Tumor bud-derived CCL5 recruits fibroblasts and promotes colorectal cancer progression via CCR5-SLC25A24 signaling

The primary antibodies used in the study are summarized in Additional file 6: Table S1. The siRNA and primer sequences used in the study are summarized in Additional file 7: Table S2 and Additional file 8: Table S3 respectively.

The human CRC cell lines LS174T (CL-188), RKO (CRL-2577), DLD-1 (CCL-221), Caco2 (HTB-37), SW620 (CCL-227), HCT-8 (CCL-244), HCT116 (CCL-247), and HCT-15 (CCL-225); the human normal colorectal epithelial cell line FHC (CRL-1831); and the human normal colorectal fibroblast cell line CCD-18Co (CRL-1459) were all purchased from the American Type Culture Collection (ATCC). All human CRC cells were cultured in RPMI-1640 medium (Gibco, C11875500BT) supplemented with 10% fetal bovine serum (FBS) (ExCell Bio, FND500). FHC was cultured in RPMI-1640 medium supplemented with 15% FBS, and CCD-18Co was cultured in Eagle’s Minimum Essential Medium (EMEM) (ATCC, 30–2003) supplemented with 10% FBS. All cells were cultured at 37 °C in humidified atmosphere containing 5% CO₂. Cell line certificates of analysis were obtained from the ATCC. All cell lines were negative for mycoplasma.

Fresh human normal colorectal mucosae were cut using surgical scissors and then enzymatically dissociated in a mixture of type IV collagenase (2.0 mg/ml, Sigma, C5138), hyaluronidase (0.4 mg/ml, Sigma, H1115000) and DNase (25 U/ml, Solarbio, D8071) at a constant temperature of 37 °C for 2 h. The tissues were then passed through a 40-μm cell strainer to generate a single-cell suspension. The cell suspensions were centrifuged at 300 × g for 12 min; the supernatant was discarded, and the cell pellet was resuspended in EMEM. Primary cells were then plated at a density of 1 × 10⁵ viable cells in 25 cm² adherent flasks and cultured at 37 °C in EMEM with 10% FBS in humidified atmosphere containing 5% CO₂ (enzyme digestion method). The tissues that could not pass through the strainer were transferred to 25 cm² adherent flasks. EMEM supplemented with 10% FBS was added to the adherent flasks after 24 h, when the tissues had stuck to the bottom. The tissues were incubated until fibroblasts crawled out of them (improved tissue planting method).

Clinical CRC specimens were obtained from patients who were pathologically diagnosed with CRC at Shenzhen Hospital, Southern Medical University. The study was approved by the ethics committee of Shenzhen Hospital, Southern Medical University, China.

Tumor budding was identified based on the presence of a single tumor cell or a tumor cell cluster of up to 4 cells at the invasive front of CRC tumors. The quantification of tumor budding was performed according to five procedures proposed by the International Tumor Budding Consensus Conference (ITBCC) 2016 for reporting tumor budding in CRC during daily diagnostic practice [13].

IHC was performed using paraffin-embedded sections of human CRC tissue following the standard LSAB protocol (Dako). Primary antibodies against α-SMA, CD90, FAP, CCL5, SLC25A24, CD31, and VEGFA were used for IHC. The degree of staining was observed and scored independently by three pathologists. The percent positivity of CCL5 staining was scored from 0 to 4, as follows: 0 (< 5%), 1 (5–25%), 2 (26–50%), 3 (51–75%), and 4 (> 75%). The staining intensity was scored on a 4-point scale, as follows: 0 (no staining), 1 (weak staining, light yellow), 2 (moderate staining, yellowish brown), and 3 (strong staining, brown). Subsequently, the CCL5 expression score was calculated by multiplying the percent positivity score with the staining intensity score. Accordingly, the expression of CCL5 was classified as low (0–4), medium (5–8), or high (9–12).

IF analysis of cells was performed as previously described [26]. For IF analysis of CRC tissue, the steps before primary antibody incubation were the same as those used for IHC. The steps after primary antibody incubation were the same as those used for IF analysis of cells. Images of cells were acquired using a laser scanning confocal microscope (Olympus, Japan), and images of CRC tissue were acquired using a fluorescence microscope (Olympus, Japan).

Boyden transwell chambers (Corning, 353,097) were used according to manufacturer’s instructions. Briefly, 2 × 10⁴ fibroblasts were added to the upper chambers. The lower chamber contained of the following: 1 × 10⁵ human CRC tumor cells, different concentrations of human CCL5 (Peprotech, 300–06-20), or the conditioned medium (CM) samples from stable CRC cell lines. After 48 h of incubation, fibroblasts that successfully migrated to the lower chamber were fixed with 4% paraformaldehyde and stained with hematoxylin. The number of cells was counted in five random visual fields using a light microscope (Olympus, Japan).

Serum-free CM samples from FHC and the human CRC tumor cell lines HCT-8, HCT116, HCT-15, and SW620 were collected after incubation for 24 h and filtered through a 0.22-μm mesh. The CM samples were added to arrays containing antibodies against 1000 unique cytokines (RayBio, GSH-CAA-X00) and processed according to manufacturer’s instructions.

Total cellular mRNA was extracted using the TRIzol reagent (TaKaRa, 9109). The Prime-Script RT Reagent Kit with gDNA Eraser (TaKaRa, D6110A) was used to reverse-transcribe mRNA into cDNA. Finally, the SYBR Premix Ex Taq (TaKaRa, RR420) and the Applied Biosystems™ 7500 Fast Real-Time PCR System (Thermo Fisher, USA) were used for qRT-PCR. All mRNA levels were normalized based on GAPDH levels, and the 2^−ΔΔCt method was used.

CCL5 supernatant levels in the CM samples from FHC and human CRC tumor cells were measured with ELISA using a commercially available kit (CUSABIO, CSB-E17375h), as described by the manufacturer’s instructions. The results were expressed in pg/ml, and the standard curve was based on the measured OD values of the standard.

siRNAs for human CCL5, CCR1, CCR3, CCR4, CCR5, CD44, GPR75, and SLC25A24 were purchased from GenePharma (Suzhou, China). HCT-8, SW620, CCD-18Co, and human primary normal colorectal fibroblasts were transiently transfected with siRNA using the Lipofectamine 3000 Transfection Reagent (Invitrogen, L3000-015) based on the manufacturer’s instructions.

The lentivirus vector LV17 (EF-1a/Luciferase17&Puro) carrying the human CCL5-overexpressing sequence (CCL5) and the lentivirus vector LV16 (U6/Firefly&Puro) carrying the indicated CCL5-repressing short hairpin RNA (shRNA) sequence (GGGTTCGGGAGTACATCAA) (shCCL5) were obtained from GenePharma (Suzhou, China). Empty LV17 and LV16 vectors served as controls for the overexpression (Vec) and repression vectors (shCtrl), respectively. Based on the manufacturer’s instructions, stable cell lines were established by transfecting human CRC cell lines with these lentiviral vectors.

BALB/C-nude mice (male, 3–5 weeks old) were purchased from GemPharmatech Co., Ltd (Guangdong, China). They were housed under specific pathogen-free conditions in the animal facility at the Shenzhen Hospital, Southern Medical University, China. All animal experiments were approved by the ethics committee of Shenzhen Hospital, Southern Medical University, China. First, 5 × 10⁶ HCT116, SW620/shCtrl, or SW620/shCCL5 cells were subcutaneously injected into the backs of nude mice (n = 3). After 2 weeks, the IVIS Spectrum In Vivo Imaging System (PerkinElmer, USA) was used to image tumor progression in mice with SW620/shCtrl and SW620/shCCL5 xenografts. Further, 15 mg/ml D-Luciferin potassium salts (Promega, P1043) were intraperitoneally administered to the mice (dose, 10 μl/g). The mice were sacrificed, and tumors were surgically removed, fixed in 10% formalin, embedded in paraffin, and cut into 2.5-μm-thick sections for hematoxylin–eosin (H&E) staining.

After the subcutaneous tumors of HCT116 formed, one part of tumor tissues was fixed and stained with H&E, and the remaining tumor tissues were removed and cut into 1-mm³ pieces using ophthalmic scissors. Of these pieces, five were chosen and buried inside the cecal serosal layer in nude mice using purse string sutures (n = 5). After 8 weeks, the ceca of nude mice were surgically removed and processed as mentioned above.

Total proteins were isolated from cells using the RIPA lysis buffer (FDbio, FD008), PMSF (FDbio, FD0100), Protease inhibitors (FDbio, FD1001), and protein phosphatase inhibitors (FDbio, FD1002). The concentration was determined using BCA protein assay kits (FDbio, FD2001). The total proteins were separated using 10% sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE) and transferred to polyvinylidene fluoride (PVDF) membranes. The membranes were blocked with 5% skim milk (FDbio, FD0080) or 5% BSA (FDbio, FD0030) for 1 h at 25 °C and then incubated with primary antibodies at 4 °C overnight. Subsequently, the membranes were incubated with a goat anti-rabbit or anti-mouse secondary antibody (FDbio, FDR007 and FDM007). The proteins were detected using an ECL chemiluminescence solution (FDbio, FD8030) and visualized using a chemiluminescence detection system (Universal Hood II, Bio-Rad). The intensity of each immunoblot band was quantified using the NIH Image J software (National Institutes of Health, USA).

Before the recruitment assay, fibroblasts were pre-cultured with a CCR1 inhibitor (BX471, 100 nM, MedChemExpress, HY-12080A), CCR5 inhibitor (Maraviroc, 100 nM, Selleck, S2003), or Akt inhibitor (MK-2206, 2 μM, Selleck, S1078) for 48 h. During the recruitment assay, BX471, Maraviroc, or MK-2206 was added to both the upper and lower chambers.

Total mRNA was extracted from fibroblasts before and after treatment with 40 ng/ml CCL5 for 24 h. Sequencing libraries were generated using the NEBNext® UltraTM RNA Library Prep Kit for Illumina® (NEB, USA) following the manufacturer’s recommendations, and index codes were added to attribute the sequences to each sample. The clustering of the index-coded samples was performed on a cBot Cluster Generation System using the TruSeq PE Cluster Kit v3-cBot-HS (Illumina) according to the manufacturer’s instructions.

First, the high-concentration Matrigel (BD, 354,248) was diluted to half the original concentration. Then, 10 μl of this diluted Matrigel was added to each well of a µ-Slide Angiogenesis Glass Bottom (Ibidi, 81,506) and allowed to polymerize for 30 min at 37 °C. Subsequently, 50 μl medium containing 1 × 10⁴ fibroblasts (treated or not treated with 40 ng/ml CCL5 for 24 h) was incubated in the diluted high-concentration Matrigel. After 2 h, the fibroblasts were fixed with 4% paraformaldehyde and stained with hematoxylin. The number of lumens was obtained across three random visual fields using a light microscope, and capillary tubes were quantified by counting the number of lumens.

A 4 °C centrifuge was pre-cooled, and PBS containing 0.1% BSA was prepared in advance. Fibroblasts treated with 40 ng/ml CCL5 for 24 h and untreated fibroblasts were collected and washed with PBS. These cells were incubated with the APC-VE-cadherin antibody or APC-rabbit-IgG antibody at 4 °C for 30 min under dark conditions. Then, the cells were washed again with PBS. Finally, the cells were resuspended in 100 μl of PBS, and the cell suspension was transferred to a flow cytometry tube for detection. The fluorescence intensity of fluorescein isothiocyanate was quantified using the Flow-jo software (BD, USA).

Sirius Red staining was performed on paraffin-embedded sections of human CRC tissue and tumor tissue from mouse caecum using the Sirius Red Staining Kit (LEAGENE, DC0041) based on the manufacturer’s protocol. The degree of staining was observed under a polarized light microscope (Olympus, Japan) and scored independently by three pathologists. The percent positivity of Sirius Red staining was scored from 0 to 4, as follows: 0 (< 5%), 1 (5–25%), 2 (26–50%), 3 (51–75%), and 4 (> 75%). The staining intensity was scored on a 4-point scale, as follows: 0 (no staining), 1 (weak staining, light orange and/or green), 2 (moderate staining, medium orange and/or green), and 3 (strong staining, orange and/or green). Subsequently, the Sirius Red staining score was calculated by multiplying the percent positivity score with the staining intensity score. Accordingly, the level of Sirius Red staining was categorized as low (0–4), medium (5–8), or high (9–12).

The human CRC microarray profile GSE39582 was used to analyze the correlation of COL1 and COL3 expression with relapse-free survival in CRC patients. The chip platform used in this analysis was the Affymetrix Human Genome U133 Plus 2.0 Array. Based on the median values of COL1 and COL3 mRNA expression, patients were divided into low and high expression groups. Each group contained mRNA values for 283 patients. Then, the survival curves of the two groups were obtained using the Kaplan–Meier method. In addition, the human CRC microarray profile GSE39582 was also used to analyze the correlation of CCL5 expression with COL1 and COL3 expression.

Fresh subcutaneous tumor tissues from nude mice (≤ 3 mm³) were surgically removed immediately after the mice were sacrificed. The tissues were promptly added to a fixative solution for electron microscopy (Servicebio, G1102) and incubated for 2 h at room temperature before being transferred to 4 °C for storage. Subsequently, the samples were observed and photographed using a scanning electron microscope (Hitachi, Japan).

SPSS software for Mac OS version 25.0 (IBM, USA) was used for statistical analyses. An unpaired two-tailed Student’s t test was used to analyze the differences between two groups. The Mann–Whitney U test was conducted to compare the scores of CCL5 staining or Sirius Red staining between CRC tumor tissues and adjacent normal tissues. Pearson’s χ² test and Spearman’s correlation test were applied to analyze the correlation of CCL5 expression or the level of Sirius Red staining with clinicopathological features. The log-rank test was performed to analyze Kaplan–Meier survival curves. The correlation of CCL5 expression with the level of Sirius Red staining in CRC tissues, and the correlation of CCL5 mRNA expression with COL1 or COL3 mRNA expression in the human CRC microarray profile GSE39582 were analyzed using Spearman’s correlation test. All data were expressed as the mean ± standard deviation (SD). P < 0.05 was considered statistically significant (ns, no significance; *, P < 0.05; **, P < 0.01; ***, P < 0.001).

The evaluation of 195 clinical CRC tissue samples using H&E staining showed that the tumor budding-containing invasive front of CRC tissues had more fibroblasts than other parts of the tumor without tumor budding (Fig. 1a). Next, we verified that the fibroblasts around tumor buds showed a high expression of α-SMA and CD90 but a low expression of FAP (α-SMA^high CD90^high FAP^low) (Fig. 1b). These results indicated that tumor budding at the invasive front of CRC tissues was closely related to fibroblast heterogeneity.

Recently, several studies have demonstrated that fibroblasts can be recruited through the effects of cytokines secreted from tumor cells or other cells in the TME [7, 27‐29]. While exploring the contribution of CRC cells to fibroblast recruitment, two types of fibroblasts were used: the human normal colorectal fibroblast cell line CCD-18Co and human primary normal colorectal fibroblast, NF1. The human primary normal colorectal fibroblasts were isolated through the enzyme digestion or improved tissue planting method (Additional file 1: Figure S1A). To identify these fibroblasts, the epithelial marker E-cadherin and fibroblast markers vimentin and α-SMA were detected using IF (Fig. 1c and Additional file 1: Figure S1B).

Next, a co-culture recruitment assay was conducted (Fig. 1d). In this assay, human CRC tumor cells were incubated in the lower chamber, while fibroblasts were incubated in the upper chamber. The pore size of the upper co-culture chamber was 8 μm — big enough to allow fibroblasts to pass through. As shown in Fig. 1e, FHC (human normal colorectal epithelial cell line) and LS174T, RKO, DLD-1, and Caco2 (human CRC cell lines) showed a weak ability to recruit fibroblasts. In contrast, HCT-8, HCT116, HCT-15, and SW620 (human CRC cell lines) showed a stronger recruitment ability.

Subsequently, HCT116 cells, which showed a strong ability to recruit fibroblasts in vitro, were used for further in vivo orthotopic CRC xenograft mouse experiments. Using H&E staining to examine tumors in the mouse cecum, we observed that HCT116 could also recruit fibroblasts in vivo (Fig. 1f). Taken together, these data demonstrated that tumor buds of CRC tumors can recruit fibroblasts.

To further ascertain which key cytokine secreted by CRC tumor cells is responsible for fibroblast recruitment, the CM samples from FHC, which had a weak ability to recruit fibroblasts, and the CM samples from HCT-8, HCT116, HCT-15, and SW620 (human CRC cell lines), which possessed a stronger recruitment ability, were analyzed using a human cytokine array (Fig. 2a). GO functional enrichment analysis indicated that the proteins differentially expressed between these groups were involved in various activities, including “positive regulation of cell migration” (Fig. 2b) and “extracellular matrix” (Additional file 2: Figure S2A). This suggested that cytokines secreted by CRC tumor cells could recruit other cells in the TME and could be involved in the recruitment of fibroblasts, which are the main source of extracellular matrix [1].

Among the differentially expressed proteins, CCL5 showed the most up-regulation in CM samples from human CRC tumor cells (Fig. 2c). The mRNA and secreted protein levels of CCL5 were examined in FHC and eight human CRC tumor cell lines, and the results were entirely consistent with initial findings from the co-culture recruitment assay (Fig. 2d, e). Furthermore, CCD-18Co and human primary normal colorectal fibroblasts were found to secrete lower levels of CCL5 than FHC (Fig. 2f). Accordingly, it appeared that the CCL5-mediated recruitment of fibroblasts occurred through paracrine and not autocrine signaling. Next, the expression of CCL5 was detected in human CRC tissues using IHC, and the results showed that CCL5 was highly expressed in tumor buds at the invasive front (Fig. 2g).

Then, to examine whether CCL5 is the key cytokine for fibroblast recruitment, another recruitment assay was performed (Fig. 2h). Among the different concentrations of CCL5, 40 ng/ml CCL5 showed the strongest ability to recruit fibroblasts (Additional file 2: Figure S2B). At this concentration, CCL5 could recruit CCD-18Co and different human primary normal colorectal fibroblasts (Fig. 2i and Additional file 2: Figure S2C).

Finally, to ascertain whether the CCL5 that recruits fibroblasts was secreted by CRC tumor cells, CCL5 knockdown and overexpressing cell lines were established using siRNA and lentiviral constructs, respectively. The CCL5 mRNA and protein secretion levels in these cell lines were verified (Additional file 2: Figure S2D-G). A co-culture recruitment assay (Fig. 2j) showed that the ability of tumor cells to recruit fibroblasts was significantly weakened after CCL5 knockdown (Fig. 2k and Additional file 2: Figure S2H). In contrast, this ability was enhanced when CCL5 was overexpressed (Fig. 2l and Additional file 2: Figure S2I). In the recruitment assay, the CM samples from stable cells (Fig. 2M and Additional file 2: Figure S2J, K) showed effects similar to those observed with whole cells (Fig. 2N, O and Additional file 2: Figure S2L, M). Together, these results revealed that CCL5 was highly expressed in CRC tumor buds and could recruit fibroblasts from the TME.

CCR1, CCR3, CCR4, CCR5, CD44, and GPR75 are known to be possible receptors of CCL5 [18, 19]. Accordingly, to explore whether CCL5 mediates fibroblast recruitment through these receptors, siRNA was used to knock down their expression respectively in fibroblasts, and the interference efficiency was verified (Fig. 3a, b). Next, recruitment assays were performed, as described in Fig. 3c. Surprisingly, only CCR5 downregulation in fibroblasts could significantly attenuate the ability of CCL5 to recruit fibroblasts (Fig. 3d).

Further, the CCR1 inhibitor BX471 and CCR5 inhibitor Maraviroc were used to treat fibroblasts and blocking CCR1 and CCR5, respectively, before and during the co-culture recruitment assay (Fig. 3e). Only Maraviroc treatment significantly weakened the ability of CCL5 to recruit fibroblasts (Fig. 3f). The above results strongly indicated that CCL5 was involved in fibroblast recruitment through the CCR5 receptor.

In order to explore the intracellular changes in fibroblasts after CCL5 stimulation, fibroblasts were examined using transcriptome sequencing before and after CCL5 treatment (Fig. 4a). The top 20 differentially expressed genes were selected and verified in CCD-18Co and human primary normal colorectal fibroblasts before and after CCL5 treatment. Among the selected genes, SLC25A24 was found to be consistently up-regulated (Additional file 3: Figure S3A). We subsequently verified that the levels of SLC25A24 mRNA were up-regulated in more types of human primary normal colorectal fibroblasts after CCL5 treatment (Fig. 4b).

SLC25A24, also known as ATP-Mg²⁺/phosphate carrier 1 (APC1), has a regulatory N-terminal domain containing EF-hand Ca²⁺ binding sites, which allow transport in response to cytosolic Ca²⁺ elevations [30‐32]. To detect the protein expression and localization of SLC25A24 in human CRC tissues, IF assays were performed. The results revealed that SLC25A24 was highly expressed in the fibroblasts surrounding the tumor buds at the invasive front (Fig. 4c). In vivo, in tumor xenografts of shCCL5-transfected CRC cells, reduced SLC25A24 expression was observed in fibroblasts (Fig. 4d). Furthermore, tumor progression was also inhibited (Additional file 3: Figure S3B). Next, we further verified that CCL5 could up-regulate the expression of SLC25A24 protein in vitro (Additional file 3: Figure S3C).

Functionally, to explore whether CCL5-mediated fibroblast recruitment was dependent on SLC25A24, siRNA against SLC25A24 was transfected into fibroblasts, and the recruitment assay was performed (Fig. 4e). A reduction of SLC25A24 expression in fibroblasts significantly attenuated the ability of CCL5 to recruit fibroblasts (Fig. 4f, g and Additional file 3: Figure S3D). Moreover, immunoblot results demonstrated that elevations of SLC25A24 expression in fibroblasts after CCL5 stimulation were dependent on CCR5 (Additional file 3: Figure S3E). This demonstrated CCL5-mediated fibroblast recruitment was dependent on SLC25A24 and CCR5.

To further ascertain which pathways were activated by CCL5 and SLC25A24 in fibroblasts, KEGG enrichment analysis was performed using transcriptome sequencing data. The PI3K-Akt signaling pathway was shown to be the most active after CCL5 stimulation (Fig. 4h). Immunoblot results showed that CCL5 could promote the phosphorylation of Akt and mTOR in fibroblasts (Fig. 4i, left panel). In contrast, when SLC25A24 expression was inhibited, the levels of phosphorylated Akt and mTOR decreased in fibroblasts after CCL5 stimulation (Fig. 4i, right panel). Furthermore, inhibiting Akt phosphorylation in fibroblasts reduced the fibroblast recruitment ability of CCL5 (Additional file 3: Figure S3F). However, this inhibition had no effect on the expression of SLC25A24 (Additional file 3: Figure S3G). These findings suggested that CCL5 recruited fibroblasts through the SLC25A24-pAkt-pmTOR axis within fibroblasts.

Our initial findings (Fig. 1b) proved that fibroblasts around tumor buds were α-SMA^high, CD90^high, and FAP^low. To examine whether CCL5 is the main contributor to the increase in α-SMA^high CD90^high FAP^low fibroblasts, an IF assay was performed. This assay showed that the expression of α-SMA and CD90 was elevated in fibroblasts treated with CCL5 (Fig. 5a). VEGFA is necessary for the proliferation, survival, migration, and invasion of vascular endothelial cells into surrounding tissue and the subsequent generation of lumen-containing structures [33]. It has been reported that fibroblasts are one of the main sources of VEGFA [2]. Furthermore, studies have shown that fibroblasts can transdifferentiate into vascular endothelial cells and facilitate tumor progression [34‐36]. We examined the role of CCL5 in fibroblast-mediated angiogenesis. Immunoblots showed that the levels of the vascular endothelial markers FLI1, VE-cadherin, CD31, and VEGFA as well as those of α-SMA and CD90 were elevated in fibroblasts treated with CCL5 (Fig. 5b). Flow cytometry also demonstrated that the proportion of the VE-cadherin⁺ subset of fibroblasts increased after CCL5 stimulation (Fig. 5c and Additional file 4: Figure S4A). Functionally, the proangiogenic ability of fibroblasts was also found to be enhanced after CCL5 stimulation (Fig. 5d). Serial sections from 10 clinical CRC samples were stained for CCL5, α-SMA, CD90, FAP, CD31, and VEGFA using IHC. At the invasive front, CCL5 was highly expressed in tumor buds, which were surrounded by a high number of α-SMA^high CD90^high FAP^low fibroblasts and blood vessels (Fig. 5e). Immunoblots and Matrigel angiogenesis experiments showed that when fibroblasts were transfected with SLC25A24 siRNA, there was no attenuation of the CCL5-mediated increase in angiogenesis (Additional file 4: Figure S4B, C). Based on these findings, we speculated that tumor bud-derived CCL5 increased the number of α-SMA^high CD90^high FAP^low fibroblasts and promoted tumor angiogenesis via the increase in VEGFA and the transdifferentiation of fibroblasts into vascular endothelial cells.

GO functional enrichment analysis performed using transcriptome sequencing data indicated that CCL5 was also involved in extracellular matrix-related functions (Fig. 6a). It is well-established that collagen type I (COL1) and collagen type III (COL3) are the primary components of the extracellular matrix and are mainly synthesized and secreted by fibroblasts [2]. To clarify the role of CCL5 in collagen synthesis, immunoblots were performed. The findings demonstrated that CCL5 could increase the protein expression of COL1 and COL3 in fibroblasts in vitro (Fig. 6b).

To further investigate the role of CCL5 in CRC, the expression of the CCL5 protein was examined in 195 paraffin-embedded human CRC tissue samples and 162 human adjacent normal colorectal tissue samples. CCL5 expression was significantly higher in CRC tissues than in normal tissues (Fig. 6c and Additional file 5: Figure S5A). The correlation between CCL5 expression and clinical features was analyzed using Pearson’s χ² test (Table 1). Further Spearman’s correlation tests showed that high levels of CCL5 expression were positively associated with a high risk of increased tumor buds (r = 0.583, P < 0.001, Fig. 6d), deep tumor invasion (r = 0.244, P = 0.001), lymph node metastasis (r = 0.237, P = 0.001), the presence of peri-intestinal cancer nodule deposition (r = 0.198, P = 0.005), and advanced TNM stages (r = 0.256, P < 0.001) (Additional file 5: Figure S5B-E). Moreover, Sirius Red staining for COL1 (reddish) and COL3 (greenish) was performed to detect the collagen distribution in 88 of the 195 human CRC tissue samples. The expression of COL1 and COL3 was elevated in CRC tissue (Fig. 6e; Additional file 5: Figure S5F). The correlation between Sirius Red staining and the clinical features of CRC was also analyzed using Pearson’s χ² test (Table 2). Further Spearman’s correlation tests showed that a high collagen distribution around CRC tumor cells was positively related to increased tumor buds (r = 0.297, P = 0.005, Fig. 6f), deep tumor invasion (r = 0.431, P = 0.001), lymph node metastasis (r = 0.351, P = 0.001), the presence of peri-intestinal cancer nodule deposition (r = 0.288, P = 0.007), and advanced TNM stages (r = 0.442, P = 0.001) (Additional file 5: Figure S5G-J). Intriguingly, the high expression of CCL5 in tumor buds at the invasive front was often accompanied by increased collagen synthesis (Fig. 6g). Spearman’s correlation analyses revealed a positive correlation between the levels of CCL5 expression in tumor cells and the amounts of collagen in the surrounding area from the same CRC tissue sample (n = 88; r = 0.317, P = 0.003, Fig. 6h).

Subsequently, HCT116 cells, which secrete large amounts of CCL5, were used to establish orthotopic CRC xenograft mouse models for in vivo analysis. We found that collagen synthesis was increased around tumor cells at the invasive front (Fig. 6i). Furthermore, the human CRC microarray profile GSE39582 was used to analyze the relapse-free survival of patients with low and high expression of COL1 or COL3. A higher expression of COL1 and COL3 was closely related to a poor prognosis in CRC patients (Fig. 6j). Spearman’s correlation analyses using GSE39582 also showed that CCL5 mRNA expression was correlated with COL1 mRNA (r = 0.211, P < 0.001) and COL3 mRNA expression (r = 0.228, P < 0.001) (Additional file 5: Figure S5K, L). Moreover, immunoblots showed that the transfection of SLC25A24 siRNA in fibroblasts did not attenuate the CCL5-mediated increase in collagen synthesis (Additional file 5: Figure S5M). Additionally, collagen fibers in the tumor xenografts of shCtrl-transfected CRC cells were oriented in the same direction, forming thicker collagen fiber bundles. In contrast, the collagen fibers in the tumor xenografts of shCCL5-transfected CRC cells were disorganized (Additional file 5: Figure S5N). These findings demonstrated that shCCL5 substantially attenuated collagen linearization in vivo, which was suggestive of attenuated stiffness and a subsequent reduction in tumor progression [37, 38]. Taken together, these results suggested that tumor buds secreted CCL5, which then promoted collagen synthesis via fibroblasts, thus contributing to tumor progression.