Risk prediction model construction for heart failure after myocardial infarction by immune B cells in blood

doi:10.21203/rs.3.rs-2476802/v1

Download PDF

Research Article

Risk prediction model construction for heart failure after myocardial infarction by immune B cells in blood

https://doi.org/10.21203/rs.3.rs-2476802/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

With a high prevalence of morbidity and mortality, myocardial infarction (MI) is a prevalent heart disease. The development and outcomes of post-MI heart failure (HF) continue to be a major factor in the poor post-MI prognosis despite the extensive medical treatment for MI. There are currently few indicators that can accurately predict post-MI heart failure. In this study, we re-examined single cell RNA-seq and bulk RNA-seq datasets collected from peripheral blood samples of myocardial infarction and dataset of patients who developed heart failure or heart failure did not occur after myocardial infarction. We discovered a subtype of immune-activation B cell which is associated with the discrimination of post-MI HF and nonHF patients. Such finding was further validated in independent corhorts using PCR. By incorporating the specific marker genes of the B subtype, we developed a prediction model with 13 markers that can predict the HF risk of myocardial infarction patients and provide useful clinical experience and tools.

myocardial infarction (MI)

heart failure (HF)

B cells

risk prediction model

single cell RNA-seq

Myocardial infarction (MI), the result of rupture or erosion of a vulnerable, lipid-filled, chronic atherosclerotic coronary plaque leading to acute disruption of myocardial blood flow and ischemic myocardial necrosis, remains a very common cardiac emergency with high morbidity and mortality worldwide [1, 2]. Despite significant advances in the treatment of coronary artery disease and acute myocardial infarction (AMI) over the past two decades, heart failure (HF) due to MI and the development of post-MI heart failure remains one of the most common causes of cardiovascular death [3, 4]. In acute coronary syndromes and myocardial infarction, acute inflammation caused by plaque rupture triggers a subsequent response, which may lead to more long-term cardiac injury. There are differences in the degree of inflammatory recovery after coronary interventions for stenting or thrombolytic reperfusion that have prognostic value for the development of ischemic heart failure [5].

Numerous studies analyzed circulating biomarkers in the blood of post-MI patients, which reflected different inflammatory traits, coagulation activity, endothelial dysfunction, atherogenesis, myocardial dysfunction and injury, apoptosis, renal function, glucose and lipid metabolism, to explore which factors can predict poor outcome after MI [6-8]. Although several genes, such as natriuretic peptides (NPs), galactose lectin-3 and soluble tumorigenic suppressor-2 (sST2) have been reported as biomarkers for HF, their reliability remains controversial [9, 10]. The above studies indicates that elevated levels of circulating pro-inflammatory biomarkers in patients with HF correlate with disease severity and prognosis. Experimental studies have shown that activation of cardiac immune response mechanisms triggers adverse cardiac remodeling and leads to left ventricular dysfunction [11]. Therefore, the development of novel powerful immune-related biomarkers with predictive potential to screen for the incidence of ischemic pretreatment and MI and the development of HF after MI remains a key goal of scientific advances in cardiovascular disease.

It has been demonstrated that inflammatory cells in the peripheral blood of patients with MI may play a significant role [12]. After the initiation of MI, myeloid cells and T lymphocytes in the blood displayed distinct disparities in abundance [13, 14]. Following MI, blood classical and non-classical monocytes make up the majority of the inflammatory cell types, with NK and B cells being crucial in the activation of inflammatory cells and the promotion of chemokine synthesis in plaque rupture [15]. Notably, this distinction between immune cells of the patients that experienced HF after myocardial infarction and those that did not was likewise very significant [16]. There are still no biomarkers of immune-inflammation that can be used to clinically predict the risk of post-MI HF.

In recent years, researchers have advanced the understanding of micro-environmental alterations in tissues with different etiologies of HF through scRNA-seq analysis and other experimental approaches [17-19]. However, most research have stopped at the analysis of altered cell types and pathological pathways within cardiac tissue, and few studies today have explored the prognosis of cardiac disease through alterations in immune cells in the blood. In this study, we reanalyzed published blood single-cell and bulk RNA-seq data from heart failure of post-MI to construct a prediction model. We found that a group of immune-activation B cells in blood may play an important role in the disease progress. Based on these findings, we developed an effective post-MI HF risk prediction signature with 13 genes. Through further analysis, the five genes with the most diagnostic value, STING1, HSPB1, CCL5, ACTN1, and ITGB2, were identified, which were further validated by PCR. These findings extend our understanding of the factors associated with the development of heart failure after MI and provide a potentially excellent predictive model for the development of heart failure after MI.

B cells, NK cells and monocytes are associated with the myocardial infarction

We re-analyzed a publicly available heart failure scRNA-seq datasets, from the blood of post-MI HF patients and normal human controls. Multiple sets of MI bulk RNA-seq datasets and one set of post-MI HF bulk RNA-seq dataset were used for validation. The work flow of this study is summarized in Fig. 1.

There were 6 cell types from the scRNA-seq dataset, CD4⁺T, CD8⁺T, NK, B, haematopoietic stem cells (HSC), and monocytes. Fig.S1 shows the results of visualization of the 6 immune cell types in two dimensions by Uniform Flowform Approximation and Projection (UMAP).

According to the workflow in Fig. 1, we used the scCODE on each individual cell type to identify the differentially expressed (DE) genes by comparing the post-MI HF and normal groups and obtained a total 5 DE gene lists (due to the low number of HSC cells, this group of cells were excluded from subsequent analyses). We only used the DE genes upregulated in the disease group. From the MSigDB, 5 DE genes lists of the disease group were discovered to be enriched in 200 gene sets (Fig.S3). The calculation of these 200 gene sets is described in detail in the study design.

In comparison to Hallmark, KEGG, and Reactome, we discovered that the GOBP gene collection had greater overall prediction capability (Fig.S3). We explored the enriched GOBP gene sets (gene enrichment) with ROC p-values < 0.05 in each DE gene list to find cell types with strong prediction potential. A DE gene list was considered significant if no less than half of the enriched gene sets had ROC FDR < 0.05. As a result, we screened 4 relevant cells on the GSE60993 bulk dataset, which are NK cells, CD8T cells, monocytes and B cells. We screened 3 relevant cells on the GSE61144 bulk dataset, which are NK cells, monocytes and B cells. Combining the results of the analysis of the above two datasets, we screened three cell types from them that may be associated with MI. They are NK cells, monocytes and B cells respectively (Fig. 2).

B cell are associated with post-MI heart failure

We performed subcluster analysis of the 3 cells mentioned above individually. First, B cells was divided into five subclusters by Seurat (Fig. 3a), and Fig. 3b showed a scatter plot of the distribution of cells between the disease and normal groups. Following the discovery of cluster-specific marker genes by FindAllMarkers (Seurat), Fig. 3c presents the expression heatmap for the top 10 marker genes. The classical marker genes of B cells were highly expressed in all subclusters (Fig.S4a).

There were 7 gene sets of DE gene of B cells with ROC p.adjust < 0.05 (Fig. 2).We used gene set variance analysis (GSVA) to evaluate the expression of these gene sets in different B cell subcluster and discovered that they were substantially expressed in subcluster 4 (Fig. 3d, Fig.S4b). Gene enrichment analysis of its marker genes suggests B cell associated markers enriched in graft rejection, immune interactions between lymphocytes and non-lymphoid cells, and activation of the NK cell cycle (Fig. 3e). The violin plot lists genes specifically expressed in the C4 cluster for additional investigation of the B cells subcluster (Fig. 3f). These genes specifically expressed in the C4 subcluster of B cells are associated with the activation and development of immune cells.

First, we performed gene enrichment analysis of marker genes in the C4 cluster of B cells and found that these marker genes in the C4 cluster were significantly predictive of the development of heart failure after MI (p.adjust < 0.05) (Fig.S4d). The gene sets composed by these B cell C4 cluster markers and we obtained a total of 115 genes from these gene sets. ROC curves for the 115 genes demonstrated a significant predictive power of the gene signature for the development of HF after MI, AUC = 0.9 (95% confidence interval: 0.86–0.94. p = 0.0034) (Fig. 4e). We also demonstrated that this gene signature (HF.Sig) was also a good predictor of MI (Fig.S4e, Fig.S4f). We performed gene enrichment analysis using the genes of HF.Sig. Gene enrichment showed that the first 20 pathways were mainly immune response-related pathways such as immune cell activation [20, 21] (Fig.S8a). We did ROC curve validation for the first 10 pathways and demonstrated good predictive performance for these pathways with AUCs between 0.89 and 0.78. (Fig.S8b). The GSVA scores of HF.Sig were significantly higher than those of other subclusters of B cells (Fig. 4a), Fig. 4b shows the GSVA scores of HF.Sig in different B-cell C4 subclusters in the disease and normal groups in different cells. These genes were enriched in HF patients (Fig. 4c) and the genes also had significantly higher GSVA scores in HF (Fig. 4d).

Post-MI heart failure prediction model construction and validation

To construct a more effective predictive model for post-MI HF, we executed a loop algorithm (Fig.S7) and calculated AUCs with different numbers of genes. The peak of AUC appeared in a combination of 13 genes (CCL5, B2M, CD3D, CD6, CD79B, DGKZ, STING1, LYAR, KLRG1, PRKCA, HSPB1, ACTN1, MYL12A) (Fig. 5a). The MI_HF.Sig consisting of new 13 gene can accurately distinguish between HF and nonHF in the GSE59867 cohort, with an AUC of 1 (p = 7.2e-07, 95% confidence interval (CI): 1–1) (Fig. 5b).

GSVA scoring of the above 13 gene signatures in GSE59867 (n.patients:HF = 8, nonHF = 9, total of 32 HF samples, 36 non-HF samples) showed that 8 of these genes were upregulated in HF relative to the nonHF sample for expression (Fig. 5c, Fig.S6d) and found 4 genes, HSPB1, ACTN1, STING1, CCL5, were significant (Fig. 5c). Fig.S6c shows the difference expression of 8 genes in GSE59867, which had similar trend. In addition, we took the overlap between all DE genes from dataset GSE59867 and 115 genes from B_C4 maker genes for the analysis, as shown in Fig. 5d. We obtained a total of 84 genes. And we used these 84 genes for protein interaction network analysis in the String online tool site. Among them, the upregulated gene with p < 0.05 and multiple linker proteins greater than 10 was ITGB2, which may play an important role in HF patients after MI (Fig. 5d).

Using the above algorithm, we demonstrated that the C4 subcluster of B cells can be used as a cell subcluster to predict the occurrence of HF after MI. However, monocytes and NK cells failed to accurately predict whether heart failure will occur after MI. Monocytes were divided into 5 subclusters (Fig.S5a). But monocytes were not able to localize to an exact subcluster that can make an accurate prediction of MI (Fig.S5d). NK cells were divided into 3 subclusters (Fig.S6a), Fig.S6b showed a scatter plot of the distribution of cells between the disease and normal groups. Fig.S6c presents the expression heatmap for the top 10 marker genes. We analyzed the expression of NK gene sets in the MI dataset by GSVA and found them to be highly expressed in NK C1 (Fig.S6d). However, these gene sets failed to accurately reflect the enrichment in HF caused by post-MI heart failure (p = 0.2933, p.adjust = 0.2933) (Fig.S6f). We performed GSVA analysis of gene sets in cells of the C1 of NK cells (Fig.S6h) (Wilcoxon rank sum test p = 0.92707). The GSVA scores for these gene sets failed to meet the requirement to distinguish between the two group of samples.

Expression of CCL5, STING, HSPB1, ACTN1 and ITGB2 genes in blood samples from post-MI HF patients and post-MI nonHF patients

We performed a qRT-PCR analysis of the mRNA expression of CCL5, STING, HSPB1, ACTN1 and ITGB2 in post-MI HF patients and post-MI nonHF patients. As expected, the expression of CCL5, ACTN1 and ITGB2 was significantly increased in blood samples from post-MI HF patients and post-MI nonHF patients (p < 0.05) (Fig. 6a). Although the p-values of STING and HSPB1 were greater than 0.05 after Wilcoxon rank-sum test, but they showed a significantly higher trend in the HF group than in the nonHF group when comparing patients with post-MI HF and post-MI HF (Fig. 6b).

scRNA-seq and bulk RNA-seq datasets collection

To investigate immunocyte subtypes and characteristics that are highly predictive of heart failure after MI, we obtained and reexamined a heart failure after MI scRNA-seq dataset that was previously published in the article [17],validated the results in multiple MI bulk RNA-seq datasets and a HF after MI bulk RNA-seq [12, 16]. The single-cell dataset is blood samples from patients with HF after MI and normal subjects for single-cell sequencing analysis. The processing, clustering, and cell type determination procedures of the scRNA-seq datasets, as well as the extensive clinicopathological data for every patient in the dataset, were thoroughly discussed in their original studies.

We used two bulk MI datasets (GSE61144, GSE60993), one comprises blood from 14 MI and 10 normal, and another dataset contains blood from 17 MI and 7 normal. The samples containing post-MI HF and post-MI nonHF in GSE59867 were then used for further screening, including 8 patients with post-MI HF and 9 post-MI nonHF patients. The blood samples from each patient in this dataset was selected for analysis at four time points: admission, discharge, 1 month post-MI and 6 months post-MI. The Hallmark, KEGG, GOBP, and Reactome gene sets were retrieved for GSEA from MSigDB (http://www.gsea-msigdb.org/gsea/index.jsp), which was the database used in our investigation.

Study design

We extracted a total of 6 cell types from GSE145154 scRNA-seq datasets. Due to the small number of haematopoietic stem cells (HSC), we will only target the remaining five immune cells in the subsequent analysis. We then used scCODE v1.0.0.0 R package provided by Zou et al [22] to determine the DE genes between normal and post-MI HF patients for each of the aforementioned 5 cell types. The reliability of single cell DE analysis is improved by scCODE's ability to examine the chosen DE genes using a range of testing method. Thus, we obtained a total of 5 DE gene lists. We then used the Investigate Gene Sets tool (http://www.gsea-msigdb.org/gsea/msigdb/annotate.jsp) to compute enriched gene sets between our gene lists and gene sets in MSigDB. For each of 5 DE gene lists identified by scCODE, we sorted them according to the absolute value of logFC of each DE gene from large to small and submitted the top 500 genes of each DE gene list to Investigate Gene Sets tool (all genes were submitted if less than 500). Each DE gene list can obtain several gene sets enriched in GOBP, Hallmark, KEGG, and Reactome using the tool's default parameters (display top 10 genes & FDR q-value less than 0.05). Using the Cancerclass v1.34.0 R program, the prediction power of these enriched gene sets was evaluated. Using the Cancerclass R program, we evaluated the following gene sets' capacity to predict the post-MI outcomes of the GSE59867 (n.patients:HF = 8, nonHF = 9, total of 32 HF samples, 36 non-HF samples) [23]. In terms of the ROC curve and matching AUC, the prediction sensitivity and specificity were evaluated. Cancerclass R package was used to create and validate high-dimensional molecular data categorization tests. Feature selection and nearest centroid classification were performed in sequence. In order to get continuous prediction scores, the multiple random validation methodology was used to carefully verify the categorization results. The success of the classifier's classification results is indicated by the p value of the ROC curve, which was obtained using the Welch's t test included in the Cancerclass R package. Each gene set was evaluated independently as a classifier.

Human research and ethics

Blood specimens for the validation experiment were obtained from patients admitted to the hospital with myocardial infarction, which included patients with and without heart failure. This study was approved by the Research Ethics Committee of Qilu Hospital (reference number: KYLL-2021(KS)-393), in compliance with the Declaration of Helsinki, the Code of Ethics of the World Medical Association. Written consent to participate in the study was obtained from all patients.

Single-cell RNA sequencing data processing

We extracted and defined cell types from post-MI HF single-cell dataset GSE145154 [17]. We used the Seurat v4.0.4 R package [24] to identify cell types and subtypes. The extracted dataset was first combined across samples. Prior to this, the expression of each gene was scaled to a scale factor of 10,000 and then converted logarithmically (Seurat's default method), normalized to the total expression of the associated cell. FindVariableFeatures was used to find the top 2000 variable features, which were then used in further research. Mutual Nearest Neighbor (MNN) approach was then used to correct batch effects [25], and the percentage of mitochondrial transcripts was regressed using ScaleData. The integrated assay was used only for dimension reduction and clustering, and the raw log-normalized expression data were used for all DE and gene level analyses. Then, principal component analysis (PCA) was performed on the integrated assay by using RunPCA. Cell clustering was performed using the first 20 main components, with a resolution parameter of 0.5. Finally, the visualization is carried out in two-dimensional space through unified manifold approximation and projection (UMAP) (Dims = 1:20).

Marker gene analysis

Marker gene analysis was performed for all clusters using FindAllMarkers included into the Seurat package, with the parameters min.pct = 0.1 and logfc.threshold = 0.25. Genes with p.adjust < 0.05 were selected as cluster-specific marker genes.

Gene enrichment and GSVA analysis

We used the Gene Set Variation Analysis (GSVA) method with default settings to assign specific gene signature activity scores to individual cells or samples, as implemented in the GSVA (v1.38.2) R package [26]. Using the FindMarkers built into the Seurat package, we compared all gene expression fold changes in disease and normal in C4 subclusters of B cells to obtain the list of DE genes for gene enrichment analysis by setting the parameter logfc.threshold = -Inf, min.pct = -Inf, min.diff.pct = -Inf. Gene enrichment was performed on DE gene lists based on predownloaded Hallmark, KEGG, GOBP, and Reactome gene sets using the default parameters of clusterProfiler v3.18.1 R package. This R package can also be used to examine whether a particular gene set is enriched at the top or bottom of the preordered gene list. The Benjamini-Hochberg method was used to compare two sets, and gene sets with FDR-adjusted p-values < 0.05 were deemed significantly enriched in one set.

Development of a gene signature associated with heart failure after MI

Using the Cancerclass v1.34.0 R package, we created gene signatures based on the cluster-specific genes of the C4 of B cell subtype [27]. Specifically, using the Cancerclass, the p-values of the ROC curves for the enriched GOBP gene sets of B cell were generated. They were adjusted for FDR using the Benjamini-Hochberg method. We combined all gene sets with p.adjust < 0.05 to create a total of 115 genes (The selected gene signature is referred to as HF.Sig in this study). Next, the cyclic algorithm was shown in Fig.S2. In each cycle, we chose 114 gene combinations from the signature at random. The Cancerclass was then used to estimate the AUC in the GSE59867 cohort and examine the prediction accuracy of each of these combinations. Out of these 115 combinations, the gene combinations with the highest AUC (genelist_maxAUC) were kept and applied to the following cycle. The cycle was repeated until there were no more than three distinct gene combinations remaining. Finally, the highest AUC of the gene panels in all loops were selected. A combination with appropriate gene numbers was selected as new HF.Sig for subsequent analysis.

Heatmap of differentially expressed genes and PPI

The generalized linear model approach of the Limma v3.46.0.[28] package (https://doi.org/doi:10.18129/B9.bioc.limma) was used to calculate gene expression fold changes between post-MI HF patients and nonHF patients. In order to plot the protein interaction nodes in accordance with the PPI network map, we first took the overlapped genes from MI-related cell clusters marker from a single cell data set genes and differential genes from the GSE59867 bulk dataset, uploaded them to the STRING database website (https://cn.string-db.org), set a medium confidence interval (medium confidence = 0.4), exported the TSV files, and then the circlize package was used for drawing.

Real-Time Polymerase Chain Reaction Analysis

Relative levels of target genes were measured using real-time polymerase chain reaction (qPCR). RNA was extracted from mononuclear cells of human blood using the RNeasy Fibrous Tissue Mini Kit (QIAGEN, Hilden, Germany) according to the manufacturer's instructions, and cDNA was synthesized using the HighCapacity cDNA Reverse Transcription Kit (Thermo Fisher Scientific, Waltham, MA). fluorescence quantification of target genes was performed using the ABI 7500 Fast Real-Time PCR System (Applied Biosystems, Foster City, CA) in a 20 µl reaction mixture consisting of 1 µl of cDNA, 2 µl of forward and reverse primers, Fast SYBR Green Master Mix 10 µl (Thermo Fisher Scientific) and 8 µl of nuclease-free water. qRT-PCR was performed with the following primer sequences. ITGB2: F 5'-ATGCTGGGCCTGCGCCC-3', R 5'- GATGGTGTCACACTCGCAGTA − 3'; STING1, F: 5'- ATGCCCCACTCCAGCCTG − 3', R: 5'- TCAAGCTGCCCACAGTAACCT − 3'. ACTN1, F: 5'-ATGGACCATTATGATTCTCAGCAAA − 3', R: 5'-TTAGAGGTCACTCTCGCCGTA − 3'. CCL5, F: 5'- ATGAAGGTCTCCGCGGCAG-3', R: 5'- TCAAGGAGCGGGTGGGGTA-3'. HSPB1: F: 5'- AGGAGTGGTCGCAGTGGTTAGG-3', R: 5'- CAGGGGACAGGGAGGAGGAAAC-3'. GAPDH: F:5'-ATCCCATCACCATCTTCC-3', and R, 5'-GAGTCCTTCCACGATACCA-3'.

Statistical analysis

Each classifier in this study was evaluated for its ability to predict post-MI HF outcomes by plotting the ROC, calculating the AUC, estimating the sensitivity and specificity implemented in the Cancerclass. The validity of each classifier was evaluated by the ROC curve's p-value. Using the Wilson’s method included in the Cancerclass, the 95 percent confidence intervals for sensitivity and specificity were computed, and p-values were determined using the Welch's t-test. All p-values in this research, unless otherwise noted, were corrected using the Benjamini-Hochberg method, and corrected p-values below 0.05 were regarded as statistically significant. Wilcoxon's test was used to classify the variables. All confidence intervals were reported as binomial 95% confidence intervals (CIs). All statistical analyses for this study were performed by R v3.5.3.

We reanalyzed one publicly available HF scRNA-seq dataset, to explore the subtypes and characteristics of immune cells in the blood after MI that predict the riskness of heart failure. We found that a B cell subtype (B_C4) is important for predicting whether heart failure will occur after MI. We utilized genetic markers associated with B_C4 and developed a signature HF.Sig consisted of 13 genes. Through gene enrichment analysis, we found that the B_C4 is mainly associated with the activation of NK immune cells and interaction between immune cells. The predictive ability of HF.Sig was systematically evaluated. A post-MI HF prediction model with HF.Sig was obtained. By analysis of GSVA and differential gene expression, we further screened 5 genes (STING1, HSPB1, CCL5, ACTN1, and ITGB2) that were highly expressed in post-MI HF patients. And through the verification of blood samples from clinical patients, the validity of the (Significant differences in the HF and nonHF groups) of the genes we found was examined. We developed a prediction model for prognosis in post-MI patients by combining analysis of scRNA-seq and bulk RNA-seq. Given the ease of blood testing in the clinic, our HF signature provides a powerful diagnostic tool for practical application in the clinic, for whether HF will occur after MI.

Immuno-inflammatory pathways have a significant impact on how patients recover from HF and MI [29]. B lymphocytes infiltrate the infarcted myocardium starting on day 1 and continue doing so until the healing period, according to earlier research utilizing animal models of MI (day 7) [30]. Reduced post-MI inflammation and improved functional cardiac outcomes in mice have been linked to B cell depletion using CD20-specific antibodies [31]. The production of chemokines by B cells to draw in monocytes and create pathogenic antibodies has been demonstrated to aggravate unfavorable remodeling of the myocardium, despite the fact that the exact processes by which B cells alter myocardial function are still poorly understood [31]. In a model of chronic HF, transgenic mice who were unable to secrete immunoglobulin B cells displayed less cardiac remodeling and diminished cardiac function in comparison to mice that could [32, 33].

The activation of autophagy by STING has been demonstrated to occur through a mechanism to TBK1 activation and interferon induction [34], autophagy serves as a steward for specific cytoplasmic protein breakdown and organelle recycling, both of which are essential for harmful cardiac remodeling [34]. After MI, specific small molecule STING1 inhibition may enhance the wound healing response and pathological remodeling, hence lowering the occurrence of ischemic HF [35]. Rech et al. demonstrated that specific small molecule STING1 inhibition after MI may enhance the wound healing response and pathological remodeling, thereby lowering the occurrence of ischemic HF [35]. HSPB1 is a negative regulator of the cardiomyocyte inflammatory response and is essential for the repair of injured heart tissue following myocardial infarction. These actions of HSPB1 are partially attributed to NF-B dependent regulation of leukocyte recruitment and subsequent inflammation [36]. Another indicator of a bad prognosis in patients with chronic HF is the up-regulation of HSPB1 [37]. CCL5 has been shown to be highly expressed in the peripheral circulation after MI, and blocking CCL5 lowers serum levels of neutrophil and monocyte chemo-attractants during chronic myocardial ischemia while anti-CCL5 monoclonal antibody therapy improves post-MI "clinical" outcomes, such as survival and cardiac function [38]. ITGB2 is a transmembrane adhesion and signaling receptor expressed only on leukocytes and extracellular vesicles. It promotes an inflammatory reaction and assists leukocytes adhere to tissue and move around [39, 40]. Liu et al. demonstrated that suppressing the expression of ITGB2 in macrophages in a mouse HF model decreased the infiltration of myocardial immune inflammatory cells and cardiac hypertrophy [41]. The inflammatory response after cardiac injury on the one hand provides protection for the short-term adaptation of the heart, nonetheless, this inflammatory response can also lead to left ventricular dysfunction and myocardial remodeling [42]. Although Jia et al. found multiple genes linked to myocardial contraction, cardiac hypertrophy, cardiac fibrosis, and myocardial damage, including ACTN1, the function of ACTN1 protein in the pathogenesis of HF following MI is not yet established [43]. But there are also several studies using ACTN1 as a marker of cardiogenesis [44, 45].

In conclusion, our study provides a valid predictive model for the prognosis after MI, through blood, we creatively explored the link between immune cells in the blood and outcomes. Given the ease with which blood specimens can be used in clinical practice, we offer a powerful tool for the clinical use of MI. But this study still has limitations. First, we only have a set of single-cell RNA-seq data and a set of bulk RNA-seq data related to heart failure after myocardial infarction for analysis. Second, there were no clinical features other than disease information in the dataset we used. We had described the relationship between B cells in the blood and the presence of heart failure after MI, but we had not clarified the mechanism of their association. Future experiments with larger samples of more rigorous design will be needed to explore the mechanisms between disease and cells to consolidate the findings in this study.

We determined that B cells are good predictors of post-MI heart failure by combined analysis of the single-cell dataset and the bulk dataset and we identified its enrichment in multiple immune-related pathways. We further identified a diagnostic model consisting of 13 genes and identified 5 genes that were upregulated in the heart failure group by multiple methods. We have validated the diagnostic model and concluded that the diagnostic model has excellent performance. We provide an excellent tool for the diagnosis of post-MI outcomes. However, our study also has the limitation that there are fewer sequencing data that meet the requirements of our study; we only used one single-cell sequencing dataset and three infarct-related datasets. However, we determined the significance of multiple genes in this model through validation of PCR of clinical specimens. In summary, we provide an excellent clinical diagnostic model that may be able to provide a new approach and idea for the clinical diagnosis and treatment after MI.

Funding

This work was supported by the Natural Science Foundation of Shandong Province (ZR2021MH347), funded by Clinical Research Center of Shandong University (No.2020SDUCRCA017) and Natural Science Foundation of Shandong Province (ZR2020MH029).

Data availability statement

The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found in the article. The original data supporting the conclusions of this paper will be provided by the authors without undue retention. The datasets generated during and analyzed during the current study are available in the GEO repository, (https://www.ncbi.nlm.nih.gov/geo/).The datasets generated during and analyzed during the current study are available from the corresponding author on reasonable request.

Authors' contributions

Conceptualization: GXH, SHR and XZ. Methodology: XZ. Investigation: XZ and KXJ. Project administration: SHR and XZ. Writing – original draft: KXJ, SHR and ZX. All authors conceived and designed the study.

Conflict of Interest

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Human research and ethics

Consent to participate

All subjects agreed to participate in the study and signed written informed consent.

Consent for publication

All authors have approved the final version to be published.

Conflict of interest

The authors declare no competing interests.

Code availability

The software or software package used for the data analysis is indicated in the text, and the code used for the specific analysis can be obtained from the corresponding author.

1. Reed, G.W., J.E. Rossi, and C.P. Cannon, Acute myocardial infarction. Lancet, 2017. 389(10065): p. 197-210.

2. Anderson, J.L. and D.A. Morrow, Acute Myocardial Infarction. N Engl J Med, 2017. 376(21): p. 2053-2064.

3. Roger, V.L., Epidemiology of heart failure. Circ Res, 2013. 113(6): p. 646-59.

4. Dhruva, S.S., et al., Association of Use of an Intravascular Microaxial Left Ventricular Assist Device vs Intra-aortic Balloon Pump With In-Hospital Mortality and Major Bleeding Among Patients With Acute Myocardial Infarction Complicated by Cardiogenic Shock. Jama, 2020. 323(8): p. 734-745.

5. Halade, G.V. and D.H. Lee, Inflammation and resolution signaling in cardiac repair and heart failure. EBioMedicine, 2022. 79: p. 103992.

6. Jenča, D., et al., Heart failure after myocardial infarction: incidence and predictors. ESC Heart Fail, 2021. 8(1): p. 222-237.

7. Mouton, A.J., O.J. Rivera, and M.L. Lindsey, Myocardial infarction remodeling that progresses to heart failure: a signaling misunderstanding. Am J Physiol Heart Circ Physiol, 2018. 315(1): p. H71-h79.

8. Wu, H.B., et al., Research progress of CA125 and BDNF in serum of patients with acute myocardial infarction for predicting acute heart failure. Clin Hemorheol Microcirc, 2020. 75(1): p. 99-106.

9. Filipe, M.D., et al., Galectin-3 and heart failure: prognosis, prediction & clinical utility. Clin Chim Acta, 2015. 443: p. 48-56.

10. Ataklte, F. and R.S. Vasan, Heart failure risk estimation based on novel biomarkers. Expert Rev Mol Diagn, 2021. 21(7): p. 655-672.

11. Zhang, Y., J. Bauersachs, and H.F. Langer, Immune mechanisms in heart failure. Eur J Heart Fail, 2017. 19(11): p. 1379-1389.

12. Park, H.J., et al., Assessment and diagnostic relevance of novel serum biomarkers for early decision of ST-elevation myocardial infarction. Oncotarget, 2015. 6(15): p. 12970-83.

13. Tobin, S.W., et al., Considering Cause and Effect of Immune Cell Aging on Cardiac Repair after Myocardial Infarction. Cells, 2020. 9(8).

14. Schloss, M.J., et al., 2-Arachidonoylglycerol mobilizes myeloid cells and worsens heart function after acute myocardial infarction. Cardiovasc Res, 2019. 115(3): p. 602-613.

15. Qian, J., et al., Single-Cell RNA Sequencing of Peripheral Blood Mononuclear Cells From Acute Myocardial Infarction. Front Immunol, 2022. 13: p. 908815.

16. Maciejak, A., et al., Gene expression profiling reveals potential prognostic biomarkers associated with the progression of heart failure. Genome Med, 2015. 7(1): p. 26.

17. Rao, M., et al., Resolving the intertwining of inflammation and fibrosis in human heart failure at single-cell level. Basic Res Cardiol, 2021. 116(1): p. 55.

18. Koenig, A.L., et al., Single-cell transcriptomics reveals cell-type-specific diversification in human heart failure. Nat Cardiovasc Res, 2022. 1(3): p. 263-280.

19. Díez-López, C., et al., Blood Differential Gene Expression in Patients with Chronic Heart Failure and Systemic Iron Deficiency: Pathways Involved in Pathophysiology and Impact on Clinical Outcomes. J Clin Med, 2021. 10(21).

20. Wang, T., J. Tian, and Y. Jin, VCAM1 expression in the myocardium is associated with the risk of heart failure and immune cell infiltration in myocardium. Sci Rep, 2021. 11(1): p. 19488.

21. Tang, S., Y. Liu, and B. Liu, Integrated bioinformatics analysis reveals marker genes and immune infiltration for pulmonary arterial hypertension. Sci Rep, 2022. 12(1): p. 10154.

22. Zou, J., et al., scCODE: an R package for data-specific differentially expressed gene detection on single-cell RNA-sequencing data. Brief Bioinform, 2022. 23(5).

23. Kim, H. and H. Park, Sparse non-negative matrix factorizations via alternating non-negativity-constrained least squares for microarray data analysis. Bioinformatics, 2007. 23(12): p. 1495-502.

24. Hao, Y., et al., Integrated analysis of multimodal single-cell data. Cell, 2021. 184(13): p. 3573-3587.e29.

25. Haghverdi, L., et al., Batch effects in single-cell RNA-sequencing data are corrected by matching mutual nearest neighbors. Nat Biotechnol, 2018. 36(5): p. 421-427.

26. Hänzelmann, S., R. Castelo, and J. Guinney, GSVA: gene set variation analysis for microarray and RNA-seq data. BMC Bioinformatics, 2013. 14: p. 7.

27. Jan, B., et al., cancerclass: An R Package for Development and Validation of Diagnostic Tests from High-Dimensional Molecular Data. Journal of Statistical Software, 2014. 59(1): p. 1 - 19.

28. Ritchie, M.E., et al., limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Res, 2015. 43(7): p. e47.

29. Swirski, F.K. and M. Nahrendorf, Cardioimmunology: the immune system in cardiac homeostasis and disease. Nat Rev Immunol, 2018. 18(12): p. 733-744.

30. Yan, X., et al., Temporal dynamics of cardiac immune cell accumulation following acute myocardial infarction. J Mol Cell Cardiol, 2013. 62: p. 24-35.

31. Zouggari, Y., et al., B lymphocytes trigger monocyte mobilization and impair heart function after acute myocardial infarction. Nat Med, 2013. 19(10): p. 1273-80.

32. Keppner, L., et al., Antibodies aggravate the development of ischemic heart failure. Am J Physiol Heart Circ Physiol, 2018. 315(5): p. H1358-h1367.

33. Macosko, E.Z., et al., Highly Parallel Genome-wide Expression Profiling of Individual Cells Using Nanoliter Droplets. Cell, 2015. 161(5): p. 1202-1214.

34. Gui, X., et al., Autophagy induction via STING trafficking is a primordial function of the cGAS pathway. Nature, 2019. 567(7747): p. 262-266.

35. Rech, L., et al., Small molecule STING inhibition improves myocardial infarction remodeling. Life Sci, 2022. 291: p. 120263.

36. Wang, Y., et al., Cardiomyocyte-specific deficiency of HSPB1 worsens cardiac dysfunction by activating NFκB-mediated leucocyte recruitment after myocardial infarction. Cardiovasc Res, 2019. 115(1): p. 154-167.

37. Traxler, D., et al., The inflammatory markers sST2, HSP27 and hsCRP as a prognostic biomarker panel in chronic heart failure patients. Clin Chim Acta, 2020. 510: p. 507-514.

38. Montecucco, F., et al., CC chemokine CCL5 plays a central role impacting infarct size and post-infarction heart failure in mice. Eur Heart J, 2012. 33(15): p. 1964-74.

39. Fagerholm, S.C., et al., Beta2-Integrins and Interacting Proteins in Leukocyte Trafficking, Immune Suppression, and Immunodeficiency Disease. Front Immunol, 2019. 10: p. 254.

40. Banik, S.K., et al., Network analysis of atherosclerotic genes elucidates druggable targets. BMC Med Genomics, 2022. 15(1): p. 42.

41. Liu, L., et al., Myocardin-related transcription factor A regulates integrin beta 2 transcription to promote macrophage infiltration and cardiac hypertrophy in mice. Cardiovasc Res, 2022. 118(3): p. 844-858.

42. Adamo, L., et al., Reappraising the role of inflammation in heart failure. Nat Rev Cardiol, 2020. 17(5): p. 269-285.

43. Jia, Y., et al., Cardiomyocyte-Specific Ablation of Med1 Subunit of the Mediator Complex Causes Lethal Dilated Cardiomyopathy in Mice. PLoS One, 2016. 11(8): p. e0160755.

44. Castells-Sala, C., et al., Three-Dimensional Cultures of Human Subcutaneous Adipose Tissue-Derived Progenitor Cells Based on RAD16-I Self-Assembling Peptide. Tissue Eng Part C Methods, 2016. 22(2): p. 113-124.

45. Kubin, T., et al., The MEK/ERK Module Is Reprogrammed in Remodeling Adult Cardiomyocytes. Int J Mol Sci, 2020. 21(17).

Supplementarymaterials.docx

Download PDF

Version 1

posted

You are reading this latest preprint version

Risk prediction model construction for heart failure after myocardial infarction by immune B cells in blood

Status:

Version 1

Abstract

Figures

1 Introduction

2 Results

3 Methods

4 Discussion

5 Conclusion

Declarations

References

Supplementary Files

Status:

Version 1