Epidemiological and histological findings implicate matrix Gla protein in diastolic left ventricular dysfunction

Objectives A novel paradigm of diastolic left ventricular (LV) dysfunction proposes involvement of the cardiac microvasculature. Vitamin K dependent matrix Gla protein (MGP) plays a role in preserving microcirculatory integrity. We hypothesized that LV filling pressure–a measure of diastolic LV dysfunction–increases with higher plasma level of inactive desphospho-uncarboxylated MGP (dp-ucMGP). We also studied the distribution of active and inactive MGP in human myocardium. Methods We measured echocardiographic diastolic LV function and plasma dp-ucMGP (ELISA) in 668 Flemish and for replication in 386 Swiss. Results Among Flemish and Swiss, E/e’ (6.78 vs. 6.73) and dp-ucMGP (3.94 μg/L vs. 4.20 μg/L) were similarly distributed. In multivariable-adjusted models, for each doubling of dp-ucMGP, E/e’ increased by 0.26, 0.33 and 0.31 in Flemish, Swiss and both cohorts combined (P≤0.026); the odds ratios for having E/e’ ≥ 8.5 were 1.99, 3.29 and 2.36, respectively (P≤0.017). Cardiac biopsies from patients with ischemic or dilated cardiomyopathy and healthy hearts (n = 4 for each) were stained with conformation-specific MGP antibodies. In diseased compared with normal hearts, uncarboxylated inactive MGP was more prevalent (P≤0.004) in the perivascular matrix and interstitium (204.4 vs. 8.6 μm2 per field) and phosphorylated active MGP in and around capillaries and interstitial cells (31.3 vs. 6.6 number of positive capillaries and cells per field). Conclusions Our study supports a role of activated MGP in maintaining myocardial integrity and diastolic LV performance and can potentially be translated into new strategies for managing diastolic LV dysfunction and preventing its progression to heart failure.


Introduction
In view of the demographic transition, heart failure (HF) is a major public health problem [1]. Diastolic HF, also referred to as HF with preserved ejection fraction, accounts for 50% of cases [2]. Mortality of diastolic HF is 30% within one year of the first hospital admission [3]. Subclinical diastolic left ventricular (LV) dysfunction has a prevalence of 25% in the general population [4,5], predisposes to further deterioration of LV function [6], and finally to overt HF [2]. A novel paradigm of diastolic HF focuses on proinflammatory signaling originating from the cardiac microvasculature [2]. Matrix Gla protein (MGP) is a small protein (11 kD) synthesized by vascular smooth muscle cells and the endothelium [7]. Activation of MGP requires two posttranslational modifications: vitamin-K dependent γ-glutamate carboxylation and serine phosphorylation [7]. Activated MGP is a potent locally acting inhibitor of calcification in large arteries [8] and protects against macrovascular complications [9,10].
Recent studies support a role of MGP in preserving the integrity of the microcirculation [11] and protecting against calcium deposition [12,13]. Relaxation of the heart requires sequestration of calcium ions in the endoplasmatic reticulum [14]. In view of the novel paradigm implicating the microcirculation in the pathogenesis of diastolic LV dysfunction [2] and the potential system-wide role of MGP [9][10][11][12][13], we hypothesized that echocardiographically assessed diastolic LV function might be inversely associated with inactive circulating desphospho-uncarboxylated MGP (dp-ucMGP). We investigated our hypothesis in a Flemish [13,15] and Swiss [8] population sample. At the time of writing of this manuscript, studies dealing with expression of MGP in the heart were scarce and did not provide any detail where in the myocardium MGP was expressed [16]. After having obtained our initial finding of association between diastolic LV function with circulating dp-ucMGP in Flemish and after having it replicated in Swiss, we undertook histological studies to confirm the presence of MGP in the heart and to determine its exact localization in healthy and diseased hearts, using conformation-specific MGP antibodies.

Population studies
The Flemish Study on Environment, Genes and Health Outcomes (FLEMENGHO) [13,15] and the Swiss KIdney Project On Genes in Hypertension (SKIPOGH) [8] are family-based population studies, which comply with the Helsinki declaration for research in human subjects [17] and received ethical approval. All participants provided informed written consent. FLE-MENGHO participants were recruited from a geographically defined area in Northern Belgium. After recruitment from 1985 until 2004 (initial participation rate, 78.0%), participants remained in follow-up. From 2005 until 2014, the re-examination included echocardiography (re-examination rate, 80.3%). Of 964 participants, 696 had dp-ucMGP measured concurrently with echocardiography. The SKIPOGH participants were enrolled from 2009 until 2013 in Berne, Geneva and Lausanne, Switzerland [8]. Of 1129 participants (participation rate, 25.6%), 390 had both echocardiographic images and plasma dp-ucMGP levels available for this analysis. For the current analysis, we excluded 17 Flemish, because of atrial fibrillation (n = 6), paced heart rhythm (n = 2) or poor image quality (n = 9) making assessment of diastolic LV function difficult. In addition, we excluded 11 Flemish and 4 Swiss, because indexes of LV function (n = 10) or MGP levels (n = 5) were more than 3 SDs away from the population mean. Thus, the number of participants statistically analyzed totaled 668 Flemish and 386 Swiss.
In the two cohorts, echocardiographic images were acquired and analyzed off-line according to current guidelines [18,19], following methods described in detail elsewhere [4]. In short, echocardiographic images were obtained with a Vivid7 Pro device (GE Vingmed, Horten, Norway) interfaced with a 2.5-to 3.5-MHz phased-array probe. For off-line analysis, we used EchoPac software, version 4.0.4 (GE Vingmed, Horten, Norway) and averaged measurements over three heart cycles. We determined the peak early (E) and peak late (A) diastolic velocities of the transmitral blood flow from the pulsed Doppler signal and peak early (e') and peak late (a') velocities of the mitral annular movement by tissue Doppler imaging (TDI) with velocities averaged over four acquisition sites (septal, lateral, inferior, and posterior). The intra-observer reproducibility coefficient for the single observer in FLEMENGHO was 4.8% for e' and 4.2% for a' [6]. In SKIPOGH (two observers), the intra-and inter-observer reproducibility coefficients were 4.4% and 13.8% for e' and 5.0% and 12.0% for a', respectively. To stage diastolic LV function, we combined the velocities of the transmitral blood flow and the mitral annular movement [4]. Patients with diastolic LV dysfunction had an abnormally low age-specific transmitral E/A ratio indicative of impaired relaxation, a mildly-to-moderately elevated LV filling pressure (E/e' >8.5) with normal or decreased age-specific E/A ratio. These age-specific criteria in a healthy reference sample drawn from FLEMENGHO [4] were replicated in an independent European population study [5].
Blood pressure was the average of five consecutive auscultatory readings obtained with a standard mercury sphygmomanometer in FLEMENGHO [13,15] and with a non-mercury manual auscultatory sphygmomanometer (A&D UM-101; A&D Company Ltd, Toshima Ku, Tokyo, Japan [20]) in SKIPOGH [8]. Nurses administered questionnaires inquiring into each participant's medical history, smoking and drinking habits, and intake of medications. Fasting blood samples were analyzed for glucose and total cholesterol, using automated methods in certified laboratories. In both cohorts, dp-ucMGP was measured by VitaK (Maastricht University, The Netherlands) on citrated plasma by pre-commercial ELISA kits [21].
For database management and statistical analysis, we used the SAS system, version 9.4 (SAS Institute Inc., Cary, NC). Significance was a two-tailed α-level of 0.05 or less. Means were compared using the large-sample z-test and proportions by Fisher's exact test. We normalized the distribution of dp-ucMGP by a logarithmic transformation. We compared cohort-, sex-and age-standardized echocardiographic measurements across halves of the dp-ucMGP distribution. We applied mixed models and logistic regression to model the multivariable-adjusted associations of continuous or categorical LV traits with dp-ucMGP, while accounting for clustering within families and cohort as random effects, as appropriate. Covariables with potential relevance have been described in previous publications [4][5][6] and included sex, age, body mass index, mean arterial pressure, pulse pressure, heart rate, serum total cholesterol, plasma glucose, left ventricular mass indexed to body surface area, alcohol intake and use of antihypertensive drugs. Analyses including both Flemish and Swiss were additionally adjusted for center.

Histological studies
Cardiac tissue samples were obtained from patients with end-stage ischemic (ICM, n = 4) or dilated (DCM, n = 4) cardiomyopathy undergoing cardiac transplantation at the University Hospitals Leuven, Belgium. Control tissue samples were obtained from unused donor hearts (HD, n = 4). The Ethics Committee of the University Hospitals Leuven approved the study (approval numbers B322201421186 [S56384] and B322201421045 [S56472]) and it passes ethical screening by the European Research Council Executive Agency (ERCEA). Patients undergoing transplantation provided informed written consent. Donors of healthy hearts discarded for transplantation were unconscious and could not provide consent. However, their next of kin did provide consent for removal of the donor heart. The identity of the donors remained unknown to the transplant team in Leuven. As outlined above, the Ethics Committee of the University Hospitals Leuven approved the use of discarded donor hearts for research.

Population studies
The geometric mean concentration of dp-ucMGP was 3.94 μg/L (interquartile range, 2.89-5.76 μg/L; 5-95th percentile interval, 1.50-8.08 μg/L) in Flemish and 4.20 μg/L (interquartile Table 2. Sex-and age-standardized echocardiographic measurements by median of dp-ucMGP distribution. range, 3.09-5.88 μg/L; 5-95th percentile interval, 1.65-8.74 μg/L) in Swiss (Fig 1). Table 1 lists the characteristics of participants by the cohort-specific median of the dp-ucMGP distribution. Age, body mass index, systolic and diastolic blood pressure, heart rate, the prevalence of hypertension and use of antihypertensive drugs, total cholesterol, plasma glucose, and serum creatinine all increased (P 0.023) with higher dp-ucMGP category, whereas the opposite was the case for smoking (P<0.001). The distribution of these cardiovascular risk factors across categories of dp-ucMGP was consistent in Flemish, Swiss and all participants combined (Table 1). Table 2 shows the sex-and age-standardized echocardiographic measurements by the cohort-specific median of the dp-ucMGP distributions. In Flemish and in all participants combined, the A peak velocity and the E/e' ratio increased (P 0.026) with higher dp-ucMGP category, whereas the opposite was the case (P 0.012) for the e' peak velocities and the ratios E/A and e'/a'. In Swiss, trends were similar, but significance was only reached for the decrease in e' (P = 0.002) and the increase in E/e' (P = 0.007) with higher dp-ucMGP category ( Table 2).
Single-model tests demonstrated that the multivariable-adjusted regression lines relating the echocardiographic indexes to dp-ucMGP were all coincident in Flemish and Swiss. In Flemish, Swiss and all participants combined, each doubling of dp-ucMGP was associated Odds ratios relating diastolic LV dysfunction to dp-ucMGP. For definition of the age-specific criteria of impaired relaxation and increased filling pressure, see Methods. The analyses accounted for cohort and family cluster and were adjusted for sex, age, body mass index, mean arterial pressure, pulse pressure, heart rate, total cholesterol, plasma glucose, left ventricular mass index, alcohol intake and antihypertensive drug treatment. Horizontal bars denote the 95% confidence interval. For each entry, the number of people with diastolic LV dysfunction vs. normal function are given.  (Table 3). Considering the e' denominator of the E/e' ratio, the association sizes were -0.21 cm/s (P = 0.049) in Flemish, -0.43 cm/s (P = 0.001) in Swiss and -0.31 cm/s (P<0.001) in all participants. The E numerator was not associated with dp-ucMGP in any cohort (P!0.217). For each doubling of dp-ucMGP, the transmitral A peak increased by 1.26 cm/s (P = 0.034) in Flemish and by 1.28 cm/s (P = 0.014) in both cohorts combined with a similar trend in Swiss (1.07 cm/s; P = 0.279); the e'/a' ratio decreased by 0.034 (P = 0.002) in both cohorts combined with a similar trend in Flemish (-0.029; P = 0.146) and in Swiss (-0.044; P = 0.079). The E/A ratio (P!0.211) and the a' peak (P!0.582) were not associated with dp-ucMGP in any cohort.
In multivariable categorical analyses, impaired LV relaxation was not associated with dp-ucMGP in any cohort (0.96 odds ratio 1.27; P!0.455; Fig 2). Increased LV filling pressure was associated with dp-ucMGP in all cohorts with odds ratios for a doubling of dp-ucMGP amounting to 1.99 (P = 0.017) in Flemish, 3.29 (P<0.001) in Swiss and 2.36 (P<0.001) in all participants (Fig 2). For impaired relaxation combined with increased LV filling pressure, the odds ratios were 1.40; P = 0.061) in Flemish, 1.72 (P = 0.032) in Swiss and 1.54 (P = 0.004) in all participants (Fig 2).

Histological studies
All patients with heart disease and donors were white. Age at transplantation ranged from 20 to 66 years and from 59 to 63 years in patients with dilated or ischemic cardiomyopathy, respectively (Table 4). Age at the death among the donors ranged from 63 to 82 years. The causes of death in donors were hemorrhagic stroke in three and ischemic stroke in one. Age was the main criterion why donor hearts were not implanted. All patients with cardiomyopathy wore a cardiac resynchronization, defibrillator or left ventricular assist device before being transplanted (Table 4).  The active MGP moieties, cMGP ( Fig 3B) and pMGP (Fig 3D), were predominantly distributed in the media and intima of muscular left ventricular microvessels in healthy and diseased hearts. Inactive ucMGP was abundant in fibrotic areas of diseased hearts, around the nuclei of interstitial cells and in the perivascular matrix (Fig 3A and Fig 4). ucMGP was more abundant in DCM and ICM than HD hearts (mean±SEM, 105.1±10.6 and 303.8±206.1 vs. 8.6±3.4 μm2/ field; P = 0.029; Fig 3E). In ICM hearts, ucMGP was particularly present in fibrotic areas and, depending on the degree of fibrosis, showed large variability between patients (Table 4). ucMGP staining was absent in cardiomyocytes ( Fig 3A). Furthermore, DCM and ICM myocardium showed more pMGP positive capillaries and interstitial cells than HD hearts (37.0 ±2.0 and 25.7±6.0 vs. 6.6±1.7; P = 0.029; Fig 3F and Table 4). Finally, staining for inactive dpMGP was almost absent in the vessel wall and in fibrotic areas, but was abundant in cardiomyocytes of all hearts and co-localized with active cMGP (Fig 3C). There were no differences in cMGP (Figs 5 and 6) and pMGP (Figs 7 and 8) staining between an older (61 years) and young (20 years) patient with DCM.

Discussion
Paulus and Tschöpe proposed a novel paradigm that shifts emphasis in diastolic LV dysfunction from LV afterload to inflammation of the coronary microcirculation [2]. This paradigm [2] along with the premise that MGP is promoting the integrity of the microcirculation [11] and might be a local inhibitor of soft tissue calcification [12,13] provided the rationale for the hypothesis that echocardiographically assessed diastolic LV function might be inversely associated with circulating inactive dp-ucMGP. We confirmed our hypothesis in two population cohorts, while subsequent tissue staining studies revealed the localization of the active and inactive MGP moieties in the myocardium, thereby supporting a role of active MGP in maintaining a healthy myocardium.
In diseased compared with normal hearts, uncarboxylated MGP was more prevalent in the perivascular matrix and interstitium and phosphorylated MGP in and around capillaries and interstitial cells. Cardiomyocytes showed no staining for uncarboxylated MGP. Unphosphorylated MGP was almost absent in vessel walls and fibrotic areas, but was present in cardiomyocytes of all hearts and co-localized with carboxylated MGP. The ischemia-induced fibrotic remodeling in ICM patients seemed to be accompanied by an even higher deposition of ucMGP (Fig 3A and 3F). Fig 9 illustrates the hypothetical sequence of events possibly explaining our histological observations. Once secreted into the extracellular matrix, carboxylated and phosphorylated MGP protects against calcium deposition [22,23] and inhibits trans-differentiation of vascular smooth muscle cells [24] and signaling via the bone morphogenetic protein (BMP) pathway [25][26][27]. Taken together with the literature [22][23][24][25][26][27], our current findings support the idea that activated MGP is an ubiquitous locally acting agent protecting the microcirculation and the perivascular matrix.
Several mechanisms might explain how activated MGP might promote preservation of diastolic LV function. First, MGP binds to calcium ions as well as to hydroxyapatite crystals and may thereby inhibit crystal growth [22]. Depending on the amino-acid sequence, varying phosphorylated and γ-carboxylated sequences of human MGP inhibited either nucleation or growth, or both, of hydroxyapatite and calcium oxalate monohydrate crystals [23]. Relevant to the current study, is the strong well-documented protein-protein interaction between MGP of ucMGP and pMGP among HD, DCM and ICM appear in panels E and F, respectively. n refers to the number of tissue samples included in the quantitative analysis. Labels: asterisks indicate cardiomyocytes; arrows point to interstitial cells with perinuclear MGP deposition; L lumen of muscularized microvessels.
https://doi.org/10.1371/journal.pone.0193967.g003 and BMP, including BMP-2 [25,26] and BMP-4 [27], whereby bound MGP reduces BMP signaling. In the heart, BMP pathways play a pivotal role in the embryogenesis of the LV chamber [28], the differentiation of cardiac progenitor cells into functional cardiomyocytes [29], maintenance of the balance between LV growth and apoptosis [30], initiating fibrosis [30], and Ca2 + channel remodeling [31]. In line with the paradigm that the microcirculation is involved in the pathogenesis of diastolic LV function [2], BMP signaling also promotes endothelial apoptosis [32], increases endothelial permeability [32], and is involved in the expression of vascular endothelial growth factor and angiogenesis [27].
Circulating dp-ucMGP is a biomarker reflecting vitamin K deficiency [33]. Indeed, among 60 middle-aged healthy volunteers randomized in a placebo-controlled double-blind trial, plasma dp-ucMGP dropped dose-dependently by 31% and 46% in response to daily supplementation for 12 weeks with 180 μg and 360 μg of menaquinone-7 (vitamin K2) [33]. Selectively re-introducing MGP expression in the liver of MGP knockout mice, produced circulating MGP levels 6-to 10-fold higher than in wild type animals [34]. The MGP originating from the transgene conserved its biological activity in vitro, but did not inhibit arterial calcification [34], confirming that MGP is a locally acting paracrine protein. We did not measure circulating levels of vitamin K, which is rarely done in clinical practice, because of the complexity of the assay and the lack of a high-throughput method [35] and because plasma levels only reflect dietary intake (vitamin K1; phylloquinone) and production by the gut microflora (vitamin K2;   Diastolic LV function and matrix Gla protein menaquinones) without giving any indication of functionality, i.e. the amount of MGP undergoing carboxylation [33].
The clinical implications of our current findings rest on the prognostic significance of the E/e' ratio [36][37][38][39]. Among 816 hypertensive patients with echocardiographic data randomized in the Anglo-Scandinavian Cardiac Outcomes Trial (ASCOT), 56 cardiac events occurred over 4.2 years. The E/e' ratio was the strongest predictor of a first cardiac event. Following adjustment for covariables, a one unit rise in the E/e' ratio was associated with 17% increment in risk of a cardiac event (95% confidence interval, 5% to 29%; P = 0.003) [38]. Similarly, among 406 patients with type-2 diabetes mellitus, the E/e' ratio was an independent predictor of cardiovascular events occurring over 5.6 years, including 12 myocardial infarctions and 7 strokes. In analyses adjusted for sex and age, a one unit rise in the E/e' ratio was associated with 8% increment in risk (1% to 16%; P = 0.021), whereas the association with global LV strain was not significant (P = 0.913) [39]. Along similar lines, e' predicted mortality in a study including 174 hypertensive patients and 78 age-matched controls [40]. Our current findings focusing on diastolic LV function, are in line with our previous report showing that high dp-ucMGP predicted adverse health outcomes, including total and cardiovascular mortality in 2318 FLEMENGHO participants followed-up for a median of 14.1 years [9].
Strong points of our current study are the consistency of the association of the E/e' ratio with dp-ucMGP in two population cohorts and the histological evidence supporting the epidemiological observations. However, our current study must also be interpreted within the context of its potential limitations. First, our population studies had a cross-sectional design, which precludes direct causal inference. Second, we did not measure circulating dp-ucMGP in the Step 1: After translation in the endoplasmatic reticulum (ER), vitamin K activates MGP by stimulating γ-carboxylation.
Step 3: A Golgi-associated casein kinase phosphorylates the serine residues of dp-cMGP to p-cMGP, thereby facilitating secretion.
Step 4: p-cMGP is secreted into the extracellular matrix or the circulation to inhibit soft tissue calcification, vascular smooth muscle cell (VSMC) trans-differentiation into osteochondrogenic progenitor cells (OPC) and signaling via the BMP (bone morphogenetic protein) pathway.
Step 5: Inactive desphospho-uncarboxylated MGP (dp-ucMGP), a biomarker reflecting poor vitamin K status, escapes from cells into the blood stream, but does not inhibit calcification.
https://doi.org/10.1371/journal.pone.0193967.g009 patients enrolled in the histological studies and we cannot conclude to what extent the patients with end-stage DCM or ICM are representative for early-stage diastolic LV dysfunction as observed in the general population. Third, the sample size of histological studies was relatively small, but of the same order of magnitude as in other studies of the same nature [41,42]. Fourth, the location of MGP species was not confirmed by Western blots, because the antibodies available to us had never before been validated for use in such studies. Finally, for the histological studies, we used conformation-specific MGP antibodies, but we did not apply measurement of protein expression or mRNA.

Conclusions
In the general population, E/e' increased with higher dp-ucMGP, a marker of vitamin K deficiency. The combined epidemiological and histological findings suggest that poor vitamin K status, as exemplified by higher plasma dp-ucMGP, represents a risk factor for diastolic LV dysfunction. Whether our current findings open new avenues to the prevention or treatment of diastolic LV dysfunction or its progression to HF remains to be established in randomized clinical trials of vitamin-K supplementation.