Skip to main content

NR4A3 and CCL20 clusters dominate the genetic networks in CD146+ blood cells during acute myocardial infarction in humans



CD146 is a tight junction-associated molecule involved in maintaining endothelial barrier, and balancing immune–inflammation response, in cardiovascular disease. Notably, peripheral CD146+ cells significantly upsurge under vessel dyshomeostasis such as acute myocardial injury (AMI), appearing to be a promising therapeutic target. In this study, with a new view of gene correlation, we aim at deciphering the complex underlying mechanism of CD146+ cells’ impact in the development of AMI.


Transcription dataset GSE 66,360 of CD146+ blood cells from clinical subjects was downloaded from NCBI. Pearson networks were constructed and the clustering coefficients were calculated to disclose the differential connectivity genes (DCGs). Analysis of gene connectivity and gene expression were performed to reveal the hub genes and hub gene clusters followed by gene enrichment analysis.

Results and conclusions

Among the total 23,520 genes, 27 genes out of 126 differential expression genes were identified as DCGs. These DCGs were found in the periphery of the networks under normal condition, but transferred to the functional center after AMI. Moreover, it was revealed that DCGs spontaneously crowded together into two functional models, CCL20 cluster and NR4A3 cluster, influencing the CD146-mediated signaling pathways during the pathology of AMI for the first time.


Cluster of differentiation 146 (CD146)/melanoma cell-associated molecule is an essential immunoglobulin-like protein initially discovered in metastatic melanoma [1]. It locates at endothelial tight junctions across all vessel beds, mediating physiological and pathological events under vascular dyshomeostasis [2, 3]. Pioneering researchers regard CD146 as a historical marker for isolating circulating endothelial cells that slough off inflamed vasculature [4]. Over several decades, CD146 has also been discovered in other cell types including mesenchymal stem cells [5], endothelial progenitor cells [6], macrophages [7], T helper 17 cells [8], B lymphocytes [9], T lymphocytes [9, 10], and natural killer cells [9]. The CD146+ circulating cells occupy about 2% of peripheral mononuclear cells in healthy individuals [9] and most notably, this percentage increases in certain conditions associated with vascular dysfunction like myocardial infarction, connective tissue diseases, and cancers [6, 11,12,13]. Moreover, CD146-activated T cells have shown an enhanced ability to interact with endothelium in adhesion, rolling, and transmigration, evidenced by human and murine studies [14, 15]. Given its multiple functions in vessel structure, angiogenesis, and lymphocyte activation and its enabled detection in the bloodstream, CD146 appears to be a potential target for vascular disorders [16,17,18].

Complex networks are of great interest to researchers in the fields of computational biology and bioinformatics [19,20,21]. It has been gradually extended from initial gene comparison to protein–protein network modeling, to protein–genetic investigation, and up to the disease–disease association exploration [22]. Most of the successful bioinformatics approaches that identify the initial key genes, however, have been based on only gene expression comparison and restrict the following analysis to the top differential expression genes (DEGs) without paying attention to the gene interaction rearrangement [23, 24]. Instead, the hub-structured network is an important motif that is, to our best knowledge, leading the genome-wide association characterization in complex networks [25, 26]. It generates the structure view angle to present the innermost gene–gene interaction, giving a comprehensive understanding of the underlying mechanisms of disorders.

Acute myocardial injury (AMI) dataset GSE 66,360 was focused on the performance of CD146+ populations during early AMI development [12, 27,28,29]. In this paper, we try to decipher gene reorientation within the correlation network structure parameter analysis [30, 31], extract the optimal gene collections which are termed the differential connectivity genes (DCGs), and reveal functional gene clusters which likely lead to the pathogenesis of peripheral CD146+ blood cells during the development of AMI in human for the first time.

Materials and methods


The GSE66360 [12] gene transcription profile data of human AMI in the NCBI database was selected as the primary interest. Clinical subjects including 50 healthy individuals and 49 AMI patient subjects were recruited in the original investment by the Topol group. To gather the data, CD146+ cells were obtained by CD146-based magnetic immunoisolation from the subjects’ blood samples. RNA samples were isolated from the CD146+ cells and processed by Affymetrix human U133 Plus 2.0 array. In this study, two cohorts were formed: a discovery cohort, consisting of 22 healthy subjects (control group) and 21 AMI patients (AMI group), which were used for the discovery of genes and appropriate testing methods, and a validation cohort, consisting of 28 healthy subjects and 28 AMI patients, which were used for the validation of the genes and methods discovered in the discovery cohort. No data were excluded from the original databases used during this study.

Study design

First, using the hypothesis test published by our group [33], we distinguished DEGs based on the gene expression profile in the discovery cohort and then verified in the validation cohort. Second, the gene networks of DEGs were constructed based on the Pearson coefficients, followed by the network separation assessment. Third, the clustering coefficient, which is a parameter indicating gene connectivity, was calculated for each DEG under each gene network [32]. Accordingly, genes with a clustering coefficient that represented a consistent increase in the AMI group among different cohorts were labeled as DCGs. Finally, two-dimensional analysis of gene connectivity and expression was employed, hub gene clusters were identified, and gene enrichment analysis was performed (Fig. 1).

Fig. 1

Flowchart for study design. DEGs differential expression genes; DCGs differential connectivity genes

Identify DEGs

A total of 23,520 genes were screened in each sample. The hypothesis test was used to screen the DEGs between control and AMI group [33]. The method primarily gave weight for the distribution shape of the expression spectrum. If the distribution shape was different between the two groups, then the gene expression was different with regard to significance level α1. If not, a normal distribution test (significance level α2) and homogeneity test of variance (significance level α3) would be carried out. T test or Welch’s t test (significance level α4) was used for normal distribution; rank sum test (significance level α4) was used for non-normal distribution with a similar distribution of expression spectrum [33]. We defined α1  =  0.00001, α2  =  0.00002, α3  =  0.00001, and α4  =  0.00001 as the significance levels of the hypothesis tests for DEGs selection regarding the limited input gene size of the following connectivity analysis.

Clustering coefficient

A local clustering coefficient was introduced to measure the compactness, or the connectivity of genes within a suspected cluster, of a complete array formed by the adjacent nodes within a network [34]. To clarify, assume that a node i in a network was connected to ki nodes. The ki nodes were called neighbors of node i. The ratio of the actual number of edges Ei and the total number of possible edges ki(ki − 1)/2 between ki nodes were defined as the clustering coefficient, Ci, of node i; that is, Ci  =  2Ei/[ki(ki − 1)].

Pearson network construction and assessment

Pearson correlation networks of DEGs were constructed according to the absolute value of Pearson coefficients. Two genes were considered correlated if the absolute value of the Pearson coefficient was greater than the threshold x (0  <  x  <  1), at which point a line could be drawn between the two genes. In the cases when genes were not correlated, there would be no link in the network and thus no line would be drawn. Gene clusters were determined by examining the clustering coefficients, and those with a non-zero value could be labeled clusters. Gene clusters represent functional modules as a whole with varying degrees and connectivity; while the degree describes the number of genes connected to one another. The average clustering coefficients of DEGs were calculated to evaluate the overall separation of the control and AMI networks. This analysis was performed in R i386 3.6.2.

Natural biological networks are scale-free networks and the degree distributions follow the power-law exponential distribution index range 2–3 [35, 36]. We specifically looked at the gene networks under threshold 0.5 and 0.7 because the power-law indexes of degree distribution in discovery cohort were in the range of 2–3. We then presented the corresponding networks in discovery cohort and validation cohort of this study (Additional file 2: Table S1).

Identify DCGs

In the analysis of network connection parameters, the greater the difference between the control and AMI group, the higher the correlation with AMI. The following describes our unique identifying method. Assume that the average clustering coefficient of the control group and AMI group could be separated at the threshold (0.1, 0.9). First, the clustering coefficients of each gene in the Pearson correlation networks under the threshold 0.1–0.9 of the control group and AMI group were calculated with step length 0.1. Secondly, the average clustering coefficients of each gene through threshold 0.4–0.8 were calculated to compare changes in connectivity between the two groups within the validation and discovery cohorts. Finally, if clustering coefficient differences in discovery cohort and validation cohort were consistently greater than 0.1 between the AMI group and the control group, the genes were identified as candidates for DCGs.

To test the reliability of the proposed candidate genes across different datasets, we expanded our method to a combination cohort, which included all subjects in both discovery and validation cohorts. This increased the number of subjects, but also introduced some variation in the data due to the less categorized subject population. The overall network between the control and the AMI groups were still separable through threshold 0.4–0.8 (data not shown). The 27 out of 39 candidate genes that still showed clustering coefficient differences greater than 0.1 in this combination cohort were defined as DCGs.

Gene set enrichment

Gene set enrichment was performed by the STRING server. Biological process, reactome pathways, and protein–protein association networks were generated for CCL20 cluster, NR4A3 cluster, and DCGs.


Heatmaps of DEGs and DCGs were generated by using heatmap.2 function in the gplots package. Networks were computed by using function. Layout algorithm of layout.kamada.kawai was used for visualizing the overall DEGs networks and the connections for individual genes. Layout algorithm of was used to visualize the gene connections within DEGs and DCGs in circle view. Cytoscape network function was used to generate the clustered DCGs networks.


DEGs identification

In our initial analysis, 126 out of 23,520 genes’ expression levels were significantly altered in the AMI group compared to the control group in discovery cohort, defined as DEGs, with the majority (79 of 126) demonstrated an upregulation feature (Fig. 2A). The validation cohort showed a similar expression pattern (Fig. 2B).

Fig. 2

Gene expression profiles of DEGs. 126 genes show significant differential expressions between the AMI and the control groups in the discovery cohort (A) and validation cohort (B), defined as DEGs. DEGs differential expression genes

Assessment of DEGs’ networks

The overall gene networks of DEGs in the control group and the AMI group were distinctly independent through a large range of thresholds in discovery cohort and validation cohort (Fig. 3A). The networks in discovery cohort were separable through threshold 0.1–0.9 and in validation cohort were separable through 0.1–0.8. The average separable widths of the discovery cohort and validation cohort were 0.218 and 0.0518, respectively. The validation cohort showed a narrower split range possibly due to the variations between two cohorts; for instance, the differential sample size, age, and co-morbid disorders.

Fig. 3

Assessment of DEGs’ networks. Networks in the control and AMI groups are independent and separable according to the average clustering coefficients of DEGs (A). Number of clusters within DEGs’ networks progressively decline when thresholds increase from 0.1 to 0.9 (B). The AMI group has a lower slope decline. The gene networks of DEGs in the AMI group have more complex connection, compared to that in the control group (C). Networks are present under threshold 0.5 and 0.7. Darker line represents connections under threshold 0.7; lighter line represents connections under threshold 0.5. DEGs differential expression genes

In addition, the gene connections within DEGs’ networks in the AMI group were more complex than those in the control group in both cohorts (Fig. 3C). In the discovery cohort, the number of gene clusters within AMI network gradually decreased from 125 to 67, when threshold increased from 0.4 to 0.8, while it more sharply decreased from 123 to 17 in the control group (Fig. 3B). Similarly, the clusters declined with a lower slope in the AMI group compared to the control group in validation cohort (Fig. 3B).

The data described above suggested that gene networks of DEGs were largely and consistently disturbed by AMI stimulation as seen in two independent cohorts, verifying our findings. Thus, DEGs and DEG-networks were mathematically reliable and hereafter could be set as the foundation for in-depth gene interaction data mining.

DCGs identification and connectivity analysis

Beyond DEGs, we identified 27 genes as DCGs whose clustering coefficient difference was greater than 0.1 in the discovery cohort, the validation cohort, and the extended combination cohort (Fig. 4A; Additional file 2: Table S2). The sub-networks of DCGs presented obvious tighter connections in the AMI group compared to the control group, in both discovery and validation cohorts (Fig. 4C). When threshold increased from 0.4 to 0.8, the average degree of DCGs progressively decreased from 16.0 to 4.30 in the AMI group, while it decreased from 7.93 to 0 in the control group in discovery cohort (Fig. 4D). Similarly, in the validation cohort, this number decreased from 19.9 to 1.41 in the AMI group while from 3.26 to 0 in the control group (Fig. 4D). In addition to the clustering coefficients, gene expression of those DCGs showed a steady increase in the AMI group in both the discovery and validation cohorts (Fig. 4B).

Fig. 4

Identification of DCGs. Genes for which the clustering coefficient increased by over 0.1 in the AMI group in both discovery cohort and validation cohort, are revealed as DCGs (A). Gene expression profile of DCGs shows a stable increase in the AMI group in both cohorts (B). The connections among DCGs in the AMI group are denser (C) and the average degrees of DCGs in AMI group are higher (D) compared to the control group in both cohorts. Networks are presented under threshold 0.5 and 0.7. Darker lines represent connections under threshold 0.7; lighter line represents connections under threshold 0.5. Degrees are presented as mean  ±  SEM. DCGs differential connectivity genes

Therefore, we proposed that the networks’ differential of all DEGs was largely attributed to the connection changes within DCGs. As visualized in kamada-kawai layout, the DCGs randomly participated in the DEGs’ network and connected to a few genes under a normal steady state. However, they appeared to interact with more functional genes and shift into central positions of the clusters after AMI in both discovery and validation cohorts (Fig. 5).

Fig. 5

Visualization of DCGs in DEGs’ networks. The networks of DEGs in discovery cohort (A) and in the validation cohort (B) indicate that the DCGs participate in distinctive ways in the control group and in the AMI group. DCGs switch to central functional position of networks and participate in more intricate connections under AMI situation. Yellow nodes indicate the DCGs. Red gene names indicate the hub genes. DEGs differential expression genes; DCGs differential connectivity genes

Two-dimensional analysis of gene connectivity and gene expression

As the power-law indexes of degree distribution in the discovery were in the range of natural networks, we regarded the discovery cohort as a more precise dataset and it was selected for the following analysis. The average clustering coefficient and gene expression of DCGs were plotted for two-dimensional analysis (Fig. 6A). NR4A3 and CCL20 presented high levels of clustering coefficient and gene expression changes, defined as CChighGeExphigh genes. SOCS3, FOSL2, PLIN2 were the genes that were found to have high clustering coefficient changes (fold-change  >  2) with low expression changes, defined as CChighGeExplow genes; while IL1R2, NLRP3, ANXA3 and AC079305.10 were the genes that were found to have high expression changes (fold-change  >  0.4) with low clustering coefficient changes, defined as CClowGeExphigh genes. All information on these genes is shown in Table 1. Subgraphs of NR4A3, CCL20, and other DCGs provided evidence that supported their increased gene connectivity after AMI in the discovery cohort (Fig. 6C, Additional file 1: Figure S1A, B).

Fig. 6

Analysis of gene connection and expression of DCGs in discovery cohort. The analysis of clustering coefficient and gene expression revealed CCL20 and NR4A3 as hub genes (A). The CCL20 is a chemoattractant while NR4A3 is a nuclear factor receptor (B). Subgraphs of CCL20 and NR4A3 substantiate their important roles in AMI development (C). Networks are presented under threshold 0.5 and 0.7. Darker lines represent connections under threshold 0.7; lighter lines represent connections under threshold 0.5. DCGs differential connectivity genes; CC clustering coefficient; GeExp gene expression

Table 1 Category of DCGs

NR4A3 and CCL20 clusters identification

Taking a closer analysis of the subgraphs of individual genes, we revealed that DCGs stay “non-activated” in the control group (Fig. 7A). Interestingly, even though there were several minor clusters in the gene network, they appeared to be spontaneously gathering together as two major separate clusters after AMI stimulation (Fig. 7B, Additional file 1: Figure S1A, B). The highlighted CChighGeExphigh genes, NR4A3 or CCL20, located in the center spot of each cluster, served as leading-like hub genes. CCL20 connected with SKIL, MMP9, ITPRIP, ANXA3, GLUL, CXCL16, IL1R2, TMCC3, NLRP3, PYGL, RNF144B, BCL6, LILRB2, CLEC4E, FCER1G, and AC079305.10, identified as the CCL20 cluster. NR4A3 connected with NR4A2, FOSL2, CDKN1A, SOCS3, GABARAPL1, ITPRIP, SYTL3, PELI1, MAP3K8, and PLIN2, identified as the NR4A3 cluster. While there were overlapping genes between clusters, ITPRIP, SKIL and MAPK38 were the intermediate genes that connected both clusters according to their subgraphs (Additional file 1: Figure S1B). The clustering coefficient fold-changes of NR4A3 and CCL20 were 15.7 and 10.2, respectively; and the gene expression fold-changes were 0.379 and 0.422, respectively. Gene connections are partly verified by the STRING datasets (Additional file 1: Figure S2).

Fig. 7

CCL20 cluster and NR4A3 cluster formation in early-stage AMI. CCL20 and NR4A3 stay in the peripheral position of DCGs’ network under normal state (A). However, they shift to the primary position of DCGs’ network dominating two functional clusters under AMI stimulation (B). DCGs differential connectivity genes

Gene enrichment

Biological process analysis showed that DCGs were involved in response to organic substrates, positive regulation of leukocyte activation, immune response, immune system process, response to cytokine, and regulation of cytokine production. The CCL20 cluster was essential to the immune response, immune system process, and regulation of localization while the NR4A3 cluster was essential to cellular response to corticotropin-releasing hormone stimulus, positive regulation of leukocyte activation, and regulation of apoptotic process (Table 2).

Table 2 Top six gene enrichment outputs of DCGs and hub gene clusters

Reactome pathway analysis revealed that DCGs were related to the immune system with regard to tasks such as signaling by interleukins, namely interleukin-1, interleukin-4 and interleukin-13, the innate immune system, and the dectin-2 family. The CCL20 cluster was essential to immune system, innate immune system, the dectin-2 family, and neutrophil degranulation while the NR4A3 cluster was essential to RNA Polymerase II Transcription, Generic Transcription Pathway, and MyD88 cascade initiated on plasma membrane (Table 2).


CD146 is a junction-associated adhesion molecule that participates in immune and inflammatory pathological processes in the initiation and development of vascular diseases [2]. CD146-activated leukocytes are recruited to the inflamed endothelium to induce the expression of chemokines and cytokines and, in doing so, progressively destroy the blood vessel barrier. Our study finds that following AMI stimulation, in CD146+ human blood cells, 126 out of 23,520 genes show significant differential expression (P  <  0.0001) and, among those, 27 genes show consistent connectivity changes and serve as DCGs. Unlike DEGs, DCGs are able to not only aggregate gene expression, but also encompass gene connectivity properties, internally coupling into functional gene clusters—NR4A3 cluster and CCL20 cluster—orchestrating the gene networks’ entire dynamics in CD146-associated AMI pathophysiology development. Meanwhile, NR4A3 and CCL20 are revealed as hub genes since they experienced both significant connectivity and expression changes after AMI stimuli. Furthermore, gene enrichment analysis shows that the DCGs are involved in inflammation–immune response, with CCL20 being principal to the immune response and regulation of localization; while, the NR4A3 cluster is principal to leukocyte activation, apoptotic process, and cellular response to corticotropin-releasing hormone stimulus. Such findings align with the well-known hypothesis that CD146-mediated inflammation plays an important role in the pathogenesis of AMI.

DCGs revealed by gene connectivity network

The network structural parameter analysis method is applied to weave the gene–gene correlation network. We identify DCGs which present steadily elevated connectivity under AMI conditions in both the discovery and validation cohorts, and further confirm the upregulation seen in the combination cohort. As expected, the gene expressions of DCGs are increased after AMI, but are not distinguishable from DEGs solely by expression signatures (data not shown). NR4A3 and CCL20 as Queryhighlight hub genes are also defined as AMI biomarkers after pre-filtering the comorbidity-relevant genes by the original Topol group [12]. SOCS3 tends to be the only “shared” AMI biomarker candidate revealed by other groups in which the same GSE66360 dataset is included as one of their study objects [27, 28]. Recognizing the CCL20, NR4A3, and SOCS3 as top DCGs substantiate previous outputs and, in turn, the validity of our method is enhanced. Therefore, we recommend the gene connectivity analysis, along with gene expression signature, to be used as a powerful and unbiased way for researchers to rank the importance of candidate DEGs.

The NR4A3 hub gene

NR4A3 belongs to the NR4A orphan nuclear receptor family (with NR4A2 and NR4A1), which plays an important role in AMI development. The JM Penninger group reports that NR4A3 is the highest-ranking gene in circulating human endothelial cells under atherosclerosis [37]. Transcription analysis of human left ventricular myocardium shows that NR4A3 upregulates during ischemia and reperfusion in normal and chronic ischemic myocardium [38]. Similarly, NR4A3 is found to be elevated 10 days post-left anterior descending artery ligation ischemia surgery in mice [39]. Overexpression of NR4A3 significantly reduces infarct size, preventing deterioration of left ventricular function and repression of neutrophil infiltration in the heart of mice after coronary artery ligation and relates to the activation of JAK2/STAT3 and the inhibition of STAT3-dependent NF-κB signaling pathways [40]. Additionally, it has to be pointed out that the NR4A subgroup, including NR4A3, is an immediate early response gene induced by diverse physiological stimuli, i.e., mechanical agitation, calcium, and inflammation cytokines [41]. This reinforces our data that, in very early-stage AMI, NR4A3 has a significant 16-fold clustering coefficient climb and 42% gene expression increase. Yet, despite these characteristics, implications of the nuclear factor NR4A3 in CD146+-related myocardial disorders remain a mystery.

The CCL20 hub gene

CCL20, a C–C motif chemokine, is a chemoattractant for recruiting leukocytes to sites of injury and inflammation. CCL20 secretion is induced by pro-inflammatory chemokines and cytokines, such as CXCL12, IL17, IL1β, IL6, and is in part related to JAK/STAT pathway signaling in multiple cell types [42,43,44]. IL6 and soluble IL6 receptor stimulate STAT3 binding to the CCL20 promotor, and IL17 stimulates the phosphorylated NF-κB binding to the CCL20 promoter in murine astrocytes, facilitate neuroinflammation within the central nervous system [42]. In addition, the co-expression of CCL20 receptor CCR6 and CD146 is a marker of effector memory Th17 cells which mediate migration and is thought to be essential for inflammation in human psoriasis [8]. Moreover, it is reported that CCL20 levels are elevated in the serum of clinical patients with ischemic myocardial infarction [45, 46]. In vitro study shows that CCL20 expression increases in CD146+ human mesenchymal stromal cells at the early pro-inflammatory phase in fracture healing [47]. Thus, we hypothesize that CCL20 binding its receptor CCR6 is likely what drives the CD146-mediated vessel inflammation progress in early AMI phase.

Functional gene clusters

In terms of functional models, DCGs are self-organized into two clusters, the NR4A3 and CCL20 clusters, with 18 genes and 12 genes in each cluster, respectively. All genes are directly linked to its hub gene and partly linked to adjacent genes. Protein–protein connection analyzed by STRING database produces a structure that is greatly similar to our network pattern in which CCL20 connects with CXCL16, IL1R2, MMP9, NLRP3, BCL6 LILRB2, PELI1, CLEC4E, FCER1G, and NR4A3 connects with NR4A2, FOSL2, RNF144B, CDKN1A, SOCS3. A few of the gene–gene correlations within clusters are stated in inflammatory diseases. MMP9 activation correlates with CCL20 expression in astrocytes via Notch-1/Akt/NF-κB pathway promoting leukocyte migration across the blood–brain barrier [48]. NR4A2 and NR4A3 as orphan nuclear receptors mediate neutrophil number and survival in chronic inflammatory signals in multiple hematologic disorders [49,50,51]. FOSL2 acts as an activating protein-1 transcription factor promoting hematopoietic progenitor cells transition to macrophages and neutrophils in an SOCS3-dependent manner is reported [52]. Nevertheless, most of the cluster functions are rarely reported in AMI pathogenesis.

Taken together, NR4A3 and CCL20 clusters are novel functional modules in CD146+ cell-mediated immuno­inflammatory balance, triggering increased susceptibility to vascular deterioration and accelerating myocardial injury. Meanwhile, we propose that NR4A3 and CCL20 are promising biomarkers for clinical diagnosis and potential therapeutic candidates since they largely impact the early AMI development. In-depth studies are necessary for understanding the mechanisms of peripheral CD146+ cells in cardiovascular disease.

Availability of data and materials

All data are included in the manuscript.


  1. 1.

    Lehmann JM, Holzmann B, Breitbart EW, Schmiegelow P, Riethmüller G, Johnson JP. Discrimination between benign and malignant cells of melanocytic lineage by two novel antigens, a glycoprotein with a molecular weight of 1,13,000 and a protein with a molecular weight of 76,000. Can Res. 1987;47(3):841–5.

    CAS  Google Scholar 

  2. 2.

    Shih IM. The role of CD146 (Mel-CAM) in biology and pathology. J Pathol. 1999;189(1):4–11.

    CAS  PubMed  Article  Google Scholar 

  3. 3.

    Tu T, Zhang C, Yan H, Luo Y, Kong R, Wen P, et al. CD146 acts as a novel receptor for netrin-1 in promoting angiogenesis and vascular development. Cell Res. 2015;25(3):275–87.

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  4. 4.

    Widemann A, Sabatier F, Arnaud L, Bonello L, Al-Massarani G, Paganelli F, et al. CD146-based immunomagnetic enrichment followed by multiparameter flow cytometry: a new approach to counting circulating endothelial cells. J Thromb Haemost. 2008;6(5):869–76.

    CAS  PubMed  Article  Google Scholar 

  5. 5.

    Espagnolle N, Guilloton F, Deschaseaux F, Gadelorge M, Sensébé L, Bourin P. CD 146 expression on mesenchymal stem cells is associated with their vascular smooth muscle commitment. J Cell Mol Med. 2014;18(1):104–14.

    CAS  PubMed  Article  Google Scholar 

  6. 6.

    Delorme B, Basire A, Gentile C, Sabatier F, Monsonis F, Desouches C, et al. Presence of endothelial progenitor cells, distinct from mature endothelial cells, within human CD146+ blood cells. Thromb Haemost. 2005;94(6):1270.

    CAS  PubMed  Article  Google Scholar 

  7. 7.

    Luo Y, Duan H, Qian Y, Feng L, Wu Z, Wang F, et al. Macrophagic CD146 promotes foam cell formation and retention during atherosclerosis. Cell Res. 2017;27(3):352–72.

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  8. 8.

    Kamiyama T, Watanabe H, Iijima M, Miyazaki A, Iwamoto S. Coexpression of CCR6 and CD146 (MCAM) is a marker of effector memory T-helper 17 cells. J Dermatol. 2012;39(10):838–42.

    CAS  PubMed  Article  Google Scholar 

  9. 9.

    Elshal MF, Khan SS, Takahashi Y, Solomon MA, McCoy JP. CD146 (Mel-CAM), an adhesion marker of endothelial cells, is a novel marker of lymphocyte subset activation in normal peripheral blood. Blood. 2005;106(8):2923–4.

    CAS  PubMed  Article  Google Scholar 

  10. 10.

    Duda DG, Cohen KS, di Tomaso E, Au P, Klein RJ, Scadden DT, et al. Differential CD146 expression on circulating versus tissue endothelial cells in rectal cancer patients: implications for circulating endothelial and progenitor cells as biomarkers for antiangiogenic therapy. J Clin Oncol Off J Am Soc Clin Oncol. 2006;24(9):1449.

    CAS  Article  Google Scholar 

  11. 11.

    Hadjinicolaou A, Wu L, Fang B, Watson P, Hall F, Busch R. Relationship of CD146 expression to activation of circulating T cells: exploratory studies in healthy donors and patients with connective tissue diseases. Clin Exp Immunol. 2013;174(1):73–88.

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  12. 12.

    Muse ED, Kramer ER, Wang H, Barrett P, Parviz F, Novotny MA, et al. A whole blood molecular signature for acute myocardial infarction. Sci Rep. 2017;7(1):1–9.

    CAS  Article  Google Scholar 

  13. 13.

    Fürstenberger G, Von Moos R, Senn H, Boneberg E. Real-time PCR of CD146 mRNA in peripheral blood enables the relative quantification of circulating endothelial cells and is an indicator of angiogenesis. Br J Cancer. 2005;93(7):793–8.

    PubMed  PubMed Central  Article  CAS  Google Scholar 

  14. 14.

    Guezguez B, Vigneron P, Lamerant N, Kieda C, Jaffredo T, Dunon D. Dual role of melanoma cell adhesion molecule (MCAM)/CD146 in lymphocyte endothelium interaction: MCAM/CD146 promotes rolling via microvilli induction in lymphocyte and is an endothelial adhesion receptor. J Immunol. 2007;179(10):6673–85.

    CAS  PubMed  Article  Google Scholar 

  15. 15.

    Dagur PK, McCoy JP Jr. Endothelial-binding, proinflammatory T cells identified by MCAM (CD146) expression: characterization and role in human autoimmune diseases. Autoimmun Rev. 2015;14(5):415–22.

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  16. 16.

    Leroyer AS, Blin MG, Bachelier R, Bardin N, Blot-Chabaud M, Dignat-George F. CD146 (cluster of differentiation 146) an adhesion molecule involved in vessel homeostasis. Arterioscler Thromb Vasc Biol. 2019;39(6):1026–33.

    CAS  PubMed  Article  Google Scholar 

  17. 17.

    Gallastegi T, Soto B, Romero JM, Galán M, Escudero JR, Camacho M. MCAM/CD146 which is differentially expressed in vascular diseases, is a potential biomarker in abdominal aortic aneurysm. Eur J Vasc Endovasc Surg. 2019;58(6):e454.

    Article  Google Scholar 

  18. 18.

    Wang Z, Yan X. CD146, a multi-functional molecule beyond adhesion. Cancer Lett. 2013;330(2):150–62.

    CAS  PubMed  Article  Google Scholar 

  19. 19.

    Milo R, Shen-Orr S, Itzkovitz S, Kashtan N, Chklovskii D, Alon U. Network motifs: simple building blocks of complex networks. Science. 2002;298(5594):824–7.

    CAS  PubMed  Article  Google Scholar 

  20. 20.

    Zhao Z, Li C, Zhang X, Chiclana F, Viedma EH. An incremental method to detect communities in dynamic evolving social networks. Knowl-Based Syst. 2019;163:404–15.

    Article  Google Scholar 

  21. 21.

    Liu Y-Y, Slotine J-J, Barabási A-L. Controllability of complex networks. Nature. 2011;473(7346):167–73.

    CAS  PubMed  Article  Google Scholar 

  22. 22.

    Goh K-I, Cusick ME, Valle D, Childs B, Vidal M, Barabási A-L. The human disease network. Proc Natl Acad Sci. 2007;104(21):8685–90.

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  23. 23.

    Akat KM, Morozov P, Brown M, Gogakos T, Da Rosa JC, Mihailovic A, et al. Comparative RNA-sequencing analysis of myocardial and circulating small RNAs in human heart failure and their utility as biomarkers. Proc Natl Acad Sci. 2014;111(30):11151–6.

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  24. 24.

    Eicher JD, Wakabayashi Y, Vitseva O, Esa N, Yang Y, Zhu J, et al. Characterization of the platelet transcriptome by RNA sequencing in patients with acute myocardial infarction. Platelets. 2016;27(3):230–9.

    CAS  PubMed  Article  Google Scholar 

  25. 25.

    Xiao M, Zheng WX, Jiang G, Cao J. Stability and bifurcation analysis of arbitrarily high-dimensional genetic regulatory networks with hub structure and bidirectional coupling. IEEE Trans Circuits Syst I Regul Pap. 2016;63(8):1243–54.

    Article  Google Scholar 

  26. 26.

    De Domenico M, Nicosia V, Arenas A, Latora V. Structural reducibility of multilayer networks. Nat Commun. 2015;6(1):1–9.

    Google Scholar 

  27. 27.

    Cheng M, An S, Li J. Identifying key genes associated with acute myocardial infarction. Medicine. 2017.

    Article  PubMed  PubMed Central  Google Scholar 

  28. 28.

    Ge WH, Lin Y, Li S, Zong X, Ge ZC. Identification of biomarkers for early diagnosis of acute myocardial infarction. J Cell Biochem. 2018;119(1):650–8.

    CAS  PubMed  Article  Google Scholar 

  29. 29.

    Qiu L, Liu X. Identification of key genes involved in myocardial infarction. Eur J Med Res. 2019;24(1):22.

    PubMed  PubMed Central  Article  Google Scholar 

  30. 30.

    Guo Y, Wu C, Guo M, Liu X, Keinan A. Gene-based nonparametric testing of interactions using distance correlation coefficient in case-control association studies. Genes. 2018;9(12):608.

    PubMed Central  Article  CAS  PubMed  Google Scholar 

  31. 31.

    Wang Y, Chi X, Meng D. The application of network structure analysis in the study of disease mechanisms. In: 2019 IEEE International Conference on Bioinformatics and Biomedicine (BIBM). IEEE: Piscataway; 2019.

    Google Scholar 

  32. 32.

    Opsahl T. Triadic closure in two-mode networks: redefining the global and local clustering coefficients. Soc Netw. 2013;35(2):159–67.

    Article  Google Scholar 

  33. 33.

    Wang Y, Kou Y, Meng D. Network structure analysis identifying key genes of autism and its mechanism. Comput Math Methods Med. 2020.

    Article  PubMed  PubMed Central  Google Scholar 

  34. 34.

    Galton F. Typical laws of heredity. III Nature. 1877;15(389):512–4.

    Google Scholar 

  35. 35.

    Jeong H, Tombor B, Albert R, Oltvai ZN, Barabási A-L. The large-scale organization of metabolic networks. Nature. 2000;407(6804):651–4.

    CAS  PubMed  Article  Google Scholar 

  36. 36.

    Jeong H, Mason SP, Barabási A-L, Oltvai ZN. Lethality and centrality in protein networks. Nature. 2001;411(6833):41–2.

    CAS  PubMed  Article  Google Scholar 

  37. 37.

    Barrett P, Topol EJ. TCT-639 NR4A3 as a gene expression marker of acute atherosclerotic plaque rupture in STEMI. J Am Coll Cardiol. 2013;62(18 Supplement 1):B194.

    Article  Google Scholar 

  38. 38.

    Gabrielsen A, Lawler PR, Yongzhong W, Steinbrüchel D, Blagoja D, Paulsson-Berne G, et al. Gene expression signals involved in ischemic injury, extracellular matrix composition and fibrosis defined by global mRNA profiling of the human left ventricular myocardium. J Mol Cell Cardiol. 2007;42(4):870–83.

    CAS  PubMed  Article  Google Scholar 

  39. 39.

    Haubner BJ, Adamowicz-Brice M, Khadayate S, Tiefenthaler V, Metzler B, Aitman T, et al. Complete cardiac regeneration in a mouse model of myocardial infarction. Aging. 2012;4(12):966.

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  40. 40.

    Jiang Y, Feng Y-P, Tang L-X, Yan Y-L, Bai J-W. The protective role of NR4A3 in acute myocardial infarction by suppressing inflammatory responses via JAK2-STAT3/NF-κB pathway. Biochem Biophys Res Commun. 2019;517(4):697–702.

    CAS  PubMed  Article  Google Scholar 

  41. 41.

    Maxwell MA, Muscat GE. The NR4A subgroup: immediate early response genes with pleiotropic physiological roles. Nucl Recept Signal. 2006;4(1):nrs.04002.

    Article  CAS  Google Scholar 

  42. 42.

    Meares GP, Ma X, Qin H, Benveniste EN. Regulation of CCL20 expression in astrocytes by IL-6 and IL-17. Glia. 2012;60(5):771–81.

    PubMed  Article  Google Scholar 

  43. 43.

    Beider K, Abraham M, Begin M, Wald H, Weiss ID, Wald O, et al. Interaction between CXCR4 and CCL20 pathways regulates tumor growth. PloS ONE. 2009;4(4):e5125.

    PubMed  PubMed Central  Article  CAS  Google Scholar 

  44. 44.

    Hosokawa Y, Shindo S, Hosokawa I, Ozaki K, Matsuo T. IL-6 trans-signaling enhances CCL20 production from IL-1β-stimulated human periodontal ligament cells. Inflammation. 2014;37(2):381–6.

    CAS  PubMed  Article  Google Scholar 

  45. 45.

    Lin C-F, Su C-J, Liu J-H, Chen S-T, Huang H-L, Pan S-L. Potential effects of CXCL9 and CCL20 on cardiac fibrosis in patients with myocardial infarction and isoproterenol-treated rats. J Clin Med. 2019;8(5):659.

    CAS  PubMed Central  Article  PubMed  Google Scholar 

  46. 46.

    Safa A, Rashidinejad H, Khalili M, Dabiri S, Nemati M, Mohammadi M, et al. Higher circulating levels of chemokines CXCL10, CCL20 and CCL22 in patients with ischemic heart disease. Cytokine. 2016;83:147–57.

    CAS  PubMed  Article  Google Scholar 

  47. 47.

    Herrmann M, Stanić B, Hildebrand M, Alini M, Verrier S. In vitro simulation of the early proinflammatory phase in fracture healing reveals strong immunomodulatory effects of CD146-positive mesenchymal stromal cells. J Tissue Eng Regen Med. 2019;13(8):1466–81.

    CAS  PubMed  Article  Google Scholar 

  48. 48.

    Song J, Wu C, Korpos E, Zhang X, Agrawal SM, Wang Y, et al. Focal MMP-2 and MMP-9 activity at the blood-brain barrier promotes chemokine-induced leukocyte migration. Cell Rep. 2015;10(7):1040–54.

    CAS  PubMed  Article  Google Scholar 

  49. 49.

    McMorrow JP, Murphy EP. Inflammation: a role for NR4A orphan nuclear receptors? Biochem Soc Trans. 2011;39(2):688–93.

    CAS  PubMed  Article  Google Scholar 

  50. 50.

    Prince LR, Prosseda SD, Higgins K, Carlring J, Prestwich EC, Ogryzko NV, et al. NR4A orphan nuclear receptor family members, NR4A2 and NR4A3, regulate neutrophil number and survival. Blood J Am Soc Hematol. 2017;130(8):1014–25.

    CAS  Google Scholar 

  51. 51.

    Liu H, Liu P, Shi X, Yin D, Zhao J. NR4A2 protects cardiomyocytes against myocardial infarction injury by promoting autophagy. Cell Death Discov. 2018;4(1):1–11.

    PubMed  Google Scholar 

  52. 52.

    Croker BA, Mielke LA, Wormald S, Metcalf D, Kiu H, Alexander WS, et al. Socs3 maintains the specificity of biological responses to cytokine signals during granulocyte and macrophage differentiation. Exp Hematol. 2008;36(7):786–98.

    CAS  PubMed  PubMed Central  Article  Google Scholar 

Download references


The authors would like to acknowledge the authors of primary study, the Topol group.


This research was supported by the Cultivation Project of Young and Innovative Talents in Universities of Shandong Province and by NSFC Grant 11501331. Support of T.W., J.M.S. and S.P.M. from NIH R01 NS096186.

Author information




YW, DM and TW did the overall design of the study. JMS was the language editor for the manuscript. CL contributed to the igraph figures. All authors were involved in the analysis of the data. All authors read and approved the final manuscript.

Corresponding authors

Correspondence to Yan-hui Wang, Da-zhi Meng or Ting Wu.

Ethics declarations

Ethics approval and consent to participate

Not required. The data are available in the NCBI database.

Consent for publication

Not required.

Competing interests

No competing interests.

Additional information

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary Information

Additional file 1: Figure S1.

A CCL20 cluster genes. B NR4A3 cluster genes. Figure S2. Protein-protein connection of DCGs by STRING database.

Additional file 2: Table S1.

Power-law indexes of degree distribution for the control network in the discovery and validation cohorts. Table S2. Clustering coefficients of DCGs in the discovery cohort, the validation cohort, and the discovery + validation cohort.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit The Creative Commons Public Domain Dedication waiver ( applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Wang, Yh., Li, Cx., Stephenson, J.M. et al. NR4A3 and CCL20 clusters dominate the genetic networks in CD146+ blood cells during acute myocardial infarction in humans. Eur J Med Res 26, 113 (2021).

Download citation


  • Acute myocardial infarction (AMI)
  • CD146
  • Pearson network
  • Clustering coefficient
  • Differential connectivity genes (DCGs)