The role of CD47 in immune escape of colon cancer and its correlation with heterogeneity of tumor immune microenvironment

Clinical data collection and screening

A total of 96 COAD samples were collected between January 2021 and July 2022 at Jiangjin Central Hospital Chongqing University. The study was approved by the hospital’s Ethics Committee (approval number: KY2022037, TYYLKYJJ-2023-014), and all participants provided written informed consent prior to inclusion.

The inclusion criteria for the study were: patients with a first-time diagnosis of COAD and no prior history of related malignancies or previous surgeries. (1) Patients with no previous gastrointestinal endoscopy and initial colonoscopy biopsy diagnosed with COAD; (2) Patients with previous gastrointestinal endoscopy, history of colon polyps and regular follow-up progression to colon cancer. All patients underwent histopathological confirmation of COAD. Detaild clinical characteristics of the participants are presented in Table S1.

Immunohistochemistry (IHC)

All COAD samples underwent standard histopathological processing, including fixation in 10% neutral formalin, dehydration, paraffin, embedding, and sectioning at a thickness of 3 µm. The sections were dried and pretreated using the DAKO PT Link automated immunohistochemical preconditioning system. Immunohistochemical staining was carried out with the DAKO AutostainerLink 48 automated system, employing the EnVision method. The primary antibody used was anti-CD47 (1:200, CST, 63000S). Tonsillar tissue was utilized as a positive control for external validation, while phosphate-buffered saline (PBS) was used as a negative control in place of the primary antibody. In addition, red blood cells within the blood vessels served as an internal control.

Interpretation of the IHC results

CD47 expression was predominantly localized to the cell membrane and cytoplasm of tumor cells. To assess CD47 staining in COAD samples, two independent and experienced pathologists performed a double-blind evaluation under high-power microscopy (400x magnification). The staining was scored based on both the intensity and the proportion of positive cells. Staining intensity for the cell membrane and cytoplasm was graded on a scale of 0 to 3: 0 points for no staining, 1 point for 1–25% positive staining, 2 points for 26–50%, and 3 points for 51–100%.

CD47 expression and the clinical dataset

Data on CD47 mRNA expression and clinical information for COAD patients were retrieved from The Cancer Genome Atlas (TCGA) database (https://cancergenome.nih.gov/). These data were processed using R software (v3.6.3) and Perl(version 5.3.0), and the deadline was March, 2023. The dataset included 480 COAD patients, of whom 41 had both normal and cancerous colon tissue samples, while the remaining 439 samples contained only cancerous tissue. Patients were divided into high and low CD47 expression groups based on the median CD47 mRNA expression level. ROC analysis of the data was performed using ROC (1.18.0) and the results were visualized with ggplot2 (3.3.6), correcting the outcome order of the data by default (ensuring the results are upward convex).

In addition, The GSE33113 dataset was obtained from the Gene Expression Omnibus (GEO) (https://www.ncbi.nlm.nih.gov/geo/) database via GEOquery (2.64.2). GSE33113 was located on the GPL570 platform ((HG-U133_Plus_2) Affymetrix Human Genome U133 Plus 2.0 Array). This dataset comprised primary tumors from 90 stage II CRC patients, along with 6 matched normal colon tissue samples. From this dataset, From this dataset, CD47 gene expression data were extracted for analysis. The study design and human data used in this analysis are presented in Table S2.

Survival analysis and clinical correlation analysis

The samples were stratified into high and low expression groups based on median CD47 gene expression levels. Kaplan–Meier survival curve analysis, along with data from the GEPIA database, was used to evaluate the correlation between CD47 expression and survival outcomes in COAD patients. Additionally, the association between CD47 expression, clinical parameters, and patient prognosis was further investigated using univariate and multivariate Cox regression analyses. Hazard ratios (HR) were calculated to assess the relative risk linked to CD47 expression. The multivariate or univariate Cox regression models were performed using R package “survival”. Logistic regression analysis was used to explore the correlation of CD47 expression with clinical factors of COAD using the R package “Limma”.

Multiplex immunofluorescence staining

Multiplex immunofluorescence staining was performed according to previously published protocols. The primary antibodies used included anti-CD47 (rabbit, 1:200, CST, 63000S), anti-CD68 (rabbit, 1:3000; AiFang Biological, Changsha, China), and anti-CD163 (rabbit, 1:3000, Proteintech, China). Signals were detected using horseradish peroxidase-conjugated secondary antibodies (GB23301, GB23303; Servicebio, Beijing, China), followed by tyramide signal amplification with FITC-TSA, CY3-TSA, 594-TSA, and 647-TSA (Servicebio). Multispectral images were acquired and analyzed using CaseViewer (C.V 2.3, C.V 2.0) and Panoramic Viewer (P.V 1.15.3) software to quantify positive cells at the single-cell level.

Functional enrichment analysis

To identify genes associated with CD47 expression, RNAseq data from the TCGA database were analyzed. Genes positively correlated with CD47 expression, displaying a —log2(fold change)— ≥ 2 and a corrected P-value <0.05, were selected for further analysis. Enrichment analysis of the 2,486 CD47 co-expressed genes was conducted using the “clusterProfiler” package. To explore the associated biological processes, molecular functions, and cellular localization, we employed the “ggplot2” package for visualization. In addition, KEGG pathway enrichment analysis was performed to elucidate the signaling pathways in which these genes are involved in COAD.

GSEA enrichment

To assess the impact of differential CD47 expression on cell signaling pathways, we analyzed the differentially expressed genes between the CD47 high- and low-expression groups. GSEA was performed using the KEGG.v7.4 dataset with default reference settings. An absolute net enrichment score (NES) >1, false discovery rate (FDR) q < 0.25, and P value < 0.05 were considered to represent significant enrichment.

Expression of CD47 in each cell in the COAD tumor microenvironment (TME)

The TISCH database (http://tisch.comp-genomics.org/) is a comprehensive resource integrating single-cell sequencing data from 27 different types of cancer. Based on MAESTRO v.1.1.0 , all collected datasets were processed uniformly through a standardized workflow including quality control, batch effect removal, cell clustering, differential expression analysis, cell type annotation, malignant cell classification, and GSEA. batch effects were corrected for if the median entropy was below 0.7 using Seurat v.3.1.2 . For each cluster, TISCH used the Wilcoxon test to identify differentially expressed genes (DEGs) based on log-transformed fold change (—logFC—>=0.25) and annotated the cell clusters using the marker-based annotation method employed in MAESTRO. Marker genes for each cell type were collected from published resources . For this study, we utilized the CRC_GSE139555 and CRC_GSE136394 datasets to investigate the expression of CD47 in individual cells within the TME of COAD. The results were visualized using a heatmap to present the expression patterns.

Analysis of tumor immune infiltration and macrophage polarization

We utilized the TIMER database (http://cistrome.shinyapps.io/timer) to perform molecular characterization of tumor-immune interactions. The correlation between CD47 expression levels and the abundance of six distinct tumor-infiltrating immune subpopulations (CD8 ⁺ T cells, CD4 ⁺ T cells, macrophages, B cells, dendritic cells, and neutrophils) was calculated, P < 0.05 was considered statistically significant.

Correlation analysis of biomarkers and treatment response

The Gene Set Cancer Analysis (GSCALite) (bioinfo.life.hust.edu.cn/web/GSCALite/) server (Liu et al., 2018) was employed to analyze the gene expression profile of CD47 and calculate the area under the dose–response curve (AUC) values across various cancer cell lines. To evaluate the treatment response to CD47 and overall survival (OS) in different cancer types, we compared seven standardized biomarkers using the biomarker assessment module of the TIDE server. These biomarkers included T cell clonality (T.clonality), B cell clonality (B.clonality), TIDE, microsatellite instability (MSI) score, tumor mutation burden (TMB), CD274 (PD-L1), and interferon-gamma (IFNG) (Fu et al., 2020). Their correlations with tumor immune response were analyzed to assess predictive value.

Statistical analysis

All experimental data were analyzed using SPSS 22.0 statistical software. For dichotomous data involving clinicopathological parameters, the chi-square test was applied, with continuity correction when expected counts were less than 5. Quantitative data were compared using the independent sample t-test. Factors found to be significant in univariate analysis were further evaluated by multivariate Cox regression to identify independent risk factors for survival. A P-value of <0.05 was considered statistically significant.

The mRNA expression of the CD47 gene in COAD

Using the TCGA database, we analyzed CD47 gene expression across various human cancers and their corresponding normal tissues. Our analysis revealed significantly elevated CD47 expression in nine cancer types, including colon cancer, cervical squamous cell carcinoma, bile duct carcinoma, esophageal cancer, head and neck squamous cell carcinoma, LIHC, metastatic cutaneous melanoma, gastric adenocarcinoma, and thyroid cancer (Fig. 1A). In comparisons of unpaired samples, tumor tissues consistently exhibited markedly higher CD47 levels than normal tissues (Fig. 1B). This finding was further confirmed by a paired analysis of 41 normal tissues and their corresponding tumor samples, which demonstrated significantly increased CD47 expression in cancerous tissues (P < 0.001) (Fig. 1C). Additionally, cross-validation with the GEO dataset (GSE33113) corroborated our results, showing heightened CD47 expression in COAD tissues compared to normal tissues (Fig. 1D). Receiver operating characteristic (ROC) curve analysis highlighted CD47’s strong diagnostic potential, with high sensitivity (0.818) and specificity (0.829) for predicting patient outcomes (AUC = 0.862) (Fig. 1E).

Expression of CD47 in tissue samples of COAD patients

CD47 protein expression was evaluated in 90 COAD tissue samples. The results showed that CD47 was predominantly localized to the cell membrane and/or cytoplasm of tumor cells, with a positivity rate of 91.11% (80/90). Figure 2A presents the variation in CD47 staining intensity as assessed by IHC. Statistical analysis revealed significantly higher CD47 IHC scores in T3-4 tumors compared to T1-2 tumors (Fig. 2B). Additionally, CD47 expression was significantly elevated in N1/2 COAD patients compared to N0 patients (Fig. 2C). Furthermore, patients with TNM stage III/IV disease exhibited significantly higher CD47 expression than those with TNM stage I/II disease (Fig. 2D).

Correlation between CD47 expression and the clinical characteristics of COAD patients

In this study, high CD47 expression was defined by an IHC score of 3 or higher. We examined the relationship between CD47 expression levels in tumor tissues and various clinicopathological factors, including gender, age at diagnosis, and tumor stage (T, N, and overall clinical stage). CD47 expression was significantly associated with both the N stage (P < 0.001) and the overall clinical stage of COAD (P < 0.001). However, no significant correlation was observed between CD47 expression and gender, age, or the T stage, as shown in Table 1.

To further validate our findings, we analyzed clinicopathological data from COAD patients in the TCGA database. This analysis confirmed the positive correlation between CD47 expression and the N stage (Fig. 3A) and T stage (Fig. 3B). However, no significant association was observed with the M stage (Fig. 3C), pathological stage (Fig. 3D), or the presence of colon polyps (Fig. 3E). Additionally, CD47 expression was significantly higher in patients with disease progression (PD), partial response (PR), and complete response (CR) compared to normal tissues. In contrast, there was no significant difference in CD47 expression between the stable disease (SD) group and normal tissues (Fig. 3F).

Analysis of the independent prognosis

The relationship between CD47 expression and the prognosis of COAD patients was evaluated using the Kaplan–Meier method. Notably, high CD47 expression was associated with longer progression-free survival (PFI) compared to low CD47 expression. However, no significant differences were observed in overall survival (OS) or disease-specific survival (DSS) (PFI: HR = 0.57, 95% CI [0.30–1.11], P = 0.025; OS: HR = 0.82, 95% CI [0.37–1.81], P = 0.144; DSS: HR = 0.9, 95% CI [0.35–2.29], P = 0.093) (Figs. 4A, 4B, 4C).

To further examine the impact of clinicopathological features on patient outcomes, we performed univariate and multivariate independent prognostic analyses. Univariate analysis identified CD47 expression and TNM stage as significant prognostic factors (P < 0.05), while TP53 was not significantly associated with prognosis (P = 0.256). Interestingly, multivariate analysis demonstrated that high CD47 expression had an HR below 1.0, indicating a favorable prognosis (Fig. 4D and Table 2), which contrasts with our earlier findings. Given the observed correlation between high CD47 expression and advanced TNM stage, it is perplexing that high CD47 expression would be linked to longer PFI in COAD patients. To investigate this discrepancy, we explored the immunophenotypic and functional differences between M1 and M2 tumor-associated macrophages (TAMs), focusing on the relationship between CD47 expression and M1 macrophage infiltration. Our analysis revealed a positive correlation between CD47 expression and the M1 macrophage markers CD86 (Spearman’s R = 0.356, P < 0.001), CXCL9 (Spearman’s R = 0.36, P < 0.001), and CD80 (Spearman’s R = 0.368, P < 0.001) (Figs. 4E–4G).

Based on these findings, we speculate that the prognostic significance of CD47 may be diminished in COAD patients with increased infiltration of M1 polarized macrophages. However, further studies are needed to clarify CD47’s role in COAD and to better understand its correlation with the heterogeneity of the tumor immune microenvironment.

Immune infiltration analyses in COAD

The OS rates in COAD were strongly influenced by tumor immune infiltration. Analysis of 24 different immune cell subtypes showed significant increases in T cells (P = 0.001), neutrophils (P < 0.001), macrophages (P < 0.001), cytotoxic cells (P = 0.006), and activated dendritic cells (aDCs) (P < 0.001) in the high CD47 expression group (Fig. 5A).

The CD47-SIRPα signaling pathway has been reported to negatively regulate macrophage phagocytosis. SIRPα expression on TAMs inhibits intracellular FcγR signaling, resulting in reduced phagocytic activity and allowing tumor cells to evade immune detection (Murata et al., 2014; McCracken, Cha & Weissman, 2015; Gül & Van Egmond, 2015). To explore this further, we examined the correlation between CD47 expression and macrophage infiltration levels. Our analysis revealed a positive correlation between CD47 expression and M1 macrophages (r = 0.312, P <  0.001), as well as total macrophage infiltration (r = 0.237, P < 0.001). In contrast, a negative correlation was observed between CD47 expression and M2 macrophages (r = −0.298, P < 0.001) (Fig. 5B). Additionally, CD47 expression showed a significant negative association with NK CD56bright cells (r = 0.170, P < 0.001) and plasmacytoid dendritic cells (pDCs) (r = 0.119, P <  0.001) (Fig. 5C).

Multiplex immunofluorescence analysis

To gain a deeper understanding of CD47’s role in COAD, we performed multiplex immunofluorescence staining to assess the co-expression of CD47 with M1 macrophage marker CD86 and SIRPα in COAD tissues. Six COAD specimens, including adjacent non-tumor tissues, were analyzed to visually assess CD47 expression. Immunofluorescence labeling was conducted with CD86 (red), CD47 (green), SIRPα (pink), and DAPI (blue). The results showed that CD47, SIRPα, and CD86 were co-expressed in COAD tissues, with significantly higher expression levels of CD47 and SIRPα compared to CD86 (Fig. 6). Further confirmation of intercellular interactions within the COAD TME was obtained through multiplex IHC, which revealed that CD47 exerts immunosuppressive effects primarily via the CD47/SIRP α axis. These findings suggest that targeting CD47 with specific blockers may enhance macrophage-mediated phagocytosis of tumor cells, thereby inhibiting tumor progression in COAD (Gu et al., 2018).

Correlation between CD47 and the tumor immune microenvironment heterogeneity

We utilized two datasets from the TISCH database, CRC_GSE139555 and CRC_GSE136394, to examine CD47 expression in immune cells within the TME (Fig. 7A). The CRC_GSE139555 dataset contains data from primary CRC patients, while the CRC_GSE136394 dataset focuses primarily on metastatic tumors (Wu et al., 2020; Lu et al., 2019). In the metastatic CRC_GSE136394 dataset, only four cell types were identified: CD4Tconv, CD8T, Mono/Macro cells, and Tprolif, with CD8T cells being the most abundant (Fig. 7B). In contrast, the CRC_GSE139555 dataset revealed 11 distinct cell types (Fig. 7C). Figures 7D–7G, visually present the distribution of these immune cells, highlighting significant differences in cellular composition and distribution between primary and metastatic CRC. The expression levels of CD47 were found to vary across different cell types, and these cellular components differed between primary and metastatic patients. This variation may explain the heterogeneity observed within the CRC TME.

CD47 participates in tumor immune evasion through different T-cell rejection

Among the 22 TCGA cancer types and subtypes assessed for tumor immune infiltration, a significant positive correlation was observed between CD47 expression and immune cell infiltration in three cancer types: COAD, kidney renal clear cell carcinoma (KIRC), and LIHC (Fig. 8A). Specifically, LIHC showed a strong correlation with immune cells (r = 0.28–0.47, P < 0.05), renal clear cell carcinoma exhibited a correlation of (r = 0.26–0.45, P < 0.05), and colorectal adenocarcinoma demonstrated a correlation of (r = 0.17–0.42, P < 0.05). These findings suggest a strong association between CD47 expression and immune cell infiltration in these cancers. Additionally, we analyzed the correlation between CD47 expression and four key immunosuppressive cell types known to promote T-cell rejection: myeloid-derived suppressor cells (MDSCs), cancer-associated fibroblasts (CAFs), regulatory T cells (Tregs), and M2 tumor-associated macrophages (M2-TAMs). CD47 expression was significantly correlated with these immunosuppressive cells in most cancer types, with the exception of adrenocortical carcinoma (ACC), cholangiocarcinoma (CHOL), renal chromophobe (KICH), pheochromocytoma and paraganglioma (PCPG), and rectal adenocarcinoma (PEAD). Notably, CD47 was positively correlated with MDSC and Treg infiltration, while showing a negative correlation with M2-TAM infiltration (Fig. 8B).

These results suggest that CD47 plays a role in T-cell rejection through multiple mechanisms. Furthermore, we compared CD47 with standardized biomarkers in terms of response prediction and OS biomarker correlations. Using 15 feature curves with an AUC greater than 0.5, we found that CD47 had a higher predictive value than tumor mutation burden (TMB) and T-cell clonality in seven and six immunotherapy subpopulations, respectively. Additionally, CD47 exhibited promising treatment outcomes in melanoma when combined with CTLA-4 inhibitor therapy (ICB_VanAllen2015) and PD-1 inhibitor therapy (ICB_Gide2019_PD1) (Fig. 8D).

Immune-related pathways regulated by CD47 in COAD

To explore the relationship between high CD47 expression and immune regulation, we performed KEGG, GO, and GSEA analyses. Initially, differential expression analysis identified genes associated with high CD47 expression in COAD (Fig. S1). KEGG pathway enrichment was then conducted to investigate the biological processes linked to elevated CD47 expression. The enriched KEGG pathways were primarily associated with oocyte meiosis, cellular senescence, viral protein interactions with cytokine and cytokine receptors, PPAR signaling, progesterone-mediated oocyte maturation, human T-cell leukemia virus 1 infection, p53 signaling, and chemokine signaling pathways (Fig. 9A). These results highlight a strong correlation between high CD47 expression and immune cell infiltration, as well as regulation of the immune microenvironment in COAD.

GO enrichment analysis revealed that the differentially expressed genes were involved in nine key biological processes: mitotic nuclear division, nuclear division, organelle fission, spindle formation, mitotic spindle organization, spindle microtubule assembly, microtubule binding, histone kinase activity, and tubulin binding (Fig. 9B). Furthermore, GSEA analysis showed that high CD47 expression was significantly associated with processes such as systemic lupus erythematosus (NES = 1.57, P = 0.00), graft-versus-host disease (NES = 1.60, P = 0.00), allograft rejection (NES = 1.53, P = 0.04), and several metabolic pathways, including sulfur metabolism (NES = −2.248, p = 0.00), glycan degradation (NES = −1.84, P = 0.00), steroid biosynthesis (NES = −1.93, P = 0.00), phenylalanine metabolism (NES = −2.11, p = 0.00), maturity-onset diabetes of the young (NES = −2.53, P = 0.01), pentose and glucuronate interconversions (NES = −1.87, P = 0.02), and the citrate (TCA) cycle (NES = −1.71, P = 0.02) (Figs. 9C–9D).