Comprehensive Proteomics Profiling Reveals Circulating Biomarkers of Hypertrophic Cardiomyopathy

Abstract

Background:

Hypertrophic cardiomyopathy (HCM) is caused by mutations in the genes coding for proteins essential in normal myocardial contraction. However, it remains unclear through which molecular pathways gene mutations mediate the development of HCM. The objectives were to determine plasma protein biomarkers of HCM and to reveal molecular pathways differentially regulated in HCM.

Methods:

We conducted a multicenter case-control study of cases with HCM and controls with hypertensive left ventricular hypertrophy. We carried out plasma proteomics profiling of 1,681 proteins. We performed a sparse partial least squares discriminant analysis to develop a proteomics-based discrimination model with data from 1 institution (i.e., the training set). We tested the discriminative ability in independent samples from the other institution (i.e., the test set). As an exploratory analysis, we executed pathway analysis of significantly dysregulated proteins. Pathways with false discovery rate <0.05 were declared positive.

Results:

The study included 266 cases and 167 controls (n=308 in the training set; n=125 in the test set). Using the proteomics-based model derived from the training set, the area under the receiver-operating-characteristic curve was 0.89 (95%CI 0.83–0.94) in the test set. Pathway analysis revealed that the Ras-MAPK pathway, along with its upstream and downstream pathways, was upregulated in HCM. Pathways involved in inflammation and fibrosis – e.g., the TGF-β pathway – were also upregulated.

Conclusions:

This study serves as the largest-scale investigation with the most comprehensive proteomics profiling in HCM, revealing circulating biomarkers and exhibiting both novel (e.g., Ras-MAPK) and known (e.g., TGF-β) pathways differentially regulated in HCM.

INTRODUCTION

Hypertrophic cardiomyopathy (HCM) is the most common genetic cardiac disease affecting 1 in 200–500 people in the US,¹ and constitutes the most common cause of sudden cardiac death in the young, especially in competitive athletes.² HCM is caused by mutations in the genes coding for proteins constructing contractile apparatus in the myocardium – i.e., the sarcomeric genes.^{3, 4} A striking characteristic of HCM is that the extent of disease severity, as manifested, for example, by New York Heart Association (NYHA) functional class, left atrial diameter, and left ventricular (LV) ejection fraction, is extremely variable.⁵ However, it is not well known through which signaling pathways a gene mutation causes functional and morphological changes in the heart (i.e., disease pathogenesis) and may lead to the development of more severe manifestations of HCM (i.e., disease progression).

Proteomics profiling is a recently developed tool to simultaneously determine the concentration of thousands of proteins in the tissue or fluid. Proteomics profiling has been applied to cardiovascular disease and cardiomyopathies other than HCM (e.g., dilated, muscular dystrophy-associated), revealing novel signaling pathways underlying disease pathogenesis and progression.^6–11 In HCM, our prior study of 15 patients with HCM and 22 controls with other cardiovascular disease demonstrated a high discriminative accuracy and displayed that the Ras-MAPK and associated signaling pathways were upregulated in HCM.¹² Yet, the prior study was limited by the small size, the lack of an independent validation cohort, and the relatively small number of analyzed proteins (~1,000). Furthermore, it remains unclear whether the Ras-MAPK pathway upregulation is also present in other conditions that cause left ventricular hypertrophy (LVH) – e.g., hypertension – or is unique to HCM.

A growing body of evidence indicates that it is possible to detect changes in the myocardium by the analysis of plasma proteins.^{3, 4} For example, molecules associated with fibrotic changes and inflammation in the heart have been successfully measured in plasma.^{3, 4, 12, 13} Furthermore, cardiac troponins can be detected in plasma in patients with heart failure, suggesting the release of intracellular contents.¹⁴ These data provide supporting evidence for the hypothesis that it is possible to detect intracellular components and proteins at the cell surface of cardiomyocytes by analyzing plasma samples.

We therefore designed the present study to identify plasma protein biomarkers of HCM and to specify signaling pathways that are (1) differentially regulated in HCM compared to hypertensive LVH and (2) correlated with clinical markers of disease severity.

METHODS

The data that support the findings of the present study are available from the corresponding author upon reasonable request.

Study design and sample

We performed a case-control study between cases with HCM and controls with hypertensive LVH. We chose hypertensive LVH as the control group because (1) our prior study has already demonstrated significant differences between patients with HCM and those with other cardiovascular diseases, and (2) it remains unclear whether the differences are attributable to the presence of LVH itself or are unique to HCM.¹² We enrolled patients with HCM that were evaluated in the HCM program and patients with hypertensive LVH followed in the general cardiology clinic at either Massachusetts General Hospital (Boston, MA) or Columbia University Irving Medical Center (New York, NY).^{12, 15, 16} The training set for the discrimination model derivation consisted of cases and controls from Massachusetts General Hospital. As an independent test set for validation, we collected data from patients with HCM or hypertensive LVH followed at Columbia University Irving Medical Center. We also obtained additional control samples in the test set from the Partners Biobank.¹⁷ We diagnosed HCM if there was echocardiographic evidence of LVH – i.e., maximal LV wall thickness ≥15 mm – out of proportion to systemic loading conditions and a nondilated LV.^{12, 15, 16, 18} We lowered the diagnostic threshold of LV wall thickness if there was family history of HCM.¹⁹ We confirmed the diagnosis with the use of the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) diagnosis code of 425.1x and the ICD-10 diagnosis code of I42.x. We excluded patients with HCM phenocopies – e.g., Noonan syndrome, Fabry disease, and cardiac amyloidosis – by performing thorough medical interview and physical examination, and if needed, additional tests (e.g., cardiac magnetic resonance imaging, genetic testing).^{12, 15, 16, 18} We considered variants categorized as “definitely pathogenic” or “likely pathogenic” as positive genotype and those classified as “variant of uncertain significance,” “likely benign,” or “benign” as negative genotype.^8,15,16 We diagnosed hypertensive LVH if the patient had history of hypertension, especially uncontrolled or long-standing, and echocardiographic evidence of concentric LVH in proportion to the degree of systemic loading. From the control group, we excluded patients with any suspicion for HCM and those with family history of HCM. We further confirmed the diagnosis of hypertensive LVH using the ICD-9-CM diagnosis code of 402.xx and the ICD-10 diagnosis code of I11.x. We excluded patients with aortic stenosis, subaortic membrane, and athlete’s heart. The Partners Institutional Review Board and that of Columbia University Irving Medical Center approved the study protocol, and all participants provided written informed consent.

Blood sample processing and proteomics profiling

We drew venous blood samples at outpatient clinic visits. We collected blood in K₂EDTA-treated tubes and centrifuged them at 3,100 rpm for 10 minutes. We aliquoted the supernatant plasma immediately and kept the samples in the freezer at −80°C.¹²

We performed proteomics profiling of 1,681 proteins with the use of the SomaScan assay (SomaLogic, Inc., Boulder, CO).^{12, 20, 21} This assay determines concentrations of proteins in plasma over a wide range of abundance, from femtomolar to micromolar, with a high level of reproducibility – the median coefficient of variation is 4.6%.^{12, 20, 21} The performance of the SomaScan assay is comparable to that of sandwich enzyme linked immunosorbent assay.^{12, 20, 21} The SomaScan assay is particularly useful in accurately measuring concentrations of low-abundance proteins that cannot be detected with the conventional liquid chromatography/mass spectrometry-based approach.²² Details of the SomaScan assay are provided in Supplemental Methods and have been published elsewhere.^{12, 20, 21}

Univariable analysis

We displayed continuous values as mean ± standard deviation if normally distributed and as median [interquartile range] if not normally distributed. To compare the clinical characteristics between cases and controls, we performed the unpaired Student’s t-test for normally distributed continuous variables, the Mann-Whitney-Wilcoxon test for ordinal variables (e.g., NYHA functional class, degree of mitral regurgitation), and the chi-square test for categorical variables.

Development of a proteomics-based model to distinguish cases from controls

We carried out a sparse partial least squares discriminant analysis (sPLS-DA) – a method of supervised machine learning to examine the discriminative ability of multi-dimensional data – to develop a proteomics-based model to discriminate cases from controls.¹² We chose this analytic method because proteomics data are sparse (i.e., a small proportion of analyzed proteins contributes heavily to the discrimination) and collinear (i.e., multiple proteins correlate with each other).²³ To preprocess the variables for the machine learning model, we performed sample-wise normalization using the median of all protein concentrations in each sample followed by protein-wise log transformation and Pareto scaling in each protein concentration. We then created a hyperparameter tuning grid to identify best hidden component and threshold parameter using the R spls and caret packages. We also performed a permutation test with 2,000 permutations to determine the statistical significance of the discrimination model. To measure the test performance of the discrimination model developed from the training set, we computed the discriminative ability of the model in the independent test set. As indicators of the discriminative ability, we calculated the area under the receiver-operating-characteristic curve (AUC), sensitivity, and specificity. We conducted 2 subgroup analyses by stratifying the case group by (1) the presence of hypertension and (2) genotype.

We performed several sensitivity analyses. First, we incorporated variables for medications with a significant univariable difference in the usage between the case and control groups in the sPLS-DA model. Second, to remove the potential effects of prior alcohol septal ablation on the proteomic profiles, we identified proteins whose concentrations were significantly (i.e., univariable p <0.05) associated with prior alcohol septal ablation status using the Wilcoxon rank-sum test and removed these proteins from the discrimination model. Third, to eliminate the influence of renal function, we removed proteins significantly correlated with creatinine. Finally, to adjust for cardiovascular risk factors, we have forcibly included age, sex, body mass index, systolic blood pressure, prior atrial fibrillation, prior ventricular tachycardia/fibrillation, and left atrial diameter in the discrimination model. We have compared the AUCs between the model with cardiovascular risk factors alone and that with the proteomics profiling data plus cardiovascular risk factors using the Delong’s test.

Pathway and network analyses

As an exploratory analysis, we performed pathway analysis to specify the canonical pathways that are either upregulated or downregulated in HCM compared to hypertensive LVH. To perform pathway analysis, we conducted a univariable analysis to examine the association between each protein concentration and disease status using the Mann-Whitney-Wilcoxon test. We included proteins with Bonferroni-corrected p value <0.05 (i.e., uncorrected p <0.05/1681 = 3.0×10⁻⁵) in the pathway analysis. We determined the associations among the most discriminant proteins and canonical pathways listed in the Kyoto Encyclopedia of Genes and Genomes.²⁴ We tested for significance based on the ratio of the number of proteins within the most discriminant proteins that map to a canonical pathway divided by the number of proteins that belong to the pathway.²⁴ We declared a pathway as positive (i.e., upregulated or downregulated) if the false discovery rate (FDR) was <0.05 and there were at least 5 associated proteins.²⁵ We further performed network analysis to visualize interconnections among the proteins included in the pathway analysis.²⁴ To avoid potential effects of age on the protein concentrations, we repeated the pathway analysis after removing proteins whose concentrations were significantly correlated with age (i.e., univariable p <0.05 with the Spearman’s rank correlation test). Additionally, we performed the pathway analysis by using the protein importance ranking determined by the discrimination model including the cardiovascular risk factors.

Correlation with clinical markers of disease severity

In patients with HCM, we correlated concentrations of the proteins with NYHA functional class, left atrial diameter, LV ejection fraction, duration of disease (i.e., time since diagnosis), and maximal LV wall thickness using the Spearman rank-order correlation test. Additionally, we treated LV ejection fraction as a dichotomized variable with the cutoff value of 50%. We then performed pathway analysis of proteins that were significantly (i.e., p <0.05 with univariable analysis) correlated with the clinical markers of disease severity. We also conducted the same analyses on NYHA functional class, left atrial diameter, and LV ejection fraction in the control group. We used Metaboanalyst 4.0 (University of Alberta, Canada) to perform the sPLS-DA,²⁶ STRING Version 10.5 (String Consortium, Europe) for the pathway and network analyses,²⁴ and Stata Statistical Software: Release 12 (StataCorp LP, College Station, TX) with the Stata command scat3 to create Figure 1. We conducted the remaining analyses with R version 3.5.1.

Figure 1. Three-dimensional score plot of proteomics profiling in cases with hypertrophic cardiomyopathy and controls with hypertensive left ventricular hypertrophy.

Each red dot represents the proteomic profile of an HCM case. Each green dot corresponds to that of a hypertensive LVH control. N = 433 (266 cases and 167 controls).

HCM = hypertrophic cardiomyopathy, LVH = left ventricular hypertrophy

RESULTS

Overall, 433 patients (266 cases and 167 controls) were included in the analysis. In this analytic cohort, 191 HCM cases and 117 controls from Massachusetts General Hospital comprised the training set to derive the discrimination model; the independent test set for validation consisted of 75 HCM cases and 50 controls from Columbia University Irving Medical Center and the Partners Biobank. Baseline patient characteristics of the training set are shown in Table 1, while those of the test set are displayed in Supplemental Table I. Patients with HCM were younger and more predominantly European ancestry and had a greater septal wall thickness than patients with hypertensive LVH. There were also differences in clinical parameters and in medications, as expected for the clinical entities. In the case group, 41 patients (21%) had an implantable cardioverter-defibrillator. Pathologic or likely pathologic mutation was found most commonly in the MYBPC3 gene (n = 13) followed by the MYH7 gene (n = 6).

Table 1.

Baseline clinical characteristics of the study sample in the training set

Characteristics*	HCM	Hypertensive LVH	P value
	(n = 191)	(n = 117)
Demographics
Age (year)	61 ± 14	67 ± 10	<0.001
Male	114 (60)	77 (66)	0.28
Body mass index (kg/m2)	31 ± 6	32 ± 7	0.15
NYHA functional class ≥2	86 (45)	9 (8)	<0.001
Race/ethnicity			0.001
European ancestry	170 (89)	84 (72)
African-American	5 (3)	5 (4)
Asian	5 (3)	9 (8)
Other or unidentified	11 (6)	19 (16)
Medical History
Hypertension	96 (50)	117 (100)	<0.001
Prior AF	53 (28)	29 (25)	0.56
Prior VT/VF	8 (4)	3 (3)	0.46
Prior non-sustained VT	36 (20)	5 (4)	-‡
Prior syncope	40 (22)	6 (5)	-‡
Family history of sudden cardiac death	17 (9)	3 (3)	-‡
Family history of HCM	51 (28)	0 (0)	<0.001
Obstructive HCM	78 (41)	-	-
Prior septal myectomy	30 (16)	-	-
Time from septal myectomy to enrollment (year)	4.1 [1.3–7.5]
Prior alcohol septal ablation	21 (11)	-	-
Time from alcohol septal ablation to enrollment (year)	6.9 [4.3–11.7]
Medications
β-blocker	130 (70)	81 (71)	0.83
Calcium channel blocker	39 (21)	37 (32)	0.03
ACE inhibitor	16 (8)	45 (39)	<0.001
ARB	26 (14)	29 (25)	0.009
Diuretic
Loop diuretic	28 (15)	32 (28)	0.004
Thiazide	15 (8)	28 (25)	<0.001
Potassium sparing diuretic	10 (5)	19 (17)	0.001
Disopyramide	17 (9)	1 (1)	0.003
Amiodarone	5 (3)	6 (5)	0.25
Blood pressure
Systolic blood pressure (mmHg)	123 ± 14	132 ± 17	<0.001
Diastolic blood pressure (mmHg)	74 ± 9	73 ± 10	0.41
Renal function
Creatinine (mg/dL)	1.0 ± 0.2	1.4 ± 1.3	<0.001
End-stage renal disease	0 (0)	6 (5)	0.006
Echocardiographic measurements
Left atrial diameter (mm)	42 ± 7	42 ± 7	0.68
Interventricular septum thickness (mm)	16 ± 4	13 ± 2	<0.001
Posterior wall thickness (mm)	12 ± 2	12 ± 1	0.31
Maximal LV wall thickness (mm)	20 ± 5	13 ± 2	<0.001
LV outflow tract gradient (mmHg) at rest among patients with obstructive HCM	38 [18–80]	-	-
LV outflow tract gradient (mmHg) with Valsalva maneuver among patients with obstructive HCM	80 [43–100]	-	-
LV ejection fraction (%)	72 ± 7	61 ± 11	<0.001
LV end-diastolic diameter (mm)	43 ± 6	48 ± 7	<0.001
LV end-systolic diameter (mm)	26 ± 5	32 ± 7	<0.001
Systolic anterior motion of mitral valve leaflet	88 (48)	1 (1)	<0.001
Degree of mitral regurgitation†	2 [1–2.5]	1 [1–2]	<0.001
LV morphology
Apical	31 (16)	-	-
Neutral	77 (40)	-	-
Reverse curve	40 (21)	-	-
Sigmoid	31 (16)	-	-
Other	1 (1)	-	-
Concentric LVH§	159 (87)	94 (83)	0.41
Genetic testing (n = 81)
Pathogenic or likely pathogenic	25 (31)	-	-

^*Data were expressed as number (percentage), mean ± standard deviation, or median [interquartile range].

^†Degree of mitral regurgitation was converted to numerical values according to the following rule: none = 0, trace = 1, trace to mild = 1.5, mild = 2, mild to moderate = 2.5, moderate = 3, moderate to severe = 3.5, severe = 4.

^‡P values were not reported as the ascertainment may have been different between the groups.

^§Defined as relative wall thickness >0.42.

ACE = angiotensin-converting enzyme; AF = atrial fibrillation; ARB = angiotensin II receptor blocker; HCM = hypertrophic cardiomyopathy; LVH = left ventricular hypertrophy; NYHA = New York Heart Association; VT/VF = ventricular tachycardia or ventricular fibrillation

As shown in Figure 1, proteomic profiles were distinctively different between HCM cases and controls. The permutation test showed p value <0.0005, indicating that the observed or more extreme differences were found only in <1 out of 2,000 random permutations assuming the null hypothesis (Supplemental Figure I). The discrimination model based on proteomics profiling in the training set exhibited a high discriminative ability in the independent test set for validation (AUC 0.89, 95% confidence interval [CI] 0.83–0.94; Figure 2). The sensitivity was 0.84 (95% CI 0.76–0.92) and the specificity was 0.78 (95% CI 0.66–0.90). Figure 3 lists 30 proteins with the highest importance in discriminating the cases from the controls according to the sPLS-DA. The list included both known (e.g., brain natriuretic peptide) and unknown cardiomyopathy biomarkers.

Figure 2. Area under the receiver-operating-characteristic curve in the independent test set, using the proteomics-based discrimination model developed in the training set.

N = 308 (191 cases and 117 controls) in the training set; N = 125 (75 cases and 50 controls) in the test set.

AUC = area under the receiver-operating-characteristic curve, CI = confidence interval

Figure 3. The 30 most discriminant proteins to distinguish cases with hypertrophic cardiomyopathy from controls with hypertensive left ventricular hypertrophy.

A red box indicates that the protein concentration was increased in HCM, while a green box means that the concentration was decreased in HCM. P values were computed using the Mann-Whitney-Wilcoxon test. Fold change was calculated by dividing the median in the cases by the median in the controls. The discrimination model was developed using the training set with N = 308 (191 cases and 117 controls).

HCM = hypertrophic cardiomyopathy

In the subgroup analyses, the discrimination model remained robust when the cases were stratified by the presence of hypertension (Supplemental Figure II) or genotype (Supplemental Figure III). The sensitivity analysis incorporating parameters reflecting medication use showed a similar AUC of 0.90 (95% CI 0.85–0.95; p = 0.17 with Delong’s test compared to the proteomics-only model; Supplemental Figure IV). The findings were materially unchanged after removing proteins associated with prior alcohol septal ablation (AUC 0.89, 95% CI 0.83–0.94; Supplemental Figure V) or correlated with creatinine (AUC 0.89, 95% CI 0.83–0.94; Supplemental Figure VI) and after adding the cardiovascular risk factors to the discrimination model (AUC 0.89, 95% CI 0.84–0.95; Supplemental Figure VII). This AUC was significantly higher than that of the model using only cardiovascular risk factors (0.53, 95% CI 0.42–0.63; p <0.001; Supplemental Figure VII).

A total of 508 proteins were significantly (i.e., Bonferroni-corrected p <0.05) associated with the disease status as shown in Supplemental Table II. Pathway analysis using the 508 most discriminant proteins revealed that the Ras-MAPK pathway was significantly upregulated in HCM along with its upstream (e.g., TNF, JAK-STAT) and downstream (e.g., PI3K-Akt, Rap1) pathways (Table 2). Pathways involved in inflammation – e.g., cytokine-cytokine receptor interaction, complement and coagulation cascades, and chemokine signaling – were also upregulated in HCM. The TGF-β signaling pathway, which has been linked to HCM, was found to be upregulated in the current analysis as well.²⁷ The findings in the pathway analysis were similar when proteins correlated with age were removed (Supplemental Table III) or when the cardiovascular risk factors were included in the model (Supplemental Table IV). Network analysis of the 508 most discriminant proteins demonstrated that there was a significantly larger number of interactions than expected (observed = 2,050; expected = 1,312; protein-protein interaction enrichment p <1.0×10⁻¹⁶). As shown in Supplemental Figure VIII, protein members of the Ras-MAPK pathway were at the hub of the interaction network.

Table 2.

Pathways that are differentially regulated between cases with hypertrophic cardiomyopathy and controls with hypertensive left ventricular hypertrophy

Pathway description	Number of matching proteins	Number of proteins in the pathway	False discovery rate
Cytokine-cytokine receptor interaction	28	263	0.0000005
MAPK signaling pathway	26	293	0.00003
PI3K-Akt signaling pathway	28	348	0.00004
Ras signaling pathway	21	228	0.0001
Complement and coagulation cascades	10	78	0.004
Apoptosis	12	135	0.01
Focal adhesion	15	197	0.01
Chemokine signaling pathway	14	181	0.01
TNF signaling pathway	10	108	0.02
Proteasome	6	43	0.03
Rap1 signaling pathway	14	203	0.03
ECM-receptor interaction	8	81	0.03
ErbB signaling pathway	8	83	0.03
TGF-β signaling pathway	8	83	0.03
Fluid shear stress and atherosclerosis	10	133	0.04
JAK-STAT signaling pathway	11	160	0.04

ECM = extracellular matrix; JAK = Janus kinase; MAPK = mitogen-activated protein kinase; PI3K = phosphoinositide 3-kinase; STAT = signal transducer and activator of transcription; TGF = transforming growth factor; TNF = tumor necrosis factor

In patients with HCM, 207 out of the 1,681 proteins were significantly correlated with NYHA functional class. This number is more than twice as large as what would be expected by chance (i.e., 84 proteins at α = 0.05 level). Pathway analysis of these 207 proteins showed upregulation of the Ras-MAPK and related pathways – e.g., PI3K-Akt, Rap1, and TNF (Table 3). Similarly, 264 of the 1,681 proteins were correlated with left atrial diameter in HCM. As shown in Table 4, pathway analysis of the 264 proteins revealed upregulation of the MAPK and the PI3K-Akt pathways, along with pathways involved in inflammation. A total of 320 out of the 1,681 proteins had significant correlation with LV ejection fraction in patients with HCM. Pathway analysis of these proteins displayed that the MAPK and related pathways (e.g., JAK-STAT, TNF), as well as pathways connected to inflammation, were upregulated (Table 5). Regarding the duration of disease, 81 proteins were significantly correlated with univariable analysis – this number is expected by chance at α = 0.05 level. Pathway analysis showed marginal significance with the cytokine-cytokine receptor interaction pathway (FDR = 0.047). The analysis using maximal LV wall thickness did not declare any significantly dysregulated pathways. The additional analysis using dichotomized LV ejection fraction did not show any significant pathway dysregulations, likely due to a small proportion (4%) of patients with LV ejection fraction of <50% and a small number of proteins included in the pathway analysis (80 proteins). The same analyses on NYHA functional class, left atrial diameter, and LV ejection fraction in the control group revealed no significantly dysregulated pathways.

Table 3.

Pathway analysis of proteins that are correlated with New York Heart Association functional class in patients with hypertrophic cardiomyopathy

Pathway description	Number of matching proteins	Number of proteins in the pathway	False discovery rate
PI3K-Akt signaling pathway	20	348	0.0000002
Cytokine-cytokine receptor interaction	17	263	0.0000003
Focal adhesion	13	197	0.00001
ECM-receptor interaction	8	81	0.0001
HIF-1 signaling pathway	8	98	0.0004
Ras signaling pathway	10	228	0.004
Rap1 signaling pathway	9	203	0.007
MAPK signaling pathway	10	293	0.02
Phagosome	7	145	0.02
TNF signaling pathway	6	108	0.02
Complement and coagulation cascades	5	78	0.02
Hematopoietic cell lineage	5	94	0.03

ECM = extracellular matrix; HIF = hypoxia inducible factor; MAPK = mitogen-activated protein kinase; PI3K = phosphoinositide 3-kinase; TNF = tumor necrosis factor

Table 4.

Pathway analysis of proteins that are correlated with left atrial diameter in patients with hypertrophic cardiomyopathy

Pathway description	Number of matching proteins	Number of proteins in the pathway	False discovery rate
Complement and coagulation cascades	20	78	1.5×10−17
Cytokine-cytokine receptor interaction	23	263	2.2×10−10
PI3K-Akt signaling pathway	19	348	0.00002
HIF-1 signaling pathway	9	98	0.0004
Hematopoietic cell lineage	8	94	0.002
Chemokine signaling pathway	10	181	0.005
PPAR signaling pathway	6	72	0.009
MAPK signaling pathway	12	293	0.009
Lysosome	7	123	0.02
Focal adhesion	9	197	0.02
NF-κB signaling pathway	6	93	0.02
Antigen processing and presentation	5	66	0.02
Apoptosis	7	135	0.02

HIF = hypoxia inducible factor; MAPK = mitogen-activated protein kinase; NF-κB = nuclear factor-κB; PI3K = phosphoinositide 3-kinase; PPAR = peroxisome proliferator-activated receptor

Table 5.

Pathway analysis of proteins that are correlated with left ventricular ejection fraction in patients with hypertrophic cardiomyopathy

Pathway description	Number of matching proteins	Number of proteins in the pathway	False discovery rate
Cytokine-cytokine receptor interaction	15	263	0.005
Apoptosis	11	135	0.005
Amino sugar and nucleotide sugar metabolism	6	48	0.01
NF-κB signaling pathway	8	93	0.01
MAPK signaling pathway	14	293	0.01
Metabolic pathways	35	1250	0.03
JAK-STAT signaling pathway	9	160	0.03
TNF signaling pathway	7	108	0.04

JAK = Janus kinase; MAPK = mitogen-activated protein kinase; NF-κB = nuclear factor-κB; STAT = signal transducer and activator of transcription; TNF = tumor necrosis factor

DISCUSSION

Summary of findings

In the present case-control study of 266 cases with HCM and 167 controls with hypertensive LVH, comprehensive proteomics profiling of 1,681 proteins demonstrated a high discriminative ability and displayed previously recognized (e.g., TGF-β) and newly recognized (e.g., Ras-MAPK) pathway upregulation in HCM. Furthermore, cross-sectional analysis in the HCM group showed that upregulation of the Ras-MAPK pathway was significantly correlated with both subjective symptoms (i.e., NYHA functional class) and objective signs (i.e., left atrial diameter, LV ejection fraction) of severe HCM. The current study represents the largest-scale investigation with the most comprehensive proteomics profiling to date exhibiting differential proteomic profile in patients with HCM, especially in those with more severe clinical manifestations.

Results in context

Proteomics profiling has been successfully applied to the discovery of biomarkers in non-cardiac conditions such as Duchenne muscular dystrophy,⁶ Alzheimer’s disease,⁷ influenza,⁸ cancer,⁹ tuberculosis,¹⁰ and arthritis.¹¹ In the field of cardiovascular disease, proteins that predict adverse events have been discovered based on proteomics data in patients with heart failure,^{28, 29} coronary artery disease,^30–32 and hypertension.³³ Proteomics profiling has also been applied to several types of cardiomyopathy – e.g., dilated, muscular dystrophy-associated, and diabetic.^{6, 34, 35} These studies identified biomarkers of disease development and progression, and provided important insights into the pathophysiology of these conditions. These data collectively suggest the potential role of proteomics profiling to identify novel biomarkers in various cardiovascular diseases.

Despite the apparent importance, scarce data are available on the application of proteomics profiling to HCM. A single-center study of 110 patients with HCM and 97 healthy controls using targeted liquid chromatography-tandem/mass spectrometry reported that a Ras-related protein was upregulated in HCM.¹³ However, the number of analyzed proteins was small and pathway analysis was not performed; among 26 candidate proteins that were preselected based on fold change and clinical relevance using a small subset of the cohort (~10 patients per group), only 15 proteins were analyzed because the other proteins were below the detectable concentration limit. Additionally, quantification of protein concentration with enzyme-linked immunosorbent assay was not performed. Our group has conducted a case-control study in which we proteomically profiled 1,129 proteins using the SomaScan assay in 15 patients with HCM and 22 patients with cardiovascular disease other than HCM. This study showed, along with a high discriminative ability (AUC 0.94), that the Ras-MAPK and associated pathways were upregulated in HCM and significantly correlated with NYHA functional class and left atrial diameter. Yet, these 2 single-center studies lacked validation by an independent external cohort and did not clarify whether the observed differences in the proteomic profile were attributable to LVH in general (i.e., present in patients with LVH due to other conditions) or unique to HCM.^{12, 13} In this context, our present study with the largest size and most comprehensive proteomics profiling adds to the body of knowledge by demonstrating that the discrimination model maintains a high accuracy in the independent validation cohort and that the upregulation of the Ras-MAPK pathway is unique to HCM.

The Ras-MAPK pathway and its potential role in the HCM pathogenesis and progression

Our prior and present proteomics studies showed that the Ras-MAPK pathway was upregulated in HCM.¹² Although the pathway analysis was exploratory, this finding is particularly interesting because patients with “RASopathies” – clinical syndromes caused by systemic upregulation of the Ras-MAPK pathway – often exhibit cardiac morphological changes similar to those of HCM in spite of the absence of sarcomeric gene mutations.³⁶ For instance, 95% of individuals with a gain-of-function mutation in the Raf1 gene, a key protein of the Ras-MAPK pathway, display cardiac morphological changes mimicking those of HCM.^{37, 38} There is also emerging evidence on the role of the Ras-MAPK pathway in the pathogenesis of HCM. A small study of 17 HCM cases and 7 controls using endomyocardial biopsy of the right ventricle reported upregulation of the c-H-ras gene in HCM.³⁹ In a rabbit model of HCM, statins (inhibitors of the Ras-MAPK pathway) were shown to reverse morphological changes in the heart.^{40, 41} The present study corroborates these prior reports in humans and animal models, collectively generating the hypothesis that upregulation of the Ras-MAPK pathway contributes to the pathogenesis of HCM.

Among patients with HCM, a larger number of proteins than what would occur by chance had significant correlation with clinical markers of disease severity (i.e., NYHA functional class, left atrial diameter, and LV ejection fraction) in the present study. These findings suggest that the proteomic profile may change as the disease progresses in HCM. Moreover, pathway analysis of the proteins that were correlated with each marker indicated upregulation of the Ras-MAPK pathway. These observations confirm those of our earlier pilot study.¹² Other investigators have also documented that mutations in genes of the Ras-MAPK pathway may be disease-modifying in patients with HCM.^{42, 43} Additionally, our finding that the Ras-MAPK pathway was not upregulated in the control group of patients with hypertensive LVH underscores the uniqueness of this pathway’s associations with the clinical markers of disease severity in HCM. Taken together, the prior and present studies suggest that the Ras-MAPK and associated pathways not only contribute to the pathogenesis of HCM but also play a unique role in the disease progression to more severe forms of HCM.

Strengths of the present study

We implemented a number of strategies to minimize false positive and false negative findings and to augment the internal and external validity of the present study. First, we derived the proteomics-based discrimination model from the training set and validated the discriminative ability in an independent test set with different clinical characteristics (younger with lower prevalence of hypertension and prior septal reduction therapy), thereby overcoming the limitations of the prior studies and enhancing the external validity. Second, to reduce false positive declarations, we have not only applied the strictest regulation (i.e., the Bonferroni method) to select proteins to be included in the pathway analysis but also used the FDR threshold of 0.05 to determine the significance of pathway dysregulation. The use of FDR restricts the study-wide rate of false positive findings. The FDR threshold of 0.05 ensures that <1 out of 20 pathways declared positive are false positive. The use of pathway analysis further strengthens the biological plausibility and lowers the risk of false positive discovery as the proteins are interconnected as opposed to isolated findings using a univariable analysis.²⁵ Third, with regard to false negative findings, our list of differentially regulated proteins and pathways includes those known to be dysregulated in HCM. For instance, brain natriuretic peptide, an established marker of heart failure severity, was found to have higher concentration in patients with HCM than in controls with hypertensive LVH. The TGF-β pathway – known to be upregulated in HCM²⁷ – was found to be upregulated in our analysis. These proteins and pathways function as “positive controls” in the present study and further support the robustness of plasma proteomics profiling to detect signaling pathways that are differentially regulated between HCM and non-HCM populations. Last, the present study included the largest number of patients with HCM with the most comprehensive proteomics profiling,^{12, 13} which buttresses the internal validity and reduces the risk of missing important protein biomarkers and pathways (i.e., false negatives).

Potential Limitations

Our study has several potential limitations. First, patients were enrolled in tertiary care centers, and our findings may not be generalizable to patient populations with less severe clinical manifestations of HCM. Yet, limiting enrollment to 2 centers enabled strict control and standardization of the protocol, which is indispensable for accurate proteomics profiling. Second, the present study did not assess temporality or causality between the differentially regulated pathways and HCM pathogenesis or disease progression. Third, not all patients with HCM underwent genetic testing in the present study. Although we have performed thorough history taking and physical examination with a focus on symptoms and signs associated with HCM phenocopies and made the diagnosis of HCM according to the guidelines, the prevalence of hypertension was relatively high⁴⁴ and the possibility of including an extreme case of hypertensive heart disease in the case group cannot be completely excluded. Fourth, it is possible that secreted proteins may have been preferentially included in the proteomics platform. Fifth, myocardial samples were not available; therefore, metabolic analysis with tissue specimens was not performed. Last, although the sample size was larger than that in the prior studies and we used multiple methods to reduce the chance of false positive discovery as detailed above, the possibility still remains.

Conclusions

This study serves as the first investigation with independent validation to demonstrate the role of proteomics profiling in HCM, revealing novel protein biomarkers and signaling pathways associated with disease status and severity. Our work displayed that multiple pathways – e.g., the Ras-MAPK pathway – were upregulated in patients with HCM, especially in the severe form of the disease. Our study should facilitate further investigations into the underlying molecular mechanisms through which genetic mutation leads to the HCM pathogenesis and progression and the development of targeted therapeutic strategies. The availability of inhibitors specific to the Ras-MAPK pathway further underscores the potential utility of such efforts.^{40, 41}

COMMENTARY.

What is new?

• There is limited prior experience with proteomic analysis in hypertrophic cardiomyopathy (HCM). Therefore, we used a comprehensive, unbiased approach, profiling 1681 proteins, and identified differences in plasma proteomic profiles in patients with HCM (n=266) compared with controls with left ventricular hypertrophy (LVH) due to hypertension (n=167).

• These findings help elucidate highly novel biology that may distinguish HCM from other conditions resulting in LVH, including potential upregulation of the Ras-MAPK pathway.

What are the clinical implications?

• It is sometimes challenging to distinguish HCM from other disease conditions that can cause LVH even with thorough clinical evaluation and genotyping. In the present study, we identified a number of plasma protein biomarkers differentially regulated in HCM. This list of circulating HCM biomarkers included both previously recognized (e.g., brain natriuretic peptide) and unrecognized proteins. The current analysis serves as the first step to specify a panel of plasma protein HCM biomarkers that are readily available without invasive procedures (e.g., endomyocardial biopsy). In the future, such circulating HCM biomarkers may aid in making the diagnosis in ambiguous cases.

Non-standard Abbreviations and Acronyms.

AUC

area under the receiver-operating-characteristic curve

CI

confidence interval

FDR

false discovery rate

HCM

hypertrophic cardiomyopathy

ICD-9-CM

International Classification of Diseases, Ninth Revision, Clinical Modification

LV

left ventricular

LVH

left ventricular hypertrophy

NYHA

New York Heart Association

sPLS-DA

sparse partial least squares discriminant analysis