Undercarboxylated matrix Gla protein is associated with indices of heart failure and mortality in symptomatic aortic stenosis

doi: 10.1111/j.1365-2796.2010.02264.x

Undercarboxylated matrix Gla protein is associated withindices of heart failure and mortality in symptomatic aorticstenosis

T. Ueland1,2, L. Gullestad3, C. P. Dahl1, P. Aukrust1,4, S. Aakhus3, O. G. Solberg3, C. Vermeer5 & L. J. Schurgers5

Fromthe1Research Institute for InternalMedicine; 2SectionofEndocrinology;3DepartmentofCardiology;4SectionofClinical ImmunologyandInfectiousDiseases,RikshospitaletUniversityHospital,UniversityofOslo,Oslo,Norway,and5VitaK&CardiovascularResearch InstituteCARIM(CV,LS),MaastrichtUniversity,Maastricht,TheNetherlands

Abstract. Ueland T, Gullestad L, Dahl CP, Aukrust P,Aakhus S, SolbergOG, Vermeer C, Schurgers LJ (Re-search Institute for Internal Medicine, University ofOslo, Oslo; University of Oslo, Oslo, Norway; and Vi-taK&Cardiovascular Research InstituteCARIM (CV,LS), Maastricht University, Maastricht, The Nether-lands) UndercarboxylatedmatrixGla protein is asso-ciated with indices of heart failure and mortality insymptomatic aortic stenosis. J Intern Med 2010;268: 483–492.

Objective. Matrix Gla protein (MGP) is a calcificationinhibitor and alterations in circulating MGP havebeen observed in different populations characterizedby vascular calcification. We hypothesized that pa-tientswith calcific valvular aortic stenosis (AS)wouldhavedysregulatedcirculatingMGP levels.

Design and subjects. We examined plasma levels of non-phosphorylated carboxylated and undercarboxylat-ed MGP (dp-cMGP and dp-ucMGP, respectively) in147 patients with symptomatic severe AS and inmatchedhealthycontrols.

Main outcome measures. We further investigated the rela-tionship between MGP levels and aortic pressuregradients and valve area by echocardiography andmeasures of heart failure. Finally, we assessed theprognostic value of elevated plasma dp-ucMGP levelin relation to all-cause mortality in patients withAS.

Results. We found markedly enhanced plasma levelsof dp-cMGP and in particular of dp-ucMGP in pa-tients with symptomatic AS. Although only weakcorrelations were found with the degree of AS, cir-culating dp-ucMGP was associated with cardiacfunction and long-term mortality in multivariateanalysis.

Conclusions. A dysregulated MGP system may have arole in thedevelopmentof left ventriculardysfunctioninpatientswithsymptomaticAS.

Keywords: all-cause mortality, aortic stenosis, calcifi-cation,heart failure,matrixGlaprotein.

Introduction

Calcific aortic stenosis (AS) is a progressive diseasethat has, until recently, been considered a degenera-tive and unmodifiable process induced by long-last-ing mechanical stress on the valve structures [1–3].However,histopathological studieshavenowdemon-strated that the pathogenesis of calcific AS is an ac-tive process, sharing a number of similarities withatherosclerosis [1, 4]. Inflammation, lipid infiltration,dystrophic calcification, ossification, platelet activa-tion and endothelial dysfunction have been observedin both diseases. Early in the development of AS, ac-tive microscopic areas of calcification are seen colo-calizing in areas of lipoprotein accumulation andinfiltration of inflammatory cells [1, 3–5]. As the

disease progresses, there is active bone formationinvolving cells with chondrocytic and ⁄or osteoblasticcharacteristics capable of phenotypic differentiationandspontaneouscalcification [3,6].

Major proteins involved in regulation of tissue calcifi-cation have been detected in calcified valvular tissue[3]. One of these regulatory proteins is matrix c-car-boxylated glutamate (Gla) protein (MGP), a 10-kDacirculating protein containing five Gla residues andthree serine residues, which can be phosphorylated[3, 7]. Both theGla residues andphosphorylated ser-ine residues have a high affinity for calcium and areformedinapost-translational reaction.Thecarboxyl-ation requires vitaminK and is essential for the prop-er function ofMGP [7]. The vitaminK antagonistwar-

ª 2010 The Association for the Publication of the Journal of Internal Medicine 483

Original Article |

farinmayblockvitaminK-dependent c-carboxylationwithin the liver andprevent hepatic formation of clot-ting factors. However, this process is not limited tothe liver, and carboxylation of peripherally expressedvitaminK-dependent proteinmay be affected aswell,thereby regulating calcium binding [8–10]. Thus,although the relationship between warfarin and vas-cular calcification inhumans isnot fully understood,observational studies suggest that inhibition of MGPcarboxylation by the vitamin K antagonist warfarinresults in extensive calcification of arteries in vitroand in vivobecauseof synthesis of the inactiveunder-carboxylatedMGP(ucMGP) [10,11].

In experimental models, ucMGP is associated withaortic calcification in ageing rats [12]. In the clinicalsetting,wehave recentlydemonstratedastrongasso-ciationbetweenvascular calcificationand localdepo-sition of ucMGP in arteries of non-uremic patientswithatherosclerosisandMonckeberg’s sclerosis [13].However, to the best of our knowledge, no studieshave investigated nonphosphorylated carboxylatedMGP (dp-cMGP) or dp-ucMGP levels in patients withAS. In contrast to phosphorylatedMGP (p-cMGP andp-ucMGP), which accumulates within the vessel wallat thesiteofvascularcalcification,dp-MGPhasa low-er affinity for calcium salts and increased circulatinglevels may partly reflect the degree of calcificationwithin thevesselwall [14].

Patients with symptomatic AS frequently develop leftventricular (LV) outflowobstructionandLVhypertro-phy to compensate and maintain ejection perfor-mance, and there is an increased mortality rate afterthe onset of symptoms [15]. Recent studies haveshown that plasma levels of N-terminal pro-brainnatriuretic peptide (Nt-proBNP), a marker of LVhypertrophy and dysfunction in patients with con-gestive heart failure (HF), are related to disease sever-ity and survival in patients with AS [16, 17]. Further-more, C-reactive protein (CRP), a sensitive marker ofsystemic inflammation and atherosclerosis progres-sion,may predict disease severity andprogression aswell as prognosis in patients with asymptomatic AS[18]. Thus, soluble markers may be useful for moni-toringdiseaseprogression inthesepatients.

Based on these findings, as well as the major role ofMGP in vascular calcification [11, 19], we hypothe-sized that plasma levels of dp-cMGP and dp-ucMGPwould be dysregulated in patients with AS. In addi-tion to examination of plasma levels of these proteinsin patients with AS and in healthy controls, we inves-tigated the relationship between their levels and

mean aortic pressure gradient, valve area and mea-sures ofHF in patients with AS. Finally, we evaluatedtheprognostic value of elevatedplasmadp-ucMGP inrelationtoall-causemortality in thesepatients.

Materials and methods

Patients

In total,147patientswithsymptomaticAS,evaluatedfor aortic valve surgery betweenMay 2005andJanu-ary 2007, were consecutively enrolled in the study(Table 1). Only patients with confirmed AS were in-cluded. Echocardiographic parameters were mea-sured and blood samples were obtained from eachpatient. Coronary angiography was performed in allpatients to determine the presence of concomitantcoronaryarterydisease (CAD) (at leastonevesselwith>50%narrowing of the luminal diameter). The exclu-sion criteria were severe (grade III) aortic or mitralregurgitation or serum creatinine concentration>150 lmol L)1. All investigations were carried outwithin a few days. Of the 147 patients, 106 werescheduled foraorticvalve replacement (AVR)whereassurgerywasnot performed in the remaining in41pa-tients for the following reasons: comorbidity ⁄highriskof surgical complications (n = 22),patientprefer-ence (n = 5), uncertain clinical benefit because ofmild symptoms (n = 10) or other reasons (n = 4). Inpatients accepted for AVR, one died pending surgerytherefore 105 patients underwent surgical interven-tion. Twenty-nine patients were receiving warfarinbecause of atrial fibrillation (n = 24) or deep venousthrombosis (n = 2), or for anticoagulation followingmyocardial infarction (n = 2) or coronary artery by-pass grafting (n = 1). For comparison, 93 sex-matched controls (aged 65.2 ± 5.4 years, mean ±SD, 48%male) were included in the study. Informedconsent was obtained from each study subject. Thestudy protocol was approved by the regional commit-tee forethics inmedicine.

Echocardiography

Conventional echocardiographic imagingof theheartwas performed from parasternal and apical viewsusing a GE Vivid 7 ultrasonic digital scanner (GEVingmed, Horten, Norway). We obtained two-dimen-sional,m-mode and colourDoppler images aswell aspulsed-wave Doppler recordings of blood flow veloci-ties in the LV outflow tract. Continuous-wave Dopp-ler recordings from multiple positions were used toobtainthemaximumaorticannularbloodflowveloci-ties and to calculate aortic valve area using the conti-

T. Ueland et al. | MGP in AS

484 ª 2010 The Association for the Publication of the Journal of Internal Medicine Journal of Internal Medicine 268; 483–492

nuity equation [20]. Doppler echocardiographic cal-culations of stroke volume and cardiac output wereperformed on the basis of the cross-sectional area offlowandaortic annular flowvelocity data. LVejectionfraction (LVEF) was determined using the biplaneSimpson method [9]. Peak systolic right ventricularpressure was estimated from the maximal tricuspidregurgitative blood flow velocity (TrPmax). Dimen-

sional and velocity parameterswere averaged fromatleast three (five inpatientswith atrial fibrillation) car-diaccycles.

Biochemistry and blood sampling

For NT-proBNP, CRP and MGP measurements,peripheral venousbloodwasdrawn intopyrogen-free

Table 1 Characteristicsof147patientswithsymptomaticaorticstenosis

Totalpopulation

Correlation

dp-ucMGP dp-cMGP uc ⁄ c-MGPratio

Age (years) 74 ± 10 0.42*** 0.36*** 0.29***

Male (%) 55 0.04 0.04 0.02

BMI (kg m)2) 26.3 ± 4.3 0.08 0.07 0.04

NYHAfunctional class I ⁄ II ⁄ III ⁄ IV (%) 5 ⁄33 ⁄61 ⁄1 0.06 )0.02 0.05

Coronaryarterydisease (%) 43 0.07 )0.02 0.10

Currentsmokers (%) 33 )0.06 )0.10 )0.03

DM2(%) 11 0.09 0.05 0.07

Hypertension (%) 25 0.15 0.10 0.13

Atrialfibrillation (%) 34 0.41*** 0.24** 0.40***

Haemodynamics

LVEF (%) 62 ± 12 )0.29*** )0.16 )0.26**

Aortic valvearea (cm2) 0.71 ± 0.35 )0.24** )0.19* )0.16

Meanaorticpressuregradient (mmHg) 53 ± 20 )0.11 )0.05 )0.11

Biochemistry

HDL-Ch (mmol L)1) 1.60 ± 0.46 )0.04 )0.10 0.00

LDL-Ch(mmol L)1) 3.20 ± 1.10 )0.14 )0.04 )0.17*

Creatinine (lmol L)1) 90 ± 36 0.41*** 0.48*** 0.20*

eGFR 72.0 ± 32.5 )0.43*** )0.46*** )0.26**

CRP (mg L)1)a 2.0 (1.0,4.8) 0.30*** 0.20* 0.27**

NT-proBNP(pmol L)1)a 107 (42,288) 0.39*** 0.33*** 0.30***

Medication (%)

ACEI 14 0.10 0.27** )0.05

ARB 19 0.12 )0.02 0.15

Betablocker 45 0.16* 0.14 0.12

Diuretic 31 0.32*** 0.24** 0.06

Statin 49 0.11 )0.01 0.14

Warfarin 19 0.65*** 0.27** 0.70***

Aspirin 47 )0.28** )0.03 )0.37***

Valuesgivenaspercentage,mean ± SD,or amedianand interquartile range.dp, dephosphorylated; uc, undercarboxylated; c, carboxylated; MGP,matrix Gla protein; BMI, bodymass index; DM2, type-2diabetesmellitus; eGFR, estimated glomerular filtration ratebasedon theCockcroft–Gault formula; LVEF, left ventricular ejec-tion fraction; Ch, Cholesterol; CRP, C-reactive protein; NT-proBNP, N-terminal pro-brain natriuretic peptide, ACEI, angioten-sin-converting enzyme inhibitor; ARB, angiotensin receptor blocker; LDL-Ch, low-density lipoprotein cholesterol; HDL-Ch,high-density lipoproteincholesterol.*P < 0.05, **P < 0.01, ***P < 0.001. Full datasets were not available for all measures. To convert NT-proBNP values frompmol L)1 topg mL)1,multiplyby8.47.


ª 2010 The Association for the Publication of the Journal of Internal Medicine Journal of Internal Medicine 268; 483–492 485

tubes with EDTA as anticoagulant. The tubes wereimmediately immersed inmelting ice andcentrifugedwithin30 minat2000 g for20 min toobtainplatelet-poor plasma. All samples were stored at )80 �C andthawed <3 times. NT-proBNP and CRP were assayedon a MODULAR platform (Roche Diagnostics, Basel,Switzerland).Levels of circulatingdp-ucMGPanddp-cMGPweredetermined inplasmausingsandwichen-zyme-linked immunosorbent assay (ELISA) tech-niques. The dp-ucMGP assay is based on the use oftwomonoclonal antibodies directed against the non-phosphorylated sequence 3–15 and the noncarboxy-lated sequence 35–49 in humanMGP (VitaK BV,Ma-astricht, The Netherlands). Dp-cMGP levels weremeasured by a similar sandwich ELISA in which thesecond antibody was directed against the carboxyl-atedsequence35–53inhumanMGP(VitaKBV).Plas-ma levels of LDL and HDL cholesterol and creatinineweremeasured enzymatically using a Roche ⁄Hitachi917 analyser (Roche Diagnostics, Mannheim, Ger-many). Estimated glomerular filtration rate (eGFR)wascalculatedusingtheCockcroft–Gault formula.

Statistics

We used parametric statistics on all measures be-cause theywouldultimately be included inamultipleregressionmodel. Variables not normally distributedas evaluated by the Kolmogorov–Smirnov test werelog transformed for the statistics but may be pre-sented as nontransformed data. Differences betweengroupswere analysed with the Student’s t-test. Rela-tionships between variables were tested by simplelinear (bivariate) regression analysis (Pearson corre-lation). For some measures, partial correlation wasused controlling for the use of warfarin (Table 3). Asubsequent multiple linear regression analysis wasperformedwithstepwiseadditionof thevariables that

had P values <0.05 in our a priori analysis and uponbivariate regression. Receiver-operating characteris-tics (ROC) curves were generated to evaluate theaccuracy of each marker for prediction of all-causemortality. Kaplan–Meier analysis with log-rank testwas performed to compare the number of events indifferent groups (comparisons pooled over strata).Cox proportional hazard analysis was performed toestimate hazard ratios using a forward stepwise con-ditional method. Prior to collectingmortality data forsurvival analysis, a sample size calculation was per-formed.Assuming a = 0.05, apowerof0.80anda rel-ative risk of 2.0 for exposure to greater than medianlevels of MGP, the following were determined: a fol-low-up interval of 36 months; an accrual time duringwhich patients were recruited of 24 months; and amedian survival of 29 months. This gave us a total of122 patients. Only variables with P < 0.05 upon uni-variate analysis were included in multivariate Coxanalysis. Follow-up time for all-cause mortality wascalculated from time of inclusion to death from anycause. P values are two sided and considered signifi-cantwhen<0.05.

Results

MGP levels in patients with AS and healthy controls

As can be seen in Fig. 1, patients with symptomaticAS (n = 147) had markedly elevated dp-cMGP andin particular dp-ucMGP levels compared to sex-matched healthy controls (n = 93), resulting in amolar abundance of dp-ucMGP compared to dp-cMGP (dp-ucMGP ⁄dp-cMGP molar ratio) in patientswith AS. The controls were �9 years younger thanthe patients. However, patients with AS had signifi-cantly raised plasma levels of MGP parameters evenafter adjustment for age by regression (P-values

Fig. 1 Plasma levels of dephosphorylated (dp), undercarboxylated (uc) and carboxylated (c) matrix Gla protein (MGP) in 147patients with symptomatic aortic stenosis (AS) compared to 93 sex-matched healthy controls (Ctr). The right-hand graph showsthemolar ratio ofucMGP ⁄ cMGP.Dataarepresentedasmedianand interquartile range.



indicating higher levels in patients with AS areshown: dp-ucMGP, P < 0.001; dp-cMGP, P = 0.040;ratio, P < 0.001).

MGP levels and clinical characteristics of patients with AS

In patients with AS, there was a positive associationbetween MGP levels and age and creatinine ⁄eGFR(Table 1). However, multivariate analyses showedthat the associationbetweenMGPandAS is indepen-dentofageandkidney function.Also,50patientshadatrial fibrillation and were further characterized bymarkedly increased levels of dp-ucMGP (2232 ± 248vs. 1005 ± 64, P < 0.001) and dp-cMGP (3126 ± 227vs.2481 ± 112,P = 0.003)andan increaseducMGP ⁄cMGP ratio (0.71 ± 0.07 vs. 0.41 ± 0.02, P < 0.001),comparedwith patients without atrial fibrillation. Bycontrast, there was no association between MGPlevels and a number of known cardiovascular riskfactors, includingbodymass index (BMI), presenceofCAD (combined or assessed separately in the leftanterior descending, left circumflex and right coro-nary artery), smoking, type 2 diabetes mellitus andhypertension, even in univariate analyses (Table 1).Furthermore, although there were no or only weakassociationsbetween levelsofMGPand lipidparame-ters, MGP was positively correlated with CRP (Ta-ble 1). For all these comparisons, the same patternwasseen fordp-ucMGPanddp-cMGP(Table1).

MGP levels and warfarin use

Whereas there were no or only weak associations be-tween MGP and the use of various cardiovascularmedications, including statins, as expected a verystrong positive association was found between use ofvitamin K antagonists and MGP levels (Table 1). Theassociationbetweenatrial fibrillationandMGP levelscould also be explained by warfarin use. Indeed,when controlling forwarfarinuse, the association be-tweenatrialfibrillationandMGPlevelswasnotsignif-icant (P = 0.185). Figure 2 demonstrates the influ-ence of warfarin on MGP levels in patients withsymptomatic AS. First, dp-ucMGP levels were higherin warfarin users, resulting in an abundance of thisMGPspecies compared todp-cMGP(Fig. 2a).Second,in bothwarfarinusers andnonusers, dp-ucMGPanddp-cMGP were significantly correlated to valve area(Fig. 2b).By contrast, no suchcorrelationwas seen inrelation to themeanpressuregradientacross theaor-tic valve,with the samepattern inwarfarin users andnonusers (Fig. 2c). However, when performingmulti-ple regression analyses in nonusers of warfarin (war-farin users not analysed separately because of low

numbers), we found no independent contribution ofdp-ucMGP (or dp-cMGP,datanot shown) to variationinaortic valvearea (Table2).

MGP levels and association with HF

Wenext investigatedwhether therewereanyassocia-tionsbetweenMGPandmeasures ofHF in this popu-lation of patients with AS (Fig. 3). No relationshipswere observed between circulating MGP levels andthe degree of clinical disease severity as assessed bythe New York Heart Association (NYHA) functionalclass (Fig. 3a). However, there was a lower dp-ucMGP ⁄dp-cMGP ratio in warfarin users with NYHAII–IV disease compared to those with NYHA I–II. Bycontrast, dp-ucMGPwassignificantlycorrelatedwithneurohormonal (i.e., NT-proBNP) and haemodynam-ic (i.e., LVEF) markers of HF severity in nonwarfarinusers, and with NT-proBNP in warfarin users(Fig. 3b, c). Table 3 provides further details on theassociation betweendifferent indices of cardiac func-tion andMGP levels as well as severity of AS in usersand non-users of warfarin. Themean aortic pressuregradient was correlated with interventricular septalthickness at diastole (IVSd), whereas the valve areawas correlated with TrPmax, cardiac index (CI)

Fig. 2 Plasma matrix Gla protein (MGP) and measures ofaortic stenosis (AS) in patients according to use of warfarin.(a) Plasma levels of dephosphorylated (dp), undercarboxylat-ed (uc) and carboxylated (c) MGP in 28 patients with symp-tomatic AS treated with warfarin and 119 patients withoutwarfarin treatment (nonwarfarin). The right-hand graphshows the molar ratio of ucMGP ⁄ cMGP. Correlation betweendp-ucMGPandvalve area (b) andaortic pressure gradient (c).Dataaregivenasmean ± SEM.*P < 0.001.



and LVEF in non-warfarin users. In addition to theassociation with LVEF, dp-ucMGP was correlatedwithTrPmaxand inversely correlatedwithCI, furthersuggesting an association between dp-ucMGP and

myocardial dysfunction. LV hypertrophy defined asIVSd >1.1 cm [9] was observed in 121 subjects, butthere was no significant association between any oftheMGPspeciesand IVSd.Asimilarpatternwasseen

Table 2 Predictors of mean aortic pressure gradient, valve area and LVEF in patients with symptomatic aortic stenosis (only non-

warfarinusers included,n = 118)

Meanaorticpressuregradient Valvearea LVEF

Univariate

r (P-value) Multivariate

Univariate


Univariate


Age 0.13 )0.41 (<0.001) b = )0.37,

t = )4.01,

P < 0.001

)0.26(0.005)

Gender )0.09 )0.10 0.31(<0.001) b = )0.31,

t = 3.33,

P = 0.001

DM )0.10 )0.02 )0.24(0.011)

eGFR )0.05 0.20 (0.044) )0.14

LDL 0.10 )0.12 0.23(0.021)

NT-proBNP 0.34(<0.001) b = 0.34,

t = 3.57

P = 0.001

)0.34 (<0.001) b = )0.24,

t = )2.61,

P = 0.011

)0.18

dp-ucMGP )0.07 )0.24 (0.014) )0.21(0.037) b = )0.19,

t = )1.99,

P = 0.050

Variables included were age, gender, BMI, presence of coronary artery disease, smoking, type-2 diabetes mellitus, hyperten-sion, warfarin use, atrial fibrillation, LDL-cholesterol, eGFR (based on theCockcroft–Gault formula), CRP, NT-proBNP anddp-ucMGP.Only predictorswith a univariate association P < 0.05 (shown in Table) were entered in a subsequent stepwise regres-sion (multivariate).CRP, C-reactive protein; DM, diabetes mellitus; LDL-cholesterol, low-density lipoprotein cholesterol; NT-proBNP, N-terminalpro-brain natriuretic peptide; dp-ucMGP, nonphosphorylated undercarboxylated MGP; MGP, matrix Gla protein; LVEF, leftventricularejection fraction; eGFR,estimatedglomerularfiltrationrate.

Table 3 Partial correlations between MGP and different haemodynamic echocardiographic measurements in patients with symp-

tomaticaorticstenosis

Patients

(mean ± SD)

Correlation,coefficient r (P-value)

Meanaortic

pressuregradient Valvearea dp-ucMGP dp-cMGP

ucMGP ⁄ cMGP

ratio

IVSd(cm) 1.26 ± 0.21 0.22* (0.29) )0.03 ()0.33) )0.01(0.33) 0.06(0.17) 0.05 (0.30)

TrPmax (mmHg) 33 ± 11 0.10(0.35) )0.50*** ()0.29) 0.23* (0.37*) 0.19(0.46*) 0.19 (0.07)

CI (L min)1 per m2) 2.67 ± 0.60 0.09(0.28) 0.45*** (0.16) )0.31** ()0.33) )0.19(0.21) )0.19 ()0.58**)

LVEF (%) 62 ± 12 0.05(0.27) 0.20* ()0.05) )0.21* (0.01) )0.10()0.09) )0.10 (0.09)

Topvalue iscorrelationcoefficient fornonwarfarinusers,withvalue forwarfarinusers inparentheses.IVSd, interventricular septal thickness at diastole; TrPmax,maximumpeak tricuspid regurgitation pressure gradient; CI, car-diac index; LVEF, left ventricular ejection fraction; dp, dephosphorylated;uc,undercarboxylated; c, carboxylated;MGP,matrixGlaprotein.*P < 0.05, **P < 0.01, ***P < 0.001.



for dp-cMGP, although the degree of correlationswassomewhat weaker (Table 3). Few correlations wereobservedbetweenMGP levels and echocardiographicindices inusers ofwarfarin but this analysiswas lim-ited by the number of observation (Table 3). Finally,multiple regression analysis showed that after cor-recting for other potential confounders, circulatingdp-ucMGP, but not dp-cMGP (data not shown), wassignificantly and independently associated withLVEF(Table2).

MGP levels and all-cause mortality

During a mean follow-up of 23 months (range1–36 months), a total of 25 patients died; 16 in thenonsurgical group died because of cardiovascular(n = 12) and other reasons (i.e. cancer and chronicobstructive pulmonary disease; n = 4) and eight inthe surgery group died because of cardiovascular(n = 7) and other reasons (n = 1). The concentrationof dp-ucMGP (dichotomized above and below950 pmol L)1), was closely associated with all-causemortality (Fig. 4a), with the same pattern in warfarinusers and nonusers (data not shown). However,because of the lownumber of patients in thewarfaringroup and a lack of power (see Materials and

methods), further mortality analyses were restrictedto nonusers of warfarin. When comparing high ver-sus low dp-ucMGP concentrations, the hazard ratio(HR) was 9.33 (95% confidence interval [CI] 2.67–32.51,P < 0.001). This relationship remainedsignifi-cant after adjustment for age, gender, BMI, eGFR,NT-proBNP, CRP, hypertension, diabetes and LVEF(Fig.4b).Whenanalysingdata frompatientswithandwithout AVR separately, a significant associationbetween dp-ucMGP and mortality was observedregardless of AVR (data not shown). However, theassociation was stronger in patients with AVR (HR15.02, 95% CI 1.92–117.39, P = 0.010). In fact, instepwisemultivariate regression, dp-ucMGPwas theonly factor that remained in the analysis in this sub-group of patients. Indeed none of the patients in theAVR subgroup who died had dp-ucMGP levels belowthemedianvalue.

Weaker nonsignificant associations were observedfor dp-cMGP (data not shown). ROC analysis did notreveal any alternative cut-off with significantly great-erdiagnosticaccuracy (datanot shown).

Discussion

Numerous studies have confirmed that MGP is a cal-cification inhibitor, and alterations in circulating lev-els of MGP have been observed in different popula-tions characterized by vascular calcification [13, 21–23]. In this study, we showmarkedly enhanced plas-ma levels of dp-cMGP and in particular of dp-ucMGPinpatientswith symptomaticAS.Althoughonlyweakcorrelations were found with the degree of AS as-sessed by echocardiography, the circulating dp-ucMGP level was associated with cardiac functionand long-term mortality, suggesting that a dysregu-lated MGP system could have a role in the develop-mentofLVhypertrophyandHFinthesepatients.

We and others have recently demonstrated massiveaccumulation of ucMGP in atherosclerotic lesionsand areas of calcification [12, 13, 24]. These studiessuggested that because of its high affinity forhydroxyapatite (possibly resulting from the presenceof three negatively charged phosphoserine residues),ucMGP is not released into the circulation fromcalci-fied arteries. We hypothesized that in the presence ofarterial calcification, the circulating fraction of p-ucMGP is decreased. Indeed, when investigating cir-culating levels of p-ucMGP using a competitive ELI-SA, we found that patient populations characterisedby arterial calcification, including AS, were associ-atedwith low systemic p-ucMGP levels [23]. To elimi-

Fig. 3 Plasma matrix Gla protein (MGP) and measures ofheart failure inpatientswithsymptomaticaortic stenosis (AS)according to use of warfarin. (a) Plasma levels of dephospho-rylated (dp), undercarboxylated (uc) and carboxylated (c)MGP inpatientswithNYHAclass I ⁄ II compared to class III ⁄ IVdisease. The right-hand graph shows the molar ratio ofucMGP ⁄ cMGP. (b) Correlation between dp-ucMGP and NT-proBNPandLVEF. Dataare given asmean ± SEM. *P < 0.05vs.NYHAI–II.



nate the potentially confounding effect of ucMGPbinding toarterial calcifications,wepresent inthis re-port two newly developed sandwich ELISA tech-niques that are able to evaluate only nonphosphory-lateducMGPandcMGPlevels. Incontrast topreviousstudies of ucMGP, we found that levels of both dp-cMGP and dp-ucMGP, but in particular of dp-ucMGP, were markedly increased in patients withAS, resulting in an increase in the dp-ucMGP ⁄dp-cMGP ratio in these patients. Moreover, dp-ucMGPcorrelatedwith indices of LVdysfunctionandwith in-creased long-termmortality. Thesedatademonstratethat the nonphosphorylated forms of MGP, and inparticular dp-ucMGP, are more suitable as plasmamarkers for AS and perhaps other cardiovasculardisorders thanthephosphorylated forms.

Our data suggest increased synthesis of MGP in pa-tients with AS, which is consistent with previous re-ports showing up-regulation of MGP synthesis at theonset of arterial calcification [25]. Thus, based on therole of vascular calcification in atherogenesis, it mayseem surprising that we found no association be-tween MGP and the presence of CAD. However, in arecent study including a large number of patients,O’Donnell et al. [26] found no correlation between

MGP levels and coronary artery calcification, sug-gesting a complex relation between MGP and the de-gree of atherosclerosis. It is possible that the raisedlevelsofMGP inASpatientsmayreflect a relationshipwith valvular calcification and not a general associa-tionwithvascularcalcification.

We found that plasma dp-ucMGP in particular wasassociated with cardiac function as assessed byLVEF and CI as well as with neurohormonal activa-tion as assessed by NT-proBNP. Previously, severalDNA microarray studies have demonstrated in-creased MGP mRNA expression in the LV duringacute and chronic pressure overload in mice andhumans [27–29] as well as in experimental HF inrats and mice [30, 31]. Also, MGP has been shownto bind to vitronectin, and this anchoring of MGPto the extracellular matrix may have modifying ef-fects on members of transforming growth factor bfamily that are involved in cardiac remodelling dur-ing HF [32, 33]. Most recently, Mustonen et al. [34]demonstrated that MGP is rapidly upregulated inresponse to cardiac overload long before the devel-opment of LV hypertrophy and remodelling follow-ing myocardial infarction. Thus, although the pre-cise mechanisms of action remain unclear, our

Fig. 4 Plasma dephosphorylated (dp), undercarboxylated (uc) and carboxylated (c) matrix Gla protein (MGP) and all-causemor-tality inpatientswithsymptomaticaorticstenosis (AS). (a)Kaplan–Meier curvesshowing thecumulative incidenceofdeathduringtheentire study (mean follow-up23 months, range1–36), according to themediandp-ucMGP (cut-off 950 pmol L)1) anddp-cMGP(cut-off 2400 pmol L)1) at enrolment. (b)Multivariatemodels for all-causemortality.Model 1 is dp-ucMGPalone.Model 2 includesdp-ucMGP, age, gender, BMI, eGFR, NT-proBNP, CRP, hypertension, diabetes and LVEF in a stepwise regression. Model 3includes thevariables inmodel2 ina forcedaddition (i.e.all variables remain in themodel).



findings suggest that plasma levels of dp-ucMGP inaddition to valvular calcification may reflect cardiacdysfunction in patients with AS. Figure 5 illustratesthe possible association between calcification,inflammation and decreased dp-MGP species in thestenotic valve with increased circulating dp-MGP.In time, this nonresolving inflammation and calcifi-cation will lead to cardiac dysfunction with elevatedNT-proBNP.

It isnotsurprising thatpatientswhoreceivedthevita-min K antagonist warfarin had disturbed MGPhomeostasis. Thus, warfarin users had increasedplasma levels of dp-cMGP and in particular of dp-ucMGP when compared with nonwarfarin users,resulting in an abundance of dp-ucMGP in thosereceiving warfarin. At present, the biological signifi-cance of this finding is unclear. Future studies withvitamin-K-independent anticoagulants could be ofinterest. Such studies should examine the effect ofwarfarin and anticoagulants prospectively duringlongitudinal testing and should also include mea-surementsofvitaminKstatus.

After the onset of symptoms, there is a high mortal-ity rate in patients with AS [15]. Thus, AVR is rec-ommended in symptomatic patients [35]. In thispresent study, we found that patients with abovethe median levels of dp-cMGP, and in particular ofdp-ucMGP, had a higher unadjusted mortality rate(9-fold increase for dp-ucMGP). After adjustmentfor risk factors, including NT-proBNP and LVEF,

the HR was still �7 for dp-ucMGP, suggesting thatplasma dp-ucMGP may be a novel predictor formortality in these patients. It is interesting that thehighest HR value for mortality was observed whenseparately analysing patients who underwent AVR.This latter finding suggests that dp-ucMGP couldbe of interest not only in untreated patients withAS but also as a pre-operative marker for long-termrisk evaluation in patients who undergo AVR. How-ever, the overall mortality in those who underwentAVR was relatively low, and these associationsshould be interpreted with caution. Further studiesare needed to evaluate the prognostic relevance ofincreased plasma dp-ucMGP. A larger study popu-lation, including both asymptomatic and symptom-atic patients as well as patients scheduled for AVR,assessed longitudinally, would reveal whether plas-ma dp-ucMGP assessment may benefit patientmanagement.

This study has certain limitations such as lownumbers of patients treated with warfarin, lack ofdata on vitamin K status (which could influenceMGP levels) and lack of some clinical data in thecontrol group (including echocardiographic and fullbiochemical assessment). Nonetheless, our findingssuggest that symptomatic AS is characterized byincreased plasma levels of the dephosphorylatedform of MGP. In particular, dp-ucMGP was inde-pendently associated with measures of cardiacfunction and significantly associated with all-causemortality in patients with AS. To further clarify therole of MGP in LV dysfunction, studies using longi-tudinal sampling in larger populations, includingboth asymptomatic and patients with symptomaticAS, as well as in-depth studies in patients with HF,are warranted.

Conflict of interest statement

Nonedeclared.

Acknowledgements

This work was supported by grants from the Norwe-gian Council of Cardiovascular Research, Helse Sør-øst andRikshospitalet. NattopharmaASAsponsoredtheMGPmeasurements.Theauthors thankananon-ymousbenefactor.

References

1 Akat K, Borggrefe M, Kaden JJ. Aortic valve calcification – basic

science toclinicalpractice.Heart2008;95:616–23.

Fig. 5 Possible association between inflammation, in-creased circulating dp-MGP level and cardiac dysfunction insymptomatic aortic stenosis. CRP, C-reactive protein; MGP,matrix c-carboxylatedglutamateprotein;cMGP,carboxylatedMGP; ucMGP, undercarboxylated MGP; t, total; dp, dephos-phorylated; NT-proBNP, N-terminal pro-brain natriuretic pep-tide.



2 Rosenhek R, Baumgartner H. Aortic sclerosis, aortic stenosis

and lipid-lowering therapy. Expert Rev Cardiovasc Ther 2008; 6:

385–90.

3 BostromK. Insights into themechanismofvascularcalcification.

AmJCardiol2001;88:20E–2E.

4 Freeman RV, Otto CM. Spectrum of calcific aortic valve disease:

pathogenesis, diseaseprogression,and treatmentstrategies.Cir-

culation2005;111:3316–26.

5 Kaden JJ, Bickelhaupt S, Grobholz R et al. Receptor activa-

tor of nuclear factor kappaB ligand and osteoprotegerin regu-

late aortic valve calcification. J Mol Cell Cardiol 2004; 36:

57–66.

6 Rajamannan NM. Calcific aortic stenosis: lessons learned from

experimental and clinical studies. Arterioscler Thromb Vasc Biol

2009;29:162–8.

7 Vermeer C. Gamma-carboxyglutamate-containing proteins and

the vitamin K-dependent carboxylase. Biochem J 1990; 266:

625–36.

8 Danziger J. Vitamin K-dependent proteins, warfarin, and vascu-

larcalcification.ClinJAmSocNephrol2008;3:1504–10.

9 Lang RM, Bierig M, Devereux RB et al. Recommendations for

chamber quantification: a report from the American Society of

Echocardiography’s Guidelines and Standards Committee and

the Chamber Quantification Writing Group, developed in con-

junction with the European Association of Echocardiography, a

branch of the EuropeanSociety of Cardiology. J AmSocEchocar-

diogr2005;18:1440–63.

10 Schurgers LJ, Aebert H, Vermeer C, BultmannB, Janzen J.Oral

anticoagulant treatment: friendor foe in cardiovascular disease?

Blood2004;104:3231–2.

11 Price PA, FausSA,WilliamsonMK.Warfarin causes rapid calcifi-

cationof the elastic lamellae in rat arteries andheart valves.Arte-

riosclerThrombVascBiol1998;18:1400–7.

12 Sweatt A, Sane DC, Hutson SM, Wallin R. Matrix Gla protein

(MGP) and bone morphogenetic protein-2 in aortic calcified

lesionsofagingrats.JThrombHaemost2003;1:178–85.

13 Schurgers LJ, Teunissen KJ, Knapen MH et al. Novel conforma-

tion-specific antibodies againstmatrix gamma-carboxyglutamic

acid (Gla) protein: undercarboxylatedmatrixGla protein asmar-

ker for vascular calcification.ArteriosclerThrombVascBiol2005;

25:1629–33.

14 Schurgers LJ, Cranenburg EC, Vermeer C. Matrix Gla-protein:

the calcification inhibitor in need of vitamin K. Thromb Haemost

2008;100:593–603.

15 Turina J, Hess O, Sepulcri F, Krayenbuehl HP. Spontaneous

courseofaortic valvedisease.EurHeartJ1987;8:471–83.

16 Lim P, Monin JL, Monchi M et al. Predictors of outcome in

patients with severe aortic stenosis and normal left ventricular

function: role ofB-typenatriuretic peptide.EurHeart J2004;25:

2048–53.

17 Bergler-Klein J, Klaar U, Heger M et al. Natriuretic peptides pre-

dict symptom-free survival and postoperative outcome in severe

aorticstenosis.Circulation2004;109:2302–8.

18 ImaiK,OkuraH,KumeT et al.C-Reactive proteinpredicts sever-

ity, progression, and prognosis of asymptomatic aortic valve ste-

nosis.AmHeartJ2008;156:713–8.

19 Schurgers LJ, Spronk HM, Soute BA, Schiffers PM, DeMey JG,

VermeerC.Regressionofwarfarin-inducedmedial elastocalcino-

sisbyhigh intakeofvitaminK inrats.Blood2007;109:2823–31.

20 SkjaerpeT,HegrenaesL,HatleL.Noninvasiveestimationof valve

area in patients with aortic stenosis by Doppler ultrasound and

two-dimensional echocardiography. Circulation 1985; 72:

810–8.

21 Jono S, Ikari Y, Vermeer C et al.Matrix Gla protein is associated

with coronary artery calcification as assessed by electron-beam

computed tomography.ThrombHaemost2004;91:790–4.

22 Hermans MM, Vermeer C, Kooman JP et al. Undercarboxylated

matrix GLA protein levels are decreased in dialysis patients and

relatedtoparametersofcalcium-phosphatemetabolismandaor-

ticaugmentation index.BloodPurif2007;25:395–401.

23 CranenburgEC,VermeerC,KoosR etal.The circulating inactive

formofmatrixGla Protein (ucMGP) asabiomarker for cardiovas-

cularcalcification.JVascRes2008;45:427–36.

24 Spronk HM, Soute BA, Schurgers LJ et al. Matrix Gla protein

accumulates at the border of regions of calcification and normal

tissue in the media of the arterial vessel wall. Biochem Biophys

ResCommun2001;289:485–90.

25 Proudfoot D, Skepper JN, Shanahan CM, Weissberg PL. Calcifi-

cationofhuman vascular cells in vitro is correlatedwith high lev-

elsofmatrixGlaproteinand low levels of osteopontinexpression.

ArteriosclerThrombVascBiol1998;18:379–88.

26 O’DonnellCJ, SheaMK, Price PAetal.MatrixGla protein is asso-

ciated with risk factors for atherosclerosis but not with coronary

artery calcification. Arterioscler Thromb Vasc Biol 2006; 26:

2769–74.

27 Hwang DM, Dempsey AA, Lee CY, Liew CC. Identification of dif-

ferentially expressed genes in cardiac hypertrophy by analysis of

expressedsequence tags.Genomics2000;66:1–14.

28 WeinbergEO,MirotsouM,GannonJ,DzauVJ, LeeRT, Pratt RE.

Sex dependence and temporal dependence of the left ventricular

genomic response to pressure overload. Physiol Genomics 2003;

12:113–27.

29 Mirotsou M, Dzau VJ, Pratt RE, Weinberg EO. Physiological

genomics of cardiac disease: quantitative relationships between

geneexpressionand left ventricularhypertrophy.PhysiolGenom-

ics2006;27:86–94.

30 Rysa J, Leskinen H, Ilves M, Ruskoaho H. Distinct upregulation

of extracellular matrix genes in transition from hypertrophy to

hypertensiveheart failure.Hypertension2005;45:927–33.

31 Blaxall BC, Spang R, Rockman HA, Koch WJ. Differential myo-

cardial geneexpression in thedevelopment and rescue ofmurine

heart failure.PhysiolGenomics2003;15:105–14.

32 Nishimoto SK, Nishimoto M. Matrix Gla protein C-terminal

region binds to vitronectin. Co-localization suggests binding

occursduringtissuedevelopment.MatrixBiol2005;24:353–61.

33 RosenkranzS.TGF-beta1andangiotensinnetworking incardiac

remodeling.CardiovascRes2004;63:423–32.

34 Mustonen E, Pohjolainen V, Aro J et al. Upregulation of cardiac

matrix Gla protein expression in response to hypertrophic stim-

uli.BloodPress2009;18:286–93.

35 Dal-Bianco JP, Khandheria BK, Mookadam F, Gentile F, Seng-

upta PP. Management of asymptomatic severe aortic stenosis.

JAmCollCardiol2008;52:1279–92.

Correspondence: Thor Ueland, Section of Endocrinology, Medical

Department,NationalUniversityHospital,N-0027Oslo,Norway.

(fax:+4723073630; e-mail: [email protected]).



Documents

Undercarboxylated matrix Gla protein is associated with indices of heart failure and mortality in symptomatic aortic stenosis