The functions and prognostic values of chemokine and chemokine receptors in gastric cancer

Abstract

Chemokine and chemokine receptors (CCRs) play a significant role in tumor infiltration of immune cells, tumor angiogenesis and distant metastasis. In this study, we explored the importance of CCRs in gastric cancer (GC) by analyzing the datasets from TCGA database. First, we analyzed the characteristics of the CCRs mutations. Then, we screened the differentially expressed CCRs and performed GO functional annotation and KEGG pathway analyses to explore their potential biological functions. Using multivariate Cox regression analyses, we constructed a prediction model based on four-CCRs (CCL15, CCL21, CCR3 and ACKR3) signature, and we found that the risk score of the model was an independent prognostic factor of GC. Next, a nomogram was constructed to assess the prognosis of GC patients. GSEA indicated that the high-risk group was significantly enriched in immune response and immune system process. Moreover, GSVA was employed to investigate the up- and down-regulated signaling pathways in the high- and low-risk groups. The correlation between risk score and immune-cell infiltration indicated that the four-CCRs signature might play a pivotal role in GC immune microenvironment. In conclusion, we revealed the potential molecular mechanisms of CCRs in GC and constructed a prediction model which might guide personalized treatment and prognosis for GC patients.

Introduction

Gastric cancer (GC) is the third leading cause of cancer deaths around the world [1,2]. About two-third GC patients were diagnosed as advanced GC, and many patients will have local recurrence and distant metastasis even after receiving the radical resection [3]. The 5-year survival rate of advanced GC patients is only about 5%-20%, with 10 months of the overall survival (OS) [4,5]. Although some biomarkers related to the tumorigenesis and prognosis of GC have been evaluated [6], the methods that could accurately predict the prognosis of GC patients remain limited. Therefore, it is an urgent need to develop predictive models for the improved prognosis and treatment of GC patients.

Chemokines are small, secreted chemoattractant molecules that can regulate immune and inflammatory response, cell migration, proliferation and survival, which play an important role in various biological and pathological processes including cancer [7]. So far, more than 50 chemokines and 20 chemokine receptors have been identified [8]. According to the primary structure of the peptide chain, chemokines are classified into 4 subfamilies: CC, CXC, XC, and CX3C. The typical chemokine receptors are also divided into four subfamilies based on their binding ligand specificity: CCR (CCR1-CCR11 and CXCR (CXCR1-CXCR6)) [9]. Recently, some atypical chemokine receptors (ACKRs), e.g., ACKR1/2/3/4, have also been identified [10]. Unlike the typical chemokine receptors, ACKRs cannot trigger the canonical G protein mediated signaling, but they can combine with typical cytokines and regulate the chemokine network [11].

Many studies have revealed that tumor and its microenvironment (TME) contain a complex chemokine network [12]. Chemokine and chemokine receptors (CCRs) may affect the degree of immunocyte infiltration and phenotype [13-15], angiogenesis [16,17], tumor cell growth [18,19], metastasis [20,21], and survival [22], which, in turn, affect the patient’s prognosis. For example, CC and CXC (e.g., CCL2, CCL5, CXCL1 and CXCL5) chemokines can recruit monocytes and neutrophils to the tumor microenvironment where monocytes and neutrophils can promote or inhibit tumor progression [13-15,23]. CXCL12 and CCL2 have been found to promote angiogenesis in ovarian cancer [24,25]. CCL28 can modulate MAPK/ERK pathway to promote breast cancer cell growth and metastasis [21]. ACKR3 promotes epithelial to mesenchymal transition (EMT) and tumor growth by mediating TGF-β1 function in lung cancer [26]. Indeed, some monoclonal antibodies and chemokine receptor inhibitors have already been used in the clinical practice for malignant tumor treatment. For example, CXCR4 antagonist (AMD3100) is used to treat relapsed or refractory acute myeloid leukemia (AML) [27]. CCL2 inhibitor (CNTO 888) is used to treat metastatic prostate cancer [28,29], and anti-CCR4 monoclonal antibody (Mogamulizumab) is used to treat T-cell leukemia-lymphoma [30]. They have shown some therapeutic effects, and more clinical studies targeting CCRs are in progress. Nevertheless, although some CCRs (e.g., CCR5 and its ligand) were found differentially expressed in GC, and their expression is associated with a poor prognosis [31], the role of most CCRs in GC progression has not been reported. Therefore, our current studies aimed to identify the potential biological function and signature of CCRs to improve prognostic evaluation of GC. By multivariate Cox regression analysis, a novel four-CCRs (CCL15, CCL21, CCR3 and ACKR3) signature was constructed.

Methods

Data extraction and data processing

Genetic mutation data, transcriptome data, and clinical data of GC samples were obtained from the TCGA database (https://portal.gdc.cancer.gov/). Mutation data were visualized using the “maftools” package in R (4.1.0).

CCRs expression in GC

Differentially expressed CCRs between tumor and normal samples were analyzed and visualized using the “limma” and “pheatmap” package in R (4.1.0). The threshold of adjusted false discovery rate was <0.05 with |log2 (fold change)| >1. P-values less than 0.05 were considered statistically significant. Spearman correlation analysis was used to evaluate the correlation of CCRs expressions. P<0.05 was selected as statistically significant.

Enrichment analysis of CCRs

To explore the functional mechanisms of CCRs in GC, we performed GO functional annotation and KEGG pathway enrichment analysis in R (4.1.0) by employing the packages “GO plot”, “Cluster Profiler”, “ggplot2”, “DOSE” and their dependency packages. The enriched GO terms, KEGG pathways and PPI network were further annotated and visualized by Metascape [32].

Establishment and verification of the prognostic model

First, Kaplan-Meier analysis was performed to screen CCRs with prognostic values. Then, multivariate Cox regression was performed to construct CCRs-related prognostic signature and calculate the risk score of each patient. Risk scores were generated based on the expression of genes multiplied by a linear combination of regression coefficients obtained from the multivariate Cox regression analysis (P<0.01 was considered statistically significant). The formula used in CCRs model was: risk score=Σ ⁿ _i=1 vi × ci. (n: the quantity of independent indicators, vi: the expression quantity of gene, ci: the regression coefficient of gene i in multivariate Cox regression analysis). All GC patients were divided into two groups by the median risk score. We plotted K-M survival curve which was tested by log rank to portion statistical significance for evaluating the discrepancy of OS between the two groups. Besides, we plotted ROC curves by the “survival ROC” package in R (4.1.0) to verify the accuracy of this model. Principal component analysis (PCA) was conducted based on the “prcomp” package in R (4.1.0). To verify whether the risk score could be used as an independent prognostic factor for GC, univariate and multivariate analyses were performed by adopting the Cox regression method. A nomogram was built to investigate the probability of prognosis in GC patients.

Immunohistochemistry

The immunohistochemistry images for the expression of ACKR3, CCR3, and CCL21 in GC tissues and normal tissues were downloaded from the Human Protein Atlas (https://www.proteinatlas.org/).

RNA extraction and real-time PCR

Total RNA was extracted by Trizol reagent (Invitrogen, Carlsbad, CA, United States) according to the manufacturer’s protocol. Subsequently, the extracted RNA was reverse transcribed using PrimeScript RT reagent Kit with gDNA Eraser (Takara, Japan). The cDNAs were subjected to SYBR Green-based real-time PCR analysis. The primers used in real-time PCR assays were listed in Table S1.

Correlation between the risk score and immune cell infiltration

The Tumor Immune Estimation Resource (TIMER) database [33] was employed to analyze the correlation between the risk score and tumor-infiltrating immune cells (B cells, CD4⁺ T cells, CD8⁺ T cells, neutrophils, macrophages and dendritic cells). The abundance of the six immune infiltrates was estimated by the TIMER algorithm in R (4.1.0).

GSEA and GSVA functional annotation

We performed GSEA by using the software Gene Set Enrichment Analysis (GSEA) v4.0.3. The Gene Set Variation Analysis (GSVA) analysis was performed using “GSVA” package in R (4.1.0) using C2 gene sets (KEGG) from MSigDB (http://www.gsea-msigdb.org/gsea/index.jsp).

Statistical analysis

All statistical analyses were performed by R (version 4.1.0). P<0.05 was considered statistically significant except for special notes.

Results

Ccrs mutation analysis in GC samples

The incidence of copy number variations and somatic mutations of 63 CCRs in GC samples were summarized in Figure 1. Among the 132 samples we examined, 82 samples contained mutations of CCRs, with the frequency of 62.88%. The top 30 frequently mutated genes were shown in the waterfall plot, and the top 10 commonly mutated genes were CX3CL1, XCR1, CCR1, CXCR1, CXCR2, CCR3, CCR5, CCR7, CXCL16 and ACKR4. Among them, CX3CL1 and XCR1 were the highest mutated genes with more than 5% mutation rates (Figure 1A). Besides, the three most common mutation types were missense mutation, followed by frameshift deletions and frameshift insertions (Figure S1A). In addition, we found that there were some co-occurring mutations (e.g., CXCR6/CCR9, CCL1/CXCR1, ACKR1/CCR7 and ACKR4/CXCL16) of CCRs in GC samples (Figure 1B).

Figure 1.

Mutation profile analysis of CCRs by using the TCGA datasets. A. Oncoplot displaying the somatic landscape of the top 30 CCRs in GC samples from TCGA database. B. The co-occurring mutations and the mutual exclusivity analysis of the top 30 CCRs. Co-occurrence, green; Exclusion, brown.

Expression of CCRs between GC and normal tissues

Since genetic variations could be the prominent factor influencing the expression of CCRs, we further compared the differential expression of 63 CCRs between cancer and normal tissues (Figure 2A). We found that 41 CCRs were differentially expressed between GC and normal tissues, and the expression of the majority of these genes were up regulated in GC (Figure 2B). The most differentially expressed 24 CCRs (P<0.01, |logFC| >1) were shown in (Table 1). Moreover, spearman correlation analysis was used to investigate the correlation between the 63 CCRs and GC, and we observed both positive and negative correlations between the expression of these genes and GC (Figure S2). The genes with the strongest positive correlation were CXCL10 and CXCL11 (R=0.92), and the genes with strongest negative correlation were CXCL16 and CCL14 (R=-0.27).

Figure 2.

Expression analysis of CCRs. A. The expression levels of CCRs in the TCGA datasets. Red indicates up-regulated genes; green indicates down-regulated genes. B. Boxplot displaying the differentially expressed CCRs in GC. Blue, normal. Red, tumor. Not significant (ns), P<0.05 (*), P<0.01 (**), and P<0.001 (***).

Table 1.

Expression of top 24 CCRs

gene	normalMean	tumorMean	logFC	P-value
CXCR2	3.77726971	1.112872577	-1.763055385	0.000365247
CXCL17	225.559568	73.02803964	-1.626986074	6.88E-05
ACKR1	25.204637	9.336958332	-1.432664629	4.25E-11
CCL21	118.318518	47.5927657	-1.313861685	1.02E-08
CCL14	0.98009266	0.411905058	-1.250606306	1.42E-07
CXCL12	17.373577	7.517509883	-1.208568054	1.03E-07
CCR10	1.56625757	0.688337184	-1.186134135	3.39E-06
CCR1	2.33214037	4.821027975	1.047686176	2.85E-05
CCL15	1.70143162	3.992240417	1.230449429	4.48E-06
CXCL16	28.7594958	72.58288985	1.335591117	3.20E-11
CXCL3	5.89865494	18.03349774	1.612221323	8.26E-07
CCL3	0.80032764	3.439417526	2.103501615	1.09E-09
CCL20	10.9999174	52.1466439	2.245081718	6.27E-05
CCL18	6.14431732	29.29875169	2.253514568	3.47E-05
CCL26	0.68780889	3.814495172	2.47141247	0.001664114
CXCL5	12.2286623	72.64480995	2.57059314	0.000747777
CCR8	0.08748828	0.585523355	2.742565112	6.26E-13
CXCL1	8.4535364	63.02696675	2.898342332	5.31E-11
CXCL6	0.50611615	4.143685123	3.033373974	3.59E-09
CXCL10	4.4314276	41.13843483	3.214643458	6.28E-10
CXCL11	1.00251453	10.61476438	3.404377294	5.41E-10
CXCL9	4.16285404	45.28027036	3.443237691	4.08E-11
CXCL8	4.26019156	49.03128271	3.524712298	4.08E-11
CCL7	0.03366889	0.777061946	4.528541442	7.91E-13

Functional annotation of the significant CCRs

To explore the biological functions and potential mechanisms of the CCRs, GO and KEGG pathway enrichment analyses were performed for the top 24 CCRs (Table 1) with the significant expression change in GC. The GO analysis showed that these CCRs were mostly enriched in chemokine-mediated signaling pathways (BP), external side of plasma membrane (CC), and the chemokine activities (MF) (Figure 3A). From the KEGG pathway analysis, we found that these differentially expressed CCRs were mainly associated with viral protein interaction with cytokine and cytokine receptor, chemokine signaling pathway and cytokine-cytokine receptor interaction (Figure 3B). Annotation of GO, KEGG pathways enrichment analysis and PPI network of these top 24 CCRs were also performed using “Metascape” website (Figure 3C; Figure S3). As shown in Figure 3D, 16 hub genes were identified in the protein - protein interaction network.

Figure 3.

Functional Annotation of the significant CCRs. (A) GO functional annotation and (B) KEGG pathway enrichment analysis of the top 24 CCRs. (C) Metascape website analysis. Nodes with the same cluster ID are typically close to each other, and the same color indicates the same cluster ID. (D) The key module of PPI network. 16 hub genes, which are labeled with red circles. The lines between the circles indicate the interaction among genes.

Construction of a prognostic risk model

To analyze the relationship between the expression of CCRs and the prognosis of GC, univariate Cox regression analysis was performed for the 63 CCRs. The results indicated that 7 CCRs were significantly correlated with the survival of GC patients (Figure 4A). In addition, multivariate Cox regression analysis was performed to construct a prognostic risk model using four genes (CCL15, CCL21, CCR3 and ACKR3) and to calculate the risk score for each patient (Table 2). Based on the median risk score, 368 GC patients were divided into high-risk group (n=184) and low-risk group (n=184). The survival overview and gene expression heatmaps of these two groups were presented in Figure 4B. By analyzing the expression data of all genes (Figure 4C) and the risk-related genes (Figure 4D), principal component analysis (PCA) was performed, and we found that our risk score could accurately distinguish the high- and low-risk groups. Kaplan-Meier analysis indicated that, compared with the low-risk group, the prognosis of the high-risk group was significantly worse (P<0.01) (Figure 4E). The AUC of the ROC curve was 0.825, suggesting our model had a good performance in predicting the prognosis of GC patients with high sensitivity and specificity (Figure 4F).

Figure 4.

Prognostic value of CCRs in GC patients. (A) Univariate Cox regression analysis between CCRs expression and OS. (B) The distribution of prognostic signature-based risk scores. Principal component analysis between low-risk and high-risk groups for all genes (C) and for risk genes (D). (E) Kaplan-Meier survival curve for the OS of GC patients. The red line represents the high-risk group, and the blue line represents the low-risk group. (F) AUC of time-dependent ROC curve analysis for evaluating the prognostic performance of the risk score for 5-year OS.

Table 2.

Four CCRs that constitute the CCRs-related prognostic model

Gene symbol	Coef	HR	HR.95L	HR.95H	P-value
CCL15	0.03720978	1.03791073	1.00674268	1.070043714	0.016759
CCL21	0.00077528	1.00077558	1.000095965	1.00145566	0.025298
ACKR3	0.03137368	1.03187102	1.009468142	1.054771079	0.005088
CCR3	0.65784562	1.93062854	1.17316883	3.177144213	0.009644

Validation and application of the prognostic risk model

To evaluate the independent prognostic ability of our prognostic model, univariate (Figure 5A; Table 3) and multivariate (Figure 5B; Table 4) Cox regression analyses were performed to explore the relationship between clinicopathological characteristics and patient prognosis. The results showed that the risk score could be used as an independent predictor of the prognosis of GC patients (P<0.001). To facilitate the clinical use of this model to obtain reliable prognostic information of each patient, we constructed a nomogram that could predict the 1-, 2- and 3-year OS probability of GC patients (Figure 5C).

Figure 5.

Validation and application of the prognostic risk model. A. Univariate Cox regression analysis. B. Multivariate Cox regression analysis. C. The nomogram for estimating the 1-, 2-, and 3-year overall survival of GC patients.

Table 3.

Univariate analyses of overall survival of GC patients by using TCGA datasets

Variables	HR	HR.95L	HR.95H	P-value
age	1.026965748	1.007829689	1.046465151	0.00556017
gender	1.483828302	0.979779065	2.247186647	0.062392212
grade	1.367875492	0.946580494	1.976676441	0.095375967
stage	1.535478376	1.221185829	1.930659353	0.000242497
T	1.297541887	1.023344886	1.645207762	0.031516711
M	2.048306027	1.096196896	3.827375896	0.024581693
N	1.267206892	1.068895108	1.5023114	0.006386528
riskScore	1.654379281	1.277259284	2.142846671	0.00013678

Table 4.

Multivariate analyses of overall survival of GC patients by using TCGA datasets

Variables	HR	HR.95L	HR.95H	P-value
age	1.038699482	1.017970559	1.059850508	0.000222768
gender	1.523418052	0.996272439	2.329485862	0.052048908
grade	1.496553445	1.018153473	2.199739307	0.040217331
stage	1.346751999	0.869066971	2.086997905	0.182848275
T	1.061399022	0.768680251	1.465587131	0.717388532
M	1.822690208	0.80192462	4.142782887	0.151850095
N	1.096270464	0.854173045	1.406985312	0.470335302
riskScore	1.664082404	1.286237132	2.152923577	0.000106385

Protein and mRNA expression of CCL15, CCL21, CCR3 and ACKR3

To reveal the protein expression profile of CCL15, CCL21, CCR3 and ACKR3 in normal tissues and GC tissues, we extracted the relevant data from the Human Protein Atlas (HPA) website. As shown in Figure 6A-F, compared with normal tissue, ACKR3 was highly expressed, whereas the expression of CCR3 was lower in GC tissues. The expression of CCL21 was not detected in both normal and GC tissues, possibly due to its low expression level. The protein expression data of CCL15 was inaccessible in HPA database; hence, CCL15 expression was not presented in this study. We also verified the mRNA expression of the four genes using our own clinical samples (paired gastric cancer tissues, n=22). The results showed that CCL15, CCL21, CCR3, and ACKR3 were expressed differentially between normal and tumor tissues (Figure 6G-J).

Figure 6.

Immunohistochemical and qPCR analysis of ACKR3, CCR3, and CCL21. The expression of ACKR3 was higher in GC tissues than in normal tissues (A, B). The expression of CCR3 was lower in GC tissues than in normal tissues (C, D). The expression of CCL21 was not detected in both GC tissues and normal tissues (E, F). All results were obtained from HPA database. (G-J) mRNA relative expression of genes in the risk model by the method of qPCR.

The risk score was related to immunocyte infiltration

As mentioned above, CCRs were closely related to inflammatory response and tumor immunity. To explore whether the risk score was related to immune cell infiltration in GC, correlation analysis was carried out by using the TIMER Database. We found that the risk score was positively correlated with the infiltration of CD4⁺ T cells, CD8⁺ T cells, neutrophils, macrophages and dendritic cells (P<0.001) (Figure 7A-E), except for B cells (Figure 7F).

Figure 7.

The correlation between risk scores and immune cell infiltrating. Neutrophils (A), CD4⁺ T cells (B), CD8⁺ T cells (C), dendritic cells (D), macrophages (E), and B cells (F) are revealed by scatter diagrams. The correlation is shown by Pearson correlation coefficient.

Functional enrichment analysis for different risk groups

GSEA was used to analyze the enrichment of high- and low-risk groups in two immune related gene sets. The results indicated that the high-risk group with poor survival was significantly enriched in these two immune related gene sets that included immune response and the process of immune system, while the low-risk group did not show such significant enrichment (Figure 8A). GSVA results indicated that, compared with the low-risk group, many metabolic pathways in the high-risk group were up-regulated or down regulated (Figure 8B, 8C). These metabolic pathways included cytokine and cytokine receptor interaction, cell adhesion molecules (CAMs) and RNA degradation. The dysregulation of these pathways may be one of the factors affecting the prognosis of GC.

Figure 8.

Differential signaling pathway enrichment in different risk groups. (A) gene sets of immune response and the process of Immune system in high-risk and low-risk groups analyzed by GSEA. (B, C) GSVA enrichment analysis showing the up and down regulated signaling pathways in different risk groups. Volcano plot (B), and Heatmap of differential expression in KEGG pathway (C). Red represents up regulated pathways, and green represents down regulated pathways.

Discussion

Tumor immune microenvironment plays a very important role in anti-tumor activity [34,35]. CCRs can affect the state of tumor immune microenvironment by regulating inflammatory response and immune cell infiltration [12,36-38]. GC is a solid tumor which consists of stromal cells, e.g., endothelial cells and fibroblasts, and is infiltrated by various immune cells, e.g., CD4⁺ T cells, CD8⁺ T cells, neutrophils, macrophages, and dendritic cells. All these cells are involved in chemokine production [39]. Although some CCRs, e.g., CCR5 and its ligand, have been found differentially expressed in GC, and their expression is associated with aggressive tumor behavior and a poor prognosis [31], the role of most CCRs in GC progression has not been reported. Therefore, a comprehensive analysis and the molecular characterization of CCRs in GC will advance our understanding of the antitumor immune response and prognosis of GC.

In our current studies, we first systematically analyzed the mutation frequency and CNV alteration of CCRs by using the information from TCGA database. We found that the expression of most CCRs was dysregulated in GC. Next, we selected the top 24 CCRs that exhibited significantly differential expression between normal and GC tissue for further investigation of their potential mechanisms and biological functions. As expected, these CCRs were significantly enriched in chemokine related gene sets and signaling pathways such as chemokine-mediated signaling pathway (BP), external side of plasma membrane (CC), chemokine activity (MF), viral protein interaction with cytokine (KEGG) and NF-kappa-B signaling pathway (KEGG). Consistent with our findings, it has recently been reported that the therapeutic concentration of 3,3’-Diindolylmethane (DIM) can upregulate the expression level of chemokines such as CCL-2 and activate NF-kappa-B signaling pathway in gastric cancer-derived mesenchymal stem cells, thus promoting the progression of GC [40]. In addition, we also conducted PPI analysis and found 16 hub genes that might play a key role in regulating the chemotaxis related molecular networks in GC.

Through multivariate Cox regression analyses, four CCRs (CCL15, CCL21, CCR3 and ACKR3) were identified to be significantly correlated with the OS of GC patients. Recently, there are studies showing the inflammatory response and pro-tumor or anti-tumor effect of these CCRs. For example, CCL15, acting mainly via CC chemokine receptor CCR1 [41], also binds to CCR3 [42] and can recruit immune cells, e.g., T-cells, monocytes, eosinophils, and neutrophils, to tumor microenvironment [42,43] to induce tumor angiogenesis [44]. It has also been reported that CCL15 is highly expressed in digestive tract and liver tissues and plays an important role in maintaining the balance of immune response in these organs. In tumor, it is found that the accumulation of CCR1⁺ tumor-associated neutrophils through CCL15-CCR1 axis can promote lung metastasis of colorectal cancer [45]. Moreover, CCL15 produced by the renal cell carcinoma cells that migrate to the bone has been shown to display a significant effect on osteoclastogenesis, and can promote the bone metastasis of tumor [46,47]. CCL21, as the ligand for CCR7, can activate T cells, preferentially naive T cells [48,49]. High expression of CCL21 in tumor can increase anti-tumor immune cell infiltration and prolong the survival of cancer patients [50,51]. However, once the tumor immune microenvironment has been completely dysregulated, CCL21 may play an opposite role [52,53] as CCL21 may recruit Treg into the tumor microenvironment, which results in tumor immune evasion. ACKR3 is the receptor for CXCL12 and CXCL11. Recent studies have shown that ACKR3 can promote tumor cell growth and metastasis [54,55] through modulating the mTOR pathway [56]. Increased expression of ACKR3 is a significant prognostic factor for the poor prognosis of patients with aggressive prostate carcinoma [57] and renal carcinoma [58]. Based on the four CCRs (CCL15, CCL21, CCR3 and ACKR3) described above, we established a four-CCRs signature-associated risk model, and our results suggested that the risk score was an independent prognostic predictor for GC patients. Importantly, our GSVA results revealed some signaling pathways that were significantly different between high- and low-risk groups, such as cytokine and cytokine receptor interaction, cell adhesion molecules (CAMs) and RNA degradation. The abnormalities of these pathways may be the cause for the difference in prognosis between high-risk and low-risk patients.

We further performed GSEA to determine the difference in immune responses between the high- and low-risk groups. Our results showed that the high-risk group was significantly enriched in two immune related gene sets including immune response and the process of immune system, while low-risk group showed no significant enrichment. This is in consistence with the fact that immune response plays an important role in anticancer immunity and is closely related to the prognosis of various human malignant tumors [59]. Tumor immune response involves many immune cells, and the role of immune cells is very complex and diverse. For example, neutrophils can promote the metastasis of tumor cells to distant organs by interacting with circulating tumor cells [60], and macrophages can promote EMT process to enhance tumor cells migration and invasion [61]. To understand the correlation between the risk score of our model and immune-cell infiltration, we conducted a correlation analysis based on the TIMER database. Our results indicated that the infiltration of CD4⁺ T cells, CD8⁺ T cells, neutrophils, macrophages and dendritic cells was positively associated with the risk score, suggesting that a high infiltration level of these immune cells might increase the mortality risk of GC patients. However, in contrary to our results, high level of CD4⁺ and CD8⁺ cells infiltration in tumor has been reported as a good prognostic factor. We speculated that this contradiction was caused by the other functions of CCL21; the high expression of CCL21 might recruit Treg and lead to tumor immune escape.

Our research constructed a four-CCRs (CCL15, CCL21, CCR3 and ACKR3) signature prognostic risk model and provided new insights into the GC immune microenvironment. However, there are some limitations in this study. Since this was a retrospective study, the results need to be further validated by large-scale prospective studies. Besides, the predictive power of this model in GC requires further testing for better prognostic stratification and treatment management. Moreover, the biological functions of these four CCRs need further experimental verification.

Conclusion

Through a series of comprehensive analyses, we identified the potential biological function and prognostic value of CCRs in GC. We further constructed a novel prognosis model for GC patients. We found that the risk score of this model could reflect the immune microenvironment status of GC patients. The current studies shed some light on our understanding of CCRs in GC and provide novel potential prognostic and therapeutic biomarkers.

Disclosure of conflict of interest

None.

Abbreviations

TCGA

The Cancer Genome Atlas

CCRs

Chemokine and chemokine receptors

GC

gastric cancer

GO

Gene Ontology

CC

Cellular Component

MF

Molecular Function

BP

Biological Process

ROC

receiver operating characteristic

OS

overall survival

GSVA

Gene Set Variation Analysis

AUC

area under the curve

KEGG

Kyoto Encyclopedia of Genes and Genomes

GSEA

Gene Set Enrichment Analysis