In‐depth proteomics approach reveals novel biomarkers of cardiac remodelling after myocardial infarction: An exploratory analysis

Abstract Cardiac remodelling following myocardial infarction (MI) is a maladaptive change associated with progressive heart failure and compromises long‐term clinical outcome. A substantial proportion of patients afflicted by MI still develop adverse outcomes associated with cardiac remodelling. Therefore, it is crucial to identify biomarkers for the early prediction of cardiac remodelling. An in‐depth proteomics approach, including both semi‐quantitative and quantitative antibody arrays, was used to identify circulating biomarkers that may be associated with detrimental cardiac remodelling. Furthermore, statistical correlation analysis was performed between the candidate biomarkers and clinical cardiac remodelling data to demonstrate their clinical utility. A systematic proteomics approach revealed that sclerostin (SOST), growth differentiation factor‐15 (GDF‐15), urokinase‐type plasminogen activator (uPA), and midkine (MK) were increased, while monocyte chemotactic protein‐3 (MCP‐3) was uniquely decreased in MI patients who developed cardiac remodelling, compared to MI patients who did not develop cardiac remodelling and healthy humen. Moreover, correlation analyses between serum proteomes and cardiac remodelling echocardiographic parameters demonstrated a moderate positive association between left ventricular end‐diastolic volume index (LVEDVi) and the three serum proteins, uPA, MK and GDF‐15 (P < .05, respectively), and a moderate negative correlation between LV ejection fraction (LVEF) and these serum proteins (P < .05, respectively). Importantly, uPA and MK were firstly identified to be associated with the development of cardiac remodelling. The present study contributes to a better understanding of the various cytokines expressed during adverse cardiac remodelling. The identified biomarkers may facilitate early identification of patients at high risk of ischaemic heart failure pending further confirmation through larger clinical trials.


| INTRODUC TI ON
Cardiac remodelling after acute myocardial infarction (MI) is characterized by progressive expansion of infarcted myocardium and disproportionate alterations in the geometry and function of the ventricular chamber. 1 The pathophysiological process of cardiac remodelling starts immediately after an acute MI and, if not attenuated or reversed through effective interventions, results in poor clinical prognosis, including ventricular arrhythmias, heart failure and subsequent cardiovascular mortality. 2,3 Moreover, once cardiac remodelling manifests with clinical heart failure symptoms, standard pharmacotherapies including angiotensin receptor-neprilysin inhibitor, natriuretic peptides, angiotensin-converting enzyme inhibitors, or β-blockers have limited effect on reversing the remodelling and/ or improving the clinical symptoms. 4 Cardiovascular magnetic resonance imaging is the gold standard for diagnosing cardiac remodelling despite being cumbersome and complex, and leading to increased risk of adverse cardiovascular events during the examination. 5 Hence, this technique is sparingly used in clinical setting. Currently, transthoracic echocardiography has been widely available to assess the segmental and overall cardiac structure and function. However, both cardiac imaging measurements do not evaluate the complex process of biochemical and neuro-endocrine changes which are at the core of the vicious cycle termed "infarct remodelling". 6 So far, N-terminal pro-brain natriuretic peptide (NT-proBNP) is established as useful non-invasive indicator of cardiac remodelling and dysfunction. [7][8][9] Elevated cardiac troponins are demonstrated as markers of increased risk for progressive left ventricular (LV) dysfunction in chronic heart failure. 10,11 Soluble suppression of tumorigenicity 2 is a predictor of reverse left ventricular remodelling. 12,13 Nevertheless, there are no early and effective means to predict the occurrence of remodelling following MI. Therefore, the development of novel and potentially more accurate approaches to predict cardiac remodelling after MI is urgently needed.
Novel biomarkers are crucial for understanding the pathophysiology, early diagnosis and prognosis of a disease, as well as for developing effective personalized pharmacological management therapies. There are several promising tools to identify such biomarkers, including clinical proteomics, an approach dedicated to the global study of secreted proteins from organs throughout the body, which may contain biomarkers for prognostication and treatment response. 14 Currently, there are two main strategies for analysing plasma/serum proteome, including microarray assays and mass spectrometry. Regrettable, mass spectrometry suffers from high false-positive rates and poor sensitivity. 15 Samples need to be treated for the remove of high-abundant proteins when detected by mass spectrometry, which would cause the depletion of low-abundant proteins such as inflammatory, angiogenic cytokines, chemokines, growth factors and so on playing a very important role in the pathophysiology of diseases. Some studies showed that mass spectrometry is valuable at the detection of lipid proteins and enzymes, whereas antibody array could detect cytokines, chemokines and receptors. 16,17 In particular, the antibody microarrays are recognized as an innovative and preferential technology for the identification of circulating cytokine biomarkers, with the advantages of being high-throughput amenable and highly sensitive for phenotype assessment. 18 In the present study, the antibody microarray technology was used to identify the serum proteome of MI patients with cardiac remodelling development. Furthermore, statistical correlation analysis was preformed between candidate cytokines and clinical cardiac remodelling data to validate the biomarkers. The comprehensive analysis of the proteome may help to better understand the pathogenesis of cardiac remodelling as well as to identify more novel serum biomarkers for early and more accurate prediction of this pathological process.

| Study design and participants
The study was conducted in accordance with the Declaration of

| Echocardiographic measurements
All echocardiographic measurements were performed by an independent cardiac sonographer who was blinded to the allocation. A commercially available ultrasound system (Philips Medical Systems) was used to obtain and store the images for subsequent off-line analysis. Standard two-dimensional short-axis view of the LV was obtained at the level of the papillary muscle to record M-mode tracings. LV end-diastolic/end-systolic internal diameter and LV anterior wall thickness were measured. Biplane end-diastolic and systolic volumes are calculated according to the modified Simpson's rule. All measurements were obtained and averaged over three consecutive cardiac cycles.

| Semi-quantitative antibody array for prescreening
A Human Cytokine Antibody Array (GSH-CAA-440, RayBiotech Company) that simultaneously and semi-quantitatively detects 440 cytokines in a single experiment was used according to the manufacturer's instructions. 3 Briefly, after a dilution with blocking buffer (1:1), serum samples (Table 1) were incubated with 440 capture antibodies previously coated onto glass slides overnight.
After washing, a biotin-conjugated anti-cytokine antibody mix was added into the arrays and further incubated for 2 hours.
Finally, Cy3-conjugated streptavidin was used to detect the intensities of the arrays which were subsequently exposed using an InnoScan 300 Microarray Scanner (Innopsys). The signal values of the 440 cytokines were normalized using a factor (the positive control values from certain array/the positive control values F I G U R E 1 Schematic of the screening strategy. Identification of serum proteomic biomarkers for prediction of cardiac remodelling following myocardial infarction. CRAMI group, cardiac remodelling after myocardial infarction; MI group, myocardial infarction without cardiac remodelling

| Quantitative antibody array for validation
To validate the differential cytokines screened from the high-throughput array GSH-CAA-440, a custom Human Cytokine Antibody Array (QAH-CUST-39, RayBiotech Company) that can simultaneously and quantitatively detect thirty-nine cytokines was used to detect more serum samples (Table 2). Briefly, the standards (gradient dilution) and serum samples (diluted two times) were added into the array pools for overnight incubation. Then, a biotin-conjugated anti-cytokine antibody mix and Cy3-conjugated streptavidin were successively added into the array pools followed by 2 hours of incubation. The slides were scanned using an InnoScan 300 Microarray Scanner (Innopsys).
Finally, the concentration of the thirty-nine cytokines in the serum was calculated using their signal values and the standard curve.

| Statistical analysis
Comparisons between groups were performed by one-way ANOVA followed by multiple comparisons performed with post hoc Bonferroni test using SPSS v20 (IBM Corp.). The relationship between serum proteome and cardiac remodelling echocardiographic data was calculated using bivariate correlation analysis. Differences were considered statistically significant when P values were < .05. All data are shown as mean ± SD (standard deviation). In addition, fold change (FC) between groups was calculated to indicate the relative levels of the cytokines.

| Differential proteins analysis
To identify the specific proteins involved in cardiac remodelling following MI, the antibody array data of the three groups were analysed by oneway ANOVA. The Bonferroni post hoc analysis showed that forty-one

TA B L E 2 Participant characteristics for validation
F I G U R E 2 Proteins screened by semi-quantitative antibody array. A, Venn diagram analysis of proteins differentially expressed between any two groups from the semi-quantitative antibody array pre-screen identified eight proteins simultaneously differential between CRAMI and MI group and between CRAMI and control group, but not differential between MI and control group, which are defined as specific biomarkers of cardiac remodelling after MI. B, The levels of these eight specific proteins of cardiac remodelling after MI are shown by histogram analysis using their fluorescence signal values from the semi-quantitative antibody array from the three groups. * means P < .05 when compared to CRAMI group. CRAMI group, cardiac remodelling after myocardial infarction; MI group, myocardial infarction without cardiac remodelling cytokines were differentially expressed between the CRAMI and control group, eighteen cytokines were differentially expressed between the CRAMI and MI group, and seventeen cytokines were differentially expressed between the MI and control groups with P values < .05 (Table S1).

| Analysis of specific biomarkers of cardiac remodelling after MI
As per definition, the specific biomarkers of cardiac remodelling following MI would be the proteins that show differential expression between CRAMI and MI group as well as between CRAMI and control group, but not between MI and control group. Therefore, a Venn diagram analysis was performed and eight specific biomarkers of cardiac remodelling following MI were identified ( Figure 2A). As shown in Table 3 Figure 2B). An unsupervised-hierarchical clustering using the signal values of these eight proteins could distinguish CRAMI group from the other two groups with 100% accuracy (Figure 3).

TA B L E 3
Specific biomarkers of cardiac remodelling after myocardial infarction from semi-quantitative antibody array assay

F I G U R E 4
Proteins validated by quantitative antibody array. A, Results of the quantitative antibody array were validated using Venn diagram analysis, and five proteins simultaneously differential between CRAMI and MI group and between CRAMI and control group, but not differential between MI and control group were identified. B, The concentration of the five specific proteins of cardiac remodelling after MI validated by the quantitative antibody array is shown using a scatter diagram analysis. * means P < .05 when compared to CRAMI group. CRAMI group, cardiac remodelling after myocardial infarction; MI group, myocardial infarction without cardiac remodelling

F I G U R E 5
The antibody array profiles of the five specific proteins of cardiac remodelling after MI. The custom quantitative antibody array for validation was prepared in two arrays. The locations of the five proteins in the arrays are noted in coloured boxes, and the levels of these proteins are proportional to their fluorescence intensity. Corresponding antibodies of thirty-nine proteins were printed in duplicate in these arrays. CRAMI group, cardiac remodelling after myocardial infarction; MI group, myocardial infarction without cardiac remodelling

| Validation of specific biomarkers
To further validate these specific biomarkers, a quantitative antibody array was used and thirty-nine proteins (including the eight specific proteins and thirty-one other cytokines) only differential in the CRAMI vs MI group, or in CRAMI vs control group, and not in MI vs control group were detected again with more samples (Table 2). After the hoc Bonferroni analysis of one-way ANOVA and Venn diagram analysis ( Figure 4A), SOST, GDF-15, uPA, MCP-3 and MK were found to be simultaneously and significantly differentially expressed between CRAMI and MI group (P < .05), and between CRAMI and control group (P < .05), but not differential between MI and control group (P > .05).
Further, they showed the same tendencies as those in the semiquantitative antibody array assay between the groups (Table 4 and Figure 4B). This suggests that these five cytokines may be robust and specific biomarkers of cardiac remodelling after MI. Additionally, their array profiles showed markedly different fluorescent intensities in the CRAMI group compared to that in the MI and control group, which were positively relevant to their expression levels, further confirming that these five proteins were uniquely differential in CRAMI ( Figure 5).

| Correlation analysis between serum proteome and cardiac remodelling
To further test the clinical utility of the five potential biomarkers

| D ISCUSS I ON
The identification of novel biomarkers for the prediction of cardiac remodelling is essential to develop timely reparative strategies against this adverse pathologic process. 21 In the present exploratory case-control study, to detect specific cytokines associated with the development of cardiac remodelling, serum samples were collected from fresh MI patients who were then followed up for uPA is a plasmin activating pro-matrix metalloproteinases in the cardiovascular system. 26  As many other novel efforts, our exploratory trial might also have certain limitations, due to the fact that the reported results have been generated only in a single-centre and involved relatively small representative population. Therefore, a future more comprehensive work is needed to confirm our exploratory, but promising data, in a multi-centre clinical trial with a larger number of participants. Also, the choice of proteins selected for microarrays assessment might not reflect all possible biomarkers that would be used as early predictors of cardiac remodelling after MI. Therefore, we cannot exclude the possibility that a number of other cytokines (not included in our panel) also participated in the process. In addition, we explored only a single time point for validation of the selected cytokine profile.
Accordingly, yet there is a possibility that other profile of original biomarkers would also be detected at other time points in a further research.

| CON CLUS ION
In this study, an in-depth proteomics approach was used to identify three specific biomarkers (GDF-15, uPA and MK) as novel serum biomarkers for the early prediction of cardiac remodelling development after MI. More importantly, uPA and MK are reported for the first time for their critical roles in predicting adverse cardiac remodelling following MI. Regardless of the study limitations, including the small cohort size, the present results are of clinical significance and clearly support the validity of the approach. As the present work was a pilot study, the findings require further confirmation in a larger trial which would also appraise the value of these novel circulating biomarkers as therapeutic response in routine clinical practice.

ACK N OWLED G EM ENTS
The authors gratefully acknowledge the contributions of all staff for their participation in the study.

CO N FLI C T O F I NTE R E S T
The authors declare that they have no competing interests.

DATA AVA I L A B I L I T Y S TAT E M E N T
All data generated or analysed during this study are included in this article.