TWIST1 is specifically upregulated in colorectal cancer peritoneal metastases
We analyzed a previously published CRC PM dataset, GSE183202, to investigate expression patterns of canonical EMT transcription factors (TWIST1, SNAI1, SNAI2, ZEB1) in CRC PM [13]. Differential gene expression profiling comparing CRC PM samples with unpaired primary tumors revealed a significant upregulation of TWIST1 (p = 0.031) and SNAI2 (p = 0.025), while SNAI1, and ZEB1 revealed no significant changes in expression (Figs. 1A, S1A–F). Western blot analysis of patient-derived specimens further validated TWIST1 was the main EMT transcription factor upregulated in CRC PM compared to primary tumors (Figs. 1B, S1G).
Fig. 1: TWIST1 is preferentially expressed in colorectal cancer (CRC) peritoneal metastases (PM) and regulates colon cancer cell migration, invasion, and stemness.
A Differential expression profiling of CRC PM patient samples reveals upregulation of TWIST1. B Western blot of EMT markers (TWIST1, SNAI1, SNAI2, ZEB1) in CRC PM samples, with NIH-3T3 as positive control. C TWIST1 expression in CRC liver metastasis (LM) samples shows no difference from normal tissue. D Western blot analysis of TWIST1 expression in CRC LM and PM samples. The bottom panel shows a summary of Western blot quantification from (B) and the upper (D) and Supplemental Fig. S1G. E Western blot of TWIST1 in MDST8 and MC38 TWIST1 knockout (KO) cells. Transwell migration and invasion assays for MDST8 (F, G) and MC38 (H, I) TWIST1 KO clones. Images (F, H) and statistics (G, I). Scale bar: 200 µm. Wound-healing assays for MDST8 (J, K) and MC38 (L, M) TWIST1 KO clones. Images (J, L) and statistics (K, M). Scale bar: 200 µm. Self-renewal capacity of MDST8 (N, O) and MC38 (P, Q) TWIST1 KO clones. Images (N, P) and statistics (O, Q). Scale bar: 500 µm. Results are expressed as mean ± SD. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001.
To assess TWIST1 specificity to peritoneal metastases, we examined transcriptomic data from GSE50760, which includes CRLM and primary tumors [26], and validated by comparing the protein expression between CRLM and PM tissues. The results showed that expression of TWIST1 was significantly upregulated only in PM (Fig. 1C, D). Collectively, these findings identify TWIST1 as the canonical EMT factor specifically upregulated in CRC PM, suggesting it may contribute to PM.
TWIST1 promotes migration, invasion, and self-renewal in CRC cells
Next, we performed CRISPR-Cas9-mediated knockout of TWIST1 (TWIST1KO) in the human CMS4 metastatic CRC cell line MDST8 and murine CRC cell lines CT26 and MC38 to elucidate the function of TWIST1 in CRC PM tumorigenesis, (Figs. 1E, S2A) [13, 27]. Additionally, we used shRNA-mediated knockdown in the human non-CMS4 metastatic CRC cell line LoVo to delineate TWIST1-mediated, non-CMS4 specific downstream targets (Figs. 1E, S2B) [27]. Analysis of TWIST1KO single-cell-derived colonies demonstrated a significant decrease in migration, invasion, and wound closure (Figs. 1F–M, S2C–F), consistent with the established role of TWIST1 in promoting cellular motility and invasive potential essential for the metastatic cascade. In addition, TWIST1 regulates the undifferentiated state in primary CD44+ CRC cells demonstrated by enhanced spheroid formation [17]. We challenged TWIST1KO cells in a three-dimensional (3D) suspension culture to evaluate self-renewal capacity via sphere formation. Notably, TWIST1KO MDST8 and MC38 cells exhibited reduced sphere formation capacity and spheroid diameter (Fig. 1N–Q). Collectively, these findings indicate that loss of TWIST1 leads to substantial in vitro deficiencies including reductions in migration/invasion and self-renewal, which are critical for distant organ metastasis.
SPON2 is a direct transcriptional target of TWIST1
To identify the dysregulated downstream target genes responsible for this phenotype, we performed RNA-seq on TWIST1KO as well as the shRNA-TWIST1 knockdown cells (Fig. S3). Concurrently, we conducted chromatin immunoprecipitation sequencing (ChIP-Seq) with anti-TWIST1 in wildtype cells to identify direct downstream target genes (Figs. 2A, S3). Through bioinformatic analysis integrating RNA-seq and ChIP-Seq results (Figs. 2B, S3), we identified 27 dysregulated TWIST1 direct target genes. Notably, the following genes were significantly upregulated in TWIST1KO and shRNA-TWIST1 knockdown cells: KCNAB2, MIR3142HG, PIK3IP1, PODXL2, SLAIN1, PCDHGC3, MGMT, PDCD4, ACTR3C, TNFRSF21, TNFAIP3, SLC12A2, and PPM1N. Conversely, the significantly downregulated genes in TWIST1KO and shRNA-TWIST1 knockdown cells included F3, CDC20, SYNE1, SLC7A5, CENPM, ASF1B, SFXN2, DBF4B, SLC7A1, RFX2, LPAR1, FOXL1, TNS1, and SPON2 (Figs. 2B, S3).
Fig. 2: TWIST1 regulates SPON2 expression.
A Western blot analysis of TWIST1 chromatin immunoprecipitation pull-down in LoVo and MDST8 cells. B ChIP-seq and RNA-seq integration in metastatic cell lines, showing TWIST1-bound genes downregulated in TWIST1-deficient LoVo and MDST8 cells; SPON2 highlighted (red). SPON2 expression in CRC PM (C) and LM (D) patient samples. Kaplan-Meier overall survival (E) and disease-free survival (F) by SPON2 expression (high vs. low) in TCGA samples. SPON2 mRNA expression in TWIST1 knockout cells, including MDST8 (G), CT26 (H), and MC38 (I). J Western blot analysis of SPON2 expression in whole-cell lysates and conditioned medium from TWIST1 knockout MDST8, MC38, and CT26 cells. K Top: Schematic of the SPON2 promoter showing the conserved E-box motif (CATCTG) in human and mouse sequences. Bottom: ChIP-qPCR analysis of TWIST1 binding to specific promoter regions (a, b, c) in MDST8 cells. Region “b” contains the predicted E-box. L Top: Schematic of the pEZX-mSpon2-GFP promoter reporter construct. Bottom: Representative fluorescence images of cells co-transfected with pEZX-mSpon2-GFP and a Doxycycline-inducible TWIST1 vector (pTK-TWIST1), treated with Vehicle or Doxycycline. M Western blot validation of GFP (reporter output) and TWIST1 expression in cells transfected with the indicated combinations of plasmids (pLenti-GFP, pEZX-mSpon2-GFP, pTK-TWIST1) and treated with Doxycycline. Results are expressed as mean ± SD. Statistical significance was determined with ***P < 0.001.
We further assessed these 27 target genes in CRC PM versus primary tumors to identify clinically relevant, dysregulated direct downstream TWIST1 target genes. SPON2 was significantly upregulated in CRC PM (P = 0.0084) (Fig. 2C). SPON2 is a secreted protein with very limited understanding of its biological function in development and tumorigenesis [28]. Although SPON2 expression has been correlated with CRC, its contribution to metastasis remains unclear. Examination of SPON2 expression in CRLM compared to primary tumors did not reveal preferential expression (Fig. 2D). TCGA analysis via GEPIA 2.0 (https://gepia2.cancer-pku.cn/) indicated a trend toward worse overall survival associated with high SPON2, with significantly reduced disease-free survival (Fig. 2E, F). To investigate the clinical relationship between TWIST1 and SPON2, we analyzed transcriptomic datasets and observed a significant positive correlation between TWIST1 and SPON2 expression in both CRC PM patient samples (Spearman R = 0.57) and the large-scale TCGA colorectal cancer cohort (Spearman R = 0.77) (Fig. S4A, B). Moreover, SPON2 levels significantly correlated with disease progression, exhibiting a dramatically increased hazard ratio in advanced stages of CRC (Stage 4 HR = 10.69) in analysis using TIMER 2.0 (http://timer.cistrome.org/) (Fig. S4C). Importantly, analyses of human and murine CRC cell lines with TWIST1KO demonstrated a significant reduction in mRNA and protein expression of SPON2 (both in lysate and secreted forms), further implicating SPON2 as a crucial direct target gene of TWIST1 (Fig. 2G–J).
To definitively confirm SPON2 as a direct transcriptional target, we analyzed the SPON2 promoter sequence and identified a conserved E-box motif (CATCTG) approximately 1 kb upstream of the transcription start site (Fig. 2K). ChIP-qPCR analysis validated that TWIST1 specifically binds to this E-box-containing region (Region b) compared to adjacent control regions (Fig. 2K). Furthermore, using a Spon2 promoter-driven GFP reporter system, we observed that Doxycycline-inducible TWIST1 overexpression significantly activated the Spon2 promoter, resulting in robust GFP expression (Fig. 2L, M). Given the specificity of SPON2 expression in CRC PM and its direct regulation by TWIST1, these data suggest SPON2 may be an important molecule of TWIST1-driven program.
SPON2 mediates TWIST1-driven metastatic phenotypes
With a goal of better understanding the biologic implication of SPON2, we performed CRISPR-Cas9 knockout of SPON2 in MDST8 and MC38 metastatic CRC cell lines (Fig. 3A). Consistent with the TWIST1KO phenotype, SPON2 knockout (SPON2KO) cells exhibited significant deficiencies in migration, invasion, wound healing, and sphere formation (Figs. 1F–Q, 3B–L). We also conducted neutralization of SPON2 protein with a monoclonal antibody, which resulted in significant and analogous reductions in migration, invasion, wound healing, and sphere formation, highlighting SPON2 as a potential therapeutic target (Figs. 3M–P, S5A–F). Conversely, treatment of MDST8 or MC38 cells with recombinant human SPON2 or murine SPON2, respectively, accelerated wound closure, invasion, and sphere formation in wildtype cells and also rescued the TWIST1KO and SPON2KO phenotypes (Fig. 3M–P, S5A–H). Moreover, overexpression of GFP-tagged SPON2 in TWIST1KO cells rescued the TWIST1KO phenotype, further solidifying that SPON2 is critical in mediating pro-metastatic cellular processes downstream of TWIST1 (Fig. 3Q–S, S5I–L).
Fig. 3: TWIST1-SPON2 cascade regulates colon cancer cell migration, invasion, and stemness.
A Western blot of SPON2 in MDST8 and MC38 SPON2 KO cells. Transwell migration and invasion assays for MDST8 (B, C) and MC38 (D, E) SPON2 KO clones. Representative images (B, D) and quantitative statistics (C, E). Scale bar: 200 µm. Wound-healing assay for MDST8 (F, G) and MC38 (H, I) cells on Matrigel. Representative images (F, H) and statistics (G, I). Scale bar: 200 µm. J–L Self-renewal assay for MDST8 and MC38 SPON2 KO clones. Representative images (J), statistics for MDST8 (K) and MC38 (L). Scale bar: 500 µm. Wound-healing assay for MDST8 (M) and MC38 (N) cells on Matrigel with 100 ng/ml SPON2 protein or 1 µg/ml SPON2 antibody. O Transwell matrigel invasion assay for MC38 TWIST1 or SPON2 KO cells with 1 mg/ml Matrigel ± 100 ng/ml SPON2 protein, 48 h, 10% FBS. P Self-renewal assay for MC38 cells with 100 ng/ml SPON2 protein or 1 µg/ml SPON2 antibody. Q Wound-healing assay for MC38 TWIST1 KO cells with GFP-SPON2 overexpression on Matrigel. R Transwell migration and invasion assays for MC38 TWIST1 KO cells with GFP-SPON2 overexpression. S Self-renewal assay for MC38 TWIST1 KO cells with GFP-SPON2 overexpression. Results are expressed as mean ± SD. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001.
Stromal SPP1 activates the TWIST1-SPON2 axis via PI3K/AKT signaling
TWIST1 expression, specifically in CRC, is an important regulator of tumorigenesis and metastasis through transcriptional regulation of dedifferentiation of the CD44+ CRC population [17]. One of the established ligands for CD44 is secreted phosphoprotein 1 (SPP1). SPP1 is a paracrine factor secreted by cells within the tumor stroma in various cancers including CRC PM [19, 29]. Therefore, we examined whether SPP1 stimulation in tumor cells could induce expression of TWIST1 and subsequent transcription of SPON2. Stimulation of MDST8 and MC38 cells with SPP1 led to a significant dose-dependent induction in TWIST1 and SPON2 transcript levels, protein expression with quantification showing a clear linear correlation, and subsequent SPON2 secretion (Fig. 4A, B). Furthermore, qPCR analysis confirmed that SPP1 treatment in WT cells significantly upregulated both Twist1 and Spon2 mRNA, an effect that was completely abolished in TWIST1KO cells (Fig. 4B). We further established stable MC38 cells transfected with pEZX-mSPON2-GFP, a vector with GFP under the control of the murine Spon2 promoter (Fig. 2L). Stimulation of these cells with recombinant Spp1 protein resulted in enhanced GFP expression, confirming the regulation of Spon2 gene transcription by Spp1 in cancer cells (Fig. 4C, D).
Fig. 4: SPP1 enhances the TWIST1-SPON2 cascade.
A TWIST1 and SPON2 expression in MC38 cell lysates and SPON2 in conditioned medium after 24-h SPP1 pulse-chase at indicated concentrations. Cells were starved for 48 h prior to treatment. Right panel: Quantification of relative protein expression and secretion. B RT-qPCR analysis of Twist1 and Spon2 mRNA levels in MC38 WT and TWIST1 KO cells treated with vehicle or 100 ng/ml SPP1 for 24 h. C, D Analysis of Spon2 promoter activity using a pEZX-mSpon2-GFP reporter in MC38 cells treated with Vehicle, SPON2, or SPP1 protein (100 ng/ml) for 24 h. Representative fluorescence images (C) and Western blot of GFP expression (D) are shown. E TWIST1, p-AKT, p-ERK, SPP1, and SPON2 in lysates after 24-h SPP1 pulse-chase at indicated concentrations in MC38, MDST8, and CT26 TWIST1 KO cells starved 48 h. Right panels: Quantification of relative SPON2 and TWIST1 expression. F TWIST1, p-AKT, p-ERK, SPP1, and SPON2 in MC38 cells treated with 100 ng/ml SPP1 and 1 µM PI3K/AKT inhibitors (MK2206 or LY294002) for 24 h. G Western blot of GFP expression in MC38 cells transfected with the pEZX-mSpon2-GFP reporter and treated with 100 ng/ml SPP1 ± 1 µM MK2206. H Expression of TWIST1 and SPON2 in MC38 cells treated with 1 or 5 µg/ml SPP1 monoclonal antibody for 24 h. Results are expressed as mean ± SD. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001.
Prior studies have reported that SPP1 regulates TWIST1 function by mediating signaling through the AKT and ERK pathways in breast cancer [30]. In CRC cells, SPP1 stimulation resulted in increased AKT phosphorylation without affecting ERK phosphorylation (Fig. 4E). This specific signal activation was consistent across multiple cell lines (MC38, MDST8, CT26), where SPP1 treatment dose-dependently increased p-AKT levels alongside TWIST1 and SPON2 upregulation (Fig. 4E, right panels and Fig. S6A). Furthermore, inhibition of the PI3K/AKT pathway utilizing small molecule inhibitors, MK2206 and LY294002, reduced TWIST1 protein levels following SPP1 stimulation, along with concurrent decreases in SPON2 mRNA and protein expression (Figs. 4F, S6B). The SPON2 promoter assay confirmed that AKT inhibition with MK2206 blocked SPP1-mediated SPON2 transcription (Figs. 4G, S6C). Additionally, neutralization of secreted SPP1 with a specific antibody suppressed SPP1-induced expression of TWIST1 and SPON2 proteins (Fig. 4H). To further explore the functional consequences of this signaling axis, we examined the impact of SPP1 on EMT. Our recent study has shown that mesothelial cells are an important source of stromal SPP1 in CRC PM. Therefore, we established an immortalized mesothelial cell line from the mesothelium of mouse omentum (OmenMeso). Co-culture with our proprietary OmenMeso cells or treatment with recombinant SPP1 induced a distinct morphological change in MC38 cells, characterized by a spindle-shaped, mesenchymal phenotype (Fig. S6D). Western blot analysis confirmed this EMT shift, showing upregulation of mesenchymal markers (NCAD, VIM, TWIST1) in both SPP1-treated and OmenMeso co-cultured cells (Fig. S6E). Collectively, these results demonstrate that the TWIST1-SPON2 regulatory cascade is induced through SPP1 in a PI3K/AKT-dependent manner and drives an EMT program conducive to metastasis.
SPON2 induces mesothelial-to-CAF transition and paracrine SPP1 secretion
We have previously shown in CRC PM, SPP1 can be derived from stromal cells and cancer cells [19]. First, we investigated whether TWIST1-SPON2 regulates the autocrine SPP1 signaling within cancer cells. Notably, comparison of wildtype cells to TWIST1KO cells revealed a significant decrease in secreted SPP1 protein levels, despite no alteration in cytoplasmic levels (Fig. S7A). Treatment of MC38 cells with recombinant SPON2 protein resulted in a significant increase in secreted SPP1 (Fig. 5A). Conversely, TWIST1KO cells exhibited suppressed secretion, even in setting of SPON2 stimulation (Fig. 5A), suggesting that the tumor cell-intrinsic SPP1 secretion mechanism may be mediated downstream of TWIST1.
Fig. 5: SPON2 regulates SPP1 expression and secretion to promote colon cancer cell migration, invasion, and stemness.
A SPP1 in MC38 cell lysates and conditioned medium after 4-h or 24-h SPON2 pulse-chase at indicated concentrations, starved 48 h prior to the SPON2 pulse-chase. Right panel: Quantification of relative SPP1 secretion. B SPP1 in PanMeso, OmenMeso, and NIH-3T3 cell lysates after 24-h 100 ng/ml SPON2 treatment. SPP1 in PanMeso and OmenMeso cell lysates after incubation with MC38 TWIST1 KO (C) and SPON2 KO (D) conditioned medium. RT-qPCR analysis of mesothelial/CAF markers (Spp1, Col1a1, Tgfb1, Pdgfrb, Upk3b, Nkain4) in OmenMeso (E) and PanMeso (F) cells treated with 100 ng/ml SPON2 protein for 24 h. G Western blot analysis of signaling pathway activation (p-FAK, p-Src, p-EGFR, p-AKT, p-ERK) in MC38 and OmenMeso cells treated with increasing concentrations of SPON2 (0–1000 ng). Right panels: Quantification of relative signaling activation. H Western blot analysis of SPP1 secretion and downstream signaling (p-FAK, p-Src, p-ERK, p-AKT) in MC38 cells treated with SPON2 protein in the presence of neutralizing antibodies (β1 Ab eBioHMb1-1, β3 Ab 2C9.G3) or SRC/FAK inhibitors (Dasatinib, Defactinib). Wound-healing assays for MC38 (I) and MDST8 (J) cells on Matrigel with 100 ng/ml SPP1 protein or 1 µg/ml SPP1 antibody. Self-renewal assay for MC38 (K) and MDST8 (L) cells with 100 ng/ml SPP1 protein or 1 µg/ml SPP1 antibody. M Left: Schematic representation of the transwell co-culture system with OmenMeso cells. Right: Quantification of migration and invasion for MC38 WT, TWIST1 KO (TK2-204), and SPON2 KO cells co-cultured with or without OmenMeso cells. N Quantification of migration and invasion for MC38 cells treated with SPON2 neutralizing antibody in the presence or absence of OmenMeso co-culture. O Schematic representation of the SPON2-SPP1 feedback loop. Results are expressed as mean ± SD. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001.
Next, we examined whether tumor-derived secreted SPON2 influences stromal cells, particularly in regulating SPP1 secretion. Our recent study identified mesothelial cells—lining the peritoneal surface—as a major cell type involved in PM. Importantly we also found that mesothelial cells are a major source of stromal SPP1 in CRC PM [31]. Given the importance of tumor-mesothelial interactions in CRC PM progression, we investigated the effect of SPON2 on mesothelial cells [19]. We treated two mesothelial cell lines, PanMeso and OmenMeso—derived from pancreatic and omental mesothelium, respectively—with recombinant SPON2 or cancer cell-conditioned medium (CM). Treatment with SPON2 or CM from wildtype MC38 cells significantly increased SPP1 levels in mesothelial cells compared to treatment with CM from TWIST1KO cells (Fig. 5B, C). Additionally, CM from SPON2KO cells resulted in reduced SPP1 expression, suggesting SPON2 as an upstream regulator of SPP1 in stromal cells (Fig. 5D). We then transfected OmenMeso cells with a Spp1 promoter-driven mCherry reporter (pEZX-mSpp1-mCherry). Upon stimulation with Spon2 protein the mesothelial cells showed enhanced mCherry expression, validating that Spon2 directly regulates Spp1 transcription in stromal cells (Fig. S7B–D). The distinct responses of cancer and stromal cells to SPON2, where SPON2 enhances cancer cell SPP1 secretion, while upregulating both intracellular and secreted SPP1 in stromal cells, may reflect cell-type-specific signaling pathways or receptor expression profiles.
To further investigate whether tumor-secreted SPON2 mediates plasticity of the mesothelial cells, we performed quantitative PCR and western blot analysis on a panel of fibroblast differentiation markers in PanMeso and OmenMeso cells following treatment with recombinant mSPON2. The results demonstrated that SPON2 significantly promoted the upregulation of genes including Col1a1, Tgfb1, Pdgfrb, and Spp1, as well as protein expression of CAF markers Col1a1 (Fig. S7E), suggesting that SPON2 can drive the differentiation of PanMeso and OmenMeso into cancer-associated fibroblasts (Fig. 5E, F). This transition was further supported by morphological changes observed in co-culture (Fig. S7F). Such novel finding highlights that tumor-derived SPON2 induces SPP1 secretion in mesothelial cells within the tumor microenvironment and is also critical in stimulation of mesothelial cells during PM.
SPON2 signaling in mesothelial cells is Src/FAK-dependent and integrin-independent
The importance of SPON2 in metastatic CRC has been previously established, with various receptors, including α5β1, proposed to mediate the downstream effects of SPON2 [28]. While the precise mechanism by which SPON2 induces SPP1 secretion remains unclear, we hypothesized it might involve integrin-mediated signaling, as described for other matricellular proteins [32]. To elucidate the precise mechanism by which SPON2 induces SPP1 secretion, we analyzed integrin downstream signaling pathways. Treatment with SPON2 induced a dose-dependent phosphorylation of Egfr (Y1068) in MC38 cancer cells, whereas it induced phosphorylation of Fak (Y397) and Src (Y416) in OmenMeso mesothelial cells (Fig. 5G), suggesting distinct SPON2 signaling pathways in tumor versus mesothelial cells. Crucially, however, the SPON2-induced upregulation of SPP1 secretion and downstream signaling activation (p-Fak, p-Src) were not affected by neutralizing antibodies against Integrin β1 or β3, but were abrogated by the Src inhibitor Dasatinib and the FAK inhibitor Defactinib (Fig. 5H). These results definitively identify that SPON2 promotes SPP1 secretion and signaling activation via a Src/Fak-dependent pathway, but independent of Integrin β1 or β3, implying that the specific SPON2 receptor on mesothelial cells remains to be identified.
To assess the functional consequences of SPON2-induced SPP1 expression, we treated wild-type MC38 and MDST8 cells with recombinant SPP1 protein and observed a significant increase in wound closure, migration, invasion, and sphere formation (Figs. 5I–L and S7G–J). Notably, neutralization of SPP1 markedly reduced these pro-tumorigenic properties, demonstrating that SPP1 is a key mediator of tumor cell aggressiveness (Figs. 5I–L and S7G–J). To confirm the role of the mesothelium in this process, we utilized a transwell co-culture system (Figs. 5M, S7E). Co-culture with OmenMeso cells significantly enhanced the migration and invasion of wild-type MC38 cells; however, this enhancement was abolished in Twist1KO and Spon2KO cells (Figs. 5M, S7K). Furthermore, treatment with a Spon2 neutralizing antibody effectively blocked the co-culture-induced migration and invasion (Figs. 5N, S7L). These findings further support that the TWIST1-SPON2 axis drives SPP1 expression and secretion in stromal cells, thereby establishing a pro-tumorigenic microenvironment that facilitates CRC metastasis. Collectively, our results highlight SPON2 as a pivotal regulator of stromal remodeling and tumor progression, with potential implications for therapeutic targeting of the TWIST1-SPON2-SPP1 signaling axis in metastatic CRC.
Furthermore, SPP1-induced upregulation of SPON2 was found to be TWIST1-dependent, indicating the presence of a positive feedback loop that reinforces TWIST1-SPON2 signaling and sustains a pro-metastatic tumor microenvironment in CRC (Fig. 4A, B). This suggests that the TWIST1-SPON2 axis actively drives SPP1 expression and secretion in stromal cells, leading to an accumulation of SPP1 within the tumor microenvironment. This enriched SPP1 pool enhances PI3K/AKT-driven TWIST1-SPON2 signaling in cancer cells, ultimately promoting cancer cell invasion, enrichment of stem-like properties, and metastasis (Fig. 5O).
The SPP1-TWIST1-SPON2 axis drives peritoneal carcinomatosis and immune exclusion in vivo
To assess the contribution of a SPP1-TWIST1-SPON2 axis in CRC PM, we utilized a syngeneic model. We established a syngeneic intraperitoneal PM model using wild-type (WT) and Spp1-deficient (Spp1-/-) mice injected intraperitoneally with 50,000 MC38-WT or MC38-Twist1KO cells. After 28 days, we assessed peritoneal tumor burden using the PCI and evaluated ascites formation [33, 34]. In WT mice injected with MC38-WT cells, the highest tumor burden was observed, as reflected by significantly elevated PCI scores (Figs. 6A, B, and S8A). Additionally, these mice exhibited significant ascites formation, occurring in 62.5% of the control group compared to complete absence (0%) in the knockout groups (Fig. 6C). Interestingly, depletion of stromal SPP1 or tumor-intrinsic TWIST1 led to a significant reduction in Spp1, Spon2, and Col1a1 expression within the tumor, while concurrently increasing Cd45 expressing cells (Figs. 6D and S8B–E). Detailed immunophenotyping revealed that this increase in immune infiltration was driven specifically by CD8+ cytotoxic T cells, which were significantly enriched in the tumors of the intervention groups, whereas CD4 + T cell and CD11b+ myeloid cell populations remained largely unchanged (Fig. 6D, E). These findings suggest that the SPP1-TWIST1-SPON2 signaling cascade likely contributes to the development of a microenvironment that promotes CRC PM. Additionally, we assessed tumor burden using PCI scores and ascites incidence in mice injected with 50,000 MC38-WT or MC38-Spon2KO cells to further investigate the impact of SPON2 on tumor progression. Loss of Spon2 in MC38 cells resulted in a significant reduction in PCI scores, ascites formation, and tumor stroma (Figs. 6F–H and S8F–G). Conversely, Spon2 depletion led to an increase in CD45-positive immune cell infiltration. Consistent with the Twist1/Spp1 depletion models, this immune infiltration in Spon2-deficient tumors was characterized by a specific and significant upregulation of CD8+ T cells, with no significant differences observed in CD4+ or CD11b+ populations (Fig. 6I, J), suggesting that Spon2 contributes to the establishment of a fibrotic and immune suppressed tumor microenvironment (Fig. 6H). Collectively, these findings demonstrate that tumor-intrinsic TWIST1-SPON2 signaling is a key driver of CRC PM by promoting stroma-derived SPP1 secretion. This axis fosters a pro-tumorigenic microenvironment characterized by enhanced fibrosis and suppression of cytotoxic T cell immunity, highlighting its potential as a therapeutic target in metastatic CRC.
Fig. 6: The TWIST1-SPON2-SPP1 cascade regulates peritoneal metastasis.
A Representative images of peritoneal metastases in mice following intraperitoneal injection of MC38 cells. C57BL/6J mice, with or without Spp1 gene knockout, were injected with 5 × 10⁴ MC38 cells, with or without Twist1 knockout, imaged at 28 days. Peritoneal carcinomatosis index (PCI) (B) and incidence of ascites (C) of mice that underwent intraperitoneal injection of MC38 Twist1 knockout cells. D Immunohistochemical (IHC) staining of SPP1, SPON2, COL1A1, CD45, CD8, CD4, and CD11b in tumor tissue sections from the indicated groups. Scale bar: 200 µm. E Quantification of immune cell infiltration (ratio of CD8+, CD4+, and CD11b+ cells) in tumor tissues based on IHC staining from (D). F Representative images of peritoneal metastases in mice following intraperitoneal injection of MC38 cells with Spon2 knockout, imaged at 28 days. Peritoneal carcinomatosis index (PCI) (G) and incidence of ascites (H) of mice that underwent intraperitoneal injection of MC38 Spon2 knockout cells. I Immunohistochemical staining of SPP1, SPON2, COL1A1, CD45, CD8, CD4, and CD11b in tissue sections of MC38 peritoneal metastases. Scale bar: 200 µm. J Quantification of immune cell infiltration (ratio of CD8+, CD4+, and CD11b+ cells) in tumor tissues based on IHC staining from (I). Results are expressed as mean ± SD. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001.
