This is a case-control study. Ten AMI, eight CAD patients, and 10 healthy individuals were recruited at the Guangdong Provincial People's Hospital from December 2018 to January 2019. The AMI and CAD patients were diagnosed by laboratory tests and coronary angiography based on the European Society of Cardiology guidelines (14, 15). Participants were 18–75 years of age and included both genders.

The healthy individuals' recruitment inclusion criteria included: normal renal and liver function; and no history of smoking, malignancy, recent cardiovascular or cerebrovascular events, rheumatologic disorders, chronic heart failure, diabetes, acute or chronic infectious disease, aortic dissection, pulmonary embolism, myocarditis, pericarditis, or congenital heart disease.

Ethical approval of human sample collection was obtained from the Ethics Committee of Guangdong Provincial People's Hospital (No. 2018160A). All patients provided informed consent.

Two milliliters of venous blood were collected into ethylene diamine tetraacetic acid (EDTA) routine blood tubes immediately at admission before coronary angiography. Plasma was separated by centrifugation at 2,000 × g for 10 min at 4°C and stored in cryogenic vials at −80°C. The exoRNeasy Serum/Plasma kit (Qiagen) was used to extract exosomal total RNA based on the manufacturer's instructions. Exosomal RNA was extracted from 1 mL of plasma. SMART technology (Clontech) was used to construct the RNA-seq libraries. RNA sequencing was performed using a illumine Nova-Seq 6000 System with the technical support of the Guangzhou Epibiotek Co., Ltd. HISAT2 was used to align sequencing reads (16). The GENCODE database was used to annotate mRNAs and lncRNAs. CircRNAs from unmapped reads were identified with the Accurate CircRNA Finder Suite (17).

Exosome morphological characteristics were assessed by transmission electron microscopy (TEM). Size and distribution were measured using NanoSight NS500 (NanoSight Ltd., Amesbury, United Kingdom). Exosomal protein markers CD9 and CD63 were detected by western blot.

The abundance of immune cells in serum of study groups were quantified using CIBERSOFT algorithm. The standardized gene expression data of the cells was uploaded to a publicly available online database (https://cibersort.stanford.edu/index.php). Overall, we characterized the serum distribution of 22 human immune cells based on 547 marker genes (18).

NetworkAnalyst (http://www.networkanalyst.ca), a visual analytics platform for comprehensive gene expression profiling (19), was used to identify exoLRs that differed between AMI, CAD, and normal samples. The platform has three steps: data filtering, normalization, and difference analysis. First, exoLRs with low expression were filtered from the dataset. Variance percentile rank lower than 15% was filtered and the minimum criterion for retaining an RNA was at least 4 counts per million. Following data filtering, counts were then normalized using the trimmed mean of M-values normalization method. These steps reduce the influence of batch effect on experimental results. Then the “DESeq2” package was applied for differential analysis of count data in order to estimate variance-mean dependence in count data from high-throughput sequencing assays. Differential expression was based on a model using the negative binomial distribution. Subsequently, the data (i.e., P-value, fold changes, or effect sizes) were extracted. Based on this overall evidence, we identified RNAs that were significantly different in expression. RNAs with P < 0.01 and log₂ fold change ≥ |1| were considered significantly different exoLRs. The “ClusterProfiler” package was used to investigate Gene Ontology (GO) enrichment analysis including biological processes, cellular components, and molecular functions (20). P < 0.05 was selected as the cut-off value for enriched function.

Exosomal mRNAs were initially filtered if expression values were <1 fragment per kilobase per million (FPKM) in at least 90% of the samples. The remaining mRNAs with standard deviations >0.2 were fed into an R package for weighted correlation network analysis (21).

Appropriate soft-threshold power was selected to ensure the co-expression network based on scale-free topology. The weighted adjacencies and correlations were transformed into a topological overlap matrix (TOM), followed by calculation of the corresponding dissimilarity (1-TOM). Next, 1-TOM, as the distance measure, was applied to a hierarchical clustering analysis of genes. A dynamic tree cut algorithm was used to identify modules. Based on module eigengene and clinical trait correlations, significant modules were identified with P < 0.05. The top 50 exosomal mRNA connections, based on topological overlap in significant modules, were used to construct a network diagram using Cytoscape (22). Exosomal mRNAs with eigengene connectivity > 0.8, within the module, were related to clinical traits and considered candidate hub genes (23).

RNA expression levels are shown as means of FPKM. Comparative Venn diagrams were constructed using Venny 2.1.0 (https://bioinfogp.cnb.csic.es/tools/venny/index.html). Principal component analysis (PCA) was applied to evaluate variables within the three groups. Data were transformed into log₂ scale and plotted using the plotPCA function in R v.3.5.2. Continuous variables that were normally distributed are displayed as means with standard deviation. One-way ANOVA followed by post-test least significant difference was performed to assess the difference among multiple groups. P < 0.05 was considered statistically significant. Pearson's correlation coefficient was calculated to test statistical correlation and r > 0.5 or r < −0.5 with a P < 0.05 was considered statistically significant. The area under the curve (AUC) of receiver operating characteristic (ROC) was applied to evaluate the specificity and sensitivity of exosomal mRNA for AMI diagnosis with 95% confidence interval (95% CI) calculated. Due to the relatively small sample size for ROC analysis, statistical power was calculated using PASS (version 15.0) with the following conditions: α = 0.05, AUC = 0.5, and n = 30. P < 0.05 was considered statistically significant. PASW Statistics 18.0 software was used for all statistical analyses.

The clinical characteristics of the three groups (10 AMI, 8 CAD, and 10 controls) are presented in Table 1. Of the 10 AMI patients, nine presented with ST-elevated myocardial infarction and one with non-ST-elevated myocardial infarction. The CAD group and control groups had similar acute inflammation levels, whereas the AMI group had significantly higher levels at study entry. MB isoenzyme of creatine kinase (CKMB) and hydroxybutyrate-dehydrogenase (HBDH) were also higher in AMI patients than in the other subjects.

Reliable exoLRs-seq data were obtained by plasma isolation, purification of exosomes, exosome RNA extraction, and RNA-seq library construction (Figure 1A). TEM results showed membrane-enclosed exosome structures without similar size or uniform distribution (Figure 1B). The average diameter of the isolated exosomes was 75.83 nm measured with a NanoSight instrument (Figure 1C). Membrane markers of exosomes, CD9 and CD63, were demonstrated by Western blot (Figure 1D). mRNA constituted 58.46% of total mapped reads. Pseudogenes and circRNAs accounted for 12.80 and 11.73%, respectively, whereas lncRNAs and antisense RNAs were 7.94 and 7.55%, respectively (Figure 1E). The number of mRNAs, circRNAs, lncRNAs, and pseudogenes for the AMI group were all significantly higher than the control group (Figure 1F). Compared with the CAD group, the AMI group had a significantly higher number of mRNAs (Figure 1F).

Twenty-two types of immune cells were assessed using the exosomal sequencing data with an established computational resource (CIBERSORT) (Figure 2A). PCA of the immunological profiles showed a non-uniform distribution (Figure 2C). Memory B cells were least in the CAD group, whereas naïve B cells of the CAD group were the highest. Neutrophils and monocytes were significantly enriched in the AMI group, which is consistent with clinical baseline characteristics (Figure 2B). Results suggest that circulating exoLRs may reflect the circulating immune cell profile.

Compared with the control group, 296 different exosomal mRNAs, consisting of 254 up- and 42 down-regulated (Supplementary Table 1), were identified for the AMI group. The top 15 up and top 15 down regulated exosomal mRNAs are shown in Table 2. Only 16 different circRNAs or lncRNAs were identified (Figures 3A,B). ClusterProfiler was used to analyze and visualize functional profiles (Gene Ontology) of the 296 different exosomal mRNAs with identification of enrichment maps of biological processes by https://www.networkanalyst.ca. Results of the functional profile analysis (Supplementary Table 2) are shown as a bar plot (Figure 3C) and the enrichment map of biological processes as a network (Figure 3D). Among the top 10 biological processes in Figures 3C,D, neutrophil degranulation, neutrophil activation, and neutrophil activation involved in immune response were significantly enriched. The enrichment map (Figure 3D) showed that the inflammatory response may be the core biological process.

By comparison of AMI and control exoLRs, we found that circulating AMI exosomal mRNAs may play a crucial role in the acute inflammation response mediated by neutrophils.

When compared with the CAD group, 230 different exosomal mRNAs (Supplementary Table 3), consisting of 120 up- and 110 down-regulated, were identified for the AMI group (Figures 4A,B). Functional profile analysis (Figure 4C and Supplementary Table 4) found neutrophil activation to play a leading role in associated biological processes. To further investigate the relationship between AMI and CAD, intersections of differently up- or down-regulated exosomal mRNAs for the AMI, CAD, and control groups are depicted in Figures 4A,B. There were 35 different exosomal mRNAs (31 up- and 4 down-regulated) that overlapped between the AMI and control groups and AMI and CAD (Supplementary Table 5). GO analysis of these 35 different exosomal mRNAs (Supplementary Table 6) found that myeloid leukocyte activation was mainly enriched (Figure 4D). ALPL mRNA is especially worth noting in that it was the only up-regulated exosomal mRNA among the three groups (Figure 4A) and may serve as a potential biomarker for AMI diagnosis.

Weighted gene co-expression network analysis was used to identify key modules and highly correlated exosomal mRNAs associated with AMI. Forty-three exosomal mRNAs modules were identified by hierarchical clustering dendrogram (Figure 5A). The association of the 43 co-expression modules was analyzed by topological overlap matrix plot that consisted of the modules and the corresponding hierarchical clustering dendrogram (Figure 5B). For module-trait analyses, only the light-yellow module containing 175 exosomal mRNAs (Supplementary Table 7) was related to AMI (Figure 5C). Although there was no significantly enriched Kyoto Encyclopedia of Genes and Genomes (KEGG) biological pathways in the light-yellow module, the results of GO enrichment analysis indicated that the light-yellow module was involved in inflammation mediated by neutrophils, i.e., neutrophil aggregation and chemokine production (Figure 5D). The association between the light-yellow module and AMI was also supported by enrichment analyses of different exosomal mRNAs between AMI (Figure 3C) and control or between AMI and CAD (Figure 4D). To further analyze the core mRNA of the light-yellow module, the top 50 exosomal mRNAs of the topological overlap matrix of this module were used to construct a network diagram (Figure 5E). In this network, exosomal mRNAs (with eigengene connectivity > 0.8) in the light-yellow module were considered core mRNAs and labeled with yellow (Figure 5E). Six mRNAs including ALPL, CXCR2, ELL2, EMC9, FAM129A, and DBF4B were identified as core mRNAs.

To estimate the potential clinical value of the six core exosomal mRNAs, ROC curve analysis was applied (Figure 6A). AUC indicated that there were two mRNAs with excellent predictive accuracy ALPL (AUC: 0.99, 95%CI, 0.9484–1.000, P = 0.0002, power = 0.99754) and CXCR2 (AUC: 0.98, 95%CI, 0.8478–1.000, P = 0.0005, power = 0.68124)]. Correlational analysis of these two exosomal mRNAs and clinical baseline characteristics (Figure 6B) indicated that ALPL and CXCR2 are associated with neutrophil count (R = 0.52 and 0.51, respectively). Statistical analysis suggested expression of exosomal mRNA ALPL and CXCR2, derived from AMI plasma, to be significantly higher than that found in the other groups (Figures 6C,D). Exosomal ALPL in the CAD group was also significantly higher than the control group. These results suggest that circulating exosomal ALPL may increase with the progression of coronary plaque to acute myocardial infarction.

Since this analysis found AMI exosomal mRNAs to be associated with acute inflammation mediated by neutrophils, we applied PCA to the neutrophil count and neutrophil ratio for the three groups. PCA of the neutrophil count and neutrophil ratio showed that although the AMI group could be separate from the other groups, there was much overlap between the CAD and control groups (Figure 6E). However, when we applied PCA to the expression of ALPL and CXCR2, the three groups could be separated (Figure 6F). In summary, exosomal ALPL and CXCR2 have the predictive potential to provide greater accuracy and identification of cardiac disease.