CCL5 and GLUT1 define leader cells in collective invasion of colorectal cancer

Abstract

Invasion and metastasis in multiple solid tumors are promoted by collective invasion which is organized by leader and follower cells. Yet in colorectal cancer (CRC), biomarkers and mechanisms that specify leader cells remain undefined. Here we identify biomarkers of leader cells and elucidate how they drive CRC collective invasion and metastasis. Using a 3D photoconvertible CRC spheroid model, we isolated leader and follower cells. Transcriptomic, functional and in vivo assays show that GLUT1, PLOD2, and CCL5 in leader cells are indispensable for collective invasion. CCL5 activates PI3K/AKT signaling to up‐regulate GLUT1 and PLOD2 in leader cells. Moreover, GLUT1 and CCL5 could serve as specific biomarkers for leader cells in CRC collective invasion, and their co‐expression correlates with poor prognosis. Pharmacologic inhibition of GLUT1 and CCL5‐CCR5 impairs collective invasion. CCL5 and GLUT1 may define the leader cells and are required for CRC collective invasion as potential key regulators of metabolic shifts and collagen deposition. © 2026 The Author(s). The Journal of Pathology published by John Wiley & Sons Ltd on behalf of The Pathological Society of Great Britain and Ireland.

Introduction

Tumor metastasis is a multi‐stage process in which invasion constitutes the critical gateway. Tumor invasion proceeds via two distinct modes: individual movement and collective migration. Individual movement, such as amoeboid or single‐cell invasion, is driven by epithelial–mesenchymal transition, enabling tumor cells to detach from the primary tumor and infiltrate adjacent stroma. Collective migration generates multicellular streams or clusters that, under the guidance of leader cells, penetrate the surrounding stroma [1]. Previous studies have shown that aggregated tumor cells exhibit greater metastatic potential than their single‐cell counterparts [2], and that collective migration and invasion are fundamental to the progression of solid tumors [3, 4].

The spatial movement of tumor cells during collective invasion resembles that observed in angiogenesis, tissue morphogenesis, and wound repair [5, 6, 7]. In solid cancers, malignant cells frequently aggregate into variably sized clusters or ‘nests’ as they infiltrate the stromal compartment at the tumor boundary. Within these collectively invading units, cells positioned at the leading edge act as ‘leader cells’: they extend actin‐rich protrusions to remodel the surrounding stroma, set the invasion trajectory and open passageways for follower cells [8, 9]. The guidance capacity of leader cells stems from their distinct epigenetic state [10, 11], genetic heterogeneity [12], and dynamic interactions with other cellular components of the tumor microenvironment [13, 14]. In breast and non‐small cell lung cancers, spontaneous epigenetic shifts generate aggressive subpopulations that initiate and sustain collective invasion of the entire population [15, 16]. Leader cells can influence collective tumor migration pathways by degrading and remodeling the extracellular matrix [17]. Leader cells steer collective movement by transmitting mechano‐chemical cues that activate traction coupling in follower cells [18]. Sustained KRT14 expression in leader cells is required to maintain the collective invasion of breast cancer [19]. However, evidence defining leader‐cell biomarkers and their mechanisms driving collective invasion in colorectal cancer (CRC) is still scarce.

This study aimed to identify specific biomarkers of leader cells that drive collective invasion in CRC and to elucidate the molecular mechanisms underlying their role in invasion and metastasis, thereby revealing potential targets and new therapeutic strategies for CRC.

Materials and methods

Ethics approval and consent to participate

Clinical studies in this study were approved by the Nanfang Hospital Ethics Review Board, conformed to the International Ethical Guidelines for Biomedical Research Involving Human Subjects (CIOMS) [20], and were conducted according to the principles of the Declaration of Helsinki [21]. Informed consent to be included in the study or equivalent was obtained from all patients. The animal experimental procedures in this work were authorized by the Animal Care and Use Committee of Nanfang Hospital, Southern Medical University, and followed the guidelines of the Laboratory Animal Centre of Southern Medical University. All institutional and national guidelines for the care and use of laboratory animals were followed.

Multicellular spheroid culture and in vitro collective invasion model construction

After counting 5,000–10,000 tumor cell resuspensions, the cells were seeded into a 96‐well black round‐bottom ultra‐low attachment (ULA) spheroid culture plate (Corning, Corning City, NY, USA, catalogue # 4442) and transferred to a cell culture incubator maintained at 37 °C with 5% CO₂ to promote the formation of densely packed multicellular spheroids. Rat tail type‐I collagen (Corning, catalogue # 354236) was prepared to a final concentration of 3 mg/ml using a pre‐cooled pipette tip, mixed with the multicellular spheroids, and quickly added to a confocal culture dish (NEST, Jiaxing, Zhejiang, PR China, catalogue # 801002). The mixture was then incubated in a 5% CO₂, 37 °C cell culture incubator to allow collagen solidification. Finally, after complete gelation, 10% complete medium was added to continue the culture.

In pathological sections, collective invasion was characterized by a directed invasion of tumor cells in the form of multicellular clusters, which maintained intercellular connections.

Photoconversion of fluorescent proteins and flow cytometric sorting

Cells within the 3D gel were selected for in vitro collective invasion assays. Continuous laser bleaching of cell fluorescence was performed on a Carl Zeiss inverted laser confocal microscope (Carl Zeiss AG, Oberkochen, Germany), with the frequency and duration of bleaching determined by the luminescence intensity of the cellular fluorescent proteins. The photoconversion process was halted when the intensity of the post‐photoconversion red fluorescence surpassed that of the pre‐photoconversion green fluorescence.

Cells from the 3D gel were digested with collagenase (Solarbio, Beijing, PR China, C8140), thoroughly pipetted to achieve a single‐cell suspension, filtered through a 200 μm sieve to remove aggregates, and resuspended in PBS containing 2% BSA. Green and red fluorescent cells were sorted using a MoFlo XDP ultra‐high speed flow sorter (Beckman Coulter, Brea, CA, USA), collected into enzyme‐free EP tubes, and submitted to Novogene Bioinformatics (Beijing, PR China) for microtranscriptome sequencing. Additional green and red bicolor fluorescent cells were collected and cultured for subsequent validation and functional experiments.

Extracellular acidification rate assay

Leader cells and follower cells were seeded into Seahorse XF96 Cell Culture Microplates (Agilent Technologies, Santa Clara, CA, USA) at a density of 1 × 10⁵ cells per well and cultured overnight in a 37 °C incubator with 5% CO₂ until they reached 80–90% confluence. The assay medium was prepared according to the manufacturer's instructions for the Seahorse XF Glycolysis Rate Assay Kit (Agilent Technologies, catalogue # 103344‐100), supplemented with 1 mM glutamine, and held at 37 °C in a water bath.

After removing the cell culture medium, the pre‐warmed assay medium was added to the cells using a multichannel pipette. The cell plate was subsequently incubated for 45–60 min in a 37 °C incubator without CO₂. Finally, the plate was loaded into the Agilent Seahorse XF96 Analyzer for real‐time measurement of the extracellular acidification rate (ECAR).

Statistical analyses

Detailed in Supplementary Materials and Methods.

Results

Construction of a 3D multicellular photoconvertible CRC spheroids model for sorting leader and follower cells

In this study, we first constructed an in vitro 3D collective invasion model that allowed for the subsequent purification of leader and follower cells. The results showed that SW620, HCT15, LoVo, and HT29 cells formed dense multicellular spheroids (supplementary material, Figure S1A), of which SW620 and HCT15 multicellular spheroids exhibited collective invasion after being cultured in a 3D gel containing type I rat tail collagen (Figure 1A). Consequently, SW620 and HCT15 cells were selected for the subsequent collective invasion experiments.

Figure 1.

Purification of leader and follower cells of colorectal cancer (CRC). (A) Type I rat tail collagen 3D culture of CRC cell lines multicellular spheroids and the statistical analysis for the invasion area of the spheroids. The solid black lines indicate the area of collective invasion. Scale bar, 100 μm; data are represented as mean ± SD (n = 3); one‐way ANOVA. **p < 0.01; ****p < 0.0001. (B) 3D multicellular spheroids culture after pDendar2‐C transfection of CRC cell lines. ULA, ultra low attachment. Scale bar, 200 μm. (C) Fluorescence conversion of SW620‐derived leader cells, followed by pDendar2‐C transfection. White arrows indicate fluorescence from green to red conversion of leader cells. Scale bar, 50 μm. (D) Flow cytometry separated bichromatic fluorescent cells. GFP, green fluorescent cells; PE, photobleached red fluorescent cells; Q2 gate, photoconverted dual‐positive (GFP⁺ and PE⁺) cells. Western blotting analysis and quantitative analysis for the expression of EMT‐related (E, F) and stemness‐related (G, H) proteins in leader and follower cells derived from SW620. Scale bar, 100 μm; data are represented as mean ± SD (n = 5); unpaired Student's t‐test. *p < 0.05; **p < 0.01; ***p < 0.001.

Currently, no specific biomarkers can distinguish leader from follower cells in CRC. A fluorescent plasmid was transfected into SW620 and HCT15 cells (supplementary material, Figure S1B,C), resulting in green fluorescence expression (supplementary material, Figure S1D). When the multicellular spheroids were collectively invaded into the three‐dimensional gel medium, the leader cells in the multicellular spheroids were bleached and transformed with a laser to change the fluorescence of the leader cells from green to red (Figure 1B,C and supplementary material, Figure S1E). Leader and follower cells with different fluorescence were then isolated by flow cytometry (Figure 1D). Both cells expressed epithelial and mesenchymal‐associated markers, indicative of a hybrid epithelial‐mesenchymal state (Figure 1E,F). Leader cells exhibited elevated expression of stemness markers CD133, OCT4, and LGR5 compared with follower cells (Figure 1G,H). Furthermore, CD26, a recognized biomarker for metastasis‐initiating cells in CRC [22], was present at greater abundance in leader cells, implying that they possess enhanced metastatic stem cell characteristics.

Leader cells are essential in the collective invasion of CRC

To investigate the role of leader cells in colorectal cancer progression, we conducted biofunctional assays on collective invasion of tumor. In vitro functional experiments revealed that leader and follower cells showed no difference in proliferative capacity (Figure 2A). As hypothesized, follower cells demonstrated greater adhesion capacity than leader cells (Figure 2B), whereas leader cells exhibited enhanced migration and invasion capabilities (Figure 2C). Furthermore, when cultured in a 3D gel, leader cells and mixtures of leader and follower cells displayed a stronger collective invasion capacity than follower cells alone (Figure 2D).

Figure 2.

Functional analysis of leader and follower cells of colorectal cancer (CRC). (A) CCK8 assay for the proliferation ability of leader and follower cells; data are represented as mean ± SD (n = 3), unpaired Student's t‐test. (B) Adhesion test of leader and follower cells. Scale bar, 200 μm; data are represented as mean ± SD (n = 5); unpaired Student's t‐test. **p < 0.01. (C) Transwell migration and invasive assay to detect migration and invasion abilities of leader and follower cells. Scale bar, 50 μm; data are represented as mean ± SD (n = 5); unpaired Student's t‐test. ****p < 0.0001. (D) 3D multicellular spheroids culture in vitro to detect collective cell invasion ability of leader and follower cells and their co‐culture (left), quantitative analysis of invasion areas in 3D multicellular spheroids (right). The solid black lines indicate the area of the collective invasion. Scale bar, 100 μm; data are represented as mean ± SD (n = 3); unpaired Student's t‐test. **p < 0.01. (E) Subcutaneous xenograft experiment of leader and follower cells derived from SW620 in nude mice. (F) Tumor growth curves of the subcutaneous nude mouse model; data are represented as mean ± SD (n = 6); two‐way ANOVA. (G) H&E staining of tumor tissue in the subcutaneous nude mouse model. The solid black lines indicate the tumor invasion boundaries. Scale bars, 100 and 20 μm. (H) Orthotopic implantation of leader and follower cells derived from SW620 into the cecum of nude mice and liver metastasis model. Red arrows indicate liver metastases. (I) H&E staining of primary and metastatic tumor tissues in the orthotopic nude mouse model. The solid black lines indicate the tumor invasion boundaries. Scale bars, 200 and 50 μm. (J–L) In vitro chemotherapy resistance assay of SW620 leader and follower cells treated with oxaliplatin and 5‐FU at 1 μmol/l, 5 μmol/l, and DMSO in 2D and 3D spheroids culture models. Scale bar, 100 μm; data are represented as mean ± SD (n = 5); unpaired Student's t‐test. *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001.

In animal experiments, subcutaneous tumor volumes did not differ significantly between the two groups of nude mice (Figure 2E,F). Tumors derived from leader cells exhibited irregular edges with collective invasion, whereas those formed by follower cells maintained a smooth, circular border (Figure 2G). Following in situ cecal inoculation, the leader cell group developed more hepatic metastases (40%, 2/5) than the follower cell group (0%, 0/5) (Figure 2H,I). After combination treatment with oxaliplatin and 5‐FU, more leader cells survived than follower cells in both 2D (Figure 2J,K) and 3D spheroid (Figure 2L) cultures, indicating that leader cells possess greater resistance to chemotherapy. These findings suggest that leader cells exhibit stem cell characteristics.

Thus, leader cells exhibit a robust capacity for collective invasion both in vivo and ex vivo, indicating their indispensable role in CRC collective invasion.

Leader cells are enriched for genes related to activations of the HIF‐1 signaling pathway, glycolysis, amino acid biosynthesis, and metabolism

To explore the phenotypic differences between leader and follower cells, we performed RNA‐Seq analysis to identify differentially expressed genes (supplementary material, Figure S2G). KEGG enrichment analysis revealed that leader cells predominantly activated the HIF‐1 signaling pathway, glycolysis, and amino acid biosynthesis metabolism (Figure 3A; supplementary material, Table S1), whereas follower cells primarily exhibited enrichment in the oxidative phosphorylation pathway (Figure 3B; supplementary material, Table S2). Validation experiments confirmed that leader cells expressed higher levels of the hypoxia‐related and glycolysis‐related gene SLC2A1 along with genes involved in amino acid biosynthesis and metabolism compared with follower cells (Figure 3C,D; supplementary material, Figure S2A). Furthermore, six oxidative phosphorylation‐related genes showed elevated expression in follower cells (Figure 3E,F).

Figure 3.

The differentially expressed genes of leader and follower cells are enriched in metabolism‐related pathways. (A) Signaling pathways for the enrichment of highly expressed genes in SW620‐derived leader cells. (B) Signaling pathways for the enrichment of highly expressed genes in SW620‐derived follower cells. (C–F) Western blotting and quantitative analysis to verify differential protein expressions in SW620‐derived leader and follower cells; data are represented as mean ± SD (n = 3); unpaired Student's t‐test. *p < 0.05, **p < 0.01. (G) 3D multicellular spheroids culture in vitro for the detection of collective invasion ability of follower cells after oligomycin treatment (left), quantitative analysis of invasion areas in 3D multicellular spheroids (right). The solid black lines indicate the area of the collective invasion. Scale bar, 100 μm; data are represented as mean ± SD (n = 3); unpaired Student's t‐test. *p < 0.05. (H) Transwell assay in vitro to detect the invasion ability of follower cells after oligomycin treatment. Scale bar, 50 μm; data are represented as mean ± SD (n = 5); unpaired Student's t‐test. *p < 0.05.

Analysis of NADH dehydrogenase complex I content, ATP levels, and extracellular acidification rates revealed that follower cells generated more NADH complex I and ATP than leader cells (supplementary material, Figure S2B,C), while leader cells exhibited greater glycolytic capacity (supplementary material, Figure S2D). Following interference with IDH2 and SAT1 (supplementary material, Figure S2E), leader cells produced significantly less ATP than control cells (supplementary material, Figure S2F). We then treated follower cells with oligomycin, an inhibitor of oxidative phosphorylation, to assess its impact on collective invasion in CRC. Oligomycin treatment markedly enhanced the collective invasion of multicellular spheres and the in vitro invasion of follower cells (Figure 3G,H); indicating that shifts in glucose metabolism may promote collective invasion. These findings collectively demonstrate that genes associated with the HIF‐1 signaling pathway, glycolysis, and amino acid biosynthesis and metabolism are enriched in CRC leader cells.

CCL5 relies on the PI3K/AKT signaling pathway to induce GLUT1 and PLOD2 expressions in leader cells, thereby promoting the collective invasion of CRC

To identify key genes driving collective invasion, we analyzed the protein expression of the top 12 most differentially expressed genes (supplementary material, Table S3). Ten of these exhibited higher expression in leader cells than in follower cells (Figure 3C; supplementary material, Figure S2A). We next assessed how silencing each of these 12 genes in leader cells influenced collective invasion (supplementary material, Figure S3A–L). Knockdown of CCL5, SLC2A1(GLUT1), PLOD2, CXCR4, ANKRD37, and SLC2A3 (GLUT3) reduced the invasive capacity of leader cells (Figure 4A,B, supplementary material, Figure S4A,B). Among these, only CCL5, GLUT1, and PLOD2 knockdown significantly impaired collective invasion (Figure 4C; supplementary material, Figure S4C). In the subcutaneous model, knockdown of CCL5, GLUT1, or PLOD2 in leader cells attenuated tumor growth (Figure 4D,E). Subcutaneous tumors derived from control leader cells displayed irregular, collectively invasive margins, whereas those from knockdown cells showed rounded, non‐invasive borders (Figure 4F). These findings indicate that CCL5, GLUT1, and PLOD2 are essential for collective invasion in CRC.

Figure 4.

CCL5‐CCR5 promotes collective invasion of colorectal cancer (CRC) cells by PI3K/AKT up‐regulating the expressions of GLUT1 and PLOD2. (A, B) Transwell assay in vitro to detect the invasion ability of leader cells after interference with CCL5, GLUT1, and PLOD2. Scale bar, 50 μm; data are represented as mean ± SD (n = 5); unpaired Student's t‐test. **p < 0.01; ***p < 0.001. (C) 3D multicellular spheroids culture in vitro to detect the collective cell invasion ability of leader cells (left), quantitative analysis of invasion areas in 3D multicellular spheroids (right). The solid black lines indicate the area of the collective invasion. Scale bar, 100 μm; data are represented as mean ± SD (n = 3); unpaired Student's t‐test. *p < 0.05. (D) Subcutaneous xenograft experiment in nude mice of leader cells after interference with CCL5, GLUT1, and PLOD2. (E) Tumor growth curves of the subcutaneous nude mouse model; data are represented as mean ± SD (n = 6); two‐way ANOVA. *p < 0.05. (F) H&E staining of tumor tissue in the subcutaneous nude mouse model. The solid black lines indicate the tumor invasion boundaries. Scale bars, 200 and 50 μm. (G) 3D multicellular spheroids culture in vitro to detect the collective invasion ability of follower cells (left), quantitative analysis of invasion areas in 3D multicellular spheroids (right). The solid black lines indicate the area of the collective invasion. Scale bar, 100 μm; data are represented as mean ± SD (n = 5); unpaired Student's t‐test. *p < 0.05, **p < 0.01; ***p < 0.001. (H and J) Western blotting analyses for the protein expressions of GLUT1 and PLOD2 in leader cells after interference with CCL5 and follower cells after overexpression of CCL5; data are represented as mean ± SD (n = 3); unpaired Student's t‐test. *p < 0.05, **p < 0.01. (I and K) Western blotting analysis for the protein expressions of p‐PI3K, p‐AKT, GLUT1, and PLOD2 in leader cells after treatment with Maraviroc; data are represented as mean ± SD (n = 5); unpaired Student's t‐test. *p < 0.05, **p < 0.01. (L) 3D multicellular spheroids culture in vitro to detect collective cell invasion ability of leader cells. (M) Quantitative analysis of invasion areas in 3D multicellular spheroids. The solid black lines indicate the area of the collective invasion. Scale bar, 100 μm; data are represented as mean ± SD (n = 5); unpaired Student's t‐test. *p < 0.05, **p < 0.01.

The CCL5/CCR5 signaling axis promotes GLUT1‐regulated glucose uptake through activation of the PI3K/AKT and AMPK pathways, which subsequently influences macrophage migration [23]. In non‐small cell lung cancer, PLOD2 is regulated by the PI3K/AKT–FOXA1 axis and facilitates metastasis both directly by enhancing migratory capacity and indirectly through collagen reorganization [24]. Based on these findings, we hypothesized that CCL5 regulates GLUT1 and PLOD2 expression via the PI3K/AKT pathway during collective cell invasion. As anticipated, CCL5 overexpression in follower cells enhanced their collective invasion capacity (Figure 4G) and was associated with upregulated GLUT1 and PLOD2 expression (Figure 4H,J). Conversely, CCL5 knockdown in leader cells attenuated their collective invasive capacity and downregulated GLUT1 and PLOD2 levels. Critically, the impaired invasion resulting from CCL5 knockdown was rescued by restoring GLUT1 or PLOD2 expression (Figure 4L,M; supplementary material, Figure S5A,B).

Abnormal expression and activity of CCL5 and its receptor CCR5 occur in several solid tumors [25]. In this study, treatment of leader cells with the CCR5 inhibitor (Maraviroc) markedly reduced the protein levels of phosphorylated PI3K, phosphorylated AKT, GLUT1, and PLOD2, although total PI3K and Akt protein abundance was unaffected (Figure 4I,K). Collectively, these findings demonstrate that CCL5 promotes collective invasion in CRC via PI3K/AKT‐mediated upregulation of GLUT1 and PLOD2 in leader cells.

CCL5 ⁺ GLUT1 ⁺ leader cells can efficiently drive collective cell invasion of CRC

Based on the high expression of CCL5, GLUT1, and PLOD2 in leader cells and their significant contribution to collective invasion, we aimed to evaluate their potential as leader cell‐specific markers. Since GLUT1 localizes to the cell membrane [26] and CCL5 can also be detected at the membrane through binding to its receptor CCR5 [27], whereas PLOD2 is primarily located within the rough endoplasmic reticulum, we consequently selected CCL5 and GLUT1 to evaluate their potential as biomarkers for CRC leader cells. After separating SW620 cells into CCL5⁺GLUT1⁺ and CCL5⁻GLUT1⁻ populations (supplementary material, Figure S5C), we observed that ZEB‐1 and E‐cadherin expression was elevated in CCL5⁺GLUT1⁺ cells relative to CCL5⁻GLUT1⁻ cells, whereas vimentin and N‐cadherin levels were reduced. These findings suggest that both populations exhibit a partial EMT phenotype (Figure 5A). In addition, CCL5⁺GLUT1⁺ cells expressed higher levels of stemness‐related markers than CCL5⁻GLUT1⁻cells (Figure 5B). The CCL5⁺GLUT1⁺ group also demonstrated enhanced migration and collective invasion in vitro compared with the CCL5⁻GLUT1⁻ group (Figure 5C,D). Although subcutaneous tumor volumes in nude mice did not differ between the two groups, the CCL5⁺GLUT1⁺ population displayed a greater capacity for collective invasion in vivo (Figure 5E,F).

Figure 5.

CCL5 and GLUT1 are specific biomarkers in the leader cells of CRC. (A, B) Western blotting and the quantitative analysis for the expression of EMT and stemness‐related markers in CCL5⁺GLUT1⁺ and CCL5⁻GLUT1⁻ cells; data are represented as mean ± SD (n = 3); unpaired Student's t‐test. *p < 0.05; **p < 0.01. (C) Transwell migration and invasive assay to detect migration and invasion ability of CCL5⁺GLUT1⁺ and CCL5⁻GLUT1⁻ cells. Scale bar, 50 μm; data are represented as, mean ± SD (n = 5); unpaired Student's t‐test. ***p < 0.001. (D) 3D multicellular spheroids culture in vitro to detect the collective cell invasion ability of CCL5⁺GLUT1⁺ and CCL5⁻GLUT1⁻ cells (left), quantitative analysis of invasion areas in 3D multicellular spheroids (right). The solid black lines indicate the area of the collective invasion. Scale bar, 100 μm; data are represented as mean ± SD (n = 3); unpaired Student's t‐test. *p < 0.05. (E) Subcutaneous xenograft experiment of SW620‐derived CCL5⁺GLUT1⁺ and CCL5⁻GLUT1⁻ cells in nude mice. (F) H&E staining of tumor tissue in the subcutaneous nude mouse model. The solid black lines indicate the tumor invasion boundaries. Scale bars, 200 μm and 50 μm. (G) Transwell assay in vitro to detect the invasion ability of leader cells after treatment with BAY876 and Maraviroc. Scale bar, 50 μm; data are represented as mean ± SD (n = 5); unpaired Student's t‐test. **p < 0.01; ***p < 0.001. (H) 3D multicellular spheroids culture in vitro to detect collective invasion ability of leader cells after treatment with BAY876 and Maraviroc (left), quantitative analysis of invasion areas in 3D multicellular spheroids (right). The solid black lines indicate the area of the collective invasion. Scale bar, 100 μm; data are represented as mean ± SD (n = 3); unpaired Student's t‐test. *p < 0.05. (I) Subcutaneous xenograft experiment of leader cells after treatment with BAY876 and Maraviroc in nude mice.(J) Tumor growth curves of subcutaneous xenograft of leader cells after treatment with BAY876 and Maraviroc in nude mice; data are represented as mean ± SD (n = 5); two‐way ANOVA. *p < 0.05. (K) H&E staining of tumor tissue in the subcutaneous nude mouse model. The solid black lines indicate the tumor invasion boundaries. Scale bars, 200 and 50 μm.

Furthermore, SW620 leader cells were treated with the CCR5 inhibitor Maraviroc and the GLUT1 inhibitor BAY‐876. As anticipated, Maraviroc, BAY‐876 (supplementary material, Figure S5D) and their mixture inhibited the invasion and collective invasion of SW620 leader cells in vitro more significantly than the controls (Figure 5G,H). Likewise, Maraviroc and BAY‐876 markedly inhibited the collective invasion of subcutaneous tumors induced by SW620 leader cells, with the combination of both inhibitors producing an even greater suppressive effect (Figure 5I–K). These results suggest that CCL5 and GLUT1 may serve as specific biomarkers for leader cells during collective invasion in CRC.

Clinicopathological analyses of CCL5, GLUT1, and PLOD2 in CRC tissues

Pathological sections from 279 CRC patients were examined, revealing collective invasion in 162 cases (58%). Patients exhibiting collective invasion demonstrated a significantly poorer prognosis than those without this feature (p < 0.0001, Figure 6A). Moreover, even among patients without distant metastasis, the presence of collective invasion remained associated with a worse prognosis (p < 0.0001, Figure 6B). Clinicopathological analysis revealed that collective invasion in CRC was strongly associated with patients' T‐stage (χ ² = 8.956, p = 0.011), recurrence (χ ² = 13.897, p = 0.000), clinical stage (χ ² = 18.525, p = 0.000), lymph node metastasis (χ ² = 10.302, p = 0.001), distant metastasis (χ ² = 81.005, p = 0.000), and MMR status (χ ² = 47.139, p = 0.000) (supplementary material, Figure S6A–F). In contrast, collective invasion showed no significant correlation with differentiation grade (χ ² = 0.842, p = 0.656), age (χ ² = 0.865, p = 0.352), gender (χ ² = 0.087, p = 0.768), and tumor size (χ ² = 1.271, p = 0.26) (supplementary material, Figure S6G–J; Table S4). Analysis of cases with N1C (tumor deposits, TD) information further demonstrated that N1C status was strongly correlated with collective invasion (χ ² = 14.539, p < 0.001) and prognosis (χ ² = 10.977, p = 0.004) (supplementary material, Figure S6K,L; Table S4).

Figure 6.

CCL5, GLUT1, and PLOD2 are highly expressed at the invasion front of CRC. (A, B) Kaplan–Meier survival analysis of patients with colorectal cancer (CRC) with or without distant metastasis. Log‐rank test. ****p < 0.0001. (C, D) IHC and statistical analysis for the expression of CCL5 and GLUT1 in tumor cells located at the front of collective invasion of human CRC primary. The image on the right is a 2× enlargement of the IHC image in the black frame on the left. Scale bars, 50 and 20 μm. (n = 65); unpaired Student's t‐test. *p < 0.05; ****p < 0.000. (E–G) IHC and statistical analysis for the expression of CCL5 and GLUT1 in tumor cells located at the front of collective invasion of human CRC primary and liver metastases tissues. The image on the right is a 2× enlargement of the IHC image in the black frame on the left. Red arrows indicate the cells of tumor nests collective invasion front. Scale bars, 50 and 20 μm; unpaired Student's t‐test. *p < 0.05; ****p < 0.000. (H) Picro‐Sirius Red staining of human CRC tissues. The solid black line indicates the tumor invasion boundaries. Blue arrows indicate collective tumor nests. Scale bars, 50 and 20 μm. The solid yellow lines indicate the tumor invasion boundaries. (I) Picro‐Sirius Red staining of nude mouse subcutaneous tumor tissues. The solid black line indicate the tumor invasion boundaries. Blue arrows indicate collective tumor nests. Scale bars, 50 and 20 μm. (J) Kaplan–Meier survival curves of patients with CRC and CCL5 and GLUT1 expressions; log‐rank test. (K) Kaplan–Meier survival curves of patients with CRC and the co‐expression of CCL5 and GLUT1 in collective invasion fronts; log‐rank test.

In human CRC tissues, CCL5 and GLUT1 were highly expressed at the irregular invasive fronts of tumor nests but not in the tumor centres, a difference that was statistically significant (Figure 6C,D). Similarly, in 15 paired samples of CRC liver metastases, the collective invasion front exhibited higher expression of CCL5 and GLUT1 than the tumor center (Figure 6E–G). In subcutaneous tumors of nude mice, CCL5, GLUT1, and PLOD2 expression was also elevated in leader cells relative to follower cells (supplementary material, Figure S6M). A marked increase in collagen deposition (visualized by Picro‐Sirius Red staining) and type I collagen expression was observed around collectively invading tumor nests in both human CRC tissues and nude mouse subcutaneous tumors (Figure 6H,I; supplementary material, Figure S6N–P). Analysis of the TCGA (The Cancer Genome Atlas) database indicated that patients with CRC and high CCL5 and GLUT1 expression had a poorer prognosis than those with low expression (p = 0.02, Figure 6J). Furthermore, in the patient subset with follow‐up data (n = 19, Figure 1C), a robust association was observed between high co‐expression of CCL5 and GLUT1 at the invasion front and poor clinical outcome (Figure 6K).

These results collectively confirm that CCL5 and GLUT1 expression strongly correlates with collective invasion in CRC.

Discussion

Clinicopathological analyses of CRC [28, 29], breast cancer [8, 29, 30], squamous cell carcinoma [31], lung cancer [32], and melanoma [33] have established that collective cancer cell invasion is defined by persistent cell–cell adhesion and a polarized invasion front. This study aimed to identify specific biomarkers of leader cells within collectively invading CRC and to elucidate their underlying molecular mechanisms. The findings offer potential new targets and strategies for the clinical treatment of CRC.

Most in vitro models of collective tumor invasion rely on 3D cultures using matrix gels such as Matrigel or type I rat tail collagen [8, 14, 34]. However, isolating leader and follower cells for further analysis remains difficult because specific molecular markers are unavailable. To address this, we introduced fluorescent plasmids into CRC cells and developed a new photoconvertible multicellular 3D spheroid model. Marcus et al previously purified leader and follower populations via SaGA, embedded lung cancer spheroids in Matrigel, and used FACS to isolate photoconverted Texas Red and FITC expressing cells for microarray analysis [9]. In our model, selected fluorescent cells are bleached with a specific laser wavelength to alter the fluorescence of tumor cells within different regions of collective invasion. Leader and follower cells were then sorted based on their distinct post‐conversion fluorescence for subsequent experiments.

Tumors expressing stem cell‐like epithelial/mesenchymal hybrid phenotypes can exhibit collective invasion [35, 36, 37, 38]. In line with these findings, we found that sorted leader and follower cells expressed epithelial and mesenchymal markers. Stemness marker expression was also higher in leader cells than in follower cells. Leader cells additionally showed elevated CD26 expression, a marker for metastasis‐initiating cells in CRC [39]. Our experiments further confirmed that leader cells possess enhanced invasive and collective invasion capacities. These results indicate that leader cells exhibit metastatic stem cell characteristics and play a predominant role in the collective invasion of CRC.

RNA‐sequencing was conducted on leader and follower subpopulations to elucidate how leader cells orchestrate collective invasion in colorectal cancer. KEGG enrichment analysis revealed predominant activation of the HIF‐1 signaling pathway, glycolysis, and amino‐acid biosynthesis/metabolism in leader cells, whereas follower cells exhibited high oxidative phosphorylation. Even in the presence of oxygen, tumor cells preferentially generate ATP through glycolysis—though less efficient, it is faster than oxidative phosphorylation—a phenomenon known as the Warburg effect [40]. How metabolism differs between leader and follower subpopulations within the tumor remains elusive. Previous studies have shown that invasive tumor cells positioned at the leading edge of collective invasion may generate ATP via oxidative phosphorylation, as these leader cells exhibit higher energy consumption [41] and greater glucose uptake than follower cells [42]. In contrast, another investigation found that leader cells preferentially utilize mitochondrial respiration through pyruvate dehydrogenase, whereas follower cells rely more on glucose uptake for energy production [43]. Our study validated that leader cells enhance glucose uptake to drive glycolysis and synergistic amino acid metabolism during collective invasion. Although oxidative phosphorylation yields more ATP, its slower energy production cannot meet the acute energetic demands of leader cells. Glucose metabolism in tumor cells is more prone to glycolysis, especially under hypoxic conditions [44]. Our data further indicate that in response to hypoxia, tumor cells upregulate glucose transporters and enhance glycolysis within leader cells [45].

Subsequently, 12 genes exhibiting the most significant upregulation in leader cells were selected for knockdown to assess their impact on collective invasion in CRC. The results indicated that GLUT1, PLOD2, and CCL5 are essential for this collective invasion process. CCL5 activates PI3K/AKT and AMPK signaling pathways through CCR5, enhances glucose uptake regulated by GLUT1, and influences macrophage migration [23]. PLOD2 is regulated by the EGFR‐PI3K/AKT–FOXA1 axis and promotes non‐small cell lung cancer metastasis through collagen remodeling [24]. CCL5 activates intracellular signaling pathways by binding to its receptor CCR5 on the cell membrane, thereby regulating the progression of various tumors [15, 16, 46]. Based on previous data, this study demonstrated that CCL5 increased the expression of GLUT1 and PLOD2 by activating the PI3K/AKT signaling pathway in leader cells and inducing glycolysis and collagen deposition to promote the collective invasion of CRC.

Molecular markers of leader cells are critical for the collective invasion of tumors, but studies on this topic are limited. In breast cancer, collective invasion is driven by specialized cancer cells with basal epithelial gene expression, such as cytokeratin‐14 (K14) and p63. Moreover, collective invasion of K14⁺ cells occurs in most human breast cancer subtypes [34]. This study identified GLUT1 and CCL5 as specific markers for leader cells, selected because both are detectable on the cell membrane and have established roles in disease prognosis. Flow sorting was used to isolate GLUT1⁺CCL5⁺ and GLUT1⁻CCL5⁻ subpopulations from SW620 cells. We then compared the two groups for their expression of EMT and stem cell markers, as well as for collective invasion capacity both in vitro and in vivo. The results demonstrated that GLUT1⁺CCL5⁺ cells exhibit a phenotype and functional characteristics similar to those of leader cells in CRC. Furthermore, treating leader cells with the CCR5 inhibitor Maraviroc and the GLUT1 inhibitor BAY‐876 markedly suppressed collective invasion in vitro and in vivo. These findings establish GLUT1 and CCL5 as specific markers and potential therapeutic targets for collective invasion in CRC.

Finally, this study confirmed that collective invasion in patients with CRC correlates with a poorer prognosis and is significantly associated with T‐stage, recurrence, clinical stage, lymph node metastasis, distant metastasis, and MMR status. CCL5, GLUT1, and PLOD2 expression levels were significantly elevated in tumor cells at the collective invasion front in both human CRC tissues and mouse subcutaneous tumors. Moreover, patients with CRC and high CCL5 and GLUT1 expression had a worse prognosis than those with low expression. Previous studies have indicated that high expression of CCL5, GLUT1, and PLOD2 is positively correlated with poor prognosis in several solid tumors [47, 48, 49]. Our findings further demonstrate that leader cells critically influence collective invasion in CRC, a process closely associated with clinical prognosis. Furthermore, the expression of CCL5, GLUT1, and PLOD2 is essential for this collective invasion process.

In this study, we established an experimental system integrating photoconvertible protein labeling, type I collagen‐based 3D culture, and flow cytometry, which enables precise isolation and enrichment of functionally active leader cells based on their spatial position at the invasive front. This provides a reliable and controllable tool for in‐depth mechanistic studies of this critical cell subpopulation. Building on this technical advance, we identified CCL5 and GLUT1 as specific surface biomarkers of leader cells during collective invasion in CRC – a significant finding that addresses the prior absence of definitive molecular markers for this critical cell population in CRC. This discovery enables direct identification and targeting of leader cells through surface markers, independent of morphological context, thereby establishing a foundation for future clinical isolation and therapeutic intervention strategies. Furthermore, our in vivo experiments demonstrated that inhibition of CCR5 (using Maraviroc) and GLUT1 (using BAY‐876) effectively suppresses collective invasion, revealing a promising therapy approach for patients with CRC exhibiting this aggressive invasive behavior.

It is also important to acknowledge the limitations of this study. First, although our collective invasion model allows for the effective isolation and enrichment of viable leader cells based on positional information, immortalized cell lines cannot fully recapitulate the complex heterogeneity and microenvironment of patient tumors. Second, the use of immunodeficient mouse models restricts a full assessment of CCL5's role. As CCL5 is a potent chemokine involved in immune cell recruitment, its potential indirect effects on collective invasion via immunomodulation remain unexplored. Despite these limitations, our validation of CCL5 and GLUT1 in the tissues of patients with CRC exhibiting collective invasion supports their clinical relevance. Future studies will focus on developing more advanced primary cell models and employing immunocompetent in vivo systems to further elucidate the role of leader cells in CRC collective invasion.

Consequently, this study employed a photoconvertible 3D CRC spheroid model to isolate leader cells, identifying CCL5 and GLUT1 as specific markers. These genes function as key regulators of metabolic shifts and collagen deposition during collective invasion in CRC, highlighting their potential as therapeutic targets.

Author contributions statement

LL, FFW and YQD designed the study. ZWZ, JYT, ZCZ and XHZ performed the experiments. FFW and LL wrote the manuscript. XHZ, ZCZ, LYM, YY, SWX and XHP performed statistical analyses. YTH, WK and HHJ assisted in collecting tissue samples and organizing clinical information. All authors read and approved the final manuscript.

Supporting information

Supplementary materials and methods

Figure S1. Establishment and photoconversion of a 3D collective invasion model

Figure S2. Validation of differentially expressed genes between leader and follower cells isolated from SW620 colorectal cancer (CRC) cells

Figure S3. Validation of gene knockdown efficiency

Figure S4. Effect on invasion ability after interference with differentially expressed genes in vitro

Figure S5. Functional validation of key targets in leader cells and flow cytometry sorting strategy

Figure S6. Clinicopathological analysis of collective invasion in patients with colorectal cancer (CRC)

Table S1. KEGG enrichment pathway list of high expression genes in SW620 leader cells

Table S2. KEGG enrichment pathway list of high expression genes in SW620 follower cells

Table S3. List of the top 12 genes with the greatest differences between leader and follower cells in SW620

Table S4. Clinicopathologic characteristics of collective invasion in CRC patients

Data availability statement

The datasets used and/or analyzed during the current study are available from the corresponding author upon reasonable request.