RAB31 orchestrates CXCL2-CXCR4-mediated neutrophil recruitment and proangiogenic niche formation in colorectal cancer

Abstract

Background

The interaction between tumour-associated neutrophils (TANs) and angiogenesis plays a crucial role in tumour progression. However, the specific regulatory mechanisms by which different TANs populations influence angiogenesis in colorectal cancer (CRC) remain poorly understood.

Methods

The study integrates our own dataset with publicly available single-cell RNA sequencing (scRNA-seq) data to analyze the infiltration of various single-cell types in colorectal cancer (CRC). Using in vitro experiments, immunofluorescence, and animal models treated with anti-VEGFR antibodies, we investigated the role of tumor-associated neutrophils (TANs) in CRC and angiogenesis. We conducted RAB31 knockdown cell line, RNA sequencing, Western blotting, quantitative PCR, ELISA, tube formation assays, CCK8, and clonogenic assays, combined with in vivo experiments, to elucidate the impact of RAB31 in the tumor microenvironment (TME) and the mechanism by which tumor-derived RAB1 regulates TANs recruitment.

Results

Single-cell analysis revealed a significant infiltration of myeloid cells, particularly TANs, in CRC tumour tissue compared to normal tissue. RAB31 was found to be involved in malignant cell pathways, promoting CRC progression. Depletion of TANs substantially suppressed tumour growth and angiogenesis, while VEGFR blockade inhibited tumour growth and altered TANs infiltration in response to angiogenesis status. In vitro, knockdown of RAB31 expression inhibited tumour cell proliferation, and CXCL2 protein secretion was significantly reduced in shRab31-transfected tumour cells. In vivo, tumour tissue from shRab31 models showed decreased angiogenesis and reduced Neutrophil Extracellular Traps (NET) production, alongside reduced TANs infiltration. Moreover, CXCR4 expression was significantly elevated in TANs, and tumour cells influenced TANs through the CXCL2–CXCR4 axis. RAB31 regulated the PI3K–Akt signalling pathway, which is linked to neutrophil recruitment in CRC.

Conclusions

Our findings suggest that tumour-derived RAB31 regulates the CXCL2–CXCR4 signalling pathway in CRC. RAB31 promotes CXCL2 secretion through the PI3K-Akt signalling pathway, facilitating TANs recruitment and angiogenesis, thereby driving tumour growth. This work enhances our understanding of the complex role of TANs and RAB31 in CRC progression and provides new insights into potential therapeutic strategies targeting the tumour microenvironment.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12967-026-07908-6.

Introduction

Colorectal cancer (CRC) is the third most common malignant tumour worldwide after breast and lung cancers [1]. Studies conducted over the past few decades have shown that tumours are not only composed of neoplastic cells but also surrounded by a dynamic tumour stroma called the tumour microenvironment (TME), which consists of numerous nonmalignant cells, such as fibroblasts, immunocytes, adipocytes, myeloid cells, connective tissue cells, and vascular components [2, 3]. Despite great advances in molecular inhibitors, surgical techniques, chemotherapy drugs, and targeted drugs, the number of patients with colorectal cancer continues to gradually increase [4]. Immunotherapy has revolutionized the treatment of cancer and is considered the most promising approach to overcome cancer [5]. Immune checkpoint blockade (ICB) therapy, represented by PD-1/PD-L1, has achieved great success in solid tumours, but only a minor fraction of patients respond to ICB. Recent studies have demonstrated that the TME in the vast majority of CRC patients is highly heterogeneous in terms of immunosuppression [6]. Therefore, elucidating the mechanisms by which tumours compromise immunity is crucial for therapeutic interventions and the improvement of the patient prognosis.

Neutrophils are the most abundant type of innate immune cell in human peripheral blood and are emerging as crucial regulators of the TME [7]. Neutrophils perform their biological functions by forming neutrophil extracellular traps (NET) that are composed of nuclear DNA and various proteins, such as myeloperoxidase (MPO), which is a marker of NET [8]. An increased neutrophil-to-lymphocyte ratio (NLR) is generally an independent predictor of the prognosis of patients with most tumours. Moreover, increased tumour-associated neutrophil (TANs) infiltration is independently correlated with poor outcomes [9–11]. Tumour angiogenesis, one of the fourteen hallmarks of cancer, is essential for tumour expansion and metastasis [12]. Hyperactivation of proangiogenic signalling in tumours disrupts normal blood vessels, leading to inconsistent perfusion, poor vascular function, and impaired vascular maturation [13]. Previous studies have reported that MMP-9 produced by bone marrow-derived neutrophils contributes to the development of squamous cell carcinoma [14]. MMP-9 supplied by neutrophils is also involved in tumour angiogenesis to promote the carcinogenesis of pancreatic and lung cancer [15]. However, the crosstalk between NET-forming extracellular structures and tumour angiogenesis remains largely unknown in colorectal cancer.

Ras-related protein 31 (Rab31, also known as Rab22B) is a 194-amino-acid protein that is involved in regulating intracellular vesicle transport from the Golgi/trans-Golgi network (TGN) to endosomes [16]. Accumulating evidence has shown that dysfunction in Rab pathways widely participates in tumour development and is associated with the prognosis of multiple cancers [17, 18]. Increased expression of Rab31 is frequently observed in breast cancer, hepatocellular carcinoma, and gastric cancer tissues and is related to shorter overall survival [19, 20]. Furthermore, elevated Rab31 expression in glioblastoma and cervical cancer promotes tumour cell invasion and proliferation [21, 22]. Conversely, Rab31 silencing inhibits tumour development in vivo. Most studies on Rab31 have focused on the tumour cells themselves, and the function of Rab31 in the TME (especially the TANs NET–angiogenesis pathway) is elusive in CRC.

Here, we reveal that TANs promote tumour angiogenesis through extracellular NET and that tumour angiogenesis in turn continues to increase the infiltration of TANs, resulting in the formation of a positive feedback loop to promote the progression of colorectal cancer. Using the LightGBM machine learning algorithm, we determined the TAN–angiogenesis subtype classification and constructed a robust signature based on ten TAN–angiogenesis-related genes to evaluate the prognosis of patients with CRC. In addition, we elucidated that RAB31 promotes neutrophil recruitment and tumour angiogenesis through the CXCL2–CXCR4 signalling axis. These findings indicate that tumour-derived RAB31 plays a new role in remodelling the TME and could serve as a potential new candidate target for therapeutic strategies regulating TANs and angiogenesis.

Materials and methods

Dataset collection

The publicly available scRNA-seq data (GSE132465 and GSE144735, referred to as the KUL3 dataset and the SMC dataset [23], clinical information are detailed in Table S1) were downloaded and subsequently subjected to debatch integration using scVI. In addition, public bulk RNA-seq datasets (GSE14333, GSE33113, GSE39582 and GSE37892) from the Gene Expression Omnibus database (GEO, https://www.ncbi.nlm.nih.gov/geo/) were also included, and these datasets were integrated into the META dataset. This information included gene expression data and clinical information for a total of 914 CRC patients. The transcriptomic data and clinical information of the Cancer Genome Atlas (TCGA) COAD cohort were obtained from the UCSC Xena data portal (https://xenabrowser.net).

The total number of human tissue samples submitted locally for inspection and eventually included in the single-cell study was 1 CRC tumour sample and 1 normal sample. The samples from patients with CRC confirmed by postoperative pathology were frozen in liquid nitrogen within 30 min after lesion resection. The tumour-adjacent specimens were excised at a distance greater than 5 cm from the core lesion site.

Single-cell RNA sequencing and in vitro and in vivo experiments

The specific processes used in the single-cell RNA sequencing analysis (data screening, normalization, dimensionality reduction, clustering, cell cluster annotation, subset cell characterization, PAGA, transcription factor module analysis, and enrichment analysis) and in vitro and in vivo experiments (immunofluorescence staining, Transwell assays, etc.) are described in the Supplementary Methodology (primary antibodies and flow cytometry antibodies are detailed in Table S2).

Statistical analysis

Statistical analysis was performed using GraphPad Prism 9.0 (RRID: SCR_002798). Comparisons between two different groups were performed with unpaired or paired two-tailed Student’s t tests. The survival analysis was performed using the Kaplan–Meier method, with the log-rank test for comparison. The experiments were repeated two to three times. Tumour growth curves were compared using two-way analysis of variance (ANOVA) with Sidak’s multiple comparisons test. HRs and their 95% CIs were calculated using stratified Cox proportional hazards regression models. The cut-off point for the optimization of patients in TCGA database was calculated using R software. The data are presented as the means ± SEMs. A value of p < 0.05 was considered to indicate statistical significance: *p < 0.05, **p < 0.01, and ***p < 0.001.

Results

Single-cell RNA sequencing reveals heterogeneity in the tumour microenvironment and provides an overall single-cell atlas between colorectal cancer and normal tissues

An overview of the study design of the present study is shown in Fig. 1A. Following rigorous quality control, the analysis yielded 66,421 high-quality cells, which were subsequently analysed in further detail. We subsequently adopted the UMAP technique to visualize the high-dimensional scRNA-seq data and successfully classified cells into 10 clusters, which were annotated as epithelial cells, malignant cells, fibroblasts, endothelial cells, etc. (Fig. 1B). Furthermore, we investigated the cell counts and proportions of the above 10 cell clusters in normal cells, tumour cells, and each sample involved (Fig. 1C), and these cell subclasses were verified with well-acknowledged cell markers (Fig. 1D). The overall cell clusters specific to samples and databases is presented in UMAP plots in Fig. S1A–B. The results for the corresponding cell counts and proportions are available in Fig. S1C–D. Fig. S2 shows the results of InferCNV analysis among groups (tumour and normal) and databases (KUL3, SMC, and our data), as well as the results of the cell proportion analysis at the myeloid cell level.

Fig. 1.

Single-cell RNA sequencing reveals tumor microenvironment heterogeneity and overall single-cell atlas between colorectal cancer and normal tissues. (A) The overview flow chart of the present research. (B) Cell clustering UMAP map of normal and tumor tissues. (C-D) The cell counts and proportion of the 10 cell clusters in normal, tumor cells, and in each sample. (E) The Ro/e and cell proportion analysis of 10 cell clusters proportion in normal and tumor cells. (F-G) Myeloid cells were clustered into 13 clusters and the corresponding cell counts and proportion

The primary objective of the R_o/e analysis is to evaluate whether the tissue distribution of different cell lines significantly deviates from random expectations. Epithelial cells, NK/T cells, myeloid cells, and malignant cells were enriched in tumour tissue (with R_o/e values > 1). The boxplot of the cell proportions also revealed that the quantities of enteric glial cells, plasma cells, and myeloid cells differed significantly between normal tissues and tumour tissues (Fig. 1E). Hence, based on the above results, we classified myeloid cells into 13 clusters according to the annotated genes and investigated the corresponding cell counts and proportions in normal and tumour tissues (Fig. 1F). The cell counts and proportions of 4 cell clusters (DCs, macrophages, monocytes and TANs) are shown in Fig. 1G.

Additionally, we investigated the differences in differentially expressed genes (DEGs), pathway activity, and transcription factor (TF) activity in six cell types between the tumour and control tissues: epithelial cells, myeloid cells, NK/T cells, endothelial cells, B cells, and plasma cells (Fig. S3).

RAB31 drives malignant pathways and promotes the progression of colorectal cancer

Machine learning (ML) has been widely used in big data analysis, which can automatically learn data patterns and optimize model parameters. Gradient boosting (GB) is a member of the ensemble learning paradigm. Its basic principle is to set up a strong model by combining multiple weak learners. Hence, the predictability of it was superior to that of a single model. The ‘mlr3’ R package was applied to construct various types of gradient boosting ML models. By comparing the rankings of multiple machine learning models, we determined that the LightGBM based on gradient boosting trees algorithm performed best (Fig.S7A). The ROC and Precision-Recall(PR) of multiple machine learning models were compared, and the results showed that the LightGBM model had the best results (Fig.S7B). We employed LightGBM to preserve the top 5 most important genes RAB31, CTHRC1, TNFSF4, FRMD6, and GPNMB according to the recursive feature elimination (RFE) algorithm. ROC curve results showed that the five genes were highly effective in identifying patients with angiogenic subtypes (Fig.S7C). The prognostic accuracy of the riskscore was assessed in the train set(TCGA), test set(TCGA), and meta cohort (Fig.S7D).

Through cNMF, 6 malignant cell programs shared among patients were identified (Fig. 2A). These programs were further analysed by annotating each one according to the 100 genes with the highest weights (Fig. S4). The Leiden algorithm was used to cluster malignant tumour cells, and 6 cell clusters were generated (Fig. 2B). CytoTRACE2 analysis revealed the differentiation potential of each cell cluster, of which cell Clusters 0 and 4 had the greatest differentiation potential (Fig. 2C). PAGA showed differentiation continuity in malignant cells, with RAB31 and CXCL2 expression being high in Cluster 4, indicating that this cluster is highly capable of differentiation. Subsequently, the expression of RAB31, CXCL2, and angiogenesis-related genes was sequentially increased, and high levels of EGFR and VEGFA expression were ultimately observed (Fig. 2D, E). Through immunoblotting, we confirmed that RAB31 expression was significantly increased in the tumour tissue of CRC patients (Fig. 2F). We explored the role of RAB31 in tumour development by constructing RAB31 knockdown cell lines (shRab31), including the mouse-derived colon cancer cell line MC38 and the human-derived colon cancer cell line SW620, via transduction with a lentiviral vector expressing GFP and a Rab31 short hairpin RNA (shRNA). Western blot and qPCR results showed significant decreases in both protein and mRNA levels in the knockdown tumour cell lines, indicating that Rab31 was successfully knocked down (Fig. 2G–H). We studied the effect of RAB31 on the TME in vivo by subcutaneously (s.c.) inoculating MC38 tumour cells into C57 WT mice. Compared with those in the control group, tumour growth, the tumour volume, and tumour weight in the shRab31 group were significantly lower (Fig. 2I–K). The results of the CCK8 and colony formation assays indicated that knocking down Rab31 expression significantly inhibited the proliferation of tumour cells in vitro (Fig. 2L–N).

Fig. 2.

RAB31 drives malignant cell pathways and promotes the progression of colorectal cancer. (A) The cNMF Heatmap shows the Euclidean distance of programs across replicates. (B) Diffusion map of malignant cells clustering via Leiden algorithm. (C) The differentiation potential of each cell cluster via CytoTRACE2 analysis. (D) PAGA graph overlaid on the diffusion maps. (E). The malignant cells developmental trajectory map F Western blot analysis of RAB31 protein expression in the tissue lysates of tumor adjacent normal tissue (normal) and tumor tissues resected from colon cancer patients. (G) Real-time PCR analysis of Rab31 gene expression of MC38 and SW620 tumor cells transduced with the lentivirus expressing control (shNC) or Rab31 shRNA (shRab31).(n = 3, paired t tests). (H) Western blot analysis of RAB31 protein expression of the tumor cell lines. (I) The images of control (shNC) and RAB31 knock down (shRab31) MC38 tumors (n = 5 mice). (L). Clonal formation experiment of the indicated control (shNC) and RAB31 knock down (shRab31) tumor cells. (J) Tumor growth curve of control (shNC) and RAB31 knock down (shRab31) MC38 tumors (n = 5 mice, paired t tests). (K) Tumor weight of control (shNC) and RAB31 knock down (shRab31) MC38 tumors (n = 5 mice, paired t tests). (M-N). The cell growth of the indicated control (shNC) and RAB31 knock down (shRab31/shRAB31) tumor cells

Anti-VEGFR therapy suppresses tumour growth and remodels the immune microenvironment

Previous evidence has suggested that tumour angiogenesis induces the recruitment of TANs [24]. Vascular endothelial growth factor (VEGF, also known as VEGF-a) is an endothelial cell-specific mitogen generated by tumour cells that plays a crucial role in tumour angiogenesis. VEGF and its specific receptors (VEGFRs) are the most effective targets for antiangiogenic therapy for cancer. Therefore, we administered intraperitoneal (i.p.) injections of anti-VEGFR antibody and an IgG isotype control to tumour-bearing mice. These results revealed that the targeted blockade of the VEGFR-mediated angiogenesis pathway significantly inhibited tumour growth, resulting in a smaller tumour volume, slower tumour growth, and decreased tumour weight (Fig. 3A–C). Compared with the control group, the anti-VEGFR group presented significantly reduced infiltration of intratumoural TANs (Fig. 3D, H). In addition, improvements in the intratumoural blood vessels prominently increased the infiltration of CD8 + T cells (Fig. 3E), indicating that treatments targeting angiogenesis combined with ICB therapy increase therapeutic efficacy. Similarly, we observed no significant differences in the distribution of infiltration in other immune cell types, such as macrophages and dendritic cells (Fig. 3F–G). Next, immunofluorescence (IF) staining of MC38 tumours from IgG- or anti-VEGFR-treated samples (Fig. 3I) and Rab31-knockdown (shRab31) or control (shNC) samples (Fig. 3M) was conducted using antibodies against CD31 (red), Ly-6G (yellow), and MPO (green), followed by counterstaining with DAPI (blue) (n = 5 mice). We confirmed that the anti-VEGFR antibody remodelled intratumoural blood vessels and substantially reduced the production of NET (Fig. 3J–L). A FACS analysis of the percentage of TANs in shNC- and shRab31-transfected MC38 tumours (n = 5 mice) was performed. Angiogenesis and NET production in shRab31-treated tumour tissue were significantly decreased as TANs infiltration decreased (Fig. 3N, O). Collectively, these data suggest that targeting tumour angiogenesis to reshape the intratumoural vascular structure reverses the immunosuppressive TME.

Fig. 3.

Anti-VEGFR therapy suppresses tumor growth and remodels the immune microenvironment. (A) The images of MC38 tumors after IgG or Anti-VEGFR treatment (n = 5 mice). (B) Tumor growth curve of MC38 tumors after IgG or Anti-VEGFR treatment (n = 5 mice). (C). Tumor weight of MC38 tumors after IgG or Anti-VEGFR treatment (n = 5 mice). (D-H). FACS analysis of the percent of tumor-infiltrating neutrophil (H), T cells (G), Macrophages, and DCs (F) in MC38 tumors after IgG or Anti-VEGFR treatment (n = 5 mice). (I, M) Immunofluorescence (IF) staining of MC38 tumors after IgG or Anti-VEGFR treatment samples (I) and RAB31 knock down (shRab31) or control (shNC) samples (M) via antibodies against CD31(red), Ly-6G (yellow), and MPO (green), counterstained with DAPI(blue)(n = 5 mice). (J-L). Quantification of CD31⁺ cells, Ly6G⁺ cells, and MPO⁺ cells in MC38 tumors after IgG or Anti-VEGFR treatment samples (n = 5 mice). (N-O). FACS analysis of the percent of tumor-infiltrating neutrophil in control (shNC) and RAB31 knock down (shRab31) MC38 tumors (n = 5 mice)

Tumour-associated neutrophil depletion inhibits tumour growth and reshapes the tumour microenvironment

We investigated the role of TANs in CRC using a C57BL/6 syngeneic MC38 cell line to establish a colon tumour model and treated it with an anti-Ly6G antibody to deplete neutrophils in vivo. After TANs were depleted, tumour growth was markedly inhibited, as indicated by significant decreases in tumour volume and weight (Fig. 4A–C). The immunosuppressive effect of the TME was reversed, and the infiltration of CD8 + T cells and CD4 + T cells was significantly increased (Fig. 4F). Interestingly, no significant difference in the intratumoural infiltration of other immune cell subtypes, such as macrophages and dendritic cells, was observed (Fig. 3D, E). The results of IF staining and cellular quantification showed that the NET in the tumour were dramatically reduced and that the orderly distribution of blood vessels in the tumour was reversed after neutrophil depletion (Fig. 4G–J). In addition, the proportions of immune cell subtypes infiltrating the spleen, such as CD8 + T cells, CD4 + T cells, macrophages, and dendritic cells, did not differ significantly (Fig. 4K, M, L).

Fig. 4.

Tumor-associated neutrophil depletion inhibits tumor growth and reshapes the tumor microenvironment. (A) The images of MC38 tumors after IgG or α-Ly-6G treatment (n = 5 mice). (B) Tumor growth curve of MC38 tumors after IgG or α-Ly-6G treatment (n = 5 mice). (C) Tumor weight of MC38 tumors after IgG or α-Ly-6G treatment (n = 5 mice). (D-F). FACS analysis of the percent of spleen-infiltrating neutrophil (F), tumor-infiltrating T cells (E), Macrophages, and DCs (D) in MC38 tumors after IgG or α-Ly-6G treatment (n = 5 mice). (G). IF staining of MC38 tumors after Anti-Ly6G treatment samples via antibodies against CD31(red), Ly-6G (yellow), and MPO (green), counterstained with DAPI(blue)(n = 5 mice). (H-J). Quantification of CD31⁺ cells, Ly6G⁺ cells, and MPO⁺ cells in MC38 tumors after IgG or α-Ly-6G treatment samples (n = 5 mice). (K-M). FACS analysis of the percent of spleen-infiltrating T cells (K, L), Macrophages, and DCs (M) in MC38 tumors after IgG or α-Ly-6G treatment (n = 5 mice). (N) An integrated network across TANs and malignant cells with interactions: Malignant cells recruit TANs by mediating the CXCL2-CXCR4 axis. (O-P) Real-time PCR analysis of Cxcl2 gene expression of MC38, and SW620 tumor cells transduced with the lentivirus expressing control (shNC) or Rab31 shRNA (shRab31). (Q) Real-time PCR analysis of Cxcr2 gene expression of tumor-infiltrating neutrophil in control (shNC) and RAB31 knock down (shRab31) MC38 tumors (n = 5 mice)

In vitro experiments, we established a co-culture system of human colon cancer cells SW620 and HUVECs, and collected the supernatant as NET-enriched conditioned medium (CM). HUVECs were then treated with this CM, and the effects of NET on tube formation and migration ability were assessed via tube formation assay and scratch wound healing assay. The results demonstrated that, compared with the control group, CM treatment significantly enhanced the tube formation and migration capabilities of HUVECs. However, these promoting effects were reversed by shRab31 and DNase-1 treatment(Fig. S8).

An integrated intra- and intercellular network centred on two significant interactions is shown in Fig. 4N. Malignant cells may recruit TANs through the CXCL2–CXCR4 axis. Notably, compared with the shNC group, the mRNA and protein levels of CXCL2 released into the extracellular space were significantly lower in shRab31-transfected tumour cells (Fig. 4O, P). We also detected CXCR4 expression levels in TANs and infiltrating splenic neutrophils and found that CXCR4 expression was significantly increased in TANs (Fig. 4Q). Furthermore, we established a co-culture system of SW620 cells and neutrophils, followed by exogenous administration of CXCL2 and AMD3100 (CXCR4 antagonist). Using Transwell assays, we assessed alterations in the recruitment of neutrophils by tumor cells and found that CXCL2 promoted neutrophil migration, while AMD3100 abrogated this effect (Fig. S9).

Distinct angiogenesis patterns in CRC patients

Through an angiogenesis-related gene-based consensus clustering analysis (Fig. S5A), we identified 2 clusters with distinct TANs statuses, i.e., Angiogenesis_C1 and Angiogenesis_C2, in CRC patients in TCGA and META databases. The univariate regression analysis of risk showed that age, stage, and the angiogenesis status were prognostic factors and that the prognosis was worse in the Angiogenesis_C2 group (Fig. S5B). We then performed a multivariate risk regression analysis to determine whether the angiogenesis status can be used as an independent prognostic factor to predict the clinical outcome of CRC patients (Fig. S5C). In addition, we further assessed a variety of clinical characteristics, including methylation of the CpG island methylator phenotype (CIMP), instability of microsatellite DNA regions (MSI), and stage, in CRC patients with different angiogenesis statuses. The results revealed an apparent difference in the clinical characteristics between the Angiogenesis_C1 group and the Angiogenesis_C2 group (Fig. S5D). We investigated the infiltration of TANs in CRC patients with different angiogenesis statuses using the CIBERSORT, CIBERSORT.ABS, MCPcounter, QUANTISEQ, TIMER, and xCell algorithms were used to calculate the infiltration score of the TANs in samples from TCGA and META databases. Consistent with previous results, patients in the Angiogenesis_C2 group with a poorer prognosis had higher TANs scores (Fig. S6C, D). Next, we analysed the diversity in angiogenesis-related genes and signalling pathways among groups with distinct angiogenesis statuses. As shown in the heatmap, angiogenic genes and signalling pathways were hyperactivated in the Angiogenesis_C2 group of patients (Fig. S6A, B).

RAB31 drives colorectal cancer progression by promoting CXCL2 secretion via the PI3K–Akt signalling pathway

We investigated potential genes associated with the Angiogenesis_C2 group by analysing TCGA and META data, and volcano plots of both linear models revealed that RAB31 expression was significantly and robustly correlated (upregulated) with that of angiogenic Cluster 2 (Fig. 5A, B). The KEGG pathway analysis of the aforementioned data revealed a strong association with the PI3K–Akt signalling pathway (Fig. 5C, D). We knocked down RAB31 in human colon cancer SW620 cells and performed transcriptomic profiling using the “Illumina” platform (Fig. 5E). The subsequent KEGG pathway analysis consistently demonstrated significant enrichment of the PI3K–Akt signalling pathway, which again indicated that RAB31 expression was significantly correlated with the activity of the PI3K–Akt signalling pathway (Fig. 5F). Next, Western blot analysis showed that RAB31 knockdown significantly reduced the levels of the p-PI3K and p-AKT proteins in both the MC38 and SW620 cell lines (Fig. 5G). We added the PI3K activator 740 Y-P to the cell culture medium.That promoted the phosphorylation of PI3K and AKT, which was inhibited by shRab31, indicating that Rab31 acted mainly through the PI3K/AKT signaling pathway.We also found that the addition of 740 Y-P rescued CXCL2 expression, promoting recruitment of TAN(Fig. S10).Notably, compared with those in shNC-transfected tumour cells, the mRNA levels of CXCL2 were significantly decreased in shRab31-transfected tumour cells (Fig. 5H, I).

Fig. 5.

RAB31 drives colorectal cancer progression by promoting CXCL2 secretion via the PI3K-Akt signaling pathway. (A-B). Volcano plots of linear models of Angiogenesis_C2 group-related genes via public data. (C-D). KEGG analysis of Angiogenesis_C2 group-related pathways via public data. E. Rab31 knockdown in human colon cancer cell SW620, RNA-seq (transcriptome sequencing) using “Illumina” platform. (F). KEGG analysis of Angiogenesis_C2 group-related pathways via RNA sequencing results. (G). Western blot analysis of RAB31 knockdown in MC38 and SW620 cell lines. (H-I). Real-time PCR analysis of CXCL2 gene expression of MC38 and SW620 tumor cells transduced with the lentivirus expressing control (shNC) or Rab31/RAB31 shRNA (shRab31/shRAB31). (J-K). ELISA result of CXCL2 protein expression level of MC38 and SW620 tumor cells transduced with the lentivirus expressing control (shNC) or Rab31 shRNA (shRab31). (L). Mechanistic diagram of the present study

ELISA results for the CXCL2 protein levels in MC38 and SW620 tumour cells transduced with a lentivirus expressing shNC or shRab31 showed that the levels of the CXCL2 protein released into the extracellular space were significantly decreased in shRab31-transfected tumour cells compared with shNC-transfected cells (Fig. 5J, K). Figure 5L illustrates the central point of the present study; i.e., RAB31 drives colorectal cancer progression through the PI3K–Akt signalling pathway by promoting CXCL2 secretion, thus recruiting TANs and promoting tumour angiogenesis.

Discussion

Increasing evidence emphasizes the complex and multidirectional interactions between tumour cells and nonmalignant cells in the TME, although the mechanism by which genetic changes in tumour cells affect immune cells remains unclear. Neutrophils have long been considered to play a crucial role in the defence against infection, whereas their function in TME remodelling has only recently attracted attention. Through the release of cytokines and reprogramming of tumour metabolism, TANs promote or inhibit tumour progression, depending on the characteristics of the TME [25]. The recruitment of TANs is promoted by large amounts of CXC chemokines (such as CXCL1 and CXCL8), growth factors (such as G-CSF and GM-CSF), and inflammatory factors (such as IL-6, IL-1β, and IL-17) produced by tumour cells, tumour-associated stromal cells, and tumour-infiltrating leukocytes [26]. Previous results have shown that neutrophils exhibit formidable immunosuppressive activity in late-stage tumours [27]. The MMPs, serine proteases, and cysteine cathepsin secreted by TANs inhibit the activation of NK cells and stimulate the extravasation of circulating tumour cells (CTCs) [28]. In addition, TANs promote the colonization and expansion of metastatic cells by remodelling the extracellular matrix (ECM) and constructing an immunosuppressive TME [29]. In the innate immune response, neutrophils release NET containing DNA, histones, and MPO in the extracellular matrix to capture and clear pathogens [30]. In recent years, the role of NET in cancer progression has been studied. A recent study revealed that interleukin-17 (IL-17) induces neutrophil infiltration in pancreatic cancer and increases NET formation [31]. In this study, we focused on the interaction of NET with tumour angiogenesis in vascular endothelial cells.

Neovascularization plays a vital role in colorectal cancer, particularly in the activation of multiple signalling networks by vascular endothelial growth factor family proteins and their receptors, leading to endothelial cell survival, migration, mitosis, differentiation, and vascular permeability [32]. VEGF signalling via three receptors (VEGFR-1, 2, and 3) on the surface of vascular endothelial cells is regulated by multiple factors, including oncogene signalling, the hypoxic TME, and immunocytes [33]. VEGFR is a crucial regulator of aberrant angiogenesis in CRC. However, antiangiogenic therapies targeting the VEGR signalling pathway rarely induce a durable tumour response in mouse models and cancer patients [34]. One of the important reasons is that stromal cells in the tumour microenvironment participate in the critical process of tumour angiogenesis. Recently, tumour maintenance or recurrence after antiangiogenic treatment has been causally associated with the recruitment of bone marrow-derived myeloid cells (BMDMCs) [35]. Therefore, clarifying the mechanism underlying the crosstalk between immune cells, especially neutrophils, and angiogenesis in the tumour immune microenvironment is particularly important for the development of novel tumour treatment strategies. We used LightGBM, a gradient boosting machine learning algorithm, to construct a prognostic risk score for TAN–angiogenesis-related genes. A significant difference in the risk score was observed between nonresponders and responders, indicating that we could more accurately predict the clinical response of CRC patients to PD-1/CTLA-4 immunotherapy through the TAN–angiogenesis-related gene signature.

Among the five key TAN–angiogenesis-related genes found in CRC, we confirmed for the first time that Rab31 regulates the recruitment of TANs. Rab proteins constitute the largest family of small GTPase monomers. Many studies have confirmed that Rab proteins are distributed in different organelles and regulate the transport of intracellular substances. Aberrant expression of Rab proteins and altered GTPase expression or activity are suggested to be associated with neurodegenerative diseases, lipid storage disorders and cancer [36]. Recent studies have shown that Rab31 is an oncogene in various cancers and that the overexpression of RAB31 is associated with carcinoma progression and poor clinical outcomes. RAB31 is significantly overexpressed in breast cancer, gastric cancer, pancreatic cancer, and other malignancies [37]. In our study, we reported that tumour-derived RAB31 regulated the production of CXCL2 secreted from tumour cells. CXCL2 recruited neutrophils to the tumour microenvironment via CXCR4 expressed on the surface of neutrophils. TANs promote tumour angiogenesis, which further promotes the recruitment of neutrophils to form a positive feedback loop. These results suggest that targeting RAB31 can provide a novel research direction and therapeutic strategy for antiangiogenic agents combined with immunotherapy.

Several limitations of this research should be noted. First, we acknowledge the limitation of the sample size for the local scRNA-seq analysis, and we cannot exclude the possibility that tumour-derived RAB31 affects other cell types in the TME, such as fibrocytes, adipocytes, and T cells. Second, owing to the biological properties of neutrophils, including low mRNA abundance and high RNase activity, their capture efficiency in 10X single-cell sequencing is compromised, which has precluded us from performing more refined subtyping of TANs. Furthermore, independent external validation of our TAN–angiogenesis-related gene model was performed in our study, but when information on patients and samples was collected retrospectively from public databases, the inclusion of all the differences between patients from different geographic regions was still difficult. In addition, the role of tumour-derived RAB31 in other tumour microenvironments remains unclear. In subsequent studies, we aim to expand the sample size, refine the experimental protocols, and further investigate the role of TAN subsets in colon cancer, which will be pursued as a key extension of our research.

To date, no drugs specifically targeting RAB31 have entered clinical trials. Although a furanonaphthyridine-based small-molecule compound has been reported to bind to the GTP-binding pocket of RAB31, interfering with its GTP/GDP exchange and thereby inhibiting its activity, the “switch” region responsible for GTP/GDP binding is highly conserved across all GTPases, including RAB31. This high degree of conservation makes it extremely challenging to design a highly selective inhibitor that suppresses RAB31 without affecting other critical GTPases (such as Ras and Rho), and it also entails a significant risk of off-target toxicity. Therefore, we propose that future strategies should avoid directly targeting the GTP-binding site and instead focus on identifying other allosteric sites on the surface of the RAB31 protein. By binding to these sites, it may be possible to indirectly modulate its GTPase activity or its affinity for effector proteins. This allosteric regulation strategy holds promise for significantly improving targeting selectivity and enhancing the potential for clinical translation.

Conclusions

We demonstrate that TANs facilitate the growth of blood vessels in tumours via extracellular NET, which in turn further recruit TANs into the tumour microenvironment. This reciprocal enhancement establishes a positive feedback loop that accelerates colorectal cancer progression.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Zhen Zong conceived the study design and interpreted the results. Zhi-kun Ning provided methodological suggestions and analysed the data. Cegui Hu was in charge of analysing the data and drafting the manuscript. Ying Tang and Qiuling Luo provided many suggestions on the writing of the article. Chenwei Yuan was of immense help in the process of writing the manuscript and methodology. Xiaoping Zhu and Xin-gen Zhu were in charge of the present study. All the authors read and approved the final manuscript.

Funding

This research was funded by the National Natural Science Foundation of China (grant no. 82260596), the Outstanding Youth Fund Progrm of Jiangxi Province (grant no. 20252BAC220050), the Natural Science Foundation of Jiangxi Province (grant no. 20242BAB25506), the Science and Technology Program of Jiangxi Provincial Health and Family Planning Commission (grant no. 202410246), the China Postdoctoral Science Foundation (grant no. 2023M741523), and the Science and Technology Program of Jiangxi Provincial Administration of Traditional Chinese Medicine (grant no. 2024A0028).

Data availability

All the data needed to evaluate the conclusions in the paper are presented in the paper and/or in the materials cited herein. The scRNA-seq data reported in this paper have been deposited in OMlX, China National Center for Bioinformation/Beijing Institute of Genomics, Chinese Academy of Sciences (https://ngdc.cncb.ac.cn/omix: accession no. OMIX010119) [37]. Additional data related to this paper may be requested from the authors.

Declarations

Ethical approval and consent to participate

Studies involving human participants were reviewed and approved by the Medical Ethics Committee of the Second Affiliated Hospital of Nanchang University, and this study was conducted in accordance with the ethical standards established in the 1964 Declaration of Helsinki. This ethics committee strictly followed China’s Good Clinical Practice (GCP) and related regulations. The patients/participants provided written informed consent to participate in this study (project number: IIT-O-2024-092).

Consent for publication

Not applicable.

Competing Interests

No potential conflicts of interest were reported by the authors.