We conducted a retrospective study and recruited 103 COAD patients who underwent radical resection surgery of COAD in Nantong University Affiliated Hospital between 2010 and 2013. The clinicopathological and follow-up data of all patients were obtained and summarized in Table 1 . All specimens of COAD were examined by two independent pathologists and according to the AJCC-stage version 7 system. The overall survival (OS) was defined as the time period between surgery and death or last follow-up. The research program was approved by the Ethics Committee of Nantong University Affiliated Hospital, and formal written consent was obtained from every patient.

Tissue microarrays (TMA) was prepared from COAD specimens by histological department of Nantong University Affiliated Hospital and subjected to immunohistochemistrical analysis. Briefly, TMAs were deparaffinized, hydrated, and heated to 121°C using an autoclave in sodium citrate buffer for 20 min to retrieve the antigen. Thereafter, the sections were blocked with 10% goat serum in 0.01M phosphate buffered saline (PBS) for 2h at room temperature, followed by incubation with an anti-CSF-1R rabbit polyclonal antibody (diluted 1:50, HPA012323 SIGMA) overnight at 4°C. The section was incubated with horseradish peroxidase (HRP)-conjugated sheep anti-rabbit IgG at 37°C for 30 min. The immunoreactivity was developed with DAB reagent (DAKO, Denmark). Finally, the slides were counterstained with hematoxylin. The immunostaining of each specimen was analyzed by two independent pathologists who were blinded to the patients’ clinic pathologic data. They evaluated the staining score of each specimen using the H score. The H score ranged from 0–300, multiplying the percentage of positive cells by the staining intensity (where 0, 1, 2, and 3 indicate negative, weak moderate, and strong staining, respectively). The best cutoff value for the staining score was selected with X-title Software. The patients were divided into two groups: CSF-1R-high (n = 57) and CSF-1R-low (n = 46).

For immunofluorescence staining, paraffin sections were deparaffinized, hydrated and antigen-retrieved, as mentioned above. Next, the sections were rinsed three times with PBS, and incubated with anti-CSF-1R antibody (1:50 dilution) and CD68 antibody (1:50 dilution, ab955 Abcam) overnight at 4°C. After washing using PBS for 3 times, and the slides were incubated with the following secondary antibodies for 1h at 37°C: Alexa Fluor488 goat anti-rabbit IgG (diluted 1:200, 111-545-144 Jackson); Alexa Fluor555 donkey anti-mouse IgG (diluted 1:300, A0453 Beyotime). Finally, the sections were mounted with Antifade Mounting solution containing 10μg/ml DAPI. Representative visual fields were acquired using a Leica DM5000 B microscope (Leica Microsystems, Wetzlar, Germany). Next, the fluorescence intensity of the CSF-1R^high TAMs in five different sections was examined and analyzed using the Image J software. Cells with immunostaining scores significantly higher (2.5-fold increase) than non TAMs were defined as CSF-1R^high TAMs.

The COAD expression data and corresponding clinical data were obtained from The Cancer Genome Atlas (TCGA) official website (http://gdc.cancer.gov). The TCGA data were all analyzed by the R software (version 4.1.0) and survival package. The expression data of CSF-1R in various CRC cells were obtained from the DepMap portal (https://depmap.org/portal/) database.

We collected fresh specimens from five patients who underwent radical resection of COAD from the Nantong University Affiliated Hospital and obtained the informed consent of each patient. After the fresh tumor tissue clears the blood with normal saline, put into the tissue preservation solution (MACS Tissue Storage Solution, Miltenyi Biotec). Then put it on the ice and ship it to the laboratory. The tumor tissue was digested with human Tumor Dissociation Kit (Miltenyi biotec), and then the gentleMACS C test tube containing enzyme mixture and tissue was inverted and connected to gentleMACS Dissociator (31) (Miltenyi biotec) for mechanical dissociation. After dissociation, the sample was applied to a MACS SmartStrainer (30um) to remove any remaining larger particles from the single-cell suspension. Then wash cell MACS SmartStrainer with 20 mL of RPMI 1640. Finally, erythrocyte lysate was used to remove red blood cells. The prepared single cell suspension was used for flow cytometry analysis and cell sorting within 2 hours.

For fluorescence activated cell sorting (FACS), single cell suspension was prepared as described previously (22) from fresh COAD specimens and incubated with Human BD Fc Block (BD Bioscience). Then, the samples were stained with APCcy7 anti-CD45 (561863, BD bioscience), FITC anti-CD14 (555397, BD bioscience), and BV421 anti-CSF-1R (565347, BD bioscience) antibodies at 4°C for 30 min in dark. After washing cells 3 times with PBS, the cells were resuspended with staining buffer, and subjected to cell sorting using a BD FACS Aria3 flow cytometry. CSF-1R^low TAMs (CD45⁺ CD14⁺ CSF-1R^low) and CSF-1R^high TAMs (CD45⁺ CD14⁺ CSF-1R^high) were obtained from single-cell mixed suspension by flow cell sorting. Total RNA was isolated from the sorted samples using Qiagen RNeasy kit and subjected to cDNA library preparation using Smart-seq2 (32) scheme. The sequencing was performed using Illumina Novaseq6000 platform (33).

We used R software (version 4.0.2) and Limma package (https://www.bioconductor.org/packages/release/bioc/html/limma.html) to analyze differentially expressed genes (DEGs) between CSF-1R^high TAMs and CSF-1R^low TAMs and used ggplot2 and pheatmap packages to draw volcano map and heat map respectively. We selected | log2 (FC) | >1 and p <0.05 as the criteria for differential gene expression. Then, according to the changes of gene expression, DEGs were divided into two groups: up-regulated genes and down-regulated genes. The Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses of DEGs were performed by DAVID (34) tool (https://david.ncifcrf.gov/tools.jsp).

We used Sanger Box data analysis platform (http://sangerbox.com/) and Pearson method to evaluate the correlation between CSF-1R expression and Tumor Mutation Burden (TMB), Microsatellite Instability (MSI) and Neoantigen. And Pearson’s correlation test was also used to evaluate the correlation between CSF-1R and immune checkpoint molecules.

Statistical analyses were performed using spss20.0, GraphPad Prisim 8.0 and R 4.0.2. The relationships between CSF-1R expression and clinic pathological features were analyzed by Pearson’s chi-squared test. Survival analysis was performed using the Kaplan-Meier method and the logarithm rank tests. We used univariate and multivariate COX regression analysis to evaluate the correlation between different variables and OS. All the significance tests are bilateral, and bilateral p < 0.05 is considered statistically significant. We applied X-title (35) (Yale University version 3.6.1) to select the best cutoff point to evaluate biomarkers.

We evaluated the expression of CSF-1R in 103 COAD specimens and 72 peritumoral specimens by IHC. The results showed that CSF-1R was largely absent in normal intestinal epithelial cells ( Figure 1A ). In contrast, abundant expression of CSF-1R was observed in COAD tissues. We employed H-score to evaluate the expression of CSF-1R in COAD tissues. The expression of CSF-1R was higher in tumor tissues than in peritumoral tissues ( Figure 1B ). The patients were divided into CSF-1R-high (n = 57) and CSF-1R-low (n = 46) using a H-score cutoff of 120. The correlations among CSF-1R expression and the clinicopathological parameters of COAD patients were evaluated ( Table 1 ). CSF-1R expression was positively correlated with M stage (p=0.029) and AJCC stage (p=0.020). These data infer a role of CSF-1R in COAD progression.

In order to decipher the prognostic value of CSF-1R expression in COAD, we first investigated the survival difference between CSF-1R-low and CSF-1R-high patients using Kaplan-Meier survival analysis ( Figure 1C ). Compared with CSF-1R-low patients, patients with high expression of CSF-1R had significantly worsened prognosis. In addition to our own cohort, we also analyzed the prognostic value of CSF-1R in TCGA cohort. High expression of CSF-1R is also associated with unfavorable prognosis in COAD patients of TCGA cohort ( Figure 1D ). To determine whether CSF-1R expression and other clinicopathological variables can be used as independent prognostic indicators in patients with COAD, we also performed univariate and multivariate Cox regression analysis on 103 patients. As shown in Table 2 , CSF-1R expression, along with tumor recurrence, AJCC-stage, N-stage, and M-stage could serve as an independent prognostic factor for patients with COAD.

Because the mechanisms by which CSF-1R promotes cancer development remain under debate, we aimed to clarify whether CSF-1R drives colorectal cancer progression via directly promoting tumor malignancy or facilitating pro-tumor inflammatory environment. To this end, we firstly analyzed whether CSF-1R is mainly expressed in CRC cells in COAD specimens. The expression profiles of CSF-1R in CRC lines and COAD tissues were obtained from Depmap portal and TCGA database, respectively. Notably, while TCGA data suggest abundant expression of CSF-1R in COAD tissues, substantially lower level of CSF-1R was identified in CRC lines, indicating that CSF-1R is not mainly expressed in tumor cells ( Figure 2A ). In agreement with these data, perturbation of CSF-1R does not cause strong growth retardation in CRC lines. CSF-1R may exhibit pro-proliferative or anti-proliferative effects in CRC cells, but these effects appear to be weak in all CRC lines (using depmap’s default perturbation effect cut-off of 0.5) ( Figure 2B ). These data suggest that CSF-1R facilitates COAD progression in a manner largely independent from its function within tumor cells.

Mounting data suggest that CSF-1R is highly expressed in immune cells, particularly macrophages (30, 36). In order to verify whether CSF-1R is mainly expressed in macrophages in COAD, we performed immunofluorescence analysis to investigate the co-localization between CSF-1R and macrophage marker CD68. Immunofluorescence assay revealed a strong co-localization between CSF-1R and CD68 ( Figure 2C ). Therefore, we infer that CSF-1R is mainly expressed in TAMs in COAD specimens. In addition, the proportion of CSF-1R^high TAMs is typically higher than CSF-1R^low TAMs in COAD specimens ( Figure 2D ). Based on these results, we speculate that CSF-1R may mainly regulate CRC immune environment through its expression in TAMs.

Next, we evaluated the role of CSF-1R in regulating TAM function and CRC immune environment. As shown in Figure 2E , we sorted CSF-1R^high TAMs and CSF-1R^low TAMs using a BD FACS cell sorter and performed RNA sequencing (RNA-seq) to determine differentially expression genes (DEGs) between these two populations. According to the changes of gene expression, we identified 1298 up-regulated genes and 371 down-regulated genes. In order to explore the biological functions of these DEGs, we used DAVID tools to carry out GO and KEGG enrichment analysis. GO analysis showed that the up-regulated DEGs were mainly enriched in genes involved in inflammatory response, immune response and the regulation of TME and the down-regulated DEGs were strongly linked to respiratory chain and mitochondrial electron transport ( Figures 3A, B ). Likewise, KEGG analysis showed that pathways involved in immune response and tumor metabolism, such as Cytokine-cytokine receptor interaction, Toll-like receptor signaling pathway, PI3K-Akt signaling pathway and HIF-1 signaling pathway were enriched in upregulated DEGs ( Figure 3C ). However, down-regulation of DEGs mainly included pathways critical for some senile diseases and oxidative phosphorylation ( Figure 3D ). On the basis of the results of KEGG and GO, we speculate that CSF-1R may promote the development of COAD via modulating pro-tumor immune environments.

To validate this hypothesis, we explored the role of CSF-1R in tumor immunity in TCGA cohort by dividing patients into high CSF-1R expression and low CSF-1R expression. Using FDR<0.05 and NES>1.5 as cutoffs, we identified 10 signaling pathways that were significantly enriched in high CSF-1R expression, including Cytokine-cytokine receptor interaction, Chemokine signaling pathway, JAK-STAT signaling pathway, Toll-like receptor signaling pathway, B-cell receptor signaling pathway, T-cell receptor signaling pathway, Pathway in cancer, MAPK signaling pathway, VEGF signaling pathway, Colorectal cancer ( Table 3 and Figure 4 ). These results suggest that high expression of CSF-1R is mainly involved in the regulation of immune microenvironment and tumor metabolism.

Immune cells and stromal cells are two main types of non-tumor components in TME and have been proposed to be valuable for tumor diagnosis and prognosis evaluation (37, 38). To further explore the relationship of CSF-1R and the TME, we used R software Estimate package to calculate the scores of immune cells, stromal cells and both of them, and then evaluated the associations among CSF-1R expression, immune cells, stromal cells and both of them. The results showed that CSF-1R expression was significantly associated with immune cell infiltration, stromal cell infiltration and both ( Figure 5A ). Therefore, we reasoned that CSF-1R mainly exerts pro-tumor functions in the TME. Then we analyzed the proportion of various immune cell types in the COAD by CIBERSORT rewinding calculation. We found that CSF-1R-high tumors contained more regulatory T cells (Tregs) and Macrophages M2 than CSF-1R-low tumors ( Figure 5B ). Of note, Tregs can inhibit the anti-tumor immune effects of DC cells, NK cells and effector T cells (Teff) through various mechanisms and are an important factor in the immunosuppressive TME (39). We reasoned that increased chemokines in CSF-1R tumors might drive enhanced Tregs recruitment. Therefore, we further analyzed the RNA sequencing results. Indeed, as shown in Figures 5C, D , a variety of Treg-recruiting cytokines, including CCL3, CCL4, CCL11, and CCL13 were upregulated in CSF-1R^high TAMs (40, 41). The above results suggest that CSF-1R ⁺ TAMs may inhibit anti-tumor immunity by recruiting Tregs to regulate COAD progression in the TME.

To explore the potential associations between CSF-1R and other genes in COAD, We carried out PPI network analysis online through STRING website (https://string-db.org/). PPI network showed that 10 genes (CSF1, IL34, HRAS, GRB2, CBL, PIK3RI, TYROBP, STAP2, TNFSF11, and STAP2) were functionally related to CSF-1R ( Figure 6A ). So far, the value of targeting CSF-1R in COAD immunotherapy remains unclear. Previous studies have reported that TMB, MSI and Neoantigen can be used as biomarkers for the survival prognosis and immune checkpoint inhibitors’ efficacy (42–44). With the help of Sanger box website tool and Pearson method, we calculated the correlation between CSF-1R expression and TMB, MSI and Neoantigen. The results showed that the expression of CSF-1R in COAD was significantly correlated with TMB (p<0.001) ( Figure 6B ), MSI (p<0.001) ( Figure 6C ), and Neoantigen (p=0.004) ( Figure 6D ). In addition, we analyzed the relationship between CSF-1R and immune checkpoint molecules with the help of Sanger box website. We found that CSF-1R was significantly correlated with tumor-related immunosuppressive molecules including PDCD1, CTLA4, CD80, CD86, HAVCR2, etc. ( Figure 7 ). These data indicates that CSF-1R may serve as a novel immunotherapeutic target for COAD.