Calcium scores and matrix Gla protein levels: association with vitamin K status

Authors

  • Roger J. M. W. Rennenberg,

    1. Department of Internal Medicine, Maastricht University Medical Centre (MUMC+) and Cardiovascular Research Institute Maastricht (CARIM), Maastricht, The Netherlands
    Search for more papers by this author
  • Peter W. De Leeuw,

    1. Department of Internal Medicine, Maastricht University Medical Centre (MUMC+) and Cardiovascular Research Institute Maastricht (CARIM), Maastricht, The Netherlands
    Search for more papers by this author
  • Alphons G. H. Kessels,

    1. Clinical Epidemiology and Medical Technology Assessment, Maastricht University Medical Centre (MUMC+), Maastricht, The Netherlands
    Search for more papers by this author
  • Leon J. Schurgers,

    1. Department of Biochemistry, Maastricht University, CARIM and VitaK B.V., Maastricht, The Netherlands
    Search for more papers by this author
  • Cees Vermeer,

    1. Department of Biochemistry, Maastricht University, CARIM and VitaK B.V., Maastricht, The Netherlands
    Search for more papers by this author
  • Jos M. A. Van Engelshoven,

    1. Department of Radiology, Maastricht University Medical Centre (MUMC+) and Cardiovascular Research Institute Maastricht (CARIM), Maastricht, The Netherlands
    Search for more papers by this author
  • Gerrit J. Kemerink,

    1. Department of Radiology, Maastricht University Medical Centre (MUMC+) and Cardiovascular Research Institute Maastricht (CARIM), Maastricht, The Netherlands
    Search for more papers by this author
  • Abraham A. Kroon

    1. Department of Internal Medicine, Maastricht University Medical Centre (MUMC+) and Cardiovascular Research Institute Maastricht (CARIM), Maastricht, The Netherlands
    Search for more papers by this author

Roger J. M. W. Rennenberg, Department of Internal Medicine, Maastricht University Medical Centre (MUMC+) and Cardiovascular Research Institute Maastricht (CARIM), P. Debyelaan 25, 6229 HX Maastricht, The Netherlands. Tel.: +31 43 3877005; fax: +31 43 3875006; e-mail: r.rennenberg@mumc.nl

Abstract

Eur J Clin Invest 2010; 40 (4): 344–349

Abstract

Background  Vascular calcification in humans is associated with an increased cardiovascular risk. Carboxylated matrix Gla protein (cMGP) inhibits vascular calcification. Vitamin K is an essential cofactor for the activation of uncarboxylated matrix Gla protein (ucMGP). It has been suggested that patients on long-term treatment with vitamin K antagonists develop aortic valve calcifications because of lower levels of circulating MGP. We therefore hypothesized that arterial calcification and a low vitamin K status are associated with ucMGP. To that aim, we measured arterial calcium scores, the osteocalcin ratio (OCR), as a proxy for vitamin K status, and ucMGP.

Materials and methods  In 36 hypertensive patients, we determined the Agatston score with computer tomography scans of the abdominal aorta, carotid and coronary arteries. The total calcium score was calculated as the sum of the separate Z-scores.

Results  The total calcium Z-score was significantly correlated to age (r = 0·683, P < 0·001), smoking (r = 0·372, P = 0·026), total cholesterol (r = 0·353, P = 0·034), LDL cholesterol (r = 0·490, P = 0·003), triglycerides (r = 0·506, P = 0·002), fasting glucose (r = 0·454, P = 0·005), systolic blood pressure (r = 0·363, P = 0·029) and pulse pressure (r = 0·685, P < 0·001). In multivariate regression analyses, OCR and total calcium score were significantly associated with ucMGP.

Conclusions  We found a positive association of total arterial calcium score and a high OCR (reflecting low vitamin K status) with ucMGP serum levels. This warrants further studies to explore the pathophysiological background of this phenomenon.

Introduction

Arterial calcification in humans is a risk factor for cardiovascular complications, not only in patients with established cardiovascular disease [1], diabetes [2,3] and chronic kidney disease [4,5] but also in asymptomatic individuals [6–8]. Vascular calcification is a complex process resembling bone formation [9]. Matrix Gla protein (MGP) plays a central role in the inhibition of calcification by influencing the function of bone morphogenetic protein type 2 and preventing the deposition of calcium in the vascular matrix [10,11]. Activation of MGP requires carboxylation, which is a vitamin K-dependent process. The balance between levels of active, carboxylated MGP (cMGP) and inactive, uncarboxylated MGP (ucMGP) may, therefore, be important in counteracting arterial calcification.

Several studies have reported on an association between high vitamin K intake and reduced cardiovascular risk or less arterial calcification [12–14]. On the contrary, the use of vitamin K antagonists is associated with increased coronary and cardiac valve calcification [15,16]. Studies on the relationship between coronary artery calcification and MGP gave conflicting results [17,18]. In this respect, vitamin K status may be of importance as studies with vascular smooth muscle cells have shown that MGP production is triggered by extra cellular calcium [19]. In human, this has also been studied in isolated vascular smooth muscle cells but not in vivo [20].

As it is not clear to what extent circulating levels of MGP and vitamin K status are associated with calcified lesions in the arterial vasculature, we performed the present study. We measured ucMGP levels and the osteocalcin ratio (OCR), as a proxy for vitamin K status, in a cohort of well-characterised, moderately hypertensive patients and used computer tomography (CT) to quantify calcification of the coronary and carotid arteries, as well as of the abdominal aorta. Our hypothesis was that, subjects with the highest ucMGP levels would have the highest calcium scores and the lowest vitamin K status.

Methods

Patients

During a 10-month period, we recruited 40 consecutive patients who had been referred to our outpatient clinic for evaluation of hypertension. Subjects were not included if they had intravascular stents, which makes interpretation of coronary calcium scores difficult, had an age below 18 years, tried to achieve pregnancy or were pregnant. Assessments included data on smoking habits (never smoked, history of smoking or current smoker), body mass index (BMI, calculated as kg m−2), ambulatory blood pressure measurements (following 3 weeks cessation of antihypertensive medication), fasting levels of blood glucose, lipid profile and serum creatinine. We also took blood samples for the determination of ucMGP and the OCR. Blood samples were centrifuged immediately at 3500 g for 10 min at 4 °C. Serum was separated and stored at −80 °C until use. Additionally, we performed a CT scan of the carotid arteries, the coronary arteries and the abdominal aorta. The study protocol was approved by the local ethics committee and all patients gave written informed consent.

Measurements

We measured 24-h ambulatory blood pressure with Space Labs 90217 (Space Labs Healthcare, Inc., Issaquah, WA, USA). Pulse pressure was calculated as the difference between the systolic and diastolic value of the 24-h ambulatory measurements. Glucose levels, lipids and serum creatinine were measured with an automated chemistry analyzer (Beckmann Synchron CX 7-2, Fullerton, CA, USA). Endogenous creatinine clearance (ECC), as a proxy of glomerular filtration rate, was calculated from serum creatinine concentrations with the formula of Cockcroft and Gault [21].

As cMGP cannot be readily assayed in the laboratory, ucMGP is measured together with vitamin K in order to asses the cMGP status. The principle of the ucMGP assay is that of a competitive enzyme-linked immuno-sorbent assay (ELISA) in which micro well plates are coated with a mouse monoclonal antibody against human ucMGP (VitaK BV, Maastricht, the Netherlands) [22]. In brief, anti-ucMGP was coupled to a micro titre plate via polyclonal rabbit-anti-mouse IgG (Dako, Heeverlee, Belgium). After washing, 5 μL of serum sample or standard were mixed with tracer (biotinylated peptide consisting of residues 35–54 in human MGP), transferred to the micro titre plate and incubated overnight at 4° C. After washing, the plate was incubated with streptavidine-peroxidase (Zymed, Breda, The Netherlands) and stained with tetramethylbenzidine (KPL protein research products, Gennep, The Netherlands). The process was stopped by adding H2SO4, and the plate was read at 450 nm. The lower limit of detection was 98 nM, with intra- and inter-assay coefficients of variation of 6% and 11·4% [22].

We refrained from estimating vitamin K intake as information from dietary questionnaires is notoriously unreliable. Because it is also difficult to measure vitamin K levels reliably, the OCR, defined as uncarboxylated osteocalcin divided by carboxylated osteocalcin, was used instead as a proxy for vitamin K status [23–26]. A lower OCR reflects higher vitamin K levels. Two commercially available test kits (Takara Bio Inc., Otsu, Shiga, Japan) were used to determine the OCR. One is specific for carboxylated osteocalcin, the other measures noncarboxylated osteocalcin. The principle of the measurement is similar to that of the MGP measurements and utilises ELISA techniques. Carboxylated MGP is estimated by dividing the level of ucMGP by the OCR.

High-resolution CT was performed on a Toshiba Aquilion multi/4 scanner (Toshiba, Zoetermeer, Netherlands) using 120 kV, 2 mm slice width and a field of view between 240 and 320 mm. Cardiac gated CT of the chest, with special attention to the depiction of the coronary arteries, covered the complete heart (additional scan parameters: tube current 250 mA, rotation time 0·32 s, pitch 0·8). Carotid artery calcifications were measured from about 7 cm below the bifurcation up to 3 cm of the external and internal carotid arteries (300 mA, 0·5 s, 1·4). The abdominal aorta was scanned from the top of the highest kidney to the bottom of the lowest kidney with a minimum distance of 10 cm (250 mA, 0·5 s, 1·4). Calcification was analysed using ScImage’s volumetric cardiac scoring software, version NETRA MD v.1.06.00a (ScImage Inc., Los Altos, CA, USA). Calcium was depicted as any pixel equal to or over 130 Hounsfield units. Coronary calcification was expressed as Agatston scores [27].

Statistical analysis

Normally distributed variables are expressed as means with standard deviations; otherwise they are expressed as medians with their minimum and maximum value. Calcium scores were measured in the coronaries, both the right and the left carotid artery and the abdominal aorta. The total calcium Z-score was then calculated as the sum of the individual Z-scores (calculated by subtracting the mean calcification score from the individual test score divided by the standard deviation of the calcification score). Trend analyses using the Kruskal–Wallis test was performed with variables that appeared to be related to increasing levels of ucMGP. Known risk factors for calcification were also correlated with the total calcium score. Spearman’s Rho was used for non-normally distributed data. With ucMGP (logarithmically transformed for regression analysis because of a non-normal distribution) as the dependent variable and total calcium Z-score and OCR as the key independent variables of interest, multivariate regression analysis was performed (model 1). Age, sex and ECC (model 2 and 3) were added to the model as potential confounders. Although they are not known as potential confounders of MGP levels, we also performed this regression analysis using smoking habits, BMI and systolic blood pressure as confounders (model 4). Additionally, this was also performed with fasting glucose levels and LDL cholesterol (model 5). Beta coefficients with 95% confidence interval (CI) are reported. Statistical calculations were performed using SPSS for Windows v.16.0 (SPSS, Chicago, IL, USA). A P-value < 0·05 was considered statistically significant.

Results

Four patients had to be excluded from the analyses because they had intra-coronary stents. Table 1 shows the patient characteristics of the remaining 36 patients. The median ucMGP level was 4708 nM (range 2031–9603 nM). This is comparable to that of a healthy nonhypertensive reference population (4976 nM; 2685–8187) [22]. The median OCR was 1·2 (0·1–14·0), which suggests a relatively low vitamin K status in these patients [24]. The median coronary score was 65, indicating that these subjects had only minor amounts of calcium in their coronary arteries. Table 2 represents data according to tertiles of ucMGP. Although there seemed to be a trend for OCR, coronary score, abdominal aorta score and total calcification Z-score with ucMGP, this was not statistically significant. Total calcium Z-score, however, was significantly correlated to age (r = 0·683, P < 0·001), smoking (r = 0·372, P = 0·026), total cholesterol (r = 0·353, P = 0·034), LDL cholesterol (r = 0·490, P = 0·003), triglycerides (r = 0·506, P = 0·002), fasting glucose (r = 0·454, P = 0·005), systolic blood pressure (r = 0·363, P = 0·029) and pulse pressure (r = 0·685, P < 0·001). Carboxylated MGP and OCR were only significantly correlated to calcification of the left carotid artery (r = 0·504, P = 0·003 and r = - 0·399, P = 0·021, respectively). There was no significant univariate correlation between total calcium score and OCR, ucMGP or cMGP. However, in crude multivariate regression analyses, and also in a model adjusted for age alone, adjusted for age, sex and ECC, adjusted for smoking habits, BMI and systolic blood pressure or adjusted for fasting glucose levels and LDL, the OCR was significantly associated with (log-)ucMGP. Except for the crude analysis and model 4 and 5, this was also true for the association with total arterial calcium-Z-score. There was significant collinearity between total arterial calcium-Z-score and the confounders used in model 4 and 5, which possibly explains the nonsignificant result (Table 3). There was no significant association of the calcium sub-scores of the coronary arteries, aorta, and right or left carotid arteries with (log-)ucMGP (Table 4).

Table 1.   Patient characteristics
VariableTotal groupTertiles of ucMGP serum levels
(n = 36)T1 (12)T2 (12)T3 (12)
  1. Unless indicated otherwise data are expressed as mean ± standard deviation.

  2. SBP, systolic blood pressure; DBP, diastolic blood pressure; HDL, high density lipoprotein; LDL, low density lipoprotein.

  3. *Denotes median values with minimum and maximum between brackets.

Age (year)53 ± 1055 ± 1054 ± 951 ± 11
Male/female (n)19/1710/25/74/8
(Ever) smoking n (%)24 (67)9 (75)7 (58)8 (67)
Body mass index (kg m−2)27 ± 527 ± 426 ± 627 ± 4
Ambulatory SBP (mmHg)162 ± 22158 ± 18172 ± 18157 ± 26
Ambulatory DBP (mmHg)97 ± 1394 ± 12104 ± 1594 ± 10
Total cholesterol (mM)6·0 ± 1·16·3 ± 1·55·8 ± 0·95·9 ± 0·7
HDL cholesterol (mM)1·31 ± 0·381·3 ± 0·31·3 ± 0·41·4 ± 0·5
LDL cholesterol (mM)*3·9 (1·8–6·0)3·9 (1·8–6·0)3·8 (2·3–5·1)3·9 (3·1–4·6)
Triglycerides (mM)*1·62 (0·49–12·42)2·37 (0·57–4·16)1·18 (0·49–4·36)1·63 (0·95–12·42)
Fasting glucose (mM)*5·6 (4·5–10·5)5·6 (4·7–10·5)5·4 (4·5–8·1)5·5 (4·5–9·9)
Estimated creatinine clearance (mL min−1)98 ± 28104 ± 3094 ± 2694 ± 29
Table 2.   Results of measurements divided by tertiles of serum ucMGP levels
VariableTertiles (mean and range) of ucMGP serum levels
Tertile (n)T1 (12)T2 (12)T3 (12)
  1. Data are expressed as medians with minimum and maximum values (nonnormal distribution).

  2. UcMGP, uncarboxylated Matrix Gla Protein; OCR, osteocalcin ratio; cMGP, carboxylated Matrix Gla Protein.

ucMGP (nM)3471 (2031–4260)4708 (4351–5215)6126 (5416–9603)
OCR1·0 (0·2–2·7)1·4 (0·5–2·5)2·1 (0·1–14·0)
cMGP (nM)3237 (1257–18889)3475 (1759–10043)3199 (665–82118)
Coronary score145 (0–1546)111 (1–3866)36 (0–5951)
Abdominal aorta score1365 (0–13063)834 (0–21438)359 (0–12944)
Left carotid artery score2 (0–772)56 (0–713)36 (0–1789)
Right carotid artery score2 (0–987)71 (0–769)0 (0–2253)
Total calcium Z-score- 1·2 (- 2·0 to 5·5)- 1·4 (- 2·0 to 5·5)- 1·9 (- 2·0 to 11·9)
Table 3.   Multivariate regression analysis with (log-)ucMGP as the dependent variable
Model*Total calcification Z-score; beta 10−3 (95% CI)Osteocalcin ratio; beta 10−3 (95% CI)
  1. All results were statistically significant with the exception of those figures indicated with NS (not significant). There was significant collinearity between the confounders in models 4 and 5 and total calcification Z-score.

  2. Model 1, crude; Model 2, adjusted for age; Model 3, adjusted for age, sex and estimated creatinine clearance; Model 4, adjusted for smoking habits, body mass index and systolic blood pressure; Model 5, adjusted for fasting glucose levels and LDL-cholesterol.

  3. OCR, osteocalcin ratio; 95% CI, 95% confidence interval.

17 (- 7 to 20)NS24 (7–40)
216 (1–31)23 (4–42)
318 (4–32)25 (6–43)
49 (− 6 to 24)NS29 (11–48)
512 (− 2 to 27)NS22 (5–40)
Table 4.   Correlation of ucMGP with calcium sub-scores
LocalisationCorrelation coefficient*
  1. *Spearman’s Rho

Left carotid artey0·026 (P = 0·882)
Right carotid artery− 0·077 (P = 0·656)
Coronary arteries− 0·065 (P = 0·706)
Aorta− 0·185 (P = 0·280)

Discussion

This study shows a significant positive association between ucMGP, total arterial calcium score and low vitamin K status. Vitamin K status was, on an average, lower in these hypertensive patients than has been reported for a healthy control population with the same average age [24]. Lower ucMGP levels are clearly associated with higher vitamin K levels. Because carboxylation is an intracellular process, we assume that this is explained by clearance of ucMGP and generation of cMGP in the presence of sufficient vitamin K. However, except for the calcification score of the left carotid artery, there was no significant correlation between cMGP or OCR with any of the other calcium sub-scores or the total calcification Z-score.

To our knowledge, this is the first study showing an association between calcium scores and ucMGP in relatively healthy hypertensive subjects. Previously, only reports showing a correlation of low serum unphosphorylated MGP and ucMGP levels in populations with increased cardiovascular risk have been published [22,28]. Phosphorylation of MGP is the first step in its activation and is induced by the presence of calcium. It provides the binding site of MGP to calcium. Reports on correlations between unphosphorylated MGP serum levels and coronary calcification scores have been conflicting [17,18,29]. Recently, ucMGP also failed to show a correlation with markers of vascular stiffness in children on dialysis [30].

The MGP that we detected in the circulation of these patients represents the noncarboxylated form of MGP. A low vitamin K status prevents carboxylation of MGP and thus makes ucMGP less functional [31]. The smooth muscle cell reacts by increasing MGP production as reflected by an increase in MGP messenger RNA [32]. Probably, although ucMGP binds to calcium, it is not able to halt crystal growth. The excess ucMGP is thought to be set free in the circulation, and thus could be a potential marker in calcified patients.

Our results fit well with the theory that, in the process of vascular calcification, serum levels of ucMGP initially rise because its production by vascular smooth muscle cells is triggered by calcium. The positive beta indicates that higher Z-scores and thus more calcification is associated with higher serum ucMGP levels in this sample of relatively healthy hypertensive subjects. In other study populations, with higher cardiovascular risk and more calcification or with renal insufficiency, lower serum ucMGP levels have been found [22]. A possible explanation is that, with progressing calcification most of the ucMGP is bound to the calcium in the vascular wall and is than trapped, ultimately leading to a fall in serum levels [22,31,33,34]. In addition, in the presence of sufficient vitamin K, ucMGP is carboxylated to cMGP giving rise to lower ucMGP levels as is supported by the positive beta in our regression analysis. Interestingly, another explanation for the lower ucMGP serum levels in subjects with more advanced arterial calcification may be a decline in vascular smooth muscle cells because of transformation to osteoblast-like cells or apoptosis [35,36].

Our analyses have several limitations. Firstly, although CT is an accepted method to measure vascular calcification, our method of estimating total calcification Z-score has not been validated. Secondly, we cannot exclude that the calcium sub-scores of separate vascular regions in our analyses failed to reach significance, because of the small number of patients or confounding by age. Thirdly, our study is based on associations, and does not show a cause-and-effect relationship. Fourthly, whereas we expected lower calcification scores with higher cMGP levels, this was not found. Our method of estimating cMGP could be too imprecise to detect this correlation. Furthermore, the ratio of ucMGP and OCR is based on the assumption that the availability of vitamin K in the vascular wall is similar to that in bone. This may not be the case, giving rise to imprecise estimation of cMGP. This imprecision may have affected the association between lower calcification scores in the presence of higher levels of vitamin K and thus higher levels of cMGP. As far as this is concerned, it is important to realize that it has recently been shown that vitamin K supplementation was associated with slower progression of coronary calcium scores [12].

In conclusion, we found, adjusted for potential confounders, a significant positive association of total arterial calcium score and a high OCR (reflecting low vitamin K status) with serum ucMGP levels in mild to moderate hypertensive patients. This is consistent with the possibility that serum ucMGP levels are stimulated by the calcification process in humans. Although these data do not prove a cause-and-effect relationship between vitamin K and arterial calcifications, this possibility needs to be explored in future studies. Research in this field is important for the interpretation of future studies with serum MGP, vitamin K and cardiovascular risk. Vitamin K supplementation could be a simple way to reduce cardiovascular risk by activating MGP.

Acknowledgements

We thank Geert Wijnhoven, computer tomography specialist, for his help and advice on calcium scoring. The authors declare to have no conflict of interest. Cees Vermeer is CEO of VitaK b.v., The Netherlands. There were no commercial fundings involved.

Address

Department of Internal Medicine, Maastricht University Medical Centre (MUMC+) and Cardiovascular Research Institute Maastricht (CARIM), Maastricht, The Netherlands (R. J. M. W. Rennenberg, P. W. de Leeuw, A. A. Kroon); Clinical Epidemiology and Medical Technology Assessment, Maastricht University Medical Centre (MUMC+), Maastricht, The Netherlands (A. G. H. Kessels); Department of Biochemistry, Maastricht University, CARIM and VitaK B.V., Maastricht, The Netherlands (L. J. Schurgers, C. Vermeer); Department of Radiology, Maastricht University Medical Centre (MUMC+) and Cardiovascular Research Institute Maastricht (CARIM), Maastricht, The Netherlands (J. M. A. van Engelshoven, G. J. Kemerink).

Ancillary