Cardiovascular and Metabolic Risk ORIGINAL ARTICLE

Matrix Gla Species and Risk of Cardiovascular Events in Type 2 Diabetic Patients

1 3 GEERTJE W. DALMEIJER, PHD W.M. MONIQUE VERSCHUREN, PHD reduced coronary calcification and 1 3 YVONNE T. VAN DER SCHOUW, PHD JOLANDA M.A. BOER, PHD – 2 1 reduced risk of CVD (6 9). These effects ELKE J. MAGDELEYNS, BSC JOLINE W.J. BEULENS, PHD 2 are thought to be mediated by increased CEES VERMEER, PHD activation of MGP (10). MGP exists as various species, which OBJECTIVEd differ in their state of phosphorylation To investigate the relationship of circulating matrix Gla protein (MGP) species or : phosphorylated, non- with incident cardiovascular disease (CVD) or coronary heart disease (CHD) in type 2 diabetic phosphorylated (desphospho-MGP patients. [dpMGP]), carboxylated (cMGP), or un- RESEARCH DESIGN AND METHODSdEPIC-NL is a prospective cohort study among carboxylated (ucMGP). Total uncarboxy- 40,011 Dutch men and women. At baseline (1993–1997), 518 participants were known to have lated MGP (t-ucMGP) is thought to be type 2 diabetes. MGP levels were measured by ELISA techniques in baseline plasma samples. The the sum of desphospho-uncarboxylated incidence of fatal and nonfatal CVD and CVD subtypesdCHD, peripheral arterial disease (PAD), MGP (dp-ucMGP) and phosphorylated- d heart failure, and stroke were obtained by linkage to national registers. Cox proportional uncarboxylated MGP (p-ucMGP) and hazard models were used to calculate hazard ratios (HRs), adjusted for sex, waist-to-hip ratio, mainly consists of p-ucMGP. physical activity, and history of CVD. Development of assays to measure RESULTSdDuring a median 11.2 years of follow-up, 160 cases of CVD were documented. circulating MGP species enabled the in- Higher circulating desphospho-uncarboxylated MGP (dp-ucMGP) levels were significantly as- vestigation of these species in the circula- – sociated with higher risk of CVD, with an HR per SD (HRSD)of1.21(95%CI1.06 1.38), PAD tion (11). These studies have shown that – – (HRSD 1.32 [95% CI 1.07 1.65]), and heart failure (HRSD 1.75 [95% CI 1.42 2.17]) after dp-ucMGP is a marker for sta- adjustment. Higher circulating dp-ucMGP levels were not related to risk of CHD (HRSD 1.12 tus, with high dp-ucMGP level reflecting a – – [95% CI 0.94 1.34]) or stroke (HRSD 1.05 [95% CI 0.73 1.49]). Circulating desphospho- low vitamin K status (12–17). In line with carboxylated MGP and circulating total-uncarboxylated MGP levels were not associated with these results, several studies indeed showed CVD or CVD subtypes. that high dp-ucMGP levels were associated CONCLUSIONSdHigh dp-ucMGP levels were associated with increased CVD risk among with more calcification, though not consis- type 2 diabetic patients, especially with the subtypes PAD and heart failure, while other MGP tently (13,16,18,19). Theoretically, dp- species were not related to CVD risk. These results suggest that a poor vitamin K status is cMGP forms the mirror image of dp-ucMGP associated with increased CVD risk. and is hypothesized to be associated with lower calcification, but results from human, – Diabetes Care 36:3766 3771, 2013 observational studies are inconsistent (16,20). Finally, in cross-sectional studies, oronary artery calcification is an 1) regulation of calcification by vascular high t-ucMGP has been associated with de- independent predictor of cardiovas- smooth muscle cell (VSMC)-derived matrix creased calcification (16,21–23). C fi cular disease (CVD) (1). Matrix Gla vesicles and apoptotic body, 2) inhibition of Although vascular calci cation may protein (MGP) is a vitamin K–dependent -phosphate precipitation, and 3)in- not be causally related to CVD, it has protein and a potent inhibitor of vascular hibition of VSMC trans differentiation (4). emerged as a strong and independent risk calcification (2). The importance of MGP Vitamin K is required for the function of marker for CVD (1), but the association of for vascular health has been demon- MGP through its role as a cofactor for MGP species with CVD events has not strated in MGP-deficient animals, who the enzyme g-glutamyl carboxylase, cat- been investigated to date. Diabetes is as- all died of massive arterial calcification alyzing the carboxylation of sociated with severe cardiovascular com- within 6–8 weeks after birth (3). The cel- residues (Glu) into g-carboxyglutamate plications, including vascular calcification lular and molecular mechanisms by (Gla) at five well-defined places in the and accelerated atherosclerosis, leading to which MGP prevents ectopic calcium de- protein (5). Human studies showed that increased morbidity and mortality in dia- position are multifaceted, including high vitamin K intake is associated with betic patients (24–27) Therefore, we performed a prospective study to investi- ccccccccccccccccccccccccccccccccccccccccccccccccc gate the association between circulating From the 1Julius Center for Health Sciences and Primary Care, University Medical Center Utrecht, Utrecht, the MGP species and CVD or coronary heart Netherlands; 2VitaK, Maastricht University, Maastricht, the Netherlands; and the 3National Institute for disease (CHD) risk among a high-risk Public Health and the Environment, Bilthoven, the Netherlands. population, i.e., type 2 diabetic patients. Corresponding author: Geertje W. Dalmeijer, [email protected]. Received 9 January 2013 and accepted 17 May 2013. DOI: 10.2337/dc13-0065 RESEARCH DESIGN AND © 2013 by the American Diabetes Association. Readers may use this article as long as the work is properly cited, the use is educational and not for profit, and the work is not altered. See http://creativecommons.org/ METHODSdThe EPIC-NL cohort is licenses/by-nc-nd/3.0/ for details. the Dutch contribution to the European

3766 DIABETES CARE, VOLUME 36, NOVEMBER 2013 care.diabetesjournals.org Dalmeijer and Associates

Prospective Investigation into Cancer and assay is based on the use of detection (Bosch & Son, Jungingen, Germany) Nutrition (EPIC) and consists of the monoclonal antibody directed against (Prospect-EPIC) or on the left arm using Prospect-EPIC and MORGEN-EPIC co- the ucMGP sequence 35–49 in human a random zero sphygmomanometer horts (28). The Prospect-EPIC study in- MGP (mAb-ucMGP; VitaK BV), while (MORGEN-EPIC), from which the mean cludes 17,357 women 49–70 years of age for dp-cMGP this was directed against was taken. Height and weight were mea- living in Utrecht and vicinity who partic- the cMGP sequence 35–54 in human sured, and BMI was calculated. Time ipated in the nationwide Dutch breast MGP (mAb-cMGP; VitaK BV). All mea- since diabetes diagnosis was calculated cancer screening program between 1993 surements were performed in duplicate, by subtracting the age of diagnosis from and 1997. The MORGEN-EPIC cohort and average values are given throughout theageatbaselineexamination.HbA1c consists of 22,654 men and women 21– this article. concentrations were measured in eryth- 64 years of age selected from random rocytes using an immunoturbidimetric samples of the Dutch population in three Outcome assessment latex test. All measurements were performed different towns. Participants were re- Morbidity follow-up data on CVD events according to standard operating procedures. cruited in both studies from 1993 to were obtained from the Dutch Centre for 1997. At baseline, a general and a food- Health Care Information, which holds a Data analysis frequency questionnaire were adminis- standardized computerized register of Participant characteristics are presented tered, and a physical examination was hospital discharge diagnoses. All diagno- as means with SDs or percentages. The performed for blood pressure measure- ses were coded according to the ICD-9. duration of follow-up was calculated as ments, anthropometry, and nonfasting Follow-up was complete until the first of the interval between date of study entry blood sampling. All participants provided January 2008. The database was linked to (1993–1997) and the occurrence of a car- informed consent before study inclusion. the cohort on the basis of birth date, sex, diovascular event, death, loss to follow- The study complies with the Declaration postal code, and general practitioner up, or 1 January 2008: whichever came of Helsinki and was approved by the in- with a validated probabilistic method first. stitutional board of the University Medical (30). Information on vital status was ob- Cox proportional hazard models Center Utrecht (Prospect) and the Medical tained through linkage with the munici- were used to calculate crude and adjusted Ethical Committee of The Netherlands Or- pal registries. Causes of death were hazard ratios (HRs) and 95% CIs for the ganization for Applied ScientificResearch collected from Statistics Netherlands. associations between MGP species (con- Nutrition and Food Research (MORGEN). End points for the present analysis were tinuous) and CVD or CVD subtypes. Three sources of ascertainment of CVD and the subtypes CHD, peripheral Potential confounding factors (age, sex, diabetes were used: self-report, hospital arterial disease (PAD), heart failure, and BMI, waist-to-hip ratio, smoking habits, discharge diagnoses, and urinary strip test stroke. CVD was defined as CHD, PAD, physical activity, education, systolic and (in the Prospect part of the cohort only). heart failure, and stroke (CVD, 410–414, diastolic blood pressure, and total cho- Potential cases of diabetes ascertained 427.5, 428, 415.1, 443.9, 430–438, 440– lesterol) were selected based on univari- by these sources were verified against 442, 444, 798.1, 798.2, and 798.9; CHD, able associations of potential confounders medical and pharmacy records. Details 410–414, 427.5, 798.1, 798.2, and with both CVD or CVD subtypes and of the ascertainment sources and verifi- 798.9; PAD, 440–444; heart failure, MGP species and whether adjustment cation procedures have previously been 428; and stroke, 430–434 and 436). for the confounder changed the HR described (29). These end points included both fatal by .10%. Variables selected this way At baseline, 615 participants were and nonfatal cases of CVD, CHD, PAD, were incorporated into a multivariate verified to have type 2 diabetes. Those heart failure, and stroke. model by means of a stepwise selection who did not give permission for linkage approach. The confounders that were en- with vital status registries were excluded Other measurements tered into the model were age, sex, waist- (n = 10). After exclusion of participants The general questionnaire contained to-hip ratio, and Cambridge physical with missing data on CVD (n = 25) and questions on demographic characteristics, activity index (CPAI) (model 1). The sec- blood samples (n = 62), 518 participants the presence of chronic diseases, and risk ond model additionally included history were left for analysis. factors for chronic diseases. Smoking was of CVD or CVD subtypes (model 2). In a categorized into current, past, and never separate model, we also checked whether MGP species smoker. Level of education was categorized adjusting for duration of diabetes, HbA1c, The measurement of plasma t-ucMGP as low (primary education up to those concentrations, blood lipids, and blood was performed with a competitive mono completing advanced elementary educa- pressure influenced the results. We re- antibody ELISA as previously described tion), average (intermediate vocational edu- peated the analyses after excluding not (11). This ELISA identifies all MGP spe- cation and higher general secondary verified type 2 diabetic patients (n = cies carrying the uncarboxylated Gla do- education) or high (higher vocational edu- 67). Moreover, analyses were repeated main, irrespective of whether the MGP cation and university). Physical activity was with outcomes fatal CVD (n = 36) and species are phosphorylated or fragmented. assessed using a questionnaire validated in all-cause mortality (n =114). In contrast to t-ucMGP, analyses of dp- an elderly population and categorized ac- The possibility of a nonlinear relation ucMGP and dp-cMGP were conducted cording to the Cambridge Physical Activity was examined nonparametrically with with a dual-antibody (“sandwich”) ELISA. Index (31). During the baseline physical ex- restricted cubic splines (32), and no evi- Both assays use a monoclonal antibody amination screening, systolic and diastolic dence for nonlinear associations was against the dpMGP sequence 3–15 as cap- blood pressure measurements were per- found. ture antibody (mAb-dpMGP; VitaK BV, formed twice in the supine position on For handling missing data for con- Maastricht, the Netherlands). The dp-ucMGP the right arm using a Boso Oscillomat founders, we used multiple imputations. care.diabetesjournals.org DIABETES CARE, VOLUME 36, NOVEMBER 2013 3767 MGP and risk of CVD in diabetic patients

We assumed that the missing data were at Table 2 shows the association be- poor vitamin K status, were associated random. We generalized 10 imputed tween MGP species and CVD or CVD sub- with increased CVD risk, especially with datasets and used Rubin rules to combine types risk. After adjustment for age, sex, subtypes PAD and heart failure, in type 2 the estimates of the parameters (33). Two- waist-to-hip ratio, CPAI, and history of diabetic patients. We did not observe a sided P values ,0.05 were considered CVD, higher circulating dp-ucMGP levels significant association between dp-cMGP statistically significant. All statistical anal- were significantly associated with higher or t-ucMGP and CVD risk. yses were conducted using IBM SPSS risk of CVD, with an HR per SD (HRSD)of No previous studies have investigated (version 20 for Windows). 1.21 (95% CI 1.06–1.38, P = 0.01). the association between dp-ucMGP levels Higher circulating dp-ucMGP levels and CVD events; thus far, only the asso- RESULTSdTable 1 shows the baseline were significantly associated with higher ciation between dp-ucMGP and mortality characteristics of the study population. risk of CHD in crude analyses (HRSD 1.24 has been investigated. These studies have The mean age of the study population [95% CI 1.06–1.45], P = 0.01). After full been performed in high-risk populations was 58.1 years, and 17.8% were men. adjustment, the association attenuated to such as patients with aortic stenosis or The diabetes duration was on average nonsignificant (HRSD 1.12 [95% CI 0.94– heart failure and generally showed strong P 6.3 years, and mean HbA1c was 8.0%. 1.34], = 0.21). Higher circulating dp- positive associations between dp-ucMGP The medians of the circulating MGP levels ucMGP levels were significantly associated and risk of total or cardiovascular mor- of the different species were 156 pmol/L with higher risk of PAD (HRSD 1.32 [95% tality (18,34,35). Only a study by with an interquartile range (IQR) of 91– CI 1.07–1.65], P = 0.01) and heart failure Schlieper et al. (20) found that hemodial- 258 for dp-ucMGP, 1,062 pmol/L for (HRSD 1.75 [95% CI 1.42–2.17], P , 0.001) ysis patients with low dp-ucMGP levels dp-cMGP (IQR 716–1,240), and 4,308 but not with the risk of stroke (HRSD 1.05 had a nonsignificant HR of 1.71 (95% nmol/L for t-ucMGP (IQR 3,514–5,079). [95% CI 0.73–1.49]) after full adjustment CI 0.92–3.17) for all-cause mortality During a mean follow-up of 9.9 years 160 (Fig. 1). and a nonsignificant HR of 1.83 (95% incident cases of CVD, 99 incident cases Circulating dp-cMGP levels were not CI 0.90–3.70) for cardiovascular mortal- of CHD, 38 incident cases of PAD, 28 in- associated with CVD risk either in crude ity, but these HRs were not adjusted for fi cident cases of heart failure, and 26 inci- analyses or after full adjustment (HRSD other risk factors. Our study con rmed dent cases of stroke were documented. 0.96 [95% CI 0.81–1.14], P = 0.64). that high dp-ucMGP levels are associated They were borderline significantly associ- with higher mortality risk. In line with ated with a lower CHD risk (HRSD 0.78 most studies on mortality, several studies, Table 1dBaseline characteristics of 518 [95% CI 0.58–1.03], P = 0.08) after full but not all, have shown that high dp- diabetic patients adjustment. There was no association be- ucMGP levels are associated with in- tween circulating dp-cMGP levels and creased calcification (13,16,18,19). We – fi N PAD (HRSD 1.16 [95% CI 0.92 1.45]), now extend these ndings by showing heart failure (HRSD 1.07 [95% CI 0.80– that high circulating dp-ucMGP levels – Follow-up (years) 11.2 (7.4 12.7) 518 1.42]) or stroke (HRSD 0.85 [95% CI are also associated with increased CVD Age (years) 58.1 6 7.1 518 0.52–1.40]) after full adjustment. risk among diabetic patients. Since stud- Male sex (%) 17.8 518 No associations between circulating t- ies have consistently shown that high dp- BMI (kg/m2) 29.6 6 4.9 516 ucMGP levels and the risk of CVD or CVD ucMGP levels are associated with a poor Waist-to-hip ratio 0.89 6 0.08 515 subtypes were observed. vitamin K status (12–17), these results Currently Restricting the outcome to CVD mor- suggest that a poor vitamin K status is smoking (%) 23.4 484 tality, we observed a statistically signifi- associated with increased CVD risk. Physically inactive cant association of dp-ucMGP (HRSD 1.31 However, in order to describe clinical (%)† 18.9 510 [95% CI 1.05–1.64], P = 0.02) with CVD relevance, the added prognostic values of High education (%) 7.9 515 mortality. However, again, dp-cMGP and dp-ucMGP levels beyond traditional risk Systolic blood t-ucMGP were not associated with this factors should be further investigated. In pressure (mmHg) 144 6 22 515 end point (data not shown). Furthermore, our study, we found an association be- Diastolic blood the association of circulating dp-ucMGP tween higher circulating dp-ucMGP and pressure (mmHg) 83 6 10 498 with all-cause mortality was similar to the increased CVD risk. However, this asso- Total cholesterol association with CVD mortality with a HRSD ciation was stronger for the subtypes PAD (mmol/L) 6.0 6 1.2 479 of 1.20 (95% CI 1.04–1.39). and heart failure. This may be explained HDL cholesterol Adjusting for diabetes duration, by the presence of two types of vascular 6 fi fi (mmol/L) 1.15 0.33 466 HbA1c concentrations, blood lipids, or calci cation, i.e., intimal calci cation and Duration of diabetes blood pressure did not affect these results medial calcification. Intimal calcification (years) 6.3 6 7.0 496 (data not shown). Excluding not verified is seen in sites of atherosclerotic plaques Using OBGL type 2 diabetic patients yielded compara- in which cellular necrosis, inflammation, drugs (%) 43 396 ble results (data not shown). and cholesterol deposition occur. The 6 HbA1c (%) 8.03 1.78 496 presence of this type of calcification is dp-ucMGP (pmol/L) 217 6 210 518 CONCLUSIONSdTo our knowledge, associated with atherosclerotic burden dp-cMGP (pmol/L) 1,062 6 623 518 this is the first study investigating the (36). Medial calcification, which occurs in t-ucMGP (nmol/L) 4,354 6 1,284 518 association of circulating MGP species with the elastic lamina of large- and medium- to 6 the risk of cardiovascular events. In this small-size , is particularly present in Data are median (interquartile range), means SD, fi or % unless otherwise indicated. OBGL, oral blood prospective study, we observed that high peripheral arteries (36). Medial calci ca- glucose lowering. †Inactive according to the CPAI. circulating dp-ucMGP levels, reflecting a tion is a metabolite-induced calcification

3768 DIABETES CARE, VOLUME 36, NOVEMBER 2013 care.diabetesjournals.org Dalmeijer and Associates

Table 2dCrude and adjusted HRs (95% CI) for the association of MGP* species with incident (fatal or nonfatal) CVD and CHD among 518 diabetic subjects

CVD CHD PAD Heart failure Stroke n 16099382826 dp-ucMGP Crude 1.37 (1.21–1.54) 1.24 (1.06–1.45) 1.40 (1.15–1.71) 1.77 (1.47–2.14) 1.22 (0.90–1.65) Model 1† 1.29 (1.13–1.48) 1.17 (0.99–1.40) 1.39 (1.12–1.73) 1.75 (1.42–2.16) 1.05 (0.73–1.49) Model 2‡ 1.21 (1.06–1.38) 1.12 (0.94–1.34) 1.33 (1.07–1.65) 1.75 (1.42–2.17) 1.05 (0.73–1.50) dp-cMGP Crude 1.02 (0.88–1.18) 0.86 (0.67–1.11) 1.14 (0.93–1.40) 1.15 (0.91–1.45) 1.01 (0.68–1.48) Model 1† 0.95 (0.80–1.12) 0.78 (0.59–1.03) 1.14 (0.90–1.44) 1.07 (0.81–1.42) 0.85 (0.52–1.40) Model 2‡ 0.96 (0.81–1.14) 0.78 (0.58–1.03) 1.16 (0.92–1.45) 1.07 (0.80–1.42) 0.85 (0.52–1.40) t-ucMGP Crude 0.99 (0.85–1.16) 0.92 (0.75–1.12) 1.03 (0.75–1.41) 0.95 (0.65–1.38) 1.07 (0.73–1.56) Model 1† 0.98 (0.84–1.15) 0.91 (0.74–1.11) 0.97 (0.72–1.31) 0.96 (0.66–1.38) 1.11 (0.76–1.62) Model 2‡ 0.99 (0.85–1.16) 0.94 (0.76–1.15) 0.99 (0.73–1.35) 0.95 (0.65–1.37) 1.10 (0.75–1.60)

*HRs are expressed per SD. †Adjusted for age, sex, waist-to-hip ratio, and CPAI. ‡Model 1 adjustments plus history of CVD or history of CVD subtype.

in the absence of lipid deposits, leading to calcification is firmly established, the pre- disturbances and MGP (41–43). This an- upregulation of osteogenic and pro- cise function of MGP in the heart is un- giotensin II–mediated increase in MGP teins with subsequent matrix mineraliza- known. However, the lack of myocardial was seen in both cardiomyocytes and fi- tion and and formation calcification in the MGP-deficient mice broblasts. In addition, Ueland et al. (35) (37). This process requires a complex reg- suggests an additional effect of MGP in the found a strong correlation between dp- ulatory network involving both stimulators heart not related to inhibition of calcifica- ucMGP and N-terminal pro-B-type natri- and inhibitors of calcification such as MGP, tion (3). Furthermore, MGP has been as- uretic peptide, which further supports osteoprotegerin, and osteopontin (37). The sociated with binding such a notion. Although further studies specific association of dp-ucMGP with PAD and apoptosis, as well as tissue growth and are needed, it is plausible that the associ- suggests that this relation is mainly driven development, and all these processes ation between dp-ucMGP and heart failure by a relation of dp-ucMGP with medial cal- may be relevant to myocardial remodeling is at least partly caused through mecha- cification. However, this needs to be fur- (38–40). Previously, raised MGP concen- nisms that are independent of vascular cal- ther investigated. trations have been found as a rapid myo- cification. Theoretically, dp-cMGP forms Also, for the subtype heart failure we cardial response to pressure overload, the mirror image of dp-ucMGP, and there- found a stronger association than for CVD. potentially induced by angiotensin II, fore we expected that high levels of dp- Although the role of MGP in vascular suggesting a link between neurohormonal cMGP would be associated with lower CVD risk. No previous studies investi- gated the association between dp-cMGP levels and CVD events, but the association with mortality has been investigated. No association between dp-cMGP levels and mortality risk in patients with symptom- atic aortic stenosis or chronic heart failure was observed (34,35). This is consistent with a previous cross-sectional study from our group, where we did not find an association between dp-cMGP and cal- cification (16). However, in hemodialysis patients low levels of dp-cMGP were asso- ciated with increased CVD mortality risk (HR 2.7 [95% CI 1.2–6.2]) (20). In our study, we could not detect an association between dp-cMGP and CVD risk. How- ever, we observed a borderline significant association of high dp-cMGP levels with a reduced risk of CHD, similar to the results in hemodialysis patients (19). Because of Figure 1dAdjusted HRs (95% CI) for the association of dp-ucMGP levels with incident (fatal and the relatively small sample size of our nonfatal) CVD and CVD subtypes among 518 diabetic subjects. HRs are expressed per SD and study, we may have had limited power adjusted for age, sex, waist-to-hip ratio, CPAI, history of CVD, and CVD subtypes. to detect a somewhat weaker association. care.diabetesjournals.org DIABETES CARE, VOLUME 36, NOVEMBER 2013 3769 MGP and risk of CVD in diabetic patients

Therefore, our results and those of In conclusion, the findings of this 2. Shanahan CM, Proudfoot D, Farzaneh- Schlieper et al. (20) suggest that there prospective study among type 2 diabetic Far A, Weissberg PL. The role of Gla mightbeanassociationbetweenhigh patients shows that high circulating dp- in vascular calcification. Crit Rev – dp-cMGP and reduced CHD risk. For fur- ucMGP levels are associated with in- Eukaryot Expr 1998;8:357 375 ther clarification of the role of dp-cMGP in creased CVD risk, especially with PAD 3.LuoG,DucyP,McKeeMD,etal.Spon- taneous calcification of arteries and carti- CVD and CHD risk, more and larger pop- and heart failure. Circulating dp-cMGP t lage in mice lacking matrix GLA protein. ulation studies, as well as physiological and -ucMGP levels were not related to Nature 1997;386:78–81 studies, are warranted. CVD risk. These results suggest that a 4. Schurgers LJ, Uitto J, Reutelingsperger One previous study investigated the poor vitamin K status is associated with CP. Vitamin K-dependent carboxylation association between t-ucMGP and CVD or increased CVD risk. of matrix Gla-protein: a crucial switch to mortality risk (44). In coronary artery dis- control ectopic mineralization. Trends ease patients, each 1,000 nmol/L higher t- Mol Med 2013;19:217–226 ucMGP level was associated with a 16% AcknowledgmentsdThe EPIC-NL study was 5. Furie B, Bouchard BA, Furie BC. Vitamin lower CVD risk and a 22% lower risk of funded by the European Commission: Public K-dependent biosynthesis of gamma- all-cause mortality. However, the associ- Health and Consumer Protection Directorate carboxyglutamic acid. Blood 1999;93: 1798–1808 ation of t-ucMGP levels with mortality 1993-2004; Research Directory-General 2005; Dutch Ministry of Public Health, Welfare and 6. Beulens JW, Bots ML, Atsma F, et al. High was limited to participants without diabe- fi Sports; Netherlands Cancer Registry; LK Re- dietary menaquinone intake is associated tes, whereas there was no signi cant as- search Funds; Dutch Prevention Funds; Dutch with reduced coronary calcification. Ath- sociation in diabetic patients (HR 1.10 Zorg Onderzoek Nederland; and World Can- erosclerosis 2009;203:489–493 [95% CI 0.85–1.41]). We also found no cer Research Fund (the Netherlands). This 7. Shea MK, O’Donnell CJ, Hoffmann U, association between t-ucMGP levels and research was supported by a personal Dr. Dekker et al. Vitamin K supplementation and CVD risk in diabetic patients. This could postdoctoral grant (2008T062) from the progression of coronary artery calcium in be due to population differences: the Netherlands Heart Foundation (to J.W.J.B.). older men and women. Am J Clin Nutr study mentioned above resulted from a Part of this research was funded by an un- 2009;89:1799–1807 cohort selected for coronary artery dis- restricted grant from the Nutricia Research 8. Gast GC, de Roos NM, Sluijs I, et al. A high ease, who have coronary artery calcifica- Foundation (2011-22). menaquinone intake reduces the inci- No potential conflicts of interest relevant to dence of coronary heart disease. Nutr tion, while our participants were selected this article were reported. Metab Cardiovasc Dis 2009;19:504–510 for the presence of diabetes. We assume G.W.D. designed the research, wrote the 9. Geleijnse JM, Vermeer C, Grobbee DE, that t-ucMGP may act differently in persons manuscript, and analyzed data. Y.T.v.d.S. de- et al. Dietary intake of menaquinone is with or without calcification because signed the research and reviewed and edited associated with a reduced risk of coronary t-ucMGP mainly consists of phosphory- the manuscript. E.J.M., C.V., W.M.M.V., and heart disease: the Rotterdam Study. J Nutr lated ucMGP species. The phosphoserines J.M.A.B. reviewed the edited the manuscript. 2004;134:3100–3105 equip the molecule with strong affinity for J.W.J.B. designed the research and reviewed 10. Schurgers LJ, Cranenburg EC, Vermeer C. vascular calcification, thereby causing a de- and edited the manuscript. G.W.D. is the Matrix Gla-protein: the calcification in- guarantor of this work and, as such, had full hibitor in need of vitamin K. Thromb crease in the plasma circulating levels of – t-ucMGP, leading to lower t-ucMGP levels access to all the data in the study and takes Haemost 2008;100:593 603 fi responsibility for the integrity of the data and 11. Cranenburg EC, Koos R, Schurgers LJ, at higher levels of calci cation (21). the accuracy of the data analysis. et al. Characterisation and potential di- Our study has certain limitations to agnostic value of circulating matrix Gla consider. The assays for dp-ucMGP and protein (MGP) species. Thromb Haemost dp-cMGP only detected the nonphos- References 2010;104:811–822 phorylated fraction of MGP. Biochemical 1. Greenland P, Bonow RO, Brundage BH, 12. Rennenberg RJ, de Leeuw PW, Kessels tests for other circulating MGP species are et al.; American College of Cardiology AG, et al. Calcium scores and matrix Gla currently not available. Further, circulat- Foundation Clinical Expert Consensus protein levels: association with vitamin ing MGP levels do not necessarily reflect Task Force (ACCF/AHA Writing Commit- K status. Eur J Clin Invest 2010;40:344–349 ’ MGP tissue levels of the vasculature. We tee to Update the 2000 Expert Consensus 13. Shea MK, O Donnell CJ, Vermeer C, et al. therefore can only speculate how circu- Document on Electron Beam Computed Circulating uncarboxylated matrix gla pro- tein is associated with vitamin K nutritional latingMGPlevelsasabiomarkerare Tomography); Society of Atherosclerosis Imaging and Prevention; Society of Car- status, but not coronary artery calcium, in related to tissue MGP. Furthermore, the diovascular Computed Tomography. older adults. J Nutr 2011;141:1529–1534 dp-ucMGP levels and dp-cMGP levels in ACCF/AHA 2007 clinical expert consensus 14. Westenfeld R, Krueger T, Schlieper G, our study were low compared with a document on coronary artery calcium et al. Effect of supplementa- healthy reference group (11). This could scoringbycomputedtomographyinglobal tion on functional vitamin K deficiency in be due to the possible degradation of cardiovascular risk assessment and in eval- hemodialysis patients: a randomized trial. MGP in our samples because samples uation of patients with chest pain: a report Am J Kidney Dis 2012;59:186–195 have been stored for a long time in straws. of the American College of Cardiology 15. Theuwissen E, Cranenburg EC, Knapen Finally, our study was limited to type 2 Foundation Clinical Expert Consensus MH, et al. Low-dose menaquinone-7 sup- diabetic patients, which limits generaliz- Task Force (ACCF/AHA Writing Commit- plementation improved extra-hepatic vita- ability of our results to the general popula- tee to Update the 2000 Expert Consensus min K status, but had no effect on thrombin Document on Electron Beam Computed generation in healthy subjects. Br J Nutr tion. However, this study population is Tomography) developed in collaboration 2012;108:1652–1657 particularly relevant to exploring associa- with the Society of Atherosclerosis Im- 16. Dalmeijer GW, van der Schouw YT, tions with CVD risk because approximately aging and Prevention and the Society of Vermeer C, Magdeleyns EJ, Schurgers LJ, one-half of patients with type 2 diabetes Cardiovascular Computed Tomography. Beulens JW. Circulating matrix Gla protein will die of cardiovascular causes (45,46). J Am Coll Cardiol 2007;49:378–402 is associated with coronary artery calcification

3770 DIABETES CARE, VOLUME 36, NOVEMBER 2013 care.diabetesjournals.org Dalmeijer and Associates

and vitamin K status in healthy women. and medical therapy: Part II. Circulation mechanisms and clinical implications. J Nutr Biochem 2013;24:624–628 2003;108:1655–1661 Circ Res 2006;99:1044–1059 17. Dalmeijer GW, van der Schouw YT, 26. Wu KK, Huan Y. Diabetic atherosclerosis 37. Abedin M, Tintut Y, Demer LL. Vascular Magdeleyns E, Ahmed N, Vermeer C, mouse models. Atherosclerosis 2007;191: calcification: mechanisms and clinical Beulens JW. The effect of menaquinone-7 241–249 ramifications. Arterioscler Thromb Vasc supplementation on circulating species of 27. Schauer IE, Snell-Bergeon JK, Bergman Biol 2004;24:1161–1170 matrix Gla protein. Atherosclerosis 2012; BC, et al. Insulin resistance, defective 38. Vermeer C. Gamma-carboxyglutamate- 225:397–402 insulin-mediated fatty acid suppression, containing proteins and the vitamin 18. Schurgers LJ, Barreto DV, Barreto FC, and coronary artery calcification in subjects K-dependent carboxylase. Biochem J 1990; et al. The circulating inactive form of with and without type 1 diabetes: The 266:625–636 matrix gla protein is a surrogate marker CACTI study. Diabetes 2011;60:306–314 39. Nishimoto SK, Nishimoto M. Matrix Gla for vascular calcification in chronic kidney 28. Beulens JW, Monninkhof EM, Verschuren protein C-terminal region binds to vi- disease: a preliminary report. Clin J Am WM, et al. Cohort profile: the EPIC-NL tronectin. Co-localization suggests bind- Soc Nephrol 2010;5:568–575 study. Int J Epidemiol 2010;39:1170–1178 ing occurs during tissue development. 19. Rennenberg RJ, van Varik BJ, Schurgers 29. Sluijs I, van der A DL, Beulens JW, et al. Matrix Biol 2005;24:353–361 LJ, et al. Chronic coumarin treatment is Ascertainment and verification of diabetes 40. Cario-Toumaniantz C, Boularan C, associated with increased extracoronary in the EPIC-NL study. Neth J Med 2010; Schurgers LJ, et al. Identification of differ- arterial calcification in humans. Blood 2010; 68:333–339 entially expressed genes in human varicose 115:5121–5123 30. Herings RM, Bakker A, Stricker BH, Nap veins: involvement of matrix gla protein 20. Schlieper G, Westenfeld R, Kruger€ T, et al. G. Pharmaco-morbidity linkage: a feasi- in extracellular matrix remodeling. J Vasc Circulating nonphosphorylated carbox- bility study comparing morbidity in two Res 2007;44:444–459 ylated matrix gla protein predicts survival pharmacy based exposure cohorts. J Epi- 41. Hwang DM, Dempsey AA, Lee CY, Liew in ESRD. J Am Soc Nephrol 2011;22:387– demiol Community Health 1992;46:136– CC. Identification of differentially ex- 395 140 pressed genes in cardiac hypertrophy by 21. Cranenburg EC, Brandenburg VM, 31. Wareham NJ, Jakes RW, Rennie KL, et al. analysis of expressed sequence tags. Ge- Vermeer C, et al. Uncarboxylated matrix Validity and repeatability of a simple in- nomics 2000;66:1–14 Gla protein (ucMGP) is associated with dex derived from the short physical ac- 42. Rysä J, Leskinen H, Ilves M, Ruskoaho H. coronary artery calcification in haemo- tivity questionnaire used in the European Distinct upregulation of extracellular ma- dialysis patients. Thromb Haemost 2009; Prospective Investigation into Cancer and trix genes in transition from hypertrophy 101:359–366 Nutrition (EPIC) study. Public Health to hypertensive heart failure. Hypertension 22. Koos R, Krueger T, Westenfeld R, et al. Nutr 2003;6:407–413 2005;45:927–933 Relation of circulating Matrix Gla-Protein 32. Durrleman S, Simon R. Flexible regression 43. Mustonen E, Pohjolainen V, Aro J, et al. and anticoagulation status in patients with models with cubic splines. Stat Med 1989; Upregulation of cardiac matrix Gla protein aortic valve calcification. Thromb Haemost 8:551–561 expression in response to hypertrophic 2009;101:706–713 33. Rubin DB. Multiple Imputation for Non- stimuli. Blood Press 2009;18:286–293 23. Parker BD, Schurgers LJ, Vermeer C, response in Surveys. New York, Wiley, 1987 44. Parker BD, Schurgers LJ, Brandenburg Schiller NB, Whooley MA, Ix JH. The asso- 34. Ueland T, Gullestad L, Dahl CP, et al. VM, et al. The associations of fibroblast ciation of uncarboxylated matrix Gla protein Undercarboxylated matrix Gla protein is growth factor 23 and uncarboxylated with mitral annular calcification differs by associated with indices of heart failure and matrix Gla protein with mortality in cor- diabetes status: The Heart and Soul study. mortality in symptomatic aortic stenosis. onary artery disease: the Heart and Soul Atherosclerosis 2010;210:320–325 J Intern Med 2010;268:483–492 Study. Ann Intern Med 2010;152:640– 24. Amos AF, McCarty DJ, Zimmet P. The 35. Ueland T, Dahl CP, Gullestad L, et al. 648 rising global burden of diabetes and its Circulating levels of non-phosphorylated 45. International Diabetes Federation. IDF complications: estimates and projections undercarboxylated matrix Gla protein are Diabetes Atlas. Brussels, Belgium, 2011. to the year 2010. Diabet Med 1997;14 associated with disease severity in patients Available from http://www.idf.org/ (Suppl. 5):S1–S85 with chronic heart failure. Clin Sci (Lond) diabetesatlas. Accessed 18 January 2013 25. Luscher€ TF, Creager MA, Beckman JA, 2011;121:119–127 46. Lundman B, Engström L. Diabetes and it’s Cosentino F. Diabetes and vascular disease: 36. Johnson RC, Leopold JA, Loscalzo J. complications in a Swedish county. Di- pathophysiology, clinical consequences, Vascular calcification: pathobiological abetes Res Clin Pract 1998;39:157–164

care.diabetesjournals.org DIABETES CARE, VOLUME 36, NOVEMBER 2013 3771