Skip main navigation

Inactive Matrix Gla Protein Is Causally Related to Adverse Health Outcomes

A Mendelian Randomization Study in a Flemish Population
Originally publishedhttps://doi.org/10.1161/HYPERTENSIONAHA.114.04494Hypertension. 2015;65:463–470

Abstract

Matrix Gla-protein is a vitamin K–dependent protein that strongly inhibits arterial calcification. Vitamin K deficiency leads to production of inactive nonphosphorylated and uncarboxylated matrix Gla protein (dp–ucMGP). The risk associated with dp–ucMGP in the population is unknown. In a Flemish population study, we measured circulating dp–ucMGP at baseline (1996–2011), genotyped MGP, recorded adverse health outcomes until December 31, 2012, and assessed the multivariable-adjusted associations of adverse health outcomes with dp–ucMGP. We applied a Mendelian randomization analysis using MGP genotypes as instrumental variables. Among 2318 participants, baseline dp–ucMGP averaged 3.61 μg/L. Over 14.1 years (median), 197 deaths occurred, 58 from cancer and 70 from cardiovascular disease; 85 participants experienced a coronary event. The risk of death and non-cancer mortality curvilinearly increased (P≤0.008) by 15.0% (95% confidence interval, 6.9–25.3) and by 21.5% (11.1–32.9) for a doubling of the nadir (1.43 and 0.97 μg/L, respectively). With higher dp–ucMGP, cardiovascular mortality log-linearly increased (hazard ratio for dp–ucMGP doubling, 1.14 [1.01–1.28]; P=0.027), but coronary events log-linearly decreased (0.93 [0.88–0.99]; P=0.021). dp–ucMGP levels were associated (P≤0.001) with MGP variants rs2098435, rs4236, and rs2430692. For non-cancer mortality and coronary events (P≤0.022), but not for total and cardiovascular mortality (P≥0.13), the Mendelian randomization analysis suggested causality. Higher dp–ucMGP predicts total, non-cancer and cardiovascular mortality, but lower coronary risk. For non-cancer mortality and coronary events, these associations are likely causal.

Introduction

Cardiovascular disease remains the leading cause of mortality and is worldwide directly responsible for ≈12 million deaths and 20% of total mortality.1 Calcification of the conduit arteries is an independent risk factor for myocardial infarction,24 stroke,3,4 and cardiovascular death.2,4 Vascular smooth muscle cells synthesize a small secretory protein (11 kDa), which contains 5 γ-carboxyglutamate (Gla) amino-acids and which was therefore named matrix Gla protein (MGP).5 Activation of MGP requires 2 posttranslational modifications: γ-glutamate carboxylation in a vitamin K–dependent manner and serine phosphorylation. Carboxylated MGP is a potent inhibitor of arterial calcification.6 In the presence of vitamin K deficiency, MGP is partly synthesized in a noncarboxylated form that does not inhibit calcification. The MGP gene is located on chromosome 12.6

In animal experiments, vitamin K antagonists induce arterial calcification.7 Patients receiving warfarin, compared with matched controls, had a 2-fold increase in aortic valve calcification.8 In patients with diabetes mellitus,9 kidney disease,10 heart failure,11 aortic valve stenosis,12 or vascular disease,13 higher levels of inactive nonphosphorylated and uncarboxylated MGP (dp–ucMGP) were associated with higher cardiovascular risk,9 more severe disease,1012 and predicted all-cause mortality.1113 In a nested case–cohort with 1462 patients with coronary heart disease or stroke and 1406 matched participants, dp–ucMGP was not related to coronary events or stroke.14 One study of 577 community-dwelling elderly followed up for 5.6 years reported a 2-fold increased risk of cardiovascular disease in the highest compared with the lowest third of the dp–ucMGP distribution.15 To our knowledge, no other population study encompassing a wide age range with long-term follow-up addressed the risk associated with deficient MGP activation. We therefore examined the adverse health outcomes in the Flemish Study on Environment, Genes and Health Outcomes (FLEMENGHO) in relation to the plasma level of dp–ucMGP at baseline. To test for causality, we used genetic variation in the MGP gene as instrumental variable.

Methods

Study Population

FLEMENGHO complies with the Helsinki declaration for research in human subjects and the Belgian legislation for the protection of privacy (http://www.privacycommission.be). The Ethics Committee of the University of Leuven approved the study. At each contact, participants gave informed written consent. Previous publications16,17 provide a detailed description of the recruitment of the FLEMENGHO participants. The participation rate at enrolment was 78.0%. The participants were repeatedly followed up. Of 3343 participants, 2329 had a plasma sample available in the FLEMENGHO bio-bank and had dp–ucMGP measured (Figure S1 in the online-only Data Supplement). The date of blood collection ranged from April 29, 1996, to November 29, 2011 (5th to 95th percentile interval, October 14, 1996, to July 18, 2005; median, October 7, 1998). For analysis of the health outcomes in relation to dp–ucMGP, we excluded 11 patients on treatment with warfarin at the time of blood sampling, who had extremely elevated dp–ucMGP levels (Figure S2). This left 2318 participants for the analysis of outcome (Figure S1). The Mendelian randomization analysis included 2304 participants who had been genotyped for MGP (Figure S1).

Clinical Measurements at Baseline

Baseline was the date of blood collection. Trained nurses measured the subjects’ anthropometric characteristics and blood pressure. Body mass index was weight in kilograms divided by the square of height in meters. Each subject’s blood pressure was the average of 5 consecutive auscultatory readings obtained with a standard sphygmomanometer after the participants had rested in the sitting position for ≥5 minutes. Hypertension was a blood pressure of ≥140 mm Hg systolic or 90 mm Hg diastolic or use of antihypertensive drugs. The nurses also administered a standardized questionnaire inquiring about each participant’s medical history, smoking and drinking habits, and intake of medications. Plasma glucose and total serum cholesterol were measured using automated methods in a single certified laboratory. Diabetes mellitus was a fasting or random glucose level exceeding 126 or 200 mg/dL (7.0 or 11.1 mmol/L) or use of antidiabetic agents.

Ascertainment of Health Outcomes

At annual intervals, we ascertained the vital status till December 31, 2012, via the Belgian Population Registry. We obtained the International Classification of Disease codes for the immediate and underlying causes of death from the Flemish Registry of Death Certificates. For 1706 participants, we collected information on the incidence of nonfatal events either via face-to-face follow-up visits with repeat administration of the same standardized questionnaire as used at baseline (n=1641) or via a structured telephone interview (n=65). Follow-up data were available from 1 visit in 685 participants, from 2 in 542, from 3 in 284, and from ≥4 or more in 195 participants.

Fatal and nonfatal stroke did not include transient ischemic attacks. Coronary events included sudden death, fatal and nonfatal myocardial infarction, acute coronary syndrome, fatal ischemic cardiomyopathy, and coronary revascularization. In addition to coronary events, the cardiac end point comprised fatal and nonfatal heart failure and atrioventricular block causing death or requiring pacemaker implantation. Fatal and nonfatal cardiovascular events consisted of cardiac end points, stroke, aortic aneurysm, pulmonary heart disease, pulmonary or arterial embolism, and peripheral arterial disease. Physicians ascertained the diseases reported on the death certificates, in the questionnaires, and in the telephone interviews against the medical records of general practitioners or hospitals. In the outcome analyses, we only considered the first event within each category.

Measurement of dp–ucMGP

As extensively described in previous publications,18 circulating dp-ucMGP was quantified in citrated plasma samples, using the inaKtif MGP iSYS kit (IDS, Boldon, UK), which is a dual-antibody test based on the previously described sandwich ELISA developed by VitaK, Maastricht University, the Netherlands.18 In this assay, the first monoclonal antibody directed against the nonphosphorylated sequence 3 to 15 in human MGP is bound to the solid phase, whereas the second antibody is directed against the noncarboxylated MGP sequence 35 to 49.

Genotyping

MGP (size: 4738 base pairs) maps to a genomic area on chromosome 12 (p13.1–p12.3) characterized by high linkage disequilibrium (Figure S3). As explained in detail in the Expanded Methods available online, we selected 4 tagging single-nucleotide polymorphisms (SNPs; rs2098435, rs4236, rs1800802, and rs2430692) that are in high linkage disequilibrium (r2>0.80) with 223 other SNPs covering the entire gene with extension into the 3′ and 5′ flanking regions (Table S1). After DNA extraction from peripheral blood,19 SNPs were genotyped using the 64X TaqMan® OpenArray™ Genotyping System (Life Technologies, Foster City, CA). All DNA samples were loaded at 50 ng/mL and amplified according to the manufacturer’s instructions. For analysis of the genotypes, we used autocalling methods as implemented in the TaqMan Genotyper software version 1.3 (Life Technologies). Next, genotype clusters were evaluated manually with the call rate set >0.90. Sixteen duplicate samples gave 99.9% and 100% reproducibility for all 64 SNPs on the chip and for the 4 MGP SNPs, respectively. The overall call rate was 98%.

Statistical Analysis

For database management and statistical analysis, we used SAS software, version 9.3 (SAS Institute, Cary, NC). We normalized the dp–ucMGP distribution by a logarithmic transformation. For comparison of means and proportions, we applied the large sample z-test or ANOVA and Fisher’s exact, respectively.

Outcome Analysis

We compared the incidence of end points across thirds of the dp–ucMGP distribution using (i) rates standardized by the direct method for sex and age (>40, 40–59, ≥60 years), (ii) Kaplan–Meier survival function estimates and the log-rank test, and (iii) the cumulative incidence derived from Cox models adjusted for sex and age and the Wald χ2 statistic. Next, we assessed the prognostic value of dp–ucMGP as a continuous variable in multivariable-adjusted Cox proportional hazard regression. We checked the proportional hazard assumption by applying a Kolmogorov-type supremum test as implemented in the ASSESS statement of the PROC PHREG procedure. We assessed the functional form of dp–ucMGP by plotting the cumulative martingale residuals versus log dp–ucMGP. To account for family clusters, we used the PROC SURVIVAL procedure of the SUDAAN 11.0.1 software (Research Triangle Institute, NC). In this procedure, nonindependence among family members was taken into account by including family as a random effect along with the aforementioned covariables as fixed effects. The random effect adjusts for the clustering effect within pedigrees. We used the log likelihood statistic to evaluate the fit of nested models.

Attributable and Population Attributable Risk

We computed the positive predictive value of highest tertile of dp–ucMGP as (R×D)/([G/100]×[R–1]+1), where R is the multivariable-adjusted hazard ratio, D is the incidence of an end point in the whole population, and G is the prevalence of highest tertile.20 The attributable risk is given by ([R–1]×100)/R and the population attributable risk by ([G/100]×[R–1]×100)/([G/100]×[R–1]+1).20

Genetic Analyses

We tested the Hardy–Weinberg equilibrium in randomly selected unrelated subjects, using the exact statistics available in the PROC ALLELE procedure of the SAS package. The 4 SNPs complied with Hardy–Weinberg equilibrium (P≥0.30; Table S3). To assess the causality of the association between adverse health outcomes and dp–ucMGP, we applied a Mendelian randomization approach using the MGP variants as instrumental variables. In a 2-stage least-square procedure,21 we first regressed log dp–ucMGP on age, body mass index, smoking and drinking, and the genetic variants of MGP, while accounting for family clusters using the PROC REGRESS procedure of the SUDAAN 11.0.1 software. In the second step, we used multivariable-adjusted Cox proportional hazard models as implemented in the PROC SURVIVAL procedure of the SUDAAN package to regress health outcome on the predicted dp–ucMGP obtained in the first step. The Mendelian randomization approach controls for unmeasured confounders and reverse causality that may distort the directly assessed association between outcome and the exposure of interest (measured dp–ucMGP).22 We assessed the strength of the a-priory defined instrumental variables, the MGP genotypes, using the F-statistic.23

Results

Baseline Characteristics

All 2318 participants were White Europeans, of whom 1186 (51.2%) were women. The study population consisted of 332 singletons and 1986 related subjects, belonging to 22 single-generation families and 181 multi-generation pedigrees. Age averaged (±SD) 43.5±17.6 years; blood pressure, 125.8±16.1 mm Hg systolic and 76.5±11.2 mm Hg diastolic; body mass index, 25.3±4.6 kg/m2; and total cholesterol, 204.8±42.0 mg/dL (5.28±1.08 mmol/L). Among all participants, 641(27.6%) had hypertension, of whom 338 (52.7%) were on antihypertensive drug treatment, 66 (2.85%) had diabetes mellitus, and 58 (2.50%) reported a history of cardiovascular disease. Of 1186 women and 1132 men, 280 (23.6%) women and 301 (26.6%) men were smokers, and 305 (25.7%) women and 521 (46.0%) men reported intake of alcohol. In smokers, median tobacco use was 14 cigarettes per day (interquartile range, 8–20 cigarettes per day). In drinkers, the median alcohol consumption was 10 g per day (interquartile range, 5–20 g per day).

In all participants, the geometric mean level of dp–ucMGP was 3.61 μg/L (340 pmol/L). Table 1 lists the baseline characteristics of participants by thirds of the distribution of dp–ucMGP. Across increasing categories of dp–ucMGP, more participants had diabetes mellitus, hypertension, or a history of cardiovascular disease, whereas fewer reported smoking. Furthermore, age, body mass index, blood pressure, and total cholesterol increased (P<0.001) with higher category of dp–ucMGP. In regression analysis, while accounting for family cluster, dp–ucMGP increased with age and body mass index and decreased with smoking and drinking. The explained variance totaled 11.0%.

Table 1. Baseline Characteristics of Participants by Thirds of the dp–ucMGP Distribution

CharacteristicLowMediumHighP Value
dp–ucMGP limits, μg/L<2.932.93–4.81≥4.81
760769789
N° with characteristics, %
 Women381 (50.1)389 (50.6)416 (52.7)0.55
 Current smoker252 (33.2)204 (26.5)125 (15.8)<0.001
 Drinking alcohol284 (37.4)269 (35.0)273 (34.6)0.47
 Diabetes mellitus13 (1.71)12 (1.56)41 (5.20)<0.001
 Hypertension152 (20.0)184 (23.9)305 (38.7)<0.001
 Treated hypertension65 (8.5)96 (12.5)177 (22.4)<0.001
 History of cardiovascular event13 (1.71)10 (1.30)35 (4.44)<0.001
Mean of characteristic (±SD)
 Age, y38.5±14.441.5±16.950.3±19.1<0.001
 Body mass index, kg/m224.1±3.924.9±4.426.8±5.0<0.001
 Systolic pressure, mm Hg122.5±13.9124.9±15.7129.7±17.5<0.001
 Diastolic pressure, mm Hg75.5±11.075.9±11.278.1±11.2<0.001
 Heart rate, beats per minute67.6±8.767.5±8.967.8±9.20.85
 Total cholesterol, mg/dL199.9±40.5201.2±42.4213.0±41.8<0.001
 Plasma glucose, mg/dL5.03±1.135.16±1.345.40±1.65<0.001
dp–ucMGP, μg/L1.77 (1.47–2.50)3.80 (3.37–4.24)6.81 (5.48–7.71)<0.001

P values are for linear trend across thirds of the dp–ucMGP distribution. Diabetes mellitus was a fasting or random glucose level exceeding 126 or 200 mg/dL (7.0 or 11.1 mmol/L) or use of antidiabetic agents. Hypertension was a blood pressure of ≥140 mm Hg systolic or ≥90 mm Hg diastolic or use of antihypertensive drugs. To convert cholesterol from mg/dL to mmol/L, multiply by 0.0258. To convert glucose mg/dL to mmol/L, multiply by 0.0556. dp–ucMGP is reported as geometric mean (interquartile range). To convert dp–ucMGP from μg/L into pmol/L, multiply by 94.299.

dp–ucMGP indicates nonphosphorylated and uncarboxylated matrix Gla protein.

Incidence of Events

Median follow-up was 14.1 years (5th to 95th percentile interval, 7.1–16.2 years). Mortality included 70 (35.5%) cardiovascular deaths, 110 (55.8%) noncardiovascular deaths, including 58 (32.2%) because of cancer and 5 (2.5%) because of renal failure, and 17 deaths (8.7%) from unknown causes. Of 180 first cardiovascular events, 70 were fatal and 110 nonfatal. Details on the incidence of cardiovascular events are given in Table S4. In total, 85 coronary events occurred, including 13 fatal and 11 nonfatal cases of acute myocardial infarction, 1 acute coronary syndrome, 6 deaths caused by ischemic cardiomyopathy, 5 sudden deaths, and 49 cases of coronary revascularization. The 127 cardiac events additionally included 21 fatal and 17 nonfatal cases of heart failure and 4 deaths caused by atrioventricular block.

Analyses by Thirds of the dp–ucMGP Distribution

With higher dp–ucMGP category (Figure S4), the sex- and age-adjusted rates of total (P=0.014), non-cancer (P<0.001), and cardiovascular (P=0.001) mortality increased, whereas they slightly decreased for coronary events (P=0.10) and showed no trend for noncardiovascular mortality (P=0.62) and for fatal combined with nonfatal cardiovascular, cardiac, and cerebrovascular events (P≥0.30). The Kaplan–Meier survival function estimates and log-rank test (Figure S5) confirmed that all-cause, non-cancer, and cardiovascular mortality increased (P<0.001) with higher dp–ucMGP levels, whereas there was no association between coronary events and dp–ucMGP (P=0.21). In analyses of total and non-cancer mortality adjusted for sex and age (Figure 1), event rates were lower in the medium compared with the top third of the dp–ucMGP distribution (P<0.001), but slightly higher (P≥0.15) in the low compared with the medium third, suggesting a J-shaped association. With higher dp–ucMGP category (Figure 1), event rates increased for cardiovascular mortality (P=0.001), but decreased for coronary events (P=0.016).

Figure 1.

Figure 1. Incidence of total mortality (A), non-cancer mortality (B), cardiovascular mortality (C), and coronary events (D) by thirds of the distribution of nonphosphorylated and uncarboxylated matrix Gla protein (dp–ucMGP). Incidence was adjusted for sex and age. Vertical bars denote the standard error. P values denote the significance of the differences between the medium and extreme thirds (A and B) or the overall difference between thirds of the dp–ucMGP distribution (C and D).

Multivariable-Adjusted Analyses

In Cox regression, we accounted for family clusters as a random effect and we adjusted for sex, age, body mass index, systolic blood pressure, heart rate, smoking and drinking, total cholesterol, diabetes mellitus, antihypertensive drug treatment, and history of cardiovascular disease. Because the association of total and non-cancer mortality with dp–ucMGP was curvilinear (Figure 1), we modeled the hazard ratios for total and non-cancer mortality using both a linear and squared term of dp–ucMGP. Adding the squared term significantly improved the model fit for total mortality (χ2 statistic of the log likelihood ratio, 9.54; P=0.008) and non-cancer mortality (16.7; P<0.001), but not for any other event (χ2 statistic ≤4.83; P≥0.089). All models met the proportional hazard assumption (P≥0.21).

The risk of total mortality and non-cancer mortality were curvilinearly related to baseline dp–ucMGP (Table 2 and Figure S6), with the nadir of the relation at 1.43 μg/L (135 pmol/L) and 0.97 μg/L (91.2 pmol/L), respectively. Doubling of the nadir value increased the risk of all-cause and non-cancer death by 15.0% (95% confidence interval, 6.9–25.3) and 21.5% (11.1–32.9), respectively. Cardiovascular mortality log-linearly increased and the risk of coronary events log-linearly decreased with higher baseline dp–ucMGP (Table 2). Doubling of dp–ucMGP entailed a 14% (P=0.027) higher risk of a cardiovascular death, but a 7.0% (P=0.021) lower risk of a coronary event. The association of cardiovascular and cardiac events with dp–ucMGP did not reach significance before (P≥0.39) or after (P≥0.11) excluding coronary events from these composite end points. Sensitivity analyses accounted for dropout bias were confirmatory (Table S5). Levels of dp–ucMGP, ranging from 1.43 μg/L (135 pmol/L) to 4.63 μg/L (437 pmol/L) were not associated with increased risk of mortality (Figure S7).

Table 2. Adjusted Hazard Ratios Associated with Baseline dp–ucMGP

EventHazard Ratio (95% CI)P Value
Total mortality
 Linear term1.06 (1.01–1.11)0.014
 Squared term1.02 (1.01–1.03)<0.001
Non-cancer mortality
 Linear term1.10 (1.04–1.16)<0.001
 Squared term1.02 (1.01–1.03)<0.001
Cardiovascular mortality1.14 (1.01–1.28)0.027
Noncardiovascular mortality0.98 (0.91–1.06)0.65
Cardiovascular events
 All0.99 (0.94–1.05)0.87
 Excluding coronary events1.07 (0.98–1.16)0.11
Cardiac events
 All0.97 (0.92–1.03)0.39
 Excluding coronary events1.08 (0.94–1.25)0.28
Coronary events0.93 (0.88–0.99)0.021
Stroke0.98 (0.81–1.19)0.83

Hazard ratios (95% confidence interval) express the risk associated with a doubling of dp–ucMGP, account for family clusters, and were adjusted for sex, age, body mass index, systolic blood pressure, heart rate, smoking and drinking, total cholesterol, diabetes mellitus, antihypertensive drug treatment, and history of cardiovascular disease. P values are for the significance of the hazard ratios. Models for total and non-cancer mortality include both the linear and squared terms of dp–ucMGP because these associations were J-shaped (Figure S6). Adding the squared term significantly improved the model fit for total (χ2 statistic of the log likelihood ratio, 9.54; P=0.008) and non-cancer (16.7; P<0.001) mortality.

CI indicates confidence interval; and dp–ucMGP, nonphosphorylated and uncarboxylated matrix Gla protein.

Attributable Risk

To compute attributable risk, we chose the 66th percentile of the dp–ucMGP distribution as a cut-off point. With adjustments applied as before, the attributable and population-attributable risks associated with highest third of the dp–ucMGP distribution were 26.5% and 10.9% for total mortality, 42.4% and 20.0% for non-cancer mortality, and 54.3% and 28.8% for cardiovascular mortality. For smoking, the corresponding estimates were 54.2% and 22.9% for total mortality, 58.7% and 26.3% for non-cancer mortality, and 45.9% and 17.5% for cardiovascular mortality.

Mendelian Randomization Study

While accounting for family clusters and with adjustments applied for age, body mass index, smoking and drinking (Table S6), dp–ucMGP correlated significantly (P≤0.001) with rs2098435, rs4236, and rs2430692, major allele carriers having higher levels, whereas there was no association between dp–ucMGP and rs1800802 (P=0.44). A sensitivity analysis in 819 unrelated founders was confirmatory (Table S6). None of the health outcomes was associated with the MGP SNPs (Table S7). With adjustments applied as in Table 2, the instrumental variable analysis (Table 3) did not support a causal association of all-cause and cardiovascular mortality with dp–ucMGP (0.13≤P≤0.96). However, for non-cancer mortality, estimates of the causal hazard ratios reached significance (P≤0.01) for all 3 instrumental variables. Finally, using rs2098435 as the instrumental variable, the causal hazard ratio for coronary events was 0.75 (95% confidence interval, 0.59–0.96; P=0.022), confirming the inverse association between coronary events and dp–ucMGP. While accounting for dropout bias, the hazard ratio for coronary events was 0.76 (0.58–0.98; P=0.037). The F-statistics for rs2098435, rs4236, and rs2430692 were 8.71, 9.85, and 13.5, respectively, and 0.56 for rs1800802.

Table 3. Instrumental Variable Estimates of the Causal Hazard Ratios Relating Health Outcomes to Baseline dp–ucMGP

CharacteristicVariablers2098435rs4236rs2430692
Total mortality (N =194)Ivar0.89 (0.75–1.06)0.87 (0.74–1.02)*0.89 (0.76–1.03)
Ivar21.05 (1.00–1.09)1.04 (1.00–1.08)*1.03 (0.99–1.07)
P0.130.130.23
Non-cancer mortality (N=136)Ivar0.82 (0.67–1.01)*0.79 (0.64–0.96)0.81 (0.68–0.98)
Ivar21.09 (1.03–1.15)1.09 (1.03–1.14)1.07 (1.02–1.13)
P0.0070.0040.010
Cardiovascular mortality (N=67)Ivar0.95 (0.73–1.25)0.91 (0.70–1.18)0.89 (0.70–1.14)
Coronary events (N=85)Ivar0.75 (0.59–0.96)0.87 (0.71–1.08)0.85 (0.70–1.03)*

Ivar and Ivar2 refer to the linear and squared terms of dp–ucMGP predicted from linear regression models, including the MGP genotype as instrumental variable, and age, body mass index, and smoking and drinking. P values derived from a log likelihood test refer to the contribution of Ivar combined with Ivar2 to the model fit. Values are hazard ratios (95% confidence interval), adjusted for sex, age, body mass index, systolic blood pressure, heart rate, smoking and drinking, total cholesterol, diabetes mellitus, antihypertensive drug treatment, and history of cardiovascular disease. The analyses included 2304 participants genotyped for MGP. N refers to the number of events.

dp–ucMGP indicates nonphosphorylated and uncarboxylated matrix Gla protein.

Significance of the hazard ratios:

*P≤0.10;

P≤0.05; and

P≤0.01.

Discussion

To our knowledge, our FLEMENGHO study is the first to relate adverse health outcomes to circulating levels of dp–ucMGP in a general population. Over 14.1 years of follow-up, dp–ucMGP curvilinearly predicted all-cause and non-cancer mortality with a nadir at 1.43 μg/L (135 pmol/L) and 0.97 μg/L (91.2 pmol/L), respectively. Baseline dp–ucMGP linearly predicted cardiovascular mortality, but was inversely related to the risk of coronary events. Mendelian randomization analyses suggested that the association of non-cancer mortality and coronary events with dp–ucMGP is causal.

Five studies1013,24 reported on the association between mortality and dp–ucMGP in patients with aortic stenosis,12 chronic10 or end-stage kidney disease,24 chronic heart failure,11 or overt vascular disease.13 The associations between all-cause mortality and dp–ucMGP were positive,1013 except in patients with advanced renal dysfunction, in whom this association was not significant.24 These reports on selected patients1013,24 are difficult to generalize. A substantial proportion of patients were taking vitamin K antagonists. This explains why median dp–ucMGP levels ranged from 9.10 μg/L (859 pmol/L)13 to 30.2 μg/L (2850 pmol/L)11 and why they were substantially higher than in our study participants. These studies only captured the top end of the dose–response relation between mortality and circulating dp–ucMGP and do not contradict the J-shaped association as observed in our current study.

The inverse association between coronary events and dp–ucMGP, although probably causal, at first sight seems counterintuitive. Active MGP is a powerful inhibitor of arterial calcification.6 Intravascular ultrasound studies demonstrated that elderly patients have more severe coronary calcifications and diffuse atherosclerosis, whereas younger patients have a more unstable coronary plaque morphology.25 The higher degree of coronary calcification in the elderly is in line with the age-related increase in inactive dp–ucMGP. In morphological studies, coronary plaque calcification was not associated with instability, whereas ruptured plaques showed little calcification.26 Furthermore, the number of coronary arterial segments with noncalcified plaque as assessed by computed tomography was associated with a 30%27 to 71%28 higher incidence of coronary27 or cardiac28 events in patients with suspected27,28 or known28 coronary artery disease. In view of the currently available literature, we assume that inhibition of plaque calcification by active MGP is a risk factor in the coronary circulation,2528 whereas in large conduit arteries, such as the femoral artery,4 the opposite is the case. An alternative explanation for the inverse association between coronary events and dp–ucMGP levels might reside in the vitamin K–dependent activation of growth arrest–specific protein 6 (Gas6).29 In animal studies, Gas6 antagonism protected mice against thromboembolism and inflammation.30,31 Thus, lack of vitamin K in the arterial wall, as reflected by high level of dp–ucMGP, decreases the levels of active Gas6 and increases protection against thrombosis and inflammation, processes that both play an important role in the pathogenesis of coronary heart disease. Dalmeijer et al14 investigated the association between coronary events and dp–ucMGP in a case–cohort study consisting of 1154 patients and 1406 participants. The dp–ucMGP level was not related to the risk of coronary events. These investigators found similar result in 518 type-2 diabetic patients.9 The nonsignificant association between coronary events and dp–ucMGP might be because of overlap of coronary events between cases and controls (n=98)14 or because of the selection of diabetic patients.9

The inverse association between coronary events and dp–ucMGP explained why the association of all of the fatal and nonfatal cardiovascular events with dp–ucMGP did not reach significance. On the other hand, cardiovascular mortality log-linearly increased with circulating dp–ucMGP. However, myocardial infarction and ischemic cardiomyopathy only contributed 23.8% of the overall cardiovascular mortality. The Mendelian randomization analysis confirmed a curvilinear association between non-cancer mortality and dp–ucMGP. The differential expression of MGP in human cancer cells3235 and the role of vitamin K–dependent Gas6 as ligand of the tyrosine kinase receptors (Axl, Tyro3, and Mer)36,37 might explain why non-cancer mortality, but not all-cause and cancer mortality, are related to the predicted dp–ucMGP. MGP mRNA is downregulated in human colorectal adenocarcinomas,32 but overexpressed in human breast cancer,33 urogenital malignancies,34 and glioblastoma.35 In many systems, Gas6 inhibits cellular proliferation and promotes survival,37 whereas in other models, GAS6 stimulates proliferation,36 dependent on the expression of the Axl, Tyro3, and Mer receptors.

The clinical implications of our current findings might be viewed from the vintage point of individuals and populations. The multivariable-adjusted attributable and population attributable risks of non-cancer mortality in our participants in the high third of the dp–ucMGP distribution were 42.4% and 20.0%. These estimates might be compared with the corresponding risks of smoking, which were 58.7% and 26.3%, respectively. High levels of plasma dp–ucMGP are a proxy for vitamin K deficiency,38,39 levels ranging from ≈1.4 to ≈4.6 μg/L (≈130 to ≈437 pmol/L), probably being safe. In our population, the 4.6 μg/L threshold corresponded with 63th percentile of dp–ucMGP distribution. The recommended dietary allowance for vitamin K is 1 μg per kilogram of body weight per day, which is sufficient to ensure normal hemostasis.40 However, in apparently healthy subjects, a substantial fraction of MGP remains in the uncarboxylated forms.41,42 This raises the question as to whether the present recommended dietary allowance for vitamin K intake is sufficient to prevent non-cancer mortality.

The present study must be interpreted within the context of some potential limitations. First, our Mendelian randomization study did not confirm a causal association of cardiovascular mortality with dp–ucMGP. However, cardiovascular mortality represented only one third of total mortality and included fatal coronary events, so that our study might have been underpowered in the Mendelian analysis of cardiovascular death. Second, our estimates of attributable for dp–ucMGP might be an overestimate, but this estimate can still be compared with that of smoking obtained in the same population, using the same multivariable-adjusted models. Third, we did not measure the vitamin K status, which is usually evaluated by administering food frequency questionnaires and correlating the information so obtained with available databases of nutrients. However, dp–ucMGP is highly correlated (r=0.62; P<0.0001) with measures of vitamin K deficiency.24 Finally, we assumed that the inverse association between coronary events and dp–ucMGP might be explained by another vitamin K–dependent protein, Gas6. However, we did not measure Gas6 in our study. Further studies are required to confirm our speculation.

In conclusion, in the general population, dp–ucMGP predicts total, non-cancer and cardiovascular mortality and lower coronary risk. These associations are probably causal for non-cancer mortality and coronary events. Our findings, pending confirmation, might open new ways to improve health by vitamin K substitution in addition to the management of classical cardiovascular risk factors.

Acknowledgments

We gratefully acknowledge the clerical assistance of Annick De Soete (Studies Coordinating Centre, Leuven, Belgium) and the contribution of Linda Custers, Marie-Jeanne Jehoul, Daisy Thijs, and Hanne Truyens in data collection at the examination center (Eksel, Belgium).

Footnotes

This article was sent to David A. Calhoun, Guest Editor, for review by expert referees, editorial decision, and final disposition.

The online-only Data Supplement is available with this article at http://hyper.ahajournals.org/lookup/suppl/doi:10.1161/HYPERTENSIONAHA.114.04494/-/DC1.

Correspondence to Jan A. Staessen, Studies Coordinating Centre, Research Unit Hypertension and Cardiovascular Epidemiology, KU Leuven Department of Cardiovascular Sciences, University of Leuven, Campus Sint Rafaël, Kapucijnenvoer 35, Box 7001, BE-3000 Leuven, Belgium. E-mail and

References

  • 1. Lopez AD, Mathers CD, Ezzati M, Jamison DT, Murray CJ. Global and regional burden of disease and risk factors, 2001: systematic analysis of population health data.Lancet. 2006; 367:1747–1757.CrossrefMedlineGoogle Scholar
  • 2. Raggi P, Callister TQ, Cooil B, He ZX, Lippolis NJ, Russo DJ, Zelinger A, Mahmarian JJ. Identification of patients at increased risk of first unheralded acute myocardial infarction by electron-beam computed tomography.Circulation. 2000; 101:850–855.LinkGoogle Scholar
  • 3. Vliegenthart R, Hollander M, Breteler MM, van der Kuip DA, Hofman A, Oudkerk M, Witteman JC. Stroke is associated with coronary calcification as detected by electron-beam CT: the Rotterdam Coronary Calcification Study.Stroke. 2002; 33:462–465.LinkGoogle Scholar
  • 4. Lehto S, Niskanen L, Suhonen M, Rönnemaa T, Laakso M. Medial artery calcification. A neglected harbinger of cardiovascular complications in non-insulin-dependent diabetes mellitus.Arterioscler Thromb Vasc Biol. 1996; 16:978–983.LinkGoogle Scholar
  • 5. Hackeng TM, Rosing J, Spronk HM, Vermeer C. Total chemical synthesis of human matrix Gla protein.Protein Sci. 2001; 10:864–870.CrossrefMedlineGoogle Scholar
  • 6. Schurgers LJ, Cranenburg EC, Vermeer C. Matrix Gla-protein: the calcification inhibitor in need of vitamin K.Thromb Haemost. 2008; 100:593–603.CrossrefMedlineGoogle Scholar
  • 7. Price PA, Faus SA, Williamson MK. Warfarin causes rapid calcification of the elastic lamellae in rat arteries and heart valves.Arterioscler Thromb Vasc Biol. 1998; 18:1400–1407.LinkGoogle Scholar
  • 8. Schurgers LJ, Aebert H, Vermeer C, Bültmann B, Janzen J. Oral anticoagulant treatment: friend or foe in cardiovascular disease?Blood. 2004; 104:3231–3232.CrossrefMedlineGoogle Scholar
  • 9. Dalmeijer GW, van der Schouw YT, Magdeleyns EJ, Vermeer C, Verschuren WM, Boer JM, Beulens JW. Matrix Gla protein species and risk of cardiovascular events in type 2 diabetic patients.Diabetes Care. 2013; 36:3766–3771.CrossrefMedlineGoogle Scholar
  • 10. Schurgers LJ, Barreto DV, Barreto FC, Liabeuf S, Renard C, Magdeleyns EJ, Vermeer C, Choukroun G, Massy ZA. The circulating inactive form of matrix gla protein is a surrogate marker for vascular calcification in chronic kidney disease: a preliminary report.Clin J Am Soc Nephrol. 2010; 5:568–575.CrossrefMedlineGoogle Scholar
  • 11. Ueland T, Dahl CP, Gullestad L, Aakhus S, Broch K, Skårdal R, Vermeer C, Aukrust P, Schurgers LJ. Circulating levels of non-phosphorylated undercarboxylated matrix Gla protein are associated with disease severity in patients with chronic heart failure.Clin Sci (Lond). 2011; 121:119–127.CrossrefMedlineGoogle Scholar
  • 12. Ueland T, Gullestad L, Dahl CP, Aukrust P, Aakhus S, Solberg OG, Vermeer C, Schurgers LJ. Undercarboxylated matrix Gla protein is associated with indices of heart failure and mortality in symptomatic aortic stenosis.J Intern Med. 2010; 268:483–492.CrossrefMedlineGoogle Scholar
  • 13. Mayer O, Seidlerová J, Bruthans J, Filipovský J, Timoracká K, Vaněk J, Cerná L, Wohlfahrt P, Cífková R, Theuwissen E, Vermeer C. Desphospho-uncarboxylated matrix Gla-protein is associated with mortality risk in patients with chronic stable vascular disease.Atherosclerosis. 2014; 235:162–168.CrossrefMedlineGoogle Scholar
  • 14. Dalmeijer GW, van der Schouw YT, Magdeleyns EJ, Vermeer C, Verschuren WM, Boer JM, Beulens JW. Circulating desphospho-uncarboxylated matrix γ-carboxyglutamate protein and the risk of coronary heart disease and stroke.J Thromb Haemost. 2014; 12:1028–1034.CrossrefMedlineGoogle Scholar
  • 15. van den Heuvel EG, van Schoor NM, Lips P, Magdeleyns EJ, Deeg DJ, Vermeer C, den Heijer M. Circulating uncarboxylated matrix Gla protein, a marker of vitamin K status, as a risk factor of cardiovascular disease.Maturitas. 2014; 77:137–141.CrossrefMedlineGoogle Scholar
  • 16. Staessen JA, Wang JG, Brand E, Barlassina C, Birkenhäger WH, Herrmann SM, Fagard R, Tizzoni L, Bianchi G. Effects of three candidate genes on prevalence and incidence of hypertension in a Caucasian population.J Hypertens. 2001; 19:1349–1358.CrossrefMedlineGoogle Scholar
  • 17. Stolarz-Skrzypek K, Kuznetsova T, Thijs L, Tikhonoff V, Seidlerová J, Richart T, Jin Y, Olszanecka A, Malyutina S, Casiglia E, Filipovský J, Kawecka-Jaszcz K, Nikitin Y, Staessen JA; European Project on Genes in Hypertension (EPOGH) Investigators. Fatal and nonfatal outcomes, incidence of hypertension, and blood pressure changes in relation to urinary sodium excretion.JAMA. 2011; 305:1777–1785.CrossrefMedlineGoogle Scholar
  • 18. Cranenburg EC, Koos R, Schurgers LJ, Magdeleyns EJ, Schoonbrood TH, Landewé RB, Brandenburg VM, Bekers O, Vermeer C. Characterisation and potential diagnostic value of circulating matrix Gla protein (MGP) species.Thromb Haemost. 2010; 104:811–822.CrossrefMedlineGoogle Scholar
  • 19. Citterio L, Simonini M, Zagato L, Salvi E, Delli Carpini S, Lanzani C, Messaggio E, Casamassima N, Frau F, D’Avila F, Cusi D, Barlassina C, Manunta P. Genes involved in vasoconstriction and vasodilation system affect salt-sensitive hypertension.PLoS One. 2011; 6:e19620.CrossrefMedlineGoogle Scholar
  • 20. Holtzman NA, Marteau TM. Will genetics revolutionize medicine?N Engl J Med. 2000; 343:141–144.CrossrefMedlineGoogle Scholar
  • 21. Palmer TM, Sterne JA, Harbord RM, Lawlor DA, Sheehan NA, Meng S, Granell R, Smith GD, Didelez V. Instrumental variable estimation of causal risk ratios and causal odds ratios in Mendelian randomization analyses.Am J Epidemiol. 2011; 173:1392–1403.CrossrefMedlineGoogle Scholar
  • 22. Greenland S. An introduction to instrumental variables for epidemiologists.Int J Epidemiol. 2000; 29:722–729.CrossrefMedlineGoogle Scholar
  • 23. Burgess S, Thompson SG; CRP CHD Genetics Collaboration. Avoiding bias from weak instruments in Mendelian randomization studies.Int J Epidemiol. 2011; 40:755–764.CrossrefMedlineGoogle Scholar
  • 24. Schlieper G, Westenfeld R, Krüger T, Cranenburg EC, Magdeleyns EJ, Brandenburg VM, Djuric Z, Damjanovic T, Ketteler M, Vermeer C, Dimkovic N, Floege J, Schurgers LJ. Circulating nonphosphorylated carboxylated matrix gla protein predicts survival in ESRD.J Am Soc Nephrol. 2011; 22:387–395.CrossrefMedlineGoogle Scholar
  • 25. Hong YJ, Jeong MH, Ahn Y, Sim DS, Chung JW, Cho JS, Yoon NS, Yoon HJ, Moon JY, Kim KH, Park HW, Kim JH, Cho JG, Park JC, Kang JC. Age-related differences in intravascular ultrasound findings in 1,009 coronary artery disease patients.Circ J. 2008; 72:1270–1275.CrossrefMedlineGoogle Scholar
  • 26. Virmani R, Burke AP, Farb A, Kolodgie FD. Pathology of the unstable plaque.Prog Cardiovasc Dis. 2002; 44:349–356.CrossrefMedlineGoogle Scholar
  • 27. van Werkhoven JM, Schuijf JD, Gaemperli O, Jukema JW, Kroft LJ, Boersma E, Pazhenkottil A, Valenta I, Pundziute G, de Roos A, van der Wall EE, Kaufmann PA, Bax JJ. Incremental prognostic value of multi-slice computed tomography coronary angiography over coronary artery calcium scoring in patients with suspected coronary artery disease.Eur Heart J. 2009; 30:2622–2629.CrossrefMedlineGoogle Scholar
  • 28. Gaemperli O, Valenta I, Schepis T, Husmann L, Scheffel H, Desbiolles L, Leschka S, Alkadhi H, Kaufmann PA. Coronary 64-slice CT angiography predicts outcome in patients with known or suspected coronary artery disease.Eur Radiol. 2008; 18:1162–1173.CrossrefMedlineGoogle Scholar
  • 29. McCann JC, Ames BN. Vitamin K, an example of triage theory: is micronutrient inadequacy linked to diseases of aging?Am J Clin Nutr. 2009; 90:889–907.CrossrefMedlineGoogle Scholar
  • 30. Angelillo-Scherrer A, de Frutos P, Aparicio C, Melis E, Savi P, Lupu F, Arnout J, Dewerchin M, Hoylaerts M, Herbert J, Collen D, Dahlbäck B, Carmeliet P. Deficiency or inhibition of Gas6 causes platelet dysfunction and protects mice against thrombosis.Nat Med. 2001; 7:215–221.CrossrefMedlineGoogle Scholar
  • 31. Tjwa M, Bellido-Martin L, Lin Y, et al. Gas6 promotes inflammation by enhancing interactions between endothelial cells, platelets, and leukocytes.Blood. 2008; 111:4096–4105.CrossrefMedlineGoogle Scholar
  • 32. Fan C, Sheu D, Fan H, Hsu K, Allen Chang C, Chan E. Down-regulation of matrix Gla protein messenger RNA in human colorectal adenocarcinomas.Cancer Lett. 2001; 165:63–69.CrossrefMedlineGoogle Scholar
  • 33. Chen L, O’Bryan JP, Smith HS, Liu E. Overexpression of matrix Gla protein mRNA in malignant human breast cells: isolation by differential cDNA hybridization.Oncogene. 1990; 5:1391–1395.MedlineGoogle Scholar
  • 34. Levedakou EN, Strohmeyer TG, Effert PJ, Liu ET. Expression of the matrix Gla protein in urogenital malignancies.Int J Cancer. 1992; 52:534–537.CrossrefMedlineGoogle Scholar
  • 35. Mertsch S, Schurgers LJ, Weber K, Paulus W, Senner V. Matrix gla protein (MGP): an overexpressed and migration-promoting mesenchymal component in glioblastoma.BMC Cancer. 2009; 9:302.CrossrefMedlineGoogle Scholar
  • 36. Holland SJ, Powell MJ, Franci C, et al. Multiple roles for the receptor tyrosine kinase axl in tumor formation.Cancer Res. 2005; 65:9294–9303.CrossrefMedlineGoogle Scholar
  • 37. Neubauer A, Fiebeler A, Graham DK, O’Bryan JP, Schmidt CA, Barckow P, Serke S, Siegert W, Snodgrass HR, Huhn D. Expression of axl, a transforming receptor tyrosine kinase, in normal and malignant hematopoiesis.Blood. 1994; 84:1931–1941.CrossrefMedlineGoogle Scholar
  • 38. Dalmeijer GW, van der Schouw YT, Magdeleyns E, Ahmed N, Vermeer C, Beulens JW. The effect of menaquinone-7 supplementation on circulating species of matrix Gla protein.Atherosclerosis. 2012; 225:397–402.CrossrefMedlineGoogle Scholar
  • 39. Westenfeld R, Krueger T, Schlieper G, Cranenburg EC, Magdeleyns EJ, Heidenreich S, Holzmann S, Vermeer C, Jahnen-Dechent W, Ketteler M, Floege J, Schurgers LJ. Effect of vitamin K2 supplementation on functional vitamin K deficiency in hemodialysis patients: a randomized trial.Am J Kidney Dis. 2012; 59:186–195.CrossrefMedlineGoogle Scholar
  • 40. Cranenburg EC, Schurgers LJ, Vermeer C. Vitamin K: the coagulation vitamin that became omnipotent.Thromb Haemost. 2007; 98:120–125.CrossrefMedlineGoogle Scholar
  • 41. Schurgers LJ, Teunissen KJ, Knapen MH, Kwaijtaal M, van Diest R, Appels A, Reutelingsperger CP, Cleutjens JP, Vermeer C. Novel conformation-specific antibodies against matrix gamma-carboxyglutamic acid (Gla) protein: undercarboxylated matrix Gla protein as marker for vascular calcification.Arterioscler Thromb Vasc Biol. 2005; 25:1629–1633.LinkGoogle Scholar
  • 42. Knapen MH, Schurgers LJ, Vermeer C. Vitamin K2 supplementation improves hip bone geometry and bone strength indices in postmenopausal women.Osteoporos Int. 2007; 18:963–972.CrossrefMedlineGoogle Scholar

Novelty and Significance

What Is New?

  • Vascular smooth muscle cells synthesize matrix Gla protein (MGP). After vitamin K dependent activation, MGP behaves as a potent inhibitor of arterial calcification. No population study assessed the risk of adverse health outcomes in relation to circulating levels of inactive nonphosphorylated and uncarboxylated MGP (dp-ucMGP).

What Is Relevant?

  • In 2318 participants over 14.1 years of follow-up (median), dp-ucMGP curvilinearly predicted all-cause and non-cancer mortality with a nadir at 1.43 mg/L and 0.97 mg/L, respectively. Baseline dp-ucMGP linearly predicted cardiovascular mortality, but was inversely related to the risk of coronary events. Mendelian randomization analyses suggested that the associations of non-cancer mortality and coronary events with dpucMGP are causal.

  • Levels of dp-ucMGP between 1.43 and 4.63 mg/L were not associated with increased risk of mortality. In our population, the upper threshold corresponded with 63th percentile of dp-ucMGP distribution.

Summary

In a general population, a substantial fraction of MGP remains in the inactive forms, which is related to adverse health outcomes. Vitamin K supplementation might be beneficial.

eLetters(0)

eLetters should relate to an article recently published in the journal and are not a forum for providing unpublished data. Comments are reviewed for appropriate use of tone and language. Comments are not peer-reviewed. Acceptable comments are posted to the journal website only. Comments are not published in an issue and are not indexed in PubMed. Comments should be no longer than 500 words and will only be posted online. References are limited to 10. Authors of the article cited in the comment will be invited to reply, as appropriate.

Comments and feedback on AHA/ASA Scientific Statements and Guidelines should be directed to the AHA/ASA Manuscript Oversight Committee via its Correspondence page.