Liver regeneration–associated hepatocellular YAP1 activation prevents colorectal cancer liver metastasis through glutamine competition

Abstract

The literature suggests that hepatocellular Yes-associated protein 1 (YAP1) signaling is activated following hepatectomy and that such activation can suppress the growth of metastatic liver tumors. The prognosis of a real-world cohort of 240 patients with colorectal cancer liver metastasis (CRLM) undergoing major and minor hepatectomy was compared after adjusting for confounding factors. To model CRLM, we induced liver metastasis in mice by transsplenically injecting MC38 cells. We found that patients with CRLM and mice undergoing major hepatectomy had better survival compared to those undergoing minor hepatectomy. Mechanistically, extensive hepatectomy activates hepatocellular YAP1 by regulating the epidermal growth factor receptor, altering glutamine metabolism–related gene expression and increasing liver glutamine consumption. This metabolic shift leads to glutamine scarcity in tumor cells, causing increased reactive oxygen species production, which promotes loss of YAP1 activity in tumor cells. Consequently, the production of the chemokine CXCL5 is suppressed, which inhibits myeloid-derived suppressor cell infiltration and enhancing the immunological function of CD8⁺ T cells.

Liver YAP1 activation after partial hepatectomy prevents the survival of liver metastases originating from colorectal cancer.

INTRODUCTION

Colorectal cancer (CRC) is the third most common cancer globally and the second leading cause of cancer-related death (1). The liver is the primary site for CRC metastasis, with ~50% of patients developing liver metastases (2). Standard treatments for colorectal cancer liver metastasis (CRLM) include curative resection and chemotherapy. However, multiple CRLMs are present in up to 80% of cases at diagnosis (3), limiting surgical options. Surgery is feasible in only 10 to 20% of cases because of factors such as tumor location, size, number, and extrahepatic metastasis. The liver’s dual blood supply and immune-tolerant environment may explain its susceptibility to metastasis (4). In 1889, Paget’s “seed and soil” hypothesis suggested that tumor cells (“seeds”) thrive in specific organs (“soil”), with many “seeds” remaining in the liver postresection (5). Consequently, the recurrence rate within a few years remains high, so recurrence prevention strategies are urgently needed.

Historically, hepatectomy has been limited to patients with few and small metastatic tumors. However, advances in chemotherapy and surgical techniques have increased the availability of these options to patients with greater tumor burden. Some patients have metastases confined to several liver segments, half the liver, or an ipsilateral trilobe. Two main surgical approaches are used for these patients: anatomic segmental hepatectomy (ASH), which is defined as the resection of one or more anatomic liver segments, and parenchymal-sparing hepatectomy (PSH), which aims for oncologic resection with minimal margins while preserving as much liver parenchyma as possible. Although ASH improves the survival of patients with hepatocellular carcinoma (HCC) (6), its effectiveness in patients with CRLM remains debated. In 2000, DeMatteo and colleagues reported that ASH, which involves removing the entire Couinaud segments affected by tumors, may decrease the recurrence rate and improve outcomes compared to those with wedge resection (7). However, Zorzi et al. reported no significant difference between anatomic resection and wedge resection in terms of tumor clearance, recurrence, or survival (8).

The Hippo/Yes-associated protein 1 (YAP1) signaling pathway plays a key role in regulating cell proliferation and liver regeneration (9, 10). In adult mouse livers, YAP1 is expressed mainly in the bile duct and endothelial cells, with minimal or no expression in hepatocytes (9). The active YAP1 protein is localized in the nucleus and functions as a transcriptional coactivator by binding to transcription factors. YAP1 is activated in regenerating hepatocytes after partial hepatectomy but is suppressed once regeneration is complete (11). Moya and colleagues reported that peritumoral YAP1 activation can inhibit the growth of liver tumors, including melanoma-derived liver metastases and intrahepatic cholangiocarcinoma, in mice (12).

The difference between ASH and PSH extends beyond the anatomical structure to the proportion of liver resected, which might influence the YAP1 signaling pathway. This led to the hypothesis that prognosis may depend more on the degree of liver resection than on the anatomical structure.

RESULTS

CRLM with major hepatectomy is associated with a better prognosis

According to the Brisbane 2000 terminology (13), partial hepatectomy is categorized according to the proportion of resected liver: Major hepatectomy involves resection of three or more Couinaud segments, whereas minor hepatectomy involves resection of fewer than three segments. To maximize the biological distinction between the two groups, we refined our definitions of major and minor hepatectomy (Fig. 1A). On the basis of liver volumetric anatomy (14), the left and right hemilivers account for ~40 and ~60% of the total liver volume, respectively. Therefore, procedures such as hemihepatectomy or trilobectomy—which inherently involve resection of >40% of the liver mass—were classified as major hepatectomy. Conversely, because individual liver segments constitute <15% of the total volume (15), we defined minor hepatectomy as resection of less than a complete segment (<15% resection). This classification ensures a clear volumetric and functional separation between the two groups.

Fig. 1. Major hepatectomy for CRLM indicates a better prognosis.

(A) Illustration of two surgical approaches for patients with multiple CRLMs. (B and C) Survival and recurrence rates of patients with CRLMs in the minor hepatectomy and major hepatectomy groups. (D) Representative MRI scans (from preoperation to 24 months postsurgery) for patients in each group. (E) Schematic of the hepatectomy-related CRLM mouse model. (F) Representative MR images of CRLM model mice with/without hepatectomy. (G) Normalized tumor numbers in each group (normalized by the remnant liver proportion). (H) Survival time of the CRLM model mice in each group.

We analyzed 240 patients with CRLM who underwent either major or minor hepatectomy to clarify the inconsistent outcomes reported in previous studies (the characteristic of the patients was provided in table S1). To control for confounding factors, we included only patients with three to eight metastases, a maximum tumor size of less than 5 cm, and no lymphoid or extrahepatic metastases. All the metastases were surgically resected without the use of ablation techniques. The results revealed that the major hepatectomy group had longer survival and lower recurrence rates postsurgery (Fig. 1, B to D), whereas other factors, such as tumor size and number, did not significantly affect prognosis (fig. S1, A to D). The univariate and multivariate analyses of factors associated with overall survival and recurrence-free survival in patients with major hepatectomy or minor hepatectomy operation were shown in table S2, which indicated that the operation method and tumor size were independent predictors of prognosis of patients with CRLMs. For the mouse studies, to ensure that the density of the tumor “seeds” was equal, MC38 cells were transsplenically injected into mice according to the proportion of the remaining liver 3 days after hepatectomy or sham operation (Fig. 1E). The results indicated that mice that underwent 70% partial hepatectomy (70% PHx, regarded as major hepatectomy) had fewer tumors and longer survival times than those with 30% partial hepatectomy (30% PHx, regarded as minor hepatectomy) (Fig. 1, F to H). These findings were confirmed in another mouse model in which cancer cells were injected 3 days before hepatectomy (fig. S2). Thus, major hepatectomy is associated with better survival outcomes than minor hepatectomy in both human and animal models.

Major hepatectomy inhibits the growth of mouse CRLMs by up-regulating hepatocellular EGFR and subsequently activating YAP1

To investigate how major hepatectomy affects CRLM growth and recurrence, we performed RNA sequencing (RNA-seq) analysis of liver tissues collected 3 days postoperation from sham, 30% PHx, and 70% PHx groups. Gene Set Enrichment Analysis (GSEA) of the transcriptional profiles revealed the YAP1 conserved signature as one of the three most significantly activated pathways following major hepatectomy (Fig. 2A and table S3). This finding was particularly intriguing given our experimental demonstration of major hepatectomy’s suppressive effect on CRLM progression and prompted us to investigate whether this antitumor effect might be mediated through liver regeneration–induced YAP1 signaling activation. The differential expression patterns of genes comprising the YAP conserved signature across treatment groups were clearly visualized in the heatmap analysis, with strongest activation observed in the 70% PHx cohort (Fig. 2B). These results align with emerging evidence of YAP1’s context-dependent roles in both liver regeneration and tumor modulation, suggesting a potential mechanistic link between regenerative signaling and metastatic control that warrants further investigation. YAP1 has five serine phosphorylation sites, with the S127 site being phosphorylated by the upstream kinase LATS, reducing YAP1 nuclear import and activity (16). Immunohistochemistry (IHC) (Fig. 2C) and Western blotting (WB) (Fig. 2D) revealed a resection extent–dependent decrease in phosphorylated YAP1 in hepatocytes, indicating increased nuclear translocation and activation of YAP1. Real-time quantitative reverse transcription polymerase chain reaction (qRT-PCR) confirmed the activation of the YAP1 pathway through the increased expression of its classical downstream genes Ccn1 and Ccn2 (Fig. 2E). Subsequently, we overexpressed YAP1 in hepatocytes using hepatocyte-specific adeno-associated virus 8 (AAV8) 1 week after CRLM modeling (Fig. 2F), which resulted in smaller tumors (Fig. 2, G to I) and extended survival times (Fig. 2J). These effects were reversed by the YAP1 inhibitor verteporfin (Fig. 2, K to O), which also negated the tumor-suppressive effect of hepatectomy (fig. S3). These results collectively demonstrated that the function of major hepatectomy to CRLMs was mediated by YAP1 signaling activation in hepatocytes.

Fig. 2. Major hepatectomy increases hepatocellular EGFR levels and contributes to YAP1 activation.

(A) GSEA of RNA-seq data from mouse livers 3 days after hepatectomy. (B) Heatmap of gene expressions of YAP1 conserved signature in RNA-seq data from mouse liver in sham, 30% PHx, and 70% PHx groups 3 days after surgeries. (C and D) IHC and WB analyses of YAP1^S127 and nuclear YAP1 in the remaining livers. (E) qRT-PCR analysis of YAP1 downstream genes Ccn1 and Ccn2 levels in the remnant liver. (F) Schematic of the CRLM mouse model with/without YAP1 overexpression in hepatocytes generated via AAV8. (G and H) Liver–to–body weight ratios and tumor-to-liver volume ratios of the negative control (NC) and overexpression (OE) groups. (I) Schematic, images, IF, and hematoxylin and eosin (H&E) staining of livers from the NC or OE groups. (J) Survival curve of the CRLM model mice in the NC and OE groups. (K) Schematic of the CRLM mouse model with/without hepatocellular YAP1 overexpression and verteporfin administration. (L and M) Liver–to–body weight ratios and tumor-to-liver volume ratios of the mice in the NC, YAP1 OE, and YAP1 OE with verteporfin groups. (N) Schematic, images, MRI scans, and IF and H&E staining of livers and lungs from CRLM model mice in the above three groups. (O) Survival curve of the CRLM model mice in the above three groups. (P and Q) WB and IHC analyses of EGFR levels in the remnant liver tissue in sham, 30% PHx, and 70% PHx groups. (R to T) WB and IF analyses of YAP1 levels in AML12 cells after the addition of EGF and canertinib. (T) WB analysis of EGFR and YAP1 levels after major hepatectomy and the addition of canertinib. (U) CoIP analysis showing the combination of the YAP1 and EGFR proteins. IgG, immunoglobulin G.

Epidermal growth factor receptor (EGFR) is known to play a critical role in liver regeneration after partial hepatectomy (17–19) and is reportedly associated with YAP1 signaling (20, 21). To explore how YAP1 activation is affected by major hepatectomy, we assessed the EGFR level in hepatocytes after major hepatectomy and found that EGFR expression was significantly higher in hepatocytes of the remnant liver than in those of the normal liver (Fig. 2, P to Q). To further prove the upstream/downstream relationship between the EGFR and YAP1, we added EGF to the culture medium of AML12 cells; we found that EGF treatment increased YAP1 levels in the nuclei of the cells through WB (Fig. 2R) and cell immunofluorescence (IF) analysis (Fig. 2S). In vivo, EGFR inhibition via canertinib posthepatectomy reduced both EGFR and YAP1 levels in hepatocytes (Fig. 2T), confirming that the EGFR is an upstream regulator of YAP1.

Our previous research verified that the EGFR can translocate into the nucleus and act as a transcription factor (22, 23), so we performed a chromatin immunoprecipitation (ChIP) experiment to explore whether the EGFR protein can bind to the promoter of YAP1 but found no evidence of this interaction (fig. S4). Because the EGFR might regulate the activity of YAP1 rather than its expression, we performed a coimmunoprecipitation (CoIP) experiment and found that these two proteins interact (Fig. 2U). Thus, our findings suggest that major hepatectomy increases EGFR levels, allowing the EGFR to bind to YAP1 protein and enhance its activity.

Major hepatectomy inhibits the recruitment and function of MDSCs

YAP1 signaling is closely linked to the immune environment (24, 25). To investigate the immunomodulatory role of YAP1 activation in post–hepatectomy CRLM, we used two complementary immunodeficient mouse models: recombination activating gene 1-knockout (RAG1-KO) mice, which lack adaptive immunity (T and B cells) while maintaining intact innate immune function, and nonobese diabetic (NOD) severely combined immunodeficient interleukin-2 (IL-2) receptor γ-chain-null (NSG) mice, which exhibit combined deficiencies in both adaptive and innate immunity (26, 27). This dual-model approach enabled us to systematically dissect the contributions of different immune compartments to YAP1-mediated regulation of CRLM progression following major hepatectomy. Both 70% PHx and YAP1 overexpression reduced the number of tumors in RAG1-KO mice (Fig. 3A and fig. S5A) but not in NSG mice (Fig. 3B and fig. S5B), suggesting the involvement of innate immunity. Cytometry by time-of-flight (CyTOF) analysis of immune cell infiltration in CRLMs revealed 28 distinct cell clusters (Fig. 3, C and D). In the 70% PHx mice, the proportion of myeloid-derived suppressor cells (MDSCs), both monocytic MDSCs (M-MDSCs) and granulocytic MDSCs (G-MDSCs), decreased, whereas the proportion of CD8⁺ T cells increased compared with those in the sham-operated mice (Fig. 3, E and F). These findings indicate that YAP1 activation enhances antitumor immunity by suppressing MDSC infiltration and increasing CD8⁺ T cell infiltration. In addition, the expressions of the chemokine receptors CCR4 (28) and CCR6 (29), which are linked to MDSC recruitment and expansion, were significantly down-regulated in CD11b⁺Ly6C⁺ and CD11b⁺Ly6G⁺ cells in mice that underwent 70% PHx (Fig. 3G). CD127, a marker of MDSC immunosuppressive function (30), was also down-regulated in MDSCs from mice that underwent 70% PHx (Fig. 3G). These results suggested that the number of MDSCs decreased whereas the immunosuppressive function of MDSCs increased in tumors after 70% PHx.

Fig. 3. Major hepatectomy inhibits the recruitment and function of MDSCs.

(A and B) Schematics, representative images, and standardized tumor numbers of the sham, 30% PHx, and 70% PHx groups of RAG1-KO mice and NSG mice. (C) Heatmap of 34 markers expressed in each cell cluster. (D) Left: t-SNE (t-distributed Stochastic Neighbor Embedding) plot of the CyTOF data of all the CD45⁺ cells from the CRLMs. Each cell cluster was annotated into a specific immune cell type. Right: t-SNE plot of all CD45⁺ cells colored according to the expression levels of specific markers. (E) t-SNE plot of all CD45⁺ cells from mouse CRLMs in sham and 70% PHx groups. (F) Proportion of each cell cluster among the CD45⁺ cells in the two groups. (G) t-SNE plot of all CD45⁺ cells from mouse CRLMs in each group colored according to CCR4, CCR6, and CD127 expressions. Bottom right: Proportion of each cell cluster among MDSCs in the two groups. (H) Flow cytometry analysis and bar chart of the ratio, immature marker CD62L, and function marker PD-L1 of MDSCs separated from tumors in sham, 30% PHx, and 70% PHx groups. (I) Flow cytometry analysis and bar chart of the ratio of CD8⁺ T cells with low CFSE density when cocultured with MDSCs separated from tumors in each group. (J) Flow cytometry analysis and bar chart of the ratio, the expression of activation related surface markers in CD8⁺ T cells cocultured with MDSCs isolated from tumors in each group. (K) Flow cytometry analysis and bar chart of the ratio, the expression of activation related intercellular markers in CD8⁺ T cells cocultured with MDSCs isolated from tumors in each group. (L) Representative liver images and chart of standardized tumor numbers in the sham, 70% PHx, and sham plus Gr-1 antibody groups.

To further assess the function of MDSCs after hepatectomy, we isolated immune cells from tumors in sham, 30% PHx, and 70% PHx groups and examined the number, activation status, and function marker of MDSCs. The results showed that the ratio of MDSCs in tumors decreased after 70% PHx, and CD62L⁺ MDSCs increased whereas PD-L1⁺ MDSCs decreased after hepatectomy, especially in the 70% PHx group (Fig. 3H). These results proved that both the ratio and the function of MDSCs decreased after major hepatectomy. Besides, we isolated MDSCs from tumors in the three groups and cocultured them with splenocytes to assess their immunosuppressive effects. MDSCs from hepatectomy groups had a reduced inhibitory effect on the proliferation (Fig. 3I) and activation (Fig. 3, J and K) of CD8⁺ T cells. To investigate whether YAP1 activation mediated the observed reduction in MDSCs and increase in CD8⁺ T cells, we conducted comparative analyses of liver tissues from three experimental groups: negative controls (NC), YAP1 overexpression in hepatocytes, and YAP1 overexpression in hepatocytes with YAP1 inhibitor verteporfin. Our findings demonstrated that, although YAP1 overexpression in hepatocytes significantly reduced MDSC accumulation and enhanced CD8⁺ T cell infiltration and function in tumors, these immunomodulatory effects were effectively reversed by YAP1 inhibition (fig. S6). These results provide compelling evidence that YAP1 activation serves as a key molecular mechanism linking liver regeneration after major resection to subsequent remodeling of the tumor immune microenvironment.

Anti-mouse Gr-1 antibody, an MDSC function inhibitor, was used in vivo to block the function of MDSCs in CRLMs, with effects comparable to those observed with 70% PHx or hepatocellular YAP1 overexpression (Fig. 3L and fig. S7). In summary, these results indicate that major hepatectomy may suppress the growth of CRLMs by inhibiting the recruitment and function of MDSCs in tumors.

Hepatocellular YAP1 activation suppresses tumoral YAP1 activation and leads to CXCL5 down-regulation

To investigate how major hepatectomy affects MDSC trafficking, we tagged MC38 cells with the mCherry fluorescent protein before intrasplenic injection. Once the CRLMs had developed, mCherry-positive cells were isolated and subjected to RNA-seq (Fig. 4A). We analyzed the differentially expressed genes [DEGs; |log₂(fold change)| > 0.585, P < 0.05] in tumor samples from the sham and 70% PHx groups. Concurrently, we compiled a list of chemokine-associated genes from the GSEA website (https://gsea-msigdb.org) (table S4). Among the 1185 DEGs, 18 were chemokine associated (Fig. 4B, left). After ranking these by absolute log₂(fold change), we identified the top 5 DEGs, comprising three down-regulated and two up-regulated genes (Fig. 4B, right). qPCR analysis of tumor tissue confirmed these findings (fig. S8). Notably, CXCL5—a chemokine known to bind CXCR2 on MDSCs and drive their migration (25)—exhibited the most significant differential expression. In addition, a mouse cytokine array kit and IHC were used to assess the levels of chemokines in CRLMs from the sham and 70% PHx groups, and the results revealed that the CXCL5 level was markedly lower in the 70% PHx group (Fig. 4, C and D, and table S5). In short, major hepatectomy resulted in decreased levels of CXCL5 in tumors, which led to MDSCs dysfunction.

Fig. 4. Hepatocellular YAP1 activation suppresses tumoral YAP1 activation and leads to CXCL5 down-regulation.

(A) Schematic of sorting MC38 cells from mouse CRLMs and a heatmap of the RNA-seq results of the sham and 70% PHx groups. (B) Left: Venn diagram presenting the overlapping chemokine-related genes and DEGs from RNA-seq. Right: List of the top 5 differentially expressed chemokine-related genes. (C) Representative images of the mouse cytokine array of the tumors of the CRLM model mice in the two groups. (D) Representative images of IHC staining for CXCL5 in the tumors of the CRLM model mice in sham, 30% PHx, and 70% PHx groups. (E) GSEA of the RNA-seq data on YAP1 signaling in sham and 70% PHx groups. (F and G) Representative images of IHC staining for YAP1^S127 and YAP1 in tumors of the CRLM model mice in sham, 30% PHx, and 70% PHx groups. (H) WB analysis of YAP1 levels in the nuclei of mouse CRLMs and adjacent livers in sham, 30% PHx, and 70% PHx groups. (I) ChIP analysis of the combination of the YAP1 protein and the Cxcl5 promoter via qPCR in triplicate. (J and K) Levels of the CXCL5 protein in the culture medium of MC38 cells with YAP1 overexpression (J) or knockdown (K). (L and M) Top left: Workflow for the Transwell migration assay in which murine bone marrow–derived MDSCs were cocultured with MC38 cells. Bottom left: Representative images of the migration of MDSCs toward YAP1-regulated MC38 cells. Right: Chart of the number of MDSCs that crossed the membrane in each group. (N) Left: Representative liver images of CRLM model mice in the four groups shown in the figure. Right: Chart of tumor numbers in the four groups. Scale bars, 100 μm (10X) and 20 μm (20X).

The GSEA of the RNA-seq data also revealed that YAP1 signaling was inhibited in tumors following 70% PHx (Fig. 4E). This finding was further validated via IHC experiments, which revealed increased YAP1^S127 levels and reduced YAP1 levels (Fig. 4, F and G). WB analysis confirmed increased expression of YAP1 in the nuclei of hepatocytes and decreased expression in the nuclei of tumor cells (Fig. 4H). Wang and colleagues (25) reported that YAP1 binds to the Cxcl5 promoter, which we corroborated via a ChIP assay (Fig. 4I). To further elucidate the function of YAP1 in vitro, we overexpressed and knocked down YAP1 in MC38 cells and detected relevant CXCL5 expression through enzyme-linked immunosorbent assay (ELISA), qPCR, and WB analysis (Fig. 4, J and K, and fig. S9, A to D). In addition, Transwell experiments in which MC38 cells were cocultured with bone marrow–derived MDSCs revealed that MC38 cells overexpressing YAP1 attracted more MDSCs, whereas the opposite effect was observed when YAP1 was knocked down (Fig. 4, L and M). In vivo experiments revealed that tumoral YAP1 overexpression reversed the effect of hepatocellular YAP1 activation and that SB225002 suppressed the growth of tumors by blocking MDSC trafficking (Fig. 4N). Together, our findings indicate that hepatocellular YAP1 activation induced by major hepatectomy inhibits tumor YAP1 activation and CXCL5 expression.

Decreased tumor Gln levels suppress tumoral YAP1 activity by promoting the production of ROS

The above findings prompted us to investigate potential pathways involved in YAP1 competition between tumor cells and surrounding hepatocytes. Studies have linked YAP1 signaling and glutamine (Gln) metabolism (31–33), both of which are associated with MDSC infiltration (25, 34–36). Thus, we hypothesized that Gln metabolism plays a key role in YAP1 activation competition between hepatocytes and tumor cells. We measured Gln levels in CRLMs and detected a reduction in 70% of PHx mice in both the MC38 and CT26 models (Fig. 5A). To assess the effect of Gln depletion on YAP1, we cultured MC38 cells under Gln-deficient conditions and found that higher Gln concentrations inhibited YAP1 dephosphorylation at Ser¹²⁷ (Fig. 5B) and increased the expression of the YAP1 target genes Ccn1 and Ccn2 (Fig. 5C). Briefly, major hepatectomy reduced Gln levels in tumors and suppressed the activity of YAP1.

Fig. 5. Tumor Gln deprivation suppresses tumoral YAP1 activity via ROS.

(A) Gln levels in the tumors of CRLM mice (MC38/CT26 generated) in sham and 70% PHx groups. (B) WB analysis of YAP1 and YAP1^S127 levels in MC38 cells with different Gln levels. (C) qRT-PCR analysis of Ccn1 and Ccn2 levels in MC38 cells with different Gln levels. (D) WB analysis of YAP1, YAP1^S127, and (P-)LATS1 levels in MC38 cells treated with different CB-839 concentrations. (E) WB analysis of YAP1 and YAP1^S127 levels in MC38 cells upon Gln deprivation and/or CB-839 treatment. (F) WB analysis of YAP1 and YAP1^S127 levels in the cytosol and nucleus of MC38 cells upon Gln deprivation and/or CB-839 treatment. (G) Cytometry analysis of relative ROS levels in MC38 cells treated with or without CB-839. MFI, mean fluorescence intensity. (H) WB analysis of YAP1, YAP1^S127, and (P-)LATS1 levels in MC38 cells treated with CB-839 and/or subjected to Gln deprivation. (I) WB analysis of YAP1 and YAP1^S127 levels in MC38 cells treated with H₂O₂. (J) GSH levels in MC38 cells treated with or without BSO. (K) WB analysis of YAP1, YAP1^S127, and (P-)LATS1 levels in MC38 cells treated with different BSO concentrations. (L) NADPH levels in MC38 cells treated with/without apocynin. (M) Cytometry analysis of relative ROS levels in MC38 cells treated with/without apocynin. (N) WB analysis of YAP1 and YAP1^S127 levels in MC38 cells treated with apocynin. (O) WB analysis of YAP1, YAP1S127, and CXCL5 levels in MC38 cells under the treatment of Gln deprivation or GLS inhibition with or without NAC treatment. (P) Representative images and chart of the number of tumors in the CRLM mice in the vehicle and CB-839 groups. (Q) WB analysis of YAP1 and YAP1^S127 levels in the tumors of the CRLM mice in each group.

Because YAP1 activity is regulated primarily by LATS-mediated phosphorylation, we investigated LATS1 phosphorylation under conditions of Gln starvation in MC38 cells. We observed increased phosphorylation of both LATS1 and YAP1 under Gln deprivation conditions but not under glucose starvation conditions; we also observed reduced expression of YAP1-targeted genes (fig. S10, A and B). These findings indicate that Gln deprivation promotes LATS-mediated YAP1 phosphorylation. Furthermore, we inhibited glutaminolysis via CB-839, a glutaminase (GLS) inhibitor known for its efficacy in multiple cancers (37, 38). Notably, CB-839 treatment promotes LATS-mediated YAP1 phosphorylation (Fig. 5D) and decreased the expression of YAP1-targeted genes (fig. S10C), mimicking the effects of Gln deprivation on the suppression of YAP1 activation (Fig. 5E). Consistent with the observed increase in YAP1 phosphorylation, we detected decreased nuclear localization of YAP1 under conditions of Gln deprivation or CB-839 treatment (Fig. 5F). Other glutaminolysis inhibitors, such as Bis-2-(5-phenylacetamido-1,3,4-thiadiazol-2-yl)ethyl sulfide (BPTES) and 6-diazo-5-oxo-l-norleucine (DON), similarly increased YAP1 phosphorylation and decreased YAP1 activity (fig. S10, D and E). These WB analyses suggested that GLS inhibition also suppressed the activity of YAP1.

Owing to the essential role of Gln metabolism in maintaining redox homeostasis (39), we hypothesized that Gln metabolism might regulate tumoral YAP1 activity via reactive oxygen species (ROS). We assessed ROS levels via cytometry and found that CB-839 treatment increased ROS production in MC38 cells (Fig. 5G). For further investigation, we used N-acetylcysteine (NAC) to inhibit ROS production and found that NAC treatment reversed the phosphorylation of LATS1 and YAP1 following CB-839 treatment (Fig. 5H) and restored YAP1 activity (fig. S10F). In addition, we enhanced ROS production by adding exogenous hydrogen peroxide (H₂O₂), and the results further supported the role of ROS in promoting YAP1 phosphorylation (Fig. 5I). These experiments indicated that GLS inhibition induced YAP1 inactivation, which was mediated through increased ROS production.

Gln is reported to be a precursor for antioxidants such as glutathione (GSH) and reduced form of nicotinamide adenine dinucleotide phosphate (NADPH), which are crucial for maintaining the cellular redox balance (40). Gln deprivation is therefore expected to deplete these antioxidants and increase ROS production. To assess test this hypothesis, we used l-buthionine sulfoximine (BSO) to suppress GSH synthesis and observed decreased NADPH levels (Fig. 5J) and increased phosphorylation of LATS1 and YAP1 in MC38 cells (Fig. 5K). NADPH oxidase (NOX) transfers electrons from NADPH and generates ROS. To explore the link between NADPH and ROS production, we inhibited NOX via apocynin, which led to increased NADPH levels (Fig. 5L), reduced ROS production (Fig. 5M), and decreased phosphorylation of YAP1 at the S127 site (Fig. 5N). To determine whether Gln availability or ROS levels regulate CXCL5 expression, we performed experiments under four conditions: (i) Gln deprivation, (ii) GLS inhibition, (iii) Gln deprivation with NAC (ROS scavenger) treatment, and (iv) GLS inhibition with NAC treatment. Our results showed that both Gln deprivation and GLS inhibition significantly reduced YAP1 activity and down-regulated CXCL5 expression. These effects were completely abolished when ROS production was blocked by NAC treatment (Fig. 5O). These findings demonstrate that Gln deprivation or GLS inhibition suppresses YAP1 activity and subsequent CXCL5 expression through a ROS-dependent mechanism.

In addition, we repeated these experiments using the human colon cancer cell line HT29 and found that YAP1 activity was similarly suppressed under Gln deprivation conditions (fig. S11, A and D) or upon GLS inhibition via CB-839 (fig. S11, B to D). ROS production in HT29 cells increased with CB-839 treatment (fig. S11E). The ROS inhibitor NAC reversed YAP1 inhibition induced by CB-839 (fig. S11F), whereas the ROS-generating agent H₂O₂ inhibited YAP1 activation (fig. S11G). The GSH inhibitor BSO also decreased GSH production (fig. S11H) and inhibited YAP1 activation (fig. S11I). Furthermore, NOX inhibition increased NADPH levels (fig. S11J), decreased ROS production (fig. S11K), and increased YAP1 activity in HT29 cells (fig. S11L).

For in vivo evaluation of glutamine metabolism inhibition, we administered CB-839 and observed significant effects on tumor burden (Fig. 5P) and increased expression of tumor YAP1^S127 (Fig. 5Q). In addition, we investigated JHU083, a DON prodrug (41) known to reprogram tumor metabolism and enhance antitumor immunity (42). Treatment of CRLM-bearing mice with JHU083 produced comparable effects to CB-839 (fig. S12, A and B), further supporting our findings that GLS inhibition modulates YAP1 activity and tumor progression. These findings support the conclusion that the Gln level robustly regulates YAP1 activity through the GSH/NADPH/ROS pathway in tumors.

Major hepatectomy–mediated hepatocellular YAP1 activation promotes Gln consumption in the liver and depletes Gln in the tumor

Our results suggest that hepatocellular YAP1 activation may contribute to Gln deprivation in tumors. To further investigate this phenomenon, we performed untargeted metabolomics on normal mouse livers and livers 3 days after 70% PHx, revealing that Gln was the down-regulated metabolite with the lowest P value (Fig. 6, A and B). To investigate whether YAP1 activation mediates Gln depletion following 70% PHx in mice with CRLMs, we inhibited YAP1 activity using verteporfin. Notably, YAP1 inhibition restored Gln levels in tumor tissue (Fig. 6C), demonstrating that PHx-induced Gln down-regulation is YAP1 dependent. On the basis of the literature review, we identified a few key Gln metabolic regulators. GLS, glutamine synthetase (GLUL), glutamate oxaloacetate transaminase 1 (GOT1), phosphoserine aminotransferase 1 (PSAT1), solute carrier family 1 member 5 (SLC1A5), SLC7A5, and SLC38A1 are enzymes or amino acid transporters that take part in glutamine metabolism. SLC1A5, SLC7A5, and SLC38A1 are members of solute carrier family, and they are all reported to contribute to glutamine transport by different mechanisms (43). GLS is a glutamine-hydrolyzing enzyme that decomposes glutamine into glutamate, whereas GLUL catalyzes the synthesis of glutamine from glutamate (44). GOT1 and PSAT1 both catalyze the transformation of glutamate to α-ketoglutaric acid through transamination or oxidative deamination, respectively (45). All these genes were reported to be regulated by YAP1 (31, 45–48), and their functions can be seen in the scheme in fig. S13A. However, we only detected the up-regulation of GLS, GLUL, SLC1A5, and SLC7A5 upon 70% PHx in liver tissue (Fig. 6, D and E). In addition, these differentially expressed Gln metabolism–related genes were also down-regulated in tumor tissue of mice with 70% PHx (fig. S13, B and C). Besides, we cultured MC38 cells in the conditioned medium of negative controls or YAP1^OE AML12 cells and found that YAP1 inactivation occurred in MC38 cells (fig. S13D).

Fig. 6. Hepatocellular YAP1 activation leads to Gln scarcity in tumors.

(A) Volcano plot of untargeted metabolomics analysis of metabolites in the tumor-adjacent livers of CRLM model mice in the sham and 70% PHx groups. (B) Relative levels of Gln in the tumor-adjacent livers of the CRLM model mice in sham, 30% PHx, and 70% PHx groups. (C) Levels of Gln and WB analysis of YAP1^S127 and nuclear YAP1 in the tumor-adjacent livers of the CRLM model mice in sham, 70% PHx, and 70% PHx with verteporfin groups. (D) qRT-PCR analysis of YAP1 downstream genes and glutamine metabolism–related genes levels in the tumor-adjacent livers of the CRLM model mice in sham, 30% PHx, and 70% PHx groups. (E) WB analysis of YAP1, YAP1^S127, YAP1 downstream proteins, and glutamine metabolism–related proteins in the whole-cell lysates and nuclear fractions of the tumor-adjacent livers of the CRLM model mice in each group. (F and G) WB and qPCR analysis of YAP1, YAP1 downstream proteins, and glutamine metabolism–related proteins in the whole-cell lysates and nuclear fractions of AML12 cells with/without YAP1 overexpression. (H) Levels of Gln and glucose in AML12 cells with/without YAP1 overexpression. (I and J) WB and qPCR analysis of YAP1, YAP1 downstream proteins, and glutamine metabolism–related proteins in the whole-cell lysates and nuclei of AML12 cells with/without YAP1 knockdown. (K) Levels of Gln and glucose in AML12 cells with/without YAP1 knockdown.

To further validate YAP1’s regulatory role in Gln metabolism in vitro, we overexpressed YAP1 expression in the normal mouse liver cell line AML12, which led to increased expression of Gln metabolism genes (Fig. 6, F and G) and decreased Gln levels (Fig. 6H). Conversely, the knockdown of YAP1 in AML12 cells induced the opposite effects (Fig. 6, I to K). Similar regulatory effects were observed in the human normal liver cell line LO2 (fig. S13, E to H), demonstrating the conservation of this mechanism across species. Together, our findings demonstrate that major hepatectomy induced activation of YAP1 signaling in hepatocytes, which resulted in overexpression of several key enzymes in Gln metabolism and transport, ultimately increasing Gln consumption in the liver and depriving liver tumor cells of Gln.

A high hepatocellular YAP1–to–tumor YAP1 ratio is associated with decreased tumor MDSC infiltration and better survival in patients with CRLM

To further validate our hypothesis, we evaluated the clinical relevance of YAP1 expression in liver tissue adjacent to tumors in patients with CRLM (Fig. 7A). IHC was performed on a microarray containing samples from 150 patients with CRLM who had undergone curative resection. The clinical characteristics of these patients, along with the IHC expression scores for liver YAP1, tumor YAP1, and tumor CXCL5, are provided in table S6. The results revealed that patients with high YAP1 expression in the liver had a significantly longer median overall survival time (Fig. 7B) and a lower risk of recurrence (Fig. 7C) than those with low YAP1 expression. We also observed that high YAP1 expression was associated with lower metastases (Fig. 7D) and tumor size (Fig. 7E). The univariate and multivariate analyses of factors associated with overall survival and recurrence-free survival in the microarray of CRLM were shown in table S7. These indicated that expression of hepatocellular YAP1 could be an independent predictor of prognosis for patients with CRLMs.

Fig. 7. A high hepatocellular YAP1–to–tumor YAP1 ratio is associated with MDSC infiltration and survival in patients with CRLM.

(A) Representative IHC images of YAP1 expression patterns and scores indicating YAP1 levels in human CRLM-adjacent livers. (B and C) Survival and recurrence curves divided by liver YAP1 expression. (D and E) Chart showing the relationships between liver YAP1 expression and the number of tumors (D) and the largest tumor diameter (E). (F) Representative IHC images of YAP1 expression patterns and scores indicating YAP1 levels in the tumors of human patients with CRLM. (G and H) Survival curves of patients with CRLM divided by tumoral YAP1 expression. (I and J) Chart showing the correlation between tumor YAP1 expression and the number of tumors (I) and the largest tumor diameter (J). Statistically significant differences were identified via unpaired two-tailed Student’s t tests. (K to N) Representative IHC images of liver YAP1, tumor YAP1, and tumor CXCL5 in two standard patients (K) and the correlations between these markers (L to N). Scale bars, 200 μm (4X) and 20 μm (20X).

Next, we assessed the clinical relevance of YAP1 expression in the tumors of patients with CRLM via IHC (Fig. 7F) and found results opposite those observed for hepatocellular YAP1 expression (Fig. 7, G to J). In addition, we evaluated the combined prognostic impact of liver and tumor YAP1 expression on CRLM via Kaplan-Meier analysis of the microarray data. This analysis revealed a significantly better prognosis in the liver-YAP1^low/tumor-YAP1^high group (fig. S14, A to D).

Our findings suggest a potential competitive interaction between liver YAP1 and tumor YAP1, and our findings revealed a strong correlation between liver and tumor YAP1 expression levels (Fig. 7K). We also assessed tumor CXCL5 expression via a CRLM microarray, which revealed that hepatocellular YAP1 expression, tumor YAP1 expression, and tumor CXCL5 expression were mutually correlated (Fig. 7, L to N). In addition, we analyzed the relationships between YAP1 and CXCL5 expression and MDSC infiltration in The Cancer Genome Atlas (TCGA) database and found that tumor YAP1 and CXCL5 expression levels were correlated with MDSC infiltration in colon adenocarcinoma (COAD) (fig. S14, E and F) and liver hepatocellular carcinoma (LIHC) (fig. S14, G and H). Overall, these results confirmed the impact of hepatocellular YAP1 expression on the prognosis of CRLM and highlighted the interrelationships among hepatocellular YAP1, tumor YAP1, tumor CXCL5, and MDSC infiltration status.

Together, our findings elucidate that major hepatectomy exerts a tumor-inhibitory effect on the growth of CRLM by regulating YAP1 activity and inducing Gln competition between hepatocytes and tumor cells, which results in down-regulation of the chemokine CXCL5 and suppressed infiltration of MDSCs in tumors.

DISCUSSION

Studies have yielded conflicting results on whether ASH is superior to PSH for the treatment of patients with CRLM. However, comparing the outcomes of these two approaches is challenging because of the ambiguous definitions of ASH and PSH. For example, a segment V resection is considered ASH when compared with wedge hepatectomy, but it is classified as PSH when compared with hemihepatectomy. According to these definitions, not all parenchymal-sparing resections are strictly nonanatomic. We hypothesized that prognosis may depend more on the proportion of liver resection rather than the anatomical structure. We categorized hemihepatectomy or hepatic trilobectomy as major hepatectomy (resection proportion > 40%) and resection of no more than 15% of the liver volume as minor hepatectomy. According to data from the Zhongshan Hospital, patients in the major hepatectomy group had a significantly better prognosis than those in the minor hepatectomy when the patients presented comparable tumor burdens. Similar results were observed in animal models. Therefore, we propose that the prognosis of CRLM is correlated with the proportion of resected liver instead of the anatomical structure. In addition, we found that major hepatectomy is associated with increased hepatocellular YAP1 activation in the remnant liver, leading to competition for Gln between hepatocytes and tumor cells. Specifically, under hepatocellular YAP1 signaling activation, Gln is rapidly consumed in the liver, resulting in Gln deprivation in tumor cells and subsequent inactivation of YAP1 in tumor cells.

Although we demonstrated that major hepatectomy might benefit patients, the intention of our study was not to encourage reserving as less parenchyma as possible as major hepatectomy also increases surgical stress, operative risk, and the incidence of severe postoperative complications (49, 50). Following major hepatectomy, subsequent resections may become less feasible in the event of recurrence, significantly affecting the prognosis. Although our study revealed that the proportion of liver resected, rather than the anatomical structure resected, is critical for the prognosis of CRLM, we did not precisely evaluate the proportion of liver resected during hepatectomy in patients because of a lack of detailed operative notes. The results would be more convincing if patients with CRLM were grouped according to the actual proportion of liver resected.

Because hepatocytes and tumor cells compete for YAP1, we propose that the ratio of hepatocellular YAP1 expression to tumoral nuclear YAP1 expression may serve as a prognostic factor for CRLM and that YAP1 might be a potential therapeutic target. We are curious whether selectively promoting the activation of YAP1 in hepatocytes or inhibiting the activity of YAP1 in cancer cells could inhibit the growth of liver metastasis. However, it is important to consider whether accelerating hepatocellular YAP1 activation might increase the risk of hepatocarcinogenesis as YAP1 signaling has been extensively implicated in the development of HCC (51, 52). It has been reported that YAP1 signaling is activated in stem cells (53, 54), suggesting that stem cell–based therapies may be a strategy for inducing competitive Gln consumption in the liver and inhibiting tumor YAP1 activity. Although the GLS inhibitor JHU083 has been shown to suppress tumor growth by enhancing CD8⁺ T cell activity (42), we are also interested in exploring other strategies to selectively inhibit Gln metabolism in tumor cells instead of hepatocytes.

In summary, our study reveals a mechanism underlying the improved prognosis of CRLMs following major hepatectomy: Surgical resection enhances YAP1 activation in hepatocytes via EGFR up-regulation, creating a competitive metabolic demand for Gln between regenerating hepatocytes and tumor cells. This competition disrupts the Gln/ROS/YAP1/CXCL5/MDSC signaling axis in tumors, ultimately suppressing tumor progression. These findings may offer guidance for selecting surgical methods for certain patients with CRLM. Moreover, this study not only suggests a potential prognostic factor for CRLM following curative surgery but also points to a promising therapeutic approach for patients who are not candidates for surgery.

MATERIALS AND METHODS

Patients and specimens

A cohort of 240 patients (cohort 1) with CRLM undergoing hepatectomy was obtained from the Fudan University–affiliated Zhongshan Hospital between January 2015 and December 2020 to evaluate the prognostic differences between surgical approaches, with 207 undergoing minor hepatectomy and 33 undergoing major hepatectomy. In addition, a second cohort (cohort 2) from the Zhongshan Hospital, collected between 2015 and 2020 and comprising 10% of the samples from the initial cohort, included 150 paired human CRLM and peritumoral specimens posthepatectomy. This secondary cohort was used to confirm the role of YAP1 signaling in the progression of CRLM. Both studies were conducted in accordance with the ethical standards of the Research Ethics Committee of Zhongshan Hospital, Fudan University (approval number: B2023-302R), and written informed consent was obtained from all participants.

Mouse models

All wild-type (WT) C57BL/6J mice (male, weighted 18 to 22 g, and aged 6 to 8 weeks) were procured from Vital River Laboratory (Beijing, China) and housed under specific pathogen–free conditions with the temperature at 21° to 26°C and relative humidity at 50 to 60%. All animal procedures complied with the Guide for the Care and Use of Laboratory Animals and received approval from the Institutional Animal Care and Ethics Committee of Fudan University affiliated Zhongshan Hospital (approval number: 2022-74).

Mouse partial hepatectomy model induction

Partial hepatectomy was conducted on 8-week-old WT male mice. Our surgical protocols were carefully standardized as follows: For the 70% PHx group, we performed a midline laparotomy and sequentially dissected (i) the falciform ligament connecting the papillary lobe and left lateral lobe, followed by (ii) the triangular ligament between the median lobe and diaphragm. Subsequently, we resected the left lateral lobe, median lobe, and left medial lobe in sequence, achieving ~70% liver resection. The abdominal wall was then closed in layers using absorbable sutures. In the 30% PHx group, after identical ligament dissection, we removed only the left lateral lobe (~30% liver mass). Sham-operated controls underwent the same ligament dissections without parenchymal resection (55). Postoperatively, carprofen (5 mg/kg body weight) was administered subcutaneously every 8 hours for 2 days to manage pain.

Mouse forced liver metastasis model induction

Forced liver metastasis in mice was induced by intrasplenic injection of the murine colon cancer cell line (56). As the MC38 cell line was the most commonly used mouse colorectal cancer cell line originating from C57BL/6 mice (57–59), we used MC38 cells to introduce forced liver metastasis in mice. Under anesthesia, mice were secured and the spleen was surgically exposed. A total of 3 × 10⁵ MC38 cells, suspended in 200 μl of phosphate-buffered saline (PBS), were injected into the spleen using a 27-gauge needle. The injection was performed in the hemispleen, which was removed 3 min after cancer cell injection. The mice were euthanized after 3 weeks, and their livers and lungs were collected for subsequent analyses.

Cell models

For in vitro cell models, we adopted the AML12 cell line as mouse normal liver cell to explore the influence of YAP1 level on the expression of glutamine metabolism–related genes in hepatocytes as AML12 was the most commonly used murine normal liver cell line due to its typical hepatocyte features (60–62). The LO2 cell line was adopted as human normal hepatocytes (63). The MC38 cell line, the most commonly used mouse colorectal cancer cell line, was used to explore how glutamine scarcity affected the YAP1 activity in murine colorectal cancer cell in vitro, whereas the HT29 cell line was used as human colorectal cancer cells (64).

Cytometry by time-of-flight

Metal-conjugated antibodies were prepared using the Maxpar Antibody Labeling Kit, as listed in table S8. Liver metastasis was dissociated into single cells using the Mouse Tumor Dissociation Kit (130-096-730, Miltenyi Biotec, Germany), following the manufacturer’s instructions to analyze the tumor immune environment. Cells were treated with 250 nM cisplatin (Fluidigm, USA) to identify and exclude dead cells, followed by CD16/CD32 antibodies to block FcγR. The cells were then incubated with a cocktail of surface antibodies for 30 min on ice and subsequently fixed overnight in intercalation solution (Maxpar Fix and Perm Buffer with 250 nM 191/193Ir, Fluidigm, USA) after two washes. The following day, an intracellular antibody cocktail was applied for 30 min on ice for intracellular staining. After washing and resuspension in deionized water, 20% EQ beads (Fluidigm, USA) were added. Samples were then analyzed using a Helios3 CyTOF system (Fluidigm, USA). The CyTOF data were normalized using a methodology based on EQ Four Element Calibration Beads and analyzed on the Cytobank platform (https://cytobank.org/). Data acquisition and analyses were performed by PLTTech Inc. (Hangzhou, China).

Mouse cytokine antibody Array

For the cytokine assay, 100 mg of mouse CRLM tumor tissue was minced and transferred to 500 μl of radioimmunoprecipitation assay (RIPA) lysis buffer supplemented with a protease inhibitor cocktail. The tissue was then homogenized using a TissueMaster (Beyotime, China) and centrifuged at 14,000g for 20 min at 4°C. The supernatant was carefully collected and transferred to a new tube as the tissue lysate. This lysate was analyzed using a mouse cytokine array kit (RayBiotech, China), following the manufacturer’s instructions. Briefly, the array membranes were blocked with blocking buffer at 37°C for 60 min and then incubated with the tissue lysate overnight at 4°C. A detection antibody cocktail was applied to each membrane for 60 min, followed by three washes. Streptavidin–horseradish peroxidase (HRP) was added and incubated for 30 min. Last, membranes were treated with Chemi Reagent Mix, and signals were visualized using an electrogenerated chemiluminescence (ECL) imaging system (Bio-Rad, USA).

In vivo imaging

All in vivo imaging was conducted at Shanghai Chenguang Medical Technologies Company using a 7.0-T CG NOVILA magnetic resonance imaging (MRI) system. Animals were anesthetized with isoflurane before imaging to ensure minimal movement and stress. Postacquisition, the images were analyzed using the RadiAnt DICOM Viewer.

Coculture of mouse MDSCs and splenocytes

Mouse liver metastasis was dissociated using the Mouse Tumor Dissociation Kit, and the resulting immune cells were further purified using the Mouse MDSC Isolation Kit (19867, STEMCELL Technologies, China) to isolate mouse MDSCs. Similarly, immune cells from mouse spleens were isolated using the Mouse Spleen Dissociation Kit (130-095-926, Miltenyi Biotec, Germany). Both MDSCs and splenocytes were cultured in RPMI 1640 medium (A1049101, Gibco, USA) supplemented with 10% heat-inactivated fetal bovine serum (10099141C, Gibco, USA), penicillin/streptomycin (100 U/liter; 15140-122, Gibco, USA), 10 mM Hepes, 2 mM glutamine, 20 μM β-mercaptoethanol, and 1 mM sodium pyruvate, in a humidified 5% CO₂ incubator at 37°C. For the T cell proliferation assay, 2 × 10⁵ MDSCs from mouse liver metastasis and 2 × 10⁶ splenocytes were seeded into the upper and bottom chambers of a 24-well plate, respectively. Splenocytes were prelabeled with 10 μM carboxyfluorescein diacetate succinimidyl ester (CFSE) (65-0850-84, Invitrogen, USA) and cocultured with MDSCs in the presence of anti-CD3 (1 μg/ml, clone 145-2C11, eBioscience, USA) and soluble anti-CD28 (2 μg/ml, clone 37.51, eBioscience, USA) antibodies for 72 hours. For the T cell activation assay, splenocytes and MDSCs were activated using a Cell Stimulation Cocktail (with brefeldin A and monensin) (00-4975-93, eBioscience, USA) for 12 hours. Subsequently, splenocytes were harvested for staining and flow cytometry analysis.

Coculture of mouse MDSCs and MC38

A WT mouse was euthanized, and the tibias and femurs were harvested using sterile techniques. The ends of these bones were clipped, and the bone marrow was flushed out using a 1-ml syringe filled with RPMI 1640 complete culture medium. After centrifugation at 1000 rpm for 3 min at 4°C, the supernatant was discarded, and Red Blood Cell Lysis Buffer (B541001, Sangon Biotech, China) was applied at room temperature for 5 min. The lysis was halted with PBS, and the cell suspension was centrifuged at 2000 rpm for 3 min. The supernatant was again discarded, and the cell pellet was resuspended in medium. To culture MDSCs, 2.5 × 10⁶ cells were plated in 100-mm dishes with 10 ml of RPMI 1640 complete culture medium supplemented with granulocyte-macrophage colony-stimulating factor (40 ng/ml), granulocyte colony-stimulating factor (40 ng/ml), and IL-6 (40 ng/ml). On day 3, floating cells were removed, and fresh medium containing cytokines was added. By day 6, MDSCs were activated and harvested. For the chemotaxis assay, 5 × 10⁴ MDSCs were placed in the upper chamber in serum-free medium, whereas 2 × 10⁵ MC38 cells (WT, YAP1-OE, or shYAP1) were seeded in the lower chamber containing Dulbecco’s modified Eagle’s medium (DMEM) complete culture medium. After 24 hours of coculture, nonmigrated cells were removed with a cotton swab. Cells at the bottom of the chamber were fixed with 4% formaldehyde for 10 min and stained with 0.5% crystal violet for 30 min. After air-drying, chambers were imaged using an inverted microscope (Olympus, Tokyo, Japan). All experiments were conducted in triplicate.

Flow cytometry

Fluorochrome-conjugated antibodies against CD4 (fluorescein isothiocyanate, RM4-5, 553046), CD8 [phycoerythrin–Cyanine 7 (PE-Cy7), 53-6.7, 552877], CD25 [peridinin chlorophyll protein–Cyanine 5.5 (PerCP-Cy5.5), PC61, 551071], CD69 [allophycocyanin (APC), H1.2F3, 560689], tumor necrosis factor–α (TNFα) (PE, MP6-XT22, 561063), and interferon-γ (IFN-γ) (PerCP-Cy5.5, XMG1.2, 560660) were procured from BD Biosciences (USA). In addition, perforin (APC, S16009A, 154303) and granzyme B (BV421, GB11, 515409) were obtained from BioLegend (USA). Primary cells from mouse tumors were isolated using methods analogous to those used in CyTOF analysis. Cells were blocked on ice for 15 min with anti-mouse CD16/32 antibody (BD Biosciences, USA) and subsequently incubated with surface markers for 30 min on ice. For intracellular staining, cells underwent fixation and permeabilization with a Fixation/Permeabilization kit (BD Biosciences, USA) following the manufacturer’s instructions and were incubated with intracellular markers for 30 min on ice. Flow cytometry analysis was conducted using a FACS Aria III cytometer (BD Biosciences, USA), and the data were analyzed with FlowJo software (v10.4).

IHC and semiquantitative analysis of staining intensity

For IHC, slides were deparaffinized and rehydrated through sequential immersion in xylene and ethanol. Antigen retrieval was performed in citrate buffer at sub-boiling temperatures for 15 min following incubation with 0.3% H₂O₂ (H1009, Sigma-Aldrich, USA) for 30 min to block endogenous peroxidase activity. The slides were treated with Triton X-100 (P0096, Beyotime, China) and blocked with normal goat serum (SL038, Solarbio, UK) for 60 min. Primary antibodies were applied overnight at 4°C, followed by incubation with HRP-conjugated secondary antibodies (7074/7076, CST, USA) at 37°C for 1 hour. Color development was achieved using a 3,3′-diaminobenzidine tetrahydrochloride kit (GK600505, Gene Tech, Shanghai, China), followed by hematoxylin for nuclear counterstaining. Images were captured using a standard microscope (Olympus, Tokyo, Japan) or CaseViewer software (3DHISTECH, Budapest, Hungary). For quantification of IHC staining, the H-score method (65) was used. Two pathologists, blinded to clinical outcomes, scored the staining intensities. The intensities were categorized as 0 (negative), 1 (low), 2 (moderate), and 3 (high). The average percentage of tissue staining positive at each intensity level was calculated. The H-score was then computed using the formula: H-score = (% of tissue stained at intensity 1 × 1) + (% of tissue stained at intensity 2 × 2) + (% of tissue stained at intensity 3 × 3).

Immunofluorescence

Cells were cultured in 24-well plates, and EGF along with canertinib was added to the culture medium as per the experimental design for 24 hours. After incubation, the cells were washed three times with PBS and fixed in methanol for 10 min. Following fixation, cells were treated with Triton X-100 (P0096, Beyotime, China) for 15 min to enhance membrane permeability and then blocked using Blocking Buffer for Immunol Staining (P0260, Beyotime, China) for 30 min. Subsequently, cells were incubated overnight at 4°C with YAP1 antibody (1:100, 14074, CST, USA). The next day, they were incubated with Alexa Fluor 594–conjugated secondary antibody (1:100, A32740, Invitrogen, USA) for 1 hour, followed by DAPI (4′,6-diamidino-2-phenylindole) staining (C1005, Beyotime, China) for 15 min. Each step was accompanied by three PBS washes, 5 min each. Last, images were captured using a fluorescence microscope (Olympus, Tokyo, Japan).

qRT-PCR assay

Total RNA was isolated using the TRIzol reagent (15596026, Invitrogen, USA) and subsequently reverse-transcribed using a cDNA Synthesis Kit (11141ES, Yeasen, China). RNA quality and concentration were assessed using a spectrophotometer (Thermo Fisher Scientific, USA). The SYBR Green Master Mix Kit (11195ES03, Yeasen, China) was used according to the manufacturer’s guidelines for double-stranded DNA detection. qRT-PCR was conducted in triplicate on a LightCycler 480 system (Roche Diagnostics, Germany). The primers used for qRT-PCR are detailed in table S9. These assays were performed at least three times to ensure reproducibility. Relative gene expression levels were quantified using the 2^−ΔΔCt method, where target gene expression was normalized to the reference gene GAPDH and then compared between the treated groups and negative controls.

WB (also called immunoblotting) assay and CoIP

For immunoblotting, cells were lysed using RIPA buffer (P0013B, Beyotime, China) supplemented with protease (P1005, Beyotime, China) and phosphatase inhibitors (P1081, Beyotime, China). Samples were then denatured by boiling in 6× SDS loading buffer (P0015F, Beyotime, China) for 10 min. Ten microliters of the protein extract was loaded into each well of a Bis-Tris SDS–polyacrylamide gel electrophoresis gel for electrophoresis. Following electrophoresis, proteins were transferred to polyvinylidene difluoride membranes. The membranes were blocked with 5% nonfat milk and incubated overnight at 4°C with primary antibodies on a shaker. Subsequently, membranes were incubated with species-matched secondary antibodies (7074/7076, CST, USA) for 60 min at 37°C. Protein bands were visualized using a chemiluminescent HRP substrate (WBKLS0500, Millipore, Billerica, MA, USA) on an ECL imaging system (Bio-Rad, USA).

For CoIP, cells were lysed using NP-40 buffer (P0013F, Beyotime, China) supplemented with protease and phosphatase inhibitors. Then, we incubated the supernatant of centrifuged lysate with antibody and Protein A/G beads (88804, Thermo Fisher Scientific, USA) overnight at 4°C. We then washed beads with a magnetic separator 5× and boiled beads in 1× SDS loading buffer. Last, we proceeded to WB with antibodies against target and suspected interactors.

Antibodies for IHC, IF, and WB analysis and reagents for cell culture and animals

The primary antibodies used for IHC, IF, and WB analyses were as follows: YAP1 (14074, CST, USA), YAP1^S127 (13008, CST), CXCL5 (DF9919, Affinity, USA), glyceraldehyde-3-phosphate dehydrogenase (GAPDH) (2118, CST), LATS1 (3477, CST), p-LATS1 (9157, CST), lamin B1 (12987-1-AP, Proteintech, USA), connective tissue growth factor (CTGF; ab227180, Abcam, UK), Cystein Rich 61 (Cyr61; ab228592, Abcam), GLS (12855-1-AP, Proteintech), GOT1 (34423, CST), GLUL (80636, CST), PSAT1 (ab154055, Abcam), SLC1A5 (ab237704, Abcam), SLC7A5 (32683, CST), and SLC38A1 (28632-1-AP, Proteintech).

Most reagents used for cell culture were sourced from MedChemExpress (USA). The specific concentrations used are detailed in the respective graphs or are as follows: EGF (1 × 10⁴ U/ml), canertinib (5 μM), CB-839 (100 nM), NAC (HY-B0215, 1 mM), BSO (HY-106376, 5 mM), apocynin (HY-N0088, 10 μM), JHU083 (HY-122218, 1 mg/kg, orally), and SB225002 (HY-16711, 2 mg/kg, intraperitoneally). In addition, H₂O₂ (H1009, 200 μM) and Gln solution (G7513, 4 mM) were procured from Sigma-Aldrich (USA). The InVivoMAb anti-mouse Gr-1 antibody (BE0075, 100 μg/mouse, intraperitoneally) was obtained from BioXCell (USA).

Cell fractionation assay

For the extraction of nuclear and cytoplasmic proteins, we used the Nuclear and Cytoplasmic Protein Extraction Kit (P0028, Beyotime, China) following the manufacturer’s instructions. Briefly, cells were collected in precooled PBS and resuspended in cytoplasmic protein isolation solution A, followed by homogenization on ice for 15 min. Subsequently, 10 μl of cytoplasmic protein isolation solution B was added, and the cells were further homogenized on ice. The homogenate was then centrifuged at 16,000g for 5 min at 4°C. The resulting supernatant, containing the cytoplasmic protein fraction, was transferred to a separate tube. The pellet containing the nuclear fraction was lysed in 50 μl of nuclear protein isolation solution and incubated on ice for 30 min. This mixture was centrifuged at 16,000g for 10 min at 4°C, and the supernatant containing the nuclear protein fraction was transferred to another tube. The efficiency of the fractionation was verified by WB analysis, using lamin B1 as a marker for nuclear proteins and GAPDH as a marker for cytoplasmic proteins.

Enzyme-linked immunosorbent assay

ELISA was conducted in accordance with the manufacturer’s instructions. Briefly, MC38 cells were plated at an appropriate density into a 96-well plate and incubated for 24 hours. Subsequently, the cell culture medium was discarded and replaced with an equivalent volume of serum-free DMEM. After an additional 24 hours, supernatants were collected, and floating cells were removed using a 0.45-μm filter. The concentration of secreted CXCL5 protein in the cell culture supernatants was quantified using a mouse-specific ELISA kit (Ab100719, Abcam, UK).

Lentivirus and AAV8 construction

Lentiviral vectors targeting YAP1 were transfected into MA38 murine colon cancer cells and AML12 normal mouse liver cells to establish stable silenced cell lines. Forty-eight hours posttransfection, cells were split and subjected to selection with puromycin (2 μg/ml) for 1 to 2 weeks. The efficiency of transfection was assessed at both protein and mRNA levels. For the construction of stable cell lines, YAP1-specific overexpression lentiviral vectors and short hairpin RNA (shRNA) lentiviral vectors carrying hairpins were constructed. The RNA interference consortium numbers used for the shRNAs of YAP1 are as follows: TRCN0000238436 and TRCN0000095864. AAV8 carrying the YAP1 gene was also constructed, and the virus was administered into mice via tail vein injection at a dose of 2 × 10¹¹ viral particles per mouse.

Measurement of cellular ROS

The levels of ROS were quantified using H2DCFDA (HY-D0940, MedChemExpress, USA), following the manufacturer’s instructions. Briefly, cells were washed with PBS and incubated with 5 μM H2DCFDA at 37°C for 30 min. Subsequently, the cells were trypsinized, centrifuged, and resuspended in PBS. Fluorescence of the APC-Cy7 channel was measured using a FACS Aria III flow cytometer (BD Biosciences, USA). Data analysis was performed with FlowJo software (version 10.4).

Measurement of NADPH and GSH

The NADPH levels in cells were determined using a NADP+/NADPH Assay Kit (S0179, Beyotime, China), following the manufacturer’s instructions. Briefly, cells were lysed with the provided extraction solution and centrifuged at 12,000g for 10 min at 4°C. The supernatant was then heated at 60°C for 30 min to decompose any NADP+. After cooling to room temperature, the supernatant—along with positive and negative controls—was incubated with a working solution for 10 min in the dark. Subsequently, a color-substrate solution was added, and the mixture was incubated for an additional 20 min. The absorbance was measured at 450 nm to determine the NADPH levels.

The GSH levels in cells were assessed using a GSH and GSSG Assay Kit (S0053, Beyotime, China) following the manufacturer’s instructions. Briefly, cells were rapidly frozen in liquid nitrogen and ground into a powder. Protein removal reagent was then added, and the mixture was centrifuged at 10,000g for 10 min at 4°C. The supernatant was collected for the determination of total GSH [including GSH and glutathione disulfide (GSSG)]. Following the removal of GSH with a GSH removing auxiliary solution, GSSG levels were measured. Absorbance was recorded at 412 nm, and the GSH concentration was calculated using the formula: 2 × (total GSH − GSSG).

Measurement of Gln

The Gln levels in tissue and cells were determined using a Glutamine Colorimetric Assay Kit (E-BC-K853-M, Elabscience, China). First, reagent A working solution, reagent B working solution, accelerator working solution, substrate working solution, reaction working solution, and 2 mM standard solution were prepared following the manufacturer’s instructions. Then, a fresh set of standards was prepared to build a serial concentration: 0, 0.4, 0.6, 0.8, 1.2, 1.6, 1.8, and 2.0 mM. Third, samples were prepared. For tissue samples, 0.1 g of liver tissue was weighed out and homogenized with 0.9 ml of normal saline with a Dounce homogenizer at 4°C. Then, it was centrifuged at 10,000g for 15 min to remove insoluble materials. Then, the supernatant was centrifuged with a 50-kDa ultrafiltration tube at 10,000g for 15 min, and the filtrate was collected on ice for detection. For cell samples, 1 × 10⁶ cells were homogenized in 0.2 ml of normal saline with an ultrasonic cell disruptor at 4°C. Then, they were centrifuged with an ultrafiltration tube as mentioned above. Fourth, 30 μl of enzyme reagent A working solution and 50 μl standard solution or samples were mixed and incubated at 37°C for 20 min. Then, a 140-μl reaction working solution was added and the optical density (OD) value at 450 nm was measured with a microplate reader, recorded as A. It was incubated at 37°C for 30 min with shading light. The OD value of each well at 450 nm was measured with a microplate reader, recorded as A2, △A = A2 − A1. Fifth, the standard curve with all standard readings was calculated and the standard curve (y = ax + b) was created with EXCEL. Last, the concentration of Gln was calculated. For the tissue sample, Gln content (mmol/kg wet weight) = (ΔA − b)/a/(m/V), where m stands for the weight of liver tissue (in grams). For the cell sample, Gln content (mmol/10⁶) = (ΔA − b)/a/(n/V), where n stands for the number of cells (10⁶).

Untargeted metabolomics

Animal tissue samples, weighing ~25 ± 1 mg, were processed for analysis. Each sample was mixed with beads and 500 μl of extraction solution comprising methanol (MeOH), acetonitrile (ACN), and water in a 2:2:1 (v/v) ratio. The extraction solution was also supplemented with deuterated internal standards. The mixture was vortexed for 30 s to ensure thorough homogenization. The sample extracts were then analyzed using a dual column chromatography system (Thermo Fisher Scientific Vanquish Duo) coupled to a Thermo Fisher Scientific Q Exactive HF-X Orbitrap high-resolution mass spectrometer. Metabolites were identified based on their mass/charge ratio (m/z), retention time, and comparison with library entries of purified known standards.

ChIP assay

ChIP was conducted using a YAP1 antibody (14074, CST, USA) and EGFR antibody (4267, CST, USA). Briefly, 5 μg of either rabbit immunoglobulin G (2729, CST, USA) or the target antibody was incubated with Protein A Dynabeads magnetic beads (10001D, Invitrogen, USA) for 4 hours, followed by extensive washing to remove any unbound antibody. The antibody-bound beads were then added to chromatin extracted from MC38 cells using DNAzol (10503027, Invitrogen, USA), following the manufacturer’s instructions, and incubated overnight. Subsequently, 60 μl of Protein A Agarose/Salmon Sperm DNA was added to each tube and incubated for 2 hours on a tube rotator at 4°C. The mixtures were then centrifuged, eluted twice, and decross-linked. Last, the DNA was purified and subjected to PCR analysis using the following primers for CXCL5 (forward, 5′-CTCCAGTTTCCTGCCTGAAG-3′ and reverse, 5′-GTGTGGAGATTGGGGCTCTA-3′) and YAP1 (forward, 5′-AGACAGAGT CTCGCTGTGTTG-3′ and reverse, 5′-CCAAAATGGTGATACCCTGTC-3′).

Gene set enrichment analysis

GSEA was performed using GSEA 4.2.3 software (downloaded in https://gsea-msigdb.org/gsea). First, the expression matrix file (including the gene symbol, gene description, and amount of gene expression in each sample) was prepared using RNA-seq data according to the manufacturer’s guidelines. Then, the phenotype file was prepared to edit the sample grouping information. Third, oncogenic gene sets for Mus musculus (including 189 gene sets) were prepared. Fourth, the expression matrix file, phenotype file, and gene sets file were uploaded to GSEA software, and analysis was run. Last, the results were analyzed and enriched gene sets (signaling pathways) were identified. Genes in YAP1 conserved signature originated from the work of Wang and colleagues (25).

TCGA analysis for MDSC signature, YAP1 expression, and CXCL5 expression in tumors of patients with COAD and LIHC

The list of putative signature genes for human MDSCs was based on the work of Wang and colleagues (25). We analyzed COAD and LIHC samples from TCGA database. The analysis was conducted using gene expression data for 39 MDSC-specific genes, using GEPIA 2 (http://gepia2.cancer-pku.cn) for clustering the samples.

Statistical analysis

All statistical analyses were conducted using R software (version 4.3.3, New Zealand) and GraphPad Prism (version 10, San Diego, CA, USA). Data are presented as means ± SD. Differences between two groups were assessed using Student’s t test. When analyzing differences among three or more groups, one-way or two-way analysis of variance (ANOVA) was used, followed by post hoc pairwise comparisons with Bonferroni correction to control for multiple testing. Survival and recurrence rates were analyzed using the Kaplan-Meier analysis, log-rank test, and Cox regression analysis. Pearson’s correlation analysis was used to examine relationships between variables. All statistical tests were two-tailed, and statistical significance was denoted as not significant (n.s.), *P < 0.05, **P < 0.01, and ***P < 0.001.

Supplementary Materials

This PDF file includes:

Figs. S1 to S14

Tables S1 to S9