dominican Republic (DR; Clinicas Corazones Unidos, Santo Domingo). All subjects provided informed consent and the study was approved by the Institutional Review Boards of Columbia University, University of Miami, the National Bioethics Committee, and the Independent Ethics Committee of Instituto Oncologico Regional del Cibao in the DR.

Overall, 1362 individuals from 100 Dominican families with complete phenotype and genotype data were analyzed. Thirty percent of subjects were enrolled in the DR. Because sequential oligogenic linkage analysis routines (SOLAR) ¹⁴analyzes relative pairs in an extended family framework, these 1362 individuals were part of a larger family structure of 2184 individuals and resulted in 1460 sib pairs, 452 half-sib pairs, and 2273 avuncular pairs. Mean family size was 22±11; median 20, and range 4-87.

Data Collection

Demographic, socioeconomic and risk factor data were collected through interviews based on The Family Study of Stroke Risk and Carotid Atherosclerosis instruments.^{4, 13} Questionnaires regarding diet, vitamin use, hypertension, diabetes, smoking, alcohol use, and physical activity were administered. Vitamin intake was assessed using the Block food frequency questionnaire (Block FFQ). Dietary folate, B12, and B6 intake were calculated from questionnaire responses using Block DIETSYS version 3.0 software.¹⁵ This questionnaire was found reliable and valid in multiple epidemiological studies.^15-17 It has been validated in Hispanic populations and covers dietary habits, nutritional supplements and specific traditional foods (e.g. plantains, mango, rice).¹⁷ Measurements of height, weight, hip and waist circumference, and skin-fold thickness were also obtained, as were serial blood pressures.

Fasting blood samples were drawn into serum tubes and spun within 1 hour at 3000g at 4° C for 20 minutes and frozen at −70°C. The blood samples were processed for lipids (total cholesterol, LDL, triglyceride, HDL), glucose levels, creatinine, methylmalonic acid (MMA), as well as tHcy. Fasting serum tHcy and MMA were measured by licensed methods for commercial use.¹⁸

Genotyping and Quality Control

Extraction of DNA was done by the Columbia University Genome Center. DNA was sent to the Center for Inherited Disease Research (CIDR) for genotyping at Johns Hopkins University. A set of 405 microsatellite markers at an average interval of 10 centimorgans (cM) across the genome was genotyped. Family structure was verified and adjusted using Relpair and PREST.^{19, 20} Mendelian error checking was performed using Pedcheck.²¹

Statistical Analyses

Heritability

To minimize ascertainment bias, the SOLAR ascertainment correction was used in all analyses. Heritability was estimated using a pedigree-based maximum-likelihood method implemented in SOLAR.¹⁴ This heritability represents the genetic proportion of total phenotypic variance after the effect of all covariates has been removed (per SOLAR parameterization). Thus, the residuals of tHcy are used for analysis and checked for normality (kurtosis<0.8) before proceeding. tHcy was natural-log transformed, multiplied by 10, and observations beyond 3 SD from the mean were dropped. An initial polygenic model was used to estimate significant covariates (p<0.10) that were used in final analyses.

Covariates that were tested included age, sex, age*sex, age², B12 deficiency, B6, B12, folate, vitamin use (self report), creatinine, pack years, BMI, alcohol use, and country of enrollment. Vitamin B12 deficiency was defined by MMA level>271 nmol/l. B6, B12, and folate were defined as dietary + supplementary. Alcohol use was defined as current drinking of more than one drink per month.

Multipoint Linkage Analysis

A multipoint variance components approach was used to conduct linkage analysis on tHcy. Allele-sharing models were obtained by estimating multipoint identity-by-descent (IBD) matrices at 1 cM intervals. LOD scores were calculated using a log-10 ratio of the likelihoods of the polygenic models. One-LOD support regions, calculated as the maximum LOD score - 1.0, were used to define the region of interest. Empirical p-values were calculated based on 10,000 replicates in which a fully-informative marker, unlinked to a given trait, was simulated.

Candidate Genes

Genes located in the 1 LOD support interval surrounding each linkage peak (LOD > 1) were identified using the University of California, Santa Cruz (UCSC) human genome annotation database (genome.ucsc.edu). Genes were considered as likely candidates if they belonged to the canonical methionine metabolism pathway and were in the SAM-dependent methyltransferase family of genes.¹¹ In addition, genes related to homocysteine in the gene database for the National Center for Biotechnology Information (http://www.ncbi.nlm.nih.gov/gene) were considered. (Supplementary Table I)

Ordered Subset Linkage Analysis

Ordered subset linkage analysis (OSA)²² was performed, using proportion of family members living in the DR as the ranking phenotype. Family-specific LOD scores were output for QTLs with LOD>1 in the multipoint linkage analysis. For each peak, family-specific LOD scores were added in trait rank order (decreasing and increasing) until a maximum LOD score was obtained. A permutation procedure was implemented to test the hypothesis that ordering by family phenotype gave stronger linkage than random ordering. Specifically, 10,000 random family orderings were permuted and empirical p-values derived.

Results

Of the 1362 Dominican individuals, a total of 1246 were included in the final analysis after outliers and individuals with missing data for significant covariates were removed. The mean tHcy level was 8.9 μmol/l, and was significantly higher in individuals living in the DR (Table 1). The covariate screening identified age, sex, age*sex, age², B12 deficiency, B6, B12, folate, creatinine, vitamin intake, pack years, alcohol, and DR enrollment as significant covariates, explaining 50% of the total variance of tHcy. The heritability estimate of tHcy was 0.44 (p<0.0001). The proportion of the total variance of tHcy explained by genes was 22%.

Table 1.

Demographic and clinical characteristics by the country of enrollment

	Not DR Enroll	DR Enroll	Total	TESTING
	(N = 844)	(N = 402)	(N = 1246)	DR vs non-DR Enroll
	Mean ± SD	Mean ± SD	Mean ± SD	Wilcoxon-Rank Sum
Age	45.5 ± 17.1	47.0 ± 17.0	46.0 ± 17.1	0.1269
BMI (kg/m2)	29.3 ± 5.8	27.5 ± 5.7	28.7 ± 5.8	< 0.0001
Waist circumference (inch)	36.8 ± 5.5	35.7 ± 5.6	36.4 ± 5.6	0.0016
Total cholesterol (mg/dl)	185.5 ± 40.0	183.8 ± 42.5	185.0 ± 40.8	0.6545
LDL (mg/dl)	111.7 ± 35.4	106.4 ± 33.1	110.0 ± 34.8	0.0209
HDL (mg/dl)	48.8 ± 14.0	53.0 ± 11.8	50.2 ± 13.5	< 0.0001
TG (mg/dl)	126.7 ± 84.0	122.6 ± 81.7	125.3 ± 83.2	0.5463
SBP (mmHg)	120.2 ± 19.1	124.9 ± 21.5	121.8 ± 20.0	0.0005
DBP (mmHg)	76.0 ± 9.9	79.4 ± 12.2	77.1 ± 10.8	< 0.0001
Total Folate	530.5 ± 287.4	579.7 ± 308.9	546.4 ± 295.3	0.0054
Total B6	2.6 ± 1.7	3.0 ± 2.0	2.8 ± 1.8	0.0013
Total B12	5.9 ± 5.3	6.1 ± 5.1	6.0 ± 5.2	0.2926
Creatinine (mg/dl)	0.9 ± 0.3	0.9 ± 0.6	0.9 ± 0.4	0.2472
Pack Years (packs/day * years)	3.7 ± 9.9	4.4 ± 11.4	3.9 ± 10.4	0.3519
Homocysteine (μmol/l)	7.9 ± 3.1	10.9 ± 4.3	8.9 ± 3.8	< 0.0001

	n	%	n	%	n	%	Chi-Square
Hypertension (History or SBP ≥ 140 and DBP ≥ 90)	n 330	% 39.10	n 163	% 40.55	n 493	% 39.60	Chi-Square 0.6252
Diabetes (History or Fasting Glucose ≥ 126)	124	14.69	48	11.94	172	13.80	0.1881
Dyslipidemia (History or Cholesterol > 240)	292	34.60	105	26.12	397	31.86	0.0027
Coronary artery disease	180	21.33	88	21.89	268	21.51	0.821
≥ High School Education	417	49.41	197	49.00	614	49.28	0.8943
Sex (Male)	321	38.03	151	37.56	472	37.88	0.8727
B12 Deficient	73	8.65	154	38.31	227	18.22	< 0.0001
Take Vitamins	320	37.91	114	28.36	434	34.83	0.0009
Alcohol (Moderate to Severe use)	384	45.50	222	55.22	606	48.64	0.0013

To identify the genetic loci underlying variation in tHcy, we performed QTL linkage analysis. This identified 11 regions with a multipoint LOD>1 on chromosomes 1, 2, 3, 4, 9, 15, 17, and 22 (Table 2) (Figure 1). This included 2 distinct peaks on chromosome 3 and 3 distinct peaks on chromosome 17. There were a total of 19 candidate genes in the one-LOD support regions for all peaks (Table 3). The region on Ch. 17q24 was suggestive for linkage with a peak at D17S2193 (MLOD=2.66, empirical p value=0.0005). The one-LOD supportive interval across all 3 peaks extends approximately from 16.2 to 71.2 Mb on Ch. 17, encompassing 788 protein coding genes.

Table 2.

SOLAR multipoint LOD scores > 1 and Ordered-Subset Analysis Results ranking families by proportion of Dominican Republic enrollment

						OSA % DR (H to L) Max Lod	OSA % DR (L to H) Max Lod
Location	CM	Marker	LOD	p-value	h2q	(p-value, # family)	(p-value, # family)
1p37	18	D1S508	1.34	0.0099	0.17		1.43
							(0.64, 98)
2p21	61	D2S1356	1.77	0.0033	0.21		2.82
							(0.0097, 81)
3p25	26	D3S4545	1.59	0.0053	0.22	1.60	1.60
						(0.79, 56)	(0.79, 99)
3q27.3	200	D3S1262	1.21	0.0127	0.17	1.21	1.55
						(1, 100)	(0.30, 92)
4q15	78	D4S2367	1.13	0.0106	0.16	1.45	1.25
						(0.23, 42)	(0.58, 98)
9q34.3	164	D9S1838	1.14	0.0153	0.15	1.14	1.52
						(1, 100)	(0.34, 64)
15q11.2	1	D15S128	1.44	0.0082	0.22	(0.44, 36)	(1, 100)
						(0.44, 36)	(1, 100)
17q11.2	56	D17S1880	2.45	0.0006	0.26	3.23	2.55
						(0.06,58)	(0.60,92)
17q21.3	75	D17S2180	2.47	0.0006	0.30	3.92	2.48
						(0.0022, 58)	(0.74, 92)
17q24.2	91	D17S2193	2.66	0.0005	0.27		2.74
							(0.57, 92)
22q13.3	55	D22S1169	1.36	0.0097	0.20		1.59
							(0.45, 98)

h2q = locus specific heritability

p-value = empirical p-value based on 10,000 replicates

H to L = ranking families from high to low percent DR enrollment

L to H = ranking families from low to high percent DR enrollment

Figure 1.

Multipoint linkage plots for tHcy. Map distance in CM is depicted on the X-axis and LOD scores on the Y-axis.

Table 3.

Homocysteine Candidate Genes within 1-LOD Region of Hcy identified peaks

Marker	Position	LOD	Candidate Gene	Gene Position	Gene Association with Hcy**
D1S508	1p37	1.34	MTHFR	1p36.3	Conversion of 5, 10-methylene-tetrahydrofolate to 5-methyl- tetrahydrofolate
			ICMT	1p36	Incubation of neuroblastoma cells with S-adenosylhomocysteine results in reduced methylation of PP2A, most likely inhibiting ICMT
D2S1356	2p21	1.77	THUMPD2	2p22.1	Implicated in the transmethylation reactions involving Methionine and Hcy
			SLC3A1	2p16	Cystine, dibasic, and neutral amino acid transporter
D3S4545	3p25	1.59	VHL	3p26	Implicated in the remethylation stage of Hcy metabolism
			GHRL	3p26	Ghrelin blocks the Hcy-induced decrease in eNOS protein levels in HCAECs; thus,helping reduce oxidative stress
D3S1262	3q27.3	1.21	ADIPOQ	3q27	Diabetic children are characterized by a higher concentration of protective adiponectin and lower plasma homocysteine compared to healthy children
			ECE2	3q27.1	Encodes a SAM dependent methyltransferase fold
D4S2367	4q15	1.13	ALB	4q11-q13	Involved in homocysteine metabolism
D9S1838	9q34.3	1.14	METTL11A	9q34.11	Responsible for metabolism of methionine which dictates Hcy metabolism
D17S2180*	17q21.3	2.47	PEMT	17p11.2	involved in S-adenosylhomocysteine metabolic process
			SHMT1	17p11	Catalyzes conversion of serine to glycine and provides carbon units for synthesis of methionine
			NOS2A	17q11	homocysteine stimulates NOS2A-mediated NO production in macrophages
			PNPO	17q21	Catalyzes conversion of B6 allowing for Hcy metabolism
			PNMT	17q21-q22	Interacts with the remethylation stage of Hcy after catecholemine biosynthesis
			LPO	17q23	Hcy can generate reactive oxygen species and induce LPO
			METTL2A	17q23.2	Responsible for metabolism of methionine which dictates Hcy metabolism
D22S1169	22q13.3	1.36	TCN2	22q12.2	Implicated in the Folate cycle of Hcy metabolism
			PPARA	22q13.1	Elevated levels of Hcy may affect methylation of PPARA

^*Note the 1 LOD region for chromosome 17 encompasses all 3 peak regions

^**Taken from http://www.ncbi.nlm.nih.gov/gene

Among the 58 families with the highest proportion of DR enrollment, the LOD score significantly increased from 2.47 to 3.92 (p=0.0022) on Ch. 17q21 at D17S2180 (Table 2) (Figure 2b). The LOD score on Ch. 2p21 increased from 1.77 to 2.82 (p=0.0097) among the 81 families with the lowest proportion of DR enrollment. (Table 2) (Figure 2a). In addition, this ranking strategy also reduced the one-LOD supporting interval size on 17q from 55 Mb to 37 Mb. This narrowed critical linkage region harbors 565 protein-coding genes. The critical region on 2p extends from approximately 32.7 to 60.8 Mb and harbors 111 protein-coding genes.

Figure 2.

(a): Multipoint linkage plot for tHcy on chromosome 2, including OSA for a subset of families with the smallest proportion of DR enrollment.

(b): Multipoint linkage plot for tHcy on chromosome 17, including OSA for a subset of families with the largest proportion of DR enrollment.

Discussion

Several studies have shown that moderate and high tHcy plasma levels may play a pivotal role in increasing the risk for CVD.^{3, 23} We have shown that vascular risk associated with elevated tHcy levels differed by race-ethnicity.⁴ In the current study using QTL mapping in extended DR families, we found 11 regions with suggestive linkage (multipoint LOD>1) on eight different chromosomes after controlling for significant covariates. The highest LOD scores were found on Ch. 2 and 17. The heritability estimate of tHcy in our study was 0.44, which was similar to those reported in European populations.^{12, 24}

The metabolism of Hcy is a complex system involving several enzymes and cofactors.⁵ Genetic analysis may help us to understand the mechanisms leading to higher levels of tHcy and increased risk for CVD. The most widely studied variants have been in MTHFR, especially the MTHFR 677 C→T polymorphism;¹⁰ and 5-methyltetrahydrofolate homocysteine methyltransferse (MTR).²⁵ A number of genome-wide studies of Hcy levels have been conducted with varying results. Possible regions of linkage have been reported on chromosomes 1q42, 9q34, 11q23, 12q24, 13q, 14q32, 16q and 19p13.^{11, 12, 24} Other polymorphisms which may affect plasma tHcy include MTR 2756A→G, MTRR 66A→G, cSHMT 1420C→T, TC 67A→G, TC 776C→G, and GCPll 1561C→T.²⁶ In the current study, we report novel linkage for tHcy to Ch. 2p21 and to Ch. 17q21 in a Caribbean population. Different study populations and differences in the environmental factors may explain the lack of replication between studies.

Among the 565 protein coding genes in the one-LOD supportive interval for Ch. 17, there are three genes related to Hcy metabolism: Phenylethanolamine N-methyltransferase (PNMT), which binds the S-adenosyl-L-homocysteine and inhibits its synthesis;²⁷ pyridoxamine 5′-phoshate oxidase (PNPO), which catalyzes conversion of pyridoxine 5′-phosphate to pyridoxal 5′-phosphate (PLP), the metabolically-active form of vitamin B₆ that is required as a coenzyme for Hcy metabolism;²⁸ and methyltransferase like 2A (METTL2A) involved in the metabolism of Methionine Cycle and therefore in the Hcy metabolism.²⁹ Among the 111 protein-coding genes in the one-LOD supportive interval for Ch. 2p21 is THUMP domain containing 2 (THUMPD2). THUMPD2 is believed to be involved in Methionine metabolism based on the presence of an S-adenosylmethionine-dependent methyltransferase domain.³⁰

The improvement in LOD scores when accounting for the proportion of individuals enrolled in the DR vs. US may be explained by a variety of factors, including the differences in dietary and vitamin intake. Numerous studies have demonstrated the importance of nutritional status on tHcy levels.²⁶ Our study confirms the importance of nutritional status on tHcy levels, with p<0.1 for vitamin use, folate, B6, B12, B12 deficiency, and alcohol consumption in the polygenic model screen. In addition, we found a significant difference in vitamin use, folate, B6, B12 deficiency, and alcohol consumption between individuals living in the DR and those living in the US (Table 1). Therefore, it is not surprising that geographical location impacts tHcy levels, with p=1.49e-20 for the effect of enrollment location in the polygenic model screen. Interestingly, folate levels are higher among those living in the DR than in the US (Table 1). This seems counterintuitive as the US has fortified certain foods with folate for over a decade, whereas the DR is just starting to fortify its foods with folate. Additional analysis (not shown) reveals that this higher folate level is actually driven by higher folate levels among younger (18-40) residents in the DR compared to their US counterparts. Perhaps, despite folate supplementation in the US, the younger immigrant US population may consume less folate supplemented food. Further investigation may be needed to determine whether the younger US Caribbean Hispanic population characterized by low SES levels and recent immigration lack access to or choose a low folate diet.

We also found significant differences in B12 deficiency between the DR and US enrolled individuals (Table 1). We hypothesize that there is an additional unidentified environmental factor that causes those living in the DR to metabolize B12 poorly even though they are receiving adequate amounts in their diet and through supplements. Observations such as ours have been previously reported from the National Health and Nutrition Examination Survey (NHANES 1999-2004).³¹ In addition, a study from NHANES 1999-2002, showed that in people with B12 deficiency, higher serum folate is associated with increased tHcy levels, as seen in our subjects enrolled in DR (Table 1).³² We have also found significantly higher waist circumference and BMI in the Dominican participants enrolled in the US than in DR (Table 1). A suggestive linkage with dietary macronutrient (total calories, total proteins, total fat, saturated fat, monounsaturated fat, and polyunsaturated fat) intake and adiposity phenotypes within Ch. region 2p22 near marker D2S1346 was previously reported in extensive Mexican American families.³³ This chromosome region neighbors our strongest linkage for tHcy (D2S1356). Further investigation between nutritional and environmental factors and variation in tHcy levels in various populations is warranted.

Strengths of the present study include the large, Dominican family study with a comprehensive baseline assessment combined with rigorous phenotype measurement. By focusing on one ethnic group we have minimized the effects of heterogeneity; however, this could explain why we did not replicate the results from other related studies. Approaches to mapping quantitative phenotypes also offer efficient statistical advantages over discrete traits. We believe some of the unknown determinants of tHcy may be related to diet or other environmental information which has not yet been analyzed. One methodological limitation is that dietary intake was estimated using a single food frequency questionnaire that asked about food consumption over the prior 12 months, resulting in possible dietary misclassification.

With the unbiased genome-wide approach, we identified several likely QTLs controlling tHcy plasma levels among Dominican families. We found novel evidence for linkage between regions in Ch. 2 and 17 and tHcy levels in plasma. Our Family Study of Stroke Risk and Carotid Atherosclerosis provides essential genetic and environmental data among Dominican families not available from other studies and may help to better understand the genetic mechanisms of increased tHcy levels leading to high risk of stroke and other vascular diseases.