HOXD13-mediated immune crosstalk between cancer cells and tumor-associated macrophages drives colorectal cancer progression

Abstract

Background

The complex tumor microenvironment (TME) of colorectal cancer (CRC), composed of diverse cellular components and dynamic interactions, constitutes a major barrier to effective immunotherapy and facilitates disease progression. There is a pressing need to elucidate CRC-intrinsic factors that induce the immunosuppressive TME. Here, we explored the role of homeobox D13 (HOXD13) in shaping the immune microenvironment of CRC and its contribution to immunosuppression.

Methods

The expression level of HOXD13 was assessed using quantitative real-time PCR, immunoblotting, and immunohistochemistry. The role of HOXD13 in CRC was investigated using orthotopic allograft models and azoxymethane/dextran sulfate sodium-induced spontaneous tumor models in intestine-specific HOXD13 knockout and knock-in mice. The immune landscape of the CRC microenvironment was characterized via flow cytometry and immunofluorescence.

Results

Our study revealed the upregulation mechanism of HOXD13 in CRC and its functional role in fostering an immunosuppressive TME. HOXD13 was upregulated in CRC, particularly in metastatic cases, and patients exhibiting high HOXD13 expression showed poorer clinical outcomes. Mechanistically, HOXD13 promoted M2-type polarization of tumor-associated macrophages (TAMs) and suppressed CD8⁺ T cells mediated antitumor immunity by transcriptionally upregulating amphiregulin (AREG) and paired immunoglobulin like type 2 receptor alpha (PILRA) in CRC cells. Concurrently, transforming growth factor beta 1 released from M2-polarized TAMs further augmented HOXD13 expression in CRC cells via activation of the Smad2/3 signaling pathway. This reciprocal interaction formed a self-reinforcing loop that sustained immunosuppression and thereby accelerated tumor progression. Notably, combined inhibition of AREG and programmed cell death ligand 1 effectively disrupted this crosstalk, restored antitumor immunity, and ultimately suppressed CRC progression.

Conclusions

Our study identified HOXD13 as a pivotal regulator in the establishment of an immunosuppressive TME and suggested that targeting the HOXD13 signaling axis represents a promising strategy to sensitize CRC to immunotherapy.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• The dysregulation of key transcription factors (TFs) is critically involved in shaping an immunosuppressive tumor microenvironment (TME), thereby substantially promoting tumor progression. For patients with metastatic colorectal cancer (CRC), the TME at their primary tumor site frequently exhibits immunosuppressive characteristics, largely driven by the infiltration of myeloid cells that impair antitumor immunity. Nevertheless, a clear understanding of how TFs drive this process is still lacking.

WHAT THIS STUDY ADDS

• Our study identifies homeobox D13 (HOXD13) as a key TF in promoting the immunosuppressive TME of CRC. HOXD13 transcriptionally upregulates the expression of amphiregulin (AREG) and paired immunoglobulin like type 2 receptor alpha in CRC cells, thereby promoting M2 polarization of tumor-associated macrophages (TAMs) and suppressing CD8⁺ T cell-mediated antitumor immunity. Meanwhile, transforming growth factor beta 1 secreted by M2-type TAMs further enhances HOXD13 expression in CRC cells, forming a positive feedback loop that exacerbates immunosuppression and accelerates CRC progression.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• High expression of HOXD13 serves as a negative prognostic biomarker in CRC. Targeting AREG inhibits HOXD13-mediated CRC progression and synergizes with immune checkpoint inhibitor by disrupting the immune crosstalk between cancer cells and M2-type TAMs.

Introduction

Colorectal cancer (CRC) represents a significant global health burden, being the third most common cancer diagnosis and the second leading cause of cancer mortality.¹ CRC often presents with asymptomatic early stages, resulting in a significant proportion of cases being diagnosed at an advanced stage. 20% of patients present with metastases at initial diagnosis, with a further 25% of localized cases progressing to distant dissemination. Metastatic colorectal cancer (mCRC) is often refractory to conventional therapies, with a discouragingly low 5-year survival rate.² Despite revolutionizing cancer treatment, immunotherapy elicits a poor response in most patients with mCRC. The intricate tumor microenvironment (TME) of CRC, characterized by its unique cellular constituents and interactions, serves as a critical impediment to immunotherapy, fostering disease progression and metastasis.^{3 4} Elucidating the key mechanism driving the formation of the immunosuppressive TME in CRC is essential to develop novel therapeutic strategies.

The TME is a complex ecosystem of malignant cells, stromal components, immune cells, and signaling molecules that collectively drive immunosuppression and cancer metastasis.⁵ Central to this process are tumor-associated macrophages (TAMs), the most abundant immune cells in the TME, which are pivotal drivers of tumor progression. The M2 subset of TAMs exerts tumor-promoting effects by expressing specific cell surface proteins and secreting soluble factors.⁶ For instance, exosomal miR-934 from CRC cells induces M2 macrophages polarization. These polarized M2 macrophages promote CRC liver metastasis through the secretion of CXCL13, which in turn activates a CXCL13/CXCR5/NFκB/p65/miR-934 positive feedback loop in CRC cells.⁷ Furthermore, in CRC, the presence of robust M2 macrophages infiltration within tumors correlates with an increased risk of liver metastasis and poorer survival, and predicts a diminished response to immunotherapy.8,10 Therefore, elucidating the mechanism governing the infiltration and polarization of TAMs in CRC could enable more precise targeting of the immunosuppressive TME, thus presenting a viable therapeutic strategy for this disease.

Dysregulation of key transcription factors (TFs) plays a critical role in the formation of an immunosuppressive TME and contributes significantly to tumor progression.¹¹ The human homeobox (HOX) gene family consists of several highly conserved TFs that function as master regulators of embryonic development. These TFs play key regulatory roles in essential cellular processes such as proliferation, apoptosis, and motility. A total of 39 human HOX genes have been classified into four clusters—HOXA, HOXB, HOXC, and HOXD—based on their sequence homology and chromosomal localization.¹² The HOXD subfamily comprises nine members that can function as tumor activators or suppressors. This subfamily plays a pleiotropic role in driving tumorigenesis and cancer progression in multiple cancer types.¹³ HOXD13 dysregulation has been observed in a range of cancers, where it is implicated in tumor progression and shows a significant correlation with clinical prognosis. For example, studies in breast cancer have demonstrated that HOXD13 is downregulated in tumor tissues and that low expression levels are associated with elevated mortality risk.¹⁴ Conversely, HOXD13 has been shown to promote progression and metastasis in other malignancies, such as gastric cancer and glioma.^{15 16} Previous studies have demonstrated that HOXD13 was significantly upregulated in CRC and was positively associated with poor prognosis.^{17 18} Nevertheless, the functional mechanism of HOXD13 in CRC progression and its potential role in remodeling the TME require further elucidation.

In this study, we elucidated the role of HOXD13 in mediating immune crosstalk between cancer cells and TAMs, which contributed to immunosuppression and CRC progression. Specifically, HOXD13 transcriptionally upregulated the expression of amphiregulin (AREG) and paired immunoglobulin like type 2 receptor alpha (PILRA) in CRC cells, thereby promoting M2 polarization of TAMs and suppressing CD8⁺ T cell-mediated antitumor immunity. Meanwhile, transforming growth factor beta 1 secreted by M2-type TAMs further enhanced HOXD13 expression in CRC cells via the Smad2/3 signaling complex, forming a positive feedback loop that exacerbated immunosuppression and accelerated CRC progression. Concurrent inhibition of AREG and programmed cell death ligand 1 (PD-L1) effectively disrupted this crosstalk, restored antitumor immunity, and ultimately suppressed CRC progression.

Materials and methods

Genetically modified mice

The Hoxd13^flox/flox mouse was generated via CRISPR/Cas9-mediated genome engineering by inserting loxP sites flanking exon 1 of Hoxd13. To establish intestinal epithelial cells (IECs)-specific Hoxd13 knockout mice (Hoxd13^ΔIEC), Hoxd13^flox/flox mice were crossed with Vil1-Cre mice. The Hoxd13^f-stop-f mice were constructed by inserting the Rosa26-pCAG-loxp-STOP-loxp-mHoxd13-pA cassette into intron 1 of the Rosa locus using CRISPR/Cas9. These mice were then bred with Vil1-Cre mice to generate IECs-specific Hoxd13 knock-in mice (Hoxd13^IEC-OE). All mice were on a C57BL/6J genetic background and were purchased from GemPharmatech (Nanjing, China).

For detailed materials and methods, please refer to online supplemental information.

Results

Elevated HOXD13 expression in CRC promotes progression by shaping an immunosuppressive microenvironment

First, to investigate the expression of HOXD13 in CRC, we analyzed The Cancer Genome Atlas dataset and found that HOXD13 expression is elevated in patients with advanced T stage and M1 stage disease (online supplemental figure S1A and B). The CMS4 subtype, which is characterized by poorer clinical outcomes and activation of the transforming growth factor beta (TGFβ) signaling pathway,¹⁹ showed a significant upregulation of HOXD13 expression compared with the CMS1 (MSI immune) subtype (online supplemental figure S1C). Analysis revealed elevated HOXD13 expression in tumors exhibiting microsatellite stability or low tumor mutation burden—two features that are associated with unfavorable responses to immune checkpoint blockade20,22 (online supplemental figure S1D and E). These bioinformatic analyses indicated that elevated HOXD13 expression was correlated with poorer clinical outcomes in patients with CRC.

Next, we measured HOXD13 expression in a cohort comprising 100 paired CRC and adjacent non-tumor tissue samples, along with 20 additional normal epithelial controls. The messenger RNA (mRNA) expression level of HOXD13 was significantly elevated in CRC tissues relative to both adjacent non-tumor tissues and normal epithelial controls. Patients with CRC who developed metastasis or recurrence exhibited significantly elevated levels of HOXD13 mRNA compared with disease-free patients (figure 1A). Moreover, the mRNA and protein level of HOXD13 was significantly higher in metastatic CRC tissues compared with both primary CRC samples and adjacent non-tumor tissues (figure 1B, online supplemental figure S2).

Figure 1. Elevated HOXD13 expression in CRC promoted progression by shaping an immunosuppressive microenvironment. (A) The mRNA level of HOXD13 was quantified in the patient samples. (B) Representative IHC staining for HOXD13 and its quantitative analysis in patient tissues of cohort II. (C) Representative IHC images showing HOXD13 expression and its quantitative scores in two independent CRC cohorts. (D) The clinical outcomes of patients with CRC stratified by HOXD13 expression. (E) The expression level of HOXD13 was quantified in a panel of CRC cell lines. (F) Following lentivirus transduction, HOXD13 expression was analyzed in MC38 cells. (G–I) Following randomization into two groups (n=10 independent mice per group), C57BL/6 mice received orthotopic inoculation with either MC38-control or MC38-HOXD13 cells. Tumor progression was subsequently tracked using an IVIS imaging system. Representative in vivo BLI obtained at the experimental endpoint, along with the dynamic changes in bioluminescent intensity (G), OS (H), and representative H&E-stained liver tissue sections with corresponding quantification of hepatic metastases (I) were shown for each group. (J) t-SNE visualization of multicolor FCM data and quantitative assessment of immune cell populations (n=5 independent tumors from 5 mice per group). (K) Representative immunofluorescence micrographs depicting CD8 and granzyme B co-staining patterns alongside CD206 expression in orthotopic tumor tissue sections. (L) Representative images showing HOXD13, CD206, and CD8 expression in CRC samples, with an analysis of their correlative relationships. BLI, bioluminescence images; CRC, colorectal cancer; FCM, flow cytometry; HOXD13, homeobox D13; IHC, immunohistochemistry; IVIS, in vivo imaging system; mRNA, messenger RNA; OS, overall survival; t-SNE, t-distributed stochastic neighbor embedding.

We then investigated HOXD13 expression and assessed its prognostic value in two independent human CRC cohorts. Immunohistochemistry (IHC) analysis confirmed a pronounced upregulation of HOXD13 protein in CRC tissues compared with non-cancerous samples (figure 1C). HOXD13 expression was significantly associated with adverse clinicopathological features, such as poorer tumor differentiation, distant metastasis, and a higher American Joint Committee on Cancer (AJCC) stage (online supplemental table 1). Positive HOXD13 expression was correlated with adverse clinical outcomes, including increased recurrence and diminished overall survival (OS) in patients with CRC (figure 1D). Multivariate analysis also confirmed HOXD13 expression as an independent predictor of poor prognosis in patients with CRC (online supplemental table 2). Furthermore, HOXD13 expression was analyzed across a panel of CRC cell lines. Results revealed that HOXD13 expression was significantly upregulated in highly metastatic CRC cells (SW620, LoVo) relative to those with low metastatic potential (SW480, Caco-2)23,25 (figure 1E).

To assess the tumor-intrinsic functions of HOXD13, we evaluated its cell-autonomous effects on malignancy using cell counting kit-8 (CCK-8), colony formation, and transwell assays. The results showed that neither knockdown nor overexpression of HOXD13 significantly altered the proliferation, migration, or invasive capacity of CRC cells (online supplemental figure S3). Given the lack of a direct cell-intrinsic effect, we hypothesized that HOXD13 likely promotes CRC progression and metastasis by remodeling the immunosuppressive TME.

To elucidate the role of HOXD13 in CRC dissemination under immunocompetent conditions, orthotopic CRC murine models were generated through inoculation of MC38 cells overexpressing HOXD13 (MC38-HOXD13) into C57BL/6 mice (figure 1F). HOXD13 overexpression markedly accelerated local tumor growth, increased the incidence of liver metastasis, and elevated the number of hepatic metastatic nodules, while concurrently shortening the OS of the mice (figure 1G–I). The composition of key immune cells within the TME was assessed via flow cytometry (FCM). HOXD13 overexpression was found to suppress CD8⁺ T cells infiltration and promote the expansion of M2-type TAMs (figure 1J). Immunofluorescence (IF) analysis further confirmed that HOXD13 overexpression facilitated the accumulation of M2-type TAMs and reduced the infiltration of CD8⁺ T cells, particularly the cytotoxic subset (figure 1K). To mitigate potential bias associated with a specific cell line, we established orthotopic CRC models in BALB/c mice by injecting CT26 cells overexpressing HOXD13 (CT26-HOXD13) (online supplemental figure 4A). Consistent with previous findings, HOXD13 upregulation significantly promoted hepatic metastasis and reduced OS in tumor-bearing mice (online supplemental figure 4B-D). FCM and IF analyses further verified that HOXD13 overexpression drove the accumulation of M2-type TAMs while simultaneously inhibiting CD8⁺ T cells infiltration and effector function (online supplemental figure 4E and F).

To elucidate the clinical relevance of these findings, we used multicolor immunofluorescence to analyze the relationship between HOXD13, M2 macrophages, and CD8⁺ T cells in clinical CRC samples. Notably, a significant positive association was observed between HOXD13 expression and M2 macrophages infiltration, whereas a significant negative association was found with CD8⁺ T cells abundance (figure 1L). Taken together, these findings demonstrated that HOXD13 drove CRC progression through the establishment of an immunosuppressive TME, marked by the enrichment of M2-type TAMs and a decline in cytotoxic CD8⁺ T cells infiltration.

HOXD13 activates the transcription of AREG and PILRA

To elucidate the mechanism through which HOXD13 fosters an immunosuppressive TME, we conducted RNA sequencing (RNA-seq) analysis to compare genome-wide mRNA expression profiles on genetic manipulation of HOXD13 in SW480. RNA-seq analysis revealed 194 upregulated and 239 downregulated genes following HOXD13 overexpression (figure 2A). Gene Ontology and Kyoto Encyclopedia of Genes and Genomes pathway analysis revealed that the upregulated genes were primarily enriched in “cytokine signaling”, “regulation of immune response”, and “suppression of immune cells” (figure 2B, online supplemental figure S5A). Among these upregulated genes, several established tumor-promoting cytokines and chemokines were identified as potential HOXD13 targets (online supplemental figure S5B). This finding suggests that HOXD13 may coordinately regulate multiple immunosuppressive factors to drive immune evasion in CRC. Among the differentially expressed genes (DEGs), AREG and PILRA exhibited the most significant upregulation in response to HOXD13 overexpression (figure 2C). While PILRA is primarily a transmembrane protein expressed on myeloid cells, recent studies have detected its presence on various tumor cells. Furthermore, PILRA interacts with CD99, a membrane protein on T cells, thereby inhibiting T cells activation.^{26 27} As an epidermal growth factor receptor (EGFR) ligand, AREG orchestrates several key facets of cancer pathogenesis, such as tumor proliferation, invasion, metastasis, and the development of drug resistance.²⁸ Given their crucial role in cancer progression, PILRA and AREG were selected for further investigation. PCR and immunoblotting analyses confirmed that HOXD13 overexpression markedly increased expression levels of AREG and PILRA, while HOXD13 knockdown significantly reduced their expression (figure 2D). We subsequently performed a luciferase reporter assay. The results indicated that increased HOXD13 expression robustly enhanced the promoter activities of both AREG and PILRA (figure 2E). Bioinformatic analysis revealed multiple putative HOXD13-binding sites located in the promoter regions of AREG and PILRA (online supplemental figure 6 and 7). Subsequent serial deletion and site-directed mutagenesis of the AREG promoter demonstrated that a putative HOXD13-binding site located between –1910 bp and –777 bp was responsible for HOXD13-mediated transactivation. Similarly, in the PILRA promoter, a predicted HOXD13-binding site located between –803 bp and –123 bp was found to be essential for HOXD13-induced transactivation (figure 2F). Chromatin immunoprecipitation (ChIP) assays verified that HOXD13 directly binds to the promoter regions of AREG and PILRA (figure 2G). These findings indicated that HOXD13 transcriptionally activated the expression of both AREG and PILRA.

Figure 2. HOXD13 activated the transcription of AREG and PILRA. (A) The DEGs between SW480-HOXD13 and SW480-control cells were displayed in the volcano plot, with the top five DEGs labeled. (B) KEGG enrichment analysis was performed on genes upregulated in SW480-HOXD13 cells relative to controls. (C) The top 10 DEGs were displayed by heatmap. (D) AREG and PILRA expression was assessed in the indicated CRC cells. (E) Dual-luciferase reporter assay showing AREG or PILRA promoter activity in SW480 cells on co-transfection with pCMV-HOXD13. (F) To identify the responsive regions, serially truncated or mutated promoter constructs of AREG or PILRA were co-transfected with pCMV-HOXD13 into SW480 cells, and luciferase activity was measured. (G) ChIP analysis of HOXD13 binding to the AREG and PILRA promoters. (H) Analysis of HOXD13 and PILRA expression by IHC in a human CRC tissue microarray, with correlation analysis. (I) HOXD13 and PILRA protein levels were assessed in the indicated cell lines. (J–L) Following randomization into two groups (n=10 independent mice per group), C57BL/6 mice received orthotopic inoculation with the indicated cell. Tumor progression was subsequently tracked using an IVIS imaging system. Representative in vivo BLI obtained at the experimental endpoint, along with the dynamic changes in bioluminescent intensity (J), OS (K), and representative H&E-stained liver tissue sections with corresponding quantification of hepatic metastases (L) were shown for each group. (M–N) The indicated tumor-infiltrating immune cells were analyzed by FCM (M) and by immunofluorescence (N) (n=5 independent tumors from five mice per group). AREG, amphiregulin; BLI, bioluminescence images; ChIP, chromatin immunoprecipitation; CRC, colorectal cancer; DEG, differentially expressed gene; FCM, flow cytometry; HOXD13, homeobox D13; IHC, immunohistochemistry; IVIS, in vivo imaging system; KEGG, Kyoto Encyclopedia of Genes and Genomes; OS, overall survival; PILRA, paired immunoglobulin like type 2 receptor alpha.

To evaluate the clinical relevance of HOXD13 and PILRA expression, IHC staining was performed using human CRC cohorts. The expression of HOXD13 and PILRA showed a positive relationship in CRC samples (figure 2H). Elevated PILRA expression in CRC was associated with a poorer prognosis (online supplemental figure 8A) and a more malignant phenotype (online supplemental table 3). Patients exhibiting co-expression of HOXD13 and PILRA conferred the most unfavorable outcomes among all subgroups analyzed (online supplemental figure 8B).

We next investigated the role of PILRA in HOXD13-mediated CRC metastasis. Knockdown of PILRA in MC38-HOXD13 cells markedly suppressed local tumor growth, reduced liver metastasis, and extended OS in mice (figure 2I–L). FCM and IF analyses collectively demonstrated that PILRA downregulation in cancer cells promoted the infiltration and activity of CD8⁺ T cells, whereas the abundance of M2-type TAMs remained unchanged (figure 2M,N). The above findings indicated that membrane-bound PILRA on tumor cells primarily modulated the function of CD8⁺ T cells. We next investigated the functional impact on CD8⁺ T cells by co-culturing them with cancer cells (online supplemental figure 9A). HOXD13 overexpression in CRC cells markedly inhibited the proliferation and interferon-γ secretion of CD8⁺ T cells. Conversely, PILRA knockdown in MC38-HOXD13 cells effectively restored CD8⁺ T cells function (online supplemental figure 9B and C). The direct interaction between PILRA and CD99 was also validated using co-immunoprecipitation, ELISA, and FCM (online supplemental figure 10). These findings indicated that HOXD13 suppressed antitumor immune responses and enhanced CRC metastasis by upregulating PILRA expression.

Overexpression of HOXD13 in CRC cells drives M2 macrophages polarization through the AREG-EGFR axis

We assessed the contribution of TAMs to HOXD13-mediated CRC metastasis by depleting them with clodronate liposomes. In the orthotopic CRC model, depleting TAMs not only curtailed tumor growth and hepatic metastasis but also prolonged the OS of mice bearing HOXD13-overexpressing tumors (figure 3A–C). The effectiveness of TAMs depletion was further confirmed through both FCM and IF analyses, which also demonstrated a restoration in the abundance and activity of CD8⁺ T cells (figure 3D–E). These findings indicated that HOXD13 promoted CRC progression in a manner dependent on TAMs.

Figure 3. Overexpression of HOXD13 in CRC cells drove M2 macrophages polarization through the AREG-EGFR axis. (A–C) C57BL/6 mice bearing orthotopic MC38-HOXD13 tumors were administered either clodronate liposomes or control liposomes via injection (n=10 independent mice per group). Representative in vivo BLI obtained at the experimental endpoint, along with the dynamic changes in bioluminescent intensity (A), OS (B), and representative H&E-stained liver tissue sections with corresponding quantification of hepatic metastases (C) were shown for each group. (D–E) The indicated tumor-infiltrating immune cells were analyzed by immunofluorescence (D) and by FCM (E) (n=5 independent tumors from five mice per group). (F) The expression of inflammation-related genes and the secretion of their corresponding cytokines were quantified in macrophages following stimulation with recombinant AREG, with or without gefitinib co-treatment. (G–H) The expression of inflammation-related genes (G) and the secretion of their corresponding cytokines (H) were measured in macrophages following stimulation with CM collected from the specified CRC cell lines. (I) HOXD13 and AREG protein levels were assessed in the indicated cell lines. (J–L) C57BL/6 mice received orthotopic inoculation with indicated cells (n=10 independent mice per group). Tumor progression was subsequently tracked using an IVIS imaging system. Representative in vivo BLI obtained at the experimental endpoint, along with the dynamic changes in bioluminescent intensity (J), OS (K), and representative H&E-stained liver tissue sections with corresponding quantification of hepatic metastases (L) were shown for each group. (M–N) The indicated tumor-infiltrating immune cells were analyzed by immunofluorescence (M) and by FCM (N) (n=5 independent tumors from five mice per group). AREG, amphiregulin; BLI, bioluminescence images; CM, conditioned medium; CRC, colorectal cancer; EGFR, epidermal growth factor receptor; FCM, flow cytometry; HOXD13, homeobox D13; IVIS, in vivo imaging system; OS, overall survival.

Accumulating evidences have indicated that AREG contributes to the establishment of an immunosuppressive TME.²⁹ Based on this, we hypothesized that AREG derived from CRC cells may reprogram the functional phenotypes of TAMs. Subsequently, macrophages were stimulated with recombinant AREG, and the level of multiple inflammation-related genes was analyzed. The results demonstrated that AREG stimulation suppressed the expression of M1 markers (eg, interleukin (IL)-1β and tumor necrosis factor-α) while upregulating that of PD-L1 and M2 markers (including TGF-β1 and IL-10) in macrophages. However, all of these effects were abolished on treatment with the EGFR inhibitor gefitinib, indicating a dependence on EGFR signaling (figure 3F, online supplemental figure 11A). Next, we sought to determine whether HOXD13-overexpressing CRC cells could promote macrophages M2 polarization through a paracrine mechanism. On stimulation with conditioned medium (CM) from HOXD13-overexpressing CRC cells, macrophages assumed an M2-polarized phenotype and exhibited concomitant upregulation of PD-L1 expression. Nevertheless, these effects were eliminated following AREG knockdown in the CRC cells (figure 3G,H, online supplemental figure 11B Figure 11B). Next, we explore the mechanism of AREG-induced macrophages M2 polarization. First, we extracted TAMs from MC38-control and MC38-HOXD13 orthotopic tumors, examined the activation of multiple downstream signaling pathways of EGFR and found that the PI3K/AKT and ERK/MAPK signaling pathways in tumor-infiltrating TAMs of MC38-HOXD13 were significantly activated (online supplemental figure 12A). Next, we co-cultured macrophages with CM from CRC cells. The results demonstrated that CM from HOXD13-overexpressing cells upregulated TGF-β1 and PD-L1 expression and activated EGFR and multiple signaling pathways in macrophages. These effects were reversed by anti-AREG monoclonal antibody (mAb) treatment, which counteracted the upregulation of TGF-β1 and PD-L1 and inhibited the activation of the EGFR and PI3K/AKT pathways (online supplemental figure 12B). Previous studies have shown that activation of the PI3K/AKT pathway in macrophages promotes M2 polarization.^{7 30} Consistent with this, the AREG-induced M2 polarization of macrophages was also reversed by treatment with an AKT inhibitor (online supplemental figure 12C and D). These results demonstrated that HOXD13-overexpressing CRC cells promoted macrophages M2 polarization through AREG/EGFR/PI3K/AKT axis.

To investigate the involvement of AREG in HOXD13-mediated CRC metastasis, we downregulated AREG expression in MC38HOXD13 cells using lentivirus-mediated gene silencing (figure 3I). Knockdown of AREG in MC38-HOXD13 cells markedly suppressed local tumor growth, reduced liver metastasis, and extended OS in mice (figure 3J–L). FCM and IF analyses consistently confirmed that inhibition of AREG expression resulted in a marked decrease in M2-type TAMs, accompanied by enhanced infiltration and functional activation of CD8⁺ T cells (figure 3M,N). The clinical association of HOXD13 with AREG was elucidated via IHC analysis in two independent CRC cohorts. We observed a strong positive association linking HOXD13 and AREG protein expression in our analysis of clinical CRC samples (online supplemental figure 13A). Patients exhibiting positive AREG expression presented with more aggressive tumor behavior (online supplemental table 4). Moreover, AREG positivity was significantly correlated with a worse prognosis. Patients with co-expression of HOXD13 and AREG have the most unfavorable prognosis among all subgroups (online supplemental figure 13B and C). Taken together, these findings established that AREG secreted by CRC cells was critical for both HOXD13-induced M2 polarization of TAMs and HOXD13-mediated metastasis in CRC.

HOXD13 is a critical mediator through which TGFβ1 drives CRC progression and metastasis

Inflammation is a known driver of CRC initiation and progression.³¹ Therefore, we hypothesized that HOXD13 expression may be driven by proinflammatory cytokines within the TME. We treated SW480 and Caco-2 cells (which exhibit low basal HOXD13 expression) with a panel of pro-inflammatory cytokines implicated in CRC progression to test this hypothesis. Notably, TGF-β1 exhibited the most robust induction of HOXD13 expression in both cell lines (figure 4A). Moreover, TGF-β1 upregulated HOXD13 expression in a concentration-dependent and time-dependent manner and further enhanced the transcriptional activity of its promoter, as evidenced by increased luciferase reporter activity (figure 4B,C, online supplemental figure 14A). TGF-β1 initiates signaling by binding to TGF-β type II and type I receptors, resulting in the assembly of a heterotrimeric receptor complex. This activation triggers the phosphorylation of Smad2/3, which then forms a complex with Smad4. The resulting Smad complex translocates into the nucleus and modulates the transcription of target genes. Motif analysis of the HOXD13 promoter identified several putative Smad binding elements (SBEs), suggesting a potential mechanism for its transcriptional regulation by TGF-β1/Smad signaling. Serial deletion and site-directed mutagenesis identified putative SBE site 2 as essential for mediating TGF-β1-induced transcriptional activation (figure 4D). Silencing Smad2/3 abrogated the TGF-β1-induced upregulation of HOXD13 and its promoter activity (figure 4E,F). Inhibition of the TGF-β type I receptor with vactosertib abolished TGF-β1-induced HOXD13 upregulation, establishing its essential function in this process (figure 4G). We also found that the Smad2/3 complex specifically binds to the HOXD13 promoter, as evidenced by ChIP (figure 4H).

Figure 4. HOXD13 was a critical mediator through which TGFβ1 drove CRC progression and metastasis. (A) After exposure to the specified cytokines, HOXD13 expression in SW480 and Caco-2 cells was assessed by immunoblotting. (B) After exposure to escalating concentrations of TGF-β1, HOXD13 expression was evaluated in SW480 and Caco-2 cells. (C) Dual-luciferase reporter assay of HOXD13 promoter activity in TGF-β1-stimulated CRC cells. (D) Transfection of SW480 cells with serially truncated or mutated HOXD13 promoter constructs was followed by TGF-β1 treatment and subsequent assessment of promoter activity. (E) The transcriptional activation of HOXD13 by TGF-β1 following Smad2/3 silencing was assessed. (F) HOXD13 expression in response to TGF-β1 stimulation following Smad2/3 silencing was analyzed. (G) HOXD13 expression was assessed following inhibition of TGF-β1 signaling using the TGFβR1 inhibitor vactosertib. (H) ChIP assay demonstrating Smad2/3 complex binding to the HOXD13 promoter. (I) Protein levels of TGF-β1 and HOXD13 were evaluated in the specified cell lines. (J–L) C57BL/6 mice received orthotopic inoculation with indicated cells (n=10 independent mice per group). Representative in vivo BLI obtained at the experimental endpoint, along with the dynamic changes in bioluminescent intensity (J), representative H&E-stained liver tissue sections with corresponding quantification of hepatic metastases (K), and OS (L) were shown for each group. (M–N) The indicated tumor-infiltrating immune cells were analyzed by immunofluorescence (M) and by FCM (N) (n=5 independent tumors from five mice per group). BLI, bioluminescence images; ChIP, chromatin immunoprecipitation; CRC, colorectal cancer; FCM, flow cytometry; HOXD13, homeobox D13; OS, overall survival; TGF-β1, transforming growth factor beta 1.

We next investigated the bidirectional interaction between macrophages and CRC cells (online supplemental figure S14B). Macrophages showed a marked increase in TGF-β1 expression after stimulation with CM from HOXD13 overexpressing CRC cells. However, this effect was abolished on treatment with an anti-AREG monoclonal antibody (online supplemental figure S14C and D). We then co-cultured these stimulated macrophages with CRC cells for 24 hours, after which HOXD13 expression was evaluated. Exposure to macrophages pre-stimulated with CM from HOXD13 overexpressing CRC cells induced a significant upregulation of HOXD13 in CRC cells. This upregulation was reversed by the addition of either vactosertib or anti-TGF-β1 antibody (online supplemental figure S14E-G). These results supported the presence of a TGF-β1–HOXD13 positive feedback loop between CRC cells and M2-type TAMs.

Given that HOXD13 is induced by TGF-β1 signaling and promotes CRC dissemination, we investigated its potential role in TGF-β1-mediated immune evasion and metastasis. TGF-β1 overexpression significantly promoted local tumor progression, enhanced the frequency of liver metastasis, and raised the count of hepatic metastatic nodules, while simultaneously reducing the OS of the mice. Conversely, knockdown of HOXD13 in TGF-β1-overexpressing MC38 cells rescued the pro-metastatic phenotypes and prolonged OS in mice (figure 4I–L). FCM and IF analysis showed that MC38-TGF-β1 tumor-bearing mice exhibited a lower density of tumor-infiltrating CD8⁺ T cells and an increased abundance of M2-type TAMs. Notably, HOXD13 silencing significantly reversed these TGF-β1-mediated immunosuppressive changes (figure 4M,N). Overall, our results demonstrated that the TGF-β1–p-Smad2/3 signaling pathway drove HOXD13 upregulation and emphasized the essential function of HOXD13 in mediating TGF-β1-promoted CRC progression.

IECs-specific Hoxd13 knockout curtailed CRC initiation and progression, whereas IECs-specific Hoxd13 knock-in exacerbated this process

Using the azoxymethane/dextran sulfate sodium (AOM/DSS) model, we established a murine model to examine the roles of HOXD13 in colitis-associated cancer (CAC) (figure 5A,B). We isolated IECs from wild-type mice at various time points following AOM/DSS treatment and assessed the expression levels of HOXD13 and its associated signaling axis. We observed a sequential increase in the expression levels of HOXD13, p-Smad2/3, AREG, PILRA, and TGF-β1 throughout CRC development and progression (figure 5C). To elucidate the role of HOXD13 in CAC, we generated mice with an IECs-specific knockout of HOXD13 (Hoxd13^ΔIEC) (figure 5D, online supplemental figure 15A). Compared with Hoxd13^fl/fl mice, Hoxd13^ΔIEC mice developed significantly fewer colon tumors and had smaller tumor volumes after AOM/DSS treatment (figure 5E). Immunoblotting analysis demonstrated a marked decrease in AREG and PILRA protein levels in IECs derived from Hoxd13^ΔIEC mice following AOM/DSS administration (figure 5F). Hoxd13^ΔIEC mice exhibited a robust shift in the immunosuppressive microenvironment, featuring a striking increase in activated CD8⁺ T cells and a marked reduction in TAMs infiltration (figure 5G,H). Next, we engineered transgenic mice for conditional overexpression of HOXD13 within IECs (Hoxd13^IEC-OE) (figure 5I, online supplemental figure 15B). Compared with controls, Hoxd13^IEC-OE mice subjected to AOM/DSS treatment developed a markedly greater number of colon tumors and achieved larger tumor sizes (figure 5J). AREG and PILRA protein levels were notably elevated in IECs from Hoxd13^IEC-OE mice following AOM/DSS treatment (figure 5K). IEC-specific HOXD13 overexpression was associated with an exacerbated immunosuppressive phenotype, manifesting as increased TAMs infiltration and a reduction in cytotoxic CD8⁺ T cells in the TME (figure 5L,M). Overall, these findings demonstrated that IECs-specific HOXD13 knockout attenuated AOM/DSS-induced colorectal carcinogenesis, whereas IECs-specific HOXD13 overexpression exacerbated these processes. These results highlighted the pivotal role of HOXD13 in modulating the CRC TME.

Figure 5. IECs-specific Hoxd13 knockout curtailed CRC initiation and progression, whereas IECs-specific Hoxd13 knock-in exacerbated this process. (A) Schematic illustration of the AOM/DSS treatment. (B) Representative histological images of colon sections showing tissue morphology and H&E staining patterns across experimental groups. (C) The expression levels of HOXD13 and associated signaling axis components at the specified time points were assessed by immunoblotting. (D) Schematic of the intestinal-specific Hoxd13 knockout strategy. (E) Representative colon images with quantitative analysis of tumor burden, including number and volume, across the designated experimental groups (n=10 independent mice per group). (F) HOXD13, PILRA, and AREG protein levels were assessed in the indicated group. (G–H) The indicated tumor-infiltrating immune cells were analyzed by FCM (G) and by immunofluorescence (H) (n=5 independent tumors from five mice per group). (I) Schematic of the intestinal-specific Hoxd13 knock-in strategy. (J) Representative colon images with quantitative analysis of tumor burden, including number and volume, across the designated experimental groups (n=10 independent mice per group). (K) HOXD13, PILRA, and AREG protein levels were assessed in the indicated group. (L–M) The indicated tumor-infiltrating immune cells were analyzed by FCM (L) and by immunofluorescence (M) (n=5 independent tumors from five mice per group). AOM, aoxymethane; AREG, amphiregulin; CRC, colorectal cancer; DSS, dextran sulfate sodium; FCM, flow cytometry; HOXD13, homeobox D13; IEC, intestinal epithelial cell; PILRA, paired immunoglobin like type 2 receptor alpha.

The dual blockade of AREG and PD-L1 markedly inhibited HOXD13-mediated CRC progression and metastasis

These findings suggested that HOXD13-AREG signaling drives TAMs polarization, thereby promoting immune escape and tumor progression in HOXD13-overexpressing CRC. We postulated that anti-AREG therapy might act synergistically with anti-PD-L1 immunotherapy in HOXD13-overexpressing CRC. To evaluate this hypothesis, we generated an orthotopic mouse model using MC38-HOXD13 cells and administered the following treatments: anti-AREG monotherapy, anti-PD-L1 monotherapy, or a combination of anti-AREG and anti-PD-L1 antibodies. Both anti-AREG and anti-PD-L-1 monotherapy partially suppressed HOXD13-mediated liver metastasis and improved OS, whereas the combination treatment resulted in the most substantial suppression of metastasis and the greatest survival benefit (figure 6A–C). FCM and IF analyses further demonstrated that, relative to the other groups, the combination treatment significantly increased the infiltration, proliferation, and cytotoxic function of CD8⁺ T cells, accompanied by a substantial reduction in M2-type TAMs infiltration (figure 6D,E, online supplemental figure 17A). Consistently, in BALB/c mice bearing orthotopic CT26-HOXD13 tumors, the combination therapy induced a robust antitumor response, which was accompanied by enhanced infiltration, proliferation, and cytotoxic function of CD8⁺ T cells and a concurrent reduction in TAMs (online supplemental figure 16, 17B). To examine EGFR pathway activation and PD-L1 expression, we isolated TAMs (F4/80⁺ cells) from orthotopic tumors. We found anti-AREG treatment significantly reduced both EGFR activation and PD-L1 expression in TAMs (online supplemental figure 18A). CCK8 assays showed that the anti-AREG antibody had no effect on the proliferation of HOXD13 overexpressing CRC cells, indicating that its mechanism in inducing tumor regression is independent of cell proliferation (online supplemental figure 18B). Moreover, early changes in plasma cytokines resulting from treatment were assessed. Anti-PD-L1 combined with anti-AREG resulted in increased cytokines/chemokines associated with T cells stimulation (C-X-C motif chemokine ligand 9 (CXCL9) and IL-1α) and macrophage activation (macrophage colony stimulating factor (M-CSF), macrophage inflammatory protein (MIP)-1α, and MIP-2) (online supplemental figure 18C). The results indicated that the combination synergistically remodeled the TME by shifting it from an immunosuppressive to an immunostimulatory state.

Figure 6. The dual inhibition of AREG and PD-L1 markedly inhibited HOXD13-mediated CRC progression and metastasis. (A–C) C57BL/6 mice bearing orthotopic MC38-HOXD13 tumors were administered with anti-AREG monotherapy, anti-PD-L1 monotherapy, or a combination of anti-AREG and anti-PD-L1 antibodies (n=10 independent mice per group). Representative in vivo BLI obtained at the experimental endpoint, along with the dynamic changes in bioluminescent intensity (A), OS (B), and representative H&E-stained liver tissue sections with corresponding quantification of hepatic metastases (C) were shown for each group. (D–E) The indicated tumor-infiltrating immune cells were analyzed by immunofluorescence (D) and by FCM (E) (n=5 independent tumors from five mice per group). (F) Overview of the experimental design and timeline. (G) Representative colon images with quantitative analysis of tumor burden, including number and volume, across the designated experimental groups (n=10 independent mice per group). (H–I) The indicated tumor-infiltrating immune cells were analyzed by immunofluorescence (H) and by FCM (I) (n=5 independent tumors from five mice per group). AREG, amphiregulin; BLI, bioluminescence image; CRC, colorectal cancer; FCM, flow cytometry; HOXD13, homeobox D13; OS, overall survival; PD-L1, programmed cell death ligand 1.

We next evaluated the efficacy of the combination therapy in a model of AOM/DSS-induced CAC (figure 6F). Mice receiving the combination therapy exhibited a significant reduction in both colon tumor number and volume (figure 6G). FCM and IF analyses further demonstrated that the combination treatment group exhibited a higher density of tumor-infiltrating cytotoxic CD8⁺ T cells, concomitant with a sharp decline in M2-type TAMs accumulation (figure 6H,I). Overall, these findings demonstrated the synergistic effect of anti-AREG and anti-PD-L1 combination therapy in counteracting HOXD13-mediated CRC progression, as consistently evidenced in both immunocompetent orthotopic and AOM/DSS-induced CAC models.

Discussion

Among the diverse cellular components of the TME, TAMs hold significant importance and frequently represent the most abundant immune cell population in solid tumors. The polarization of TAMs towards an immunostimulatory or immunosuppressive fate is critically influenced by signals within the surrounding immune milieu. In mature solid tumors, TAMs predominantly undergo polarization toward the M2 phenotype.³² These cells exhibit immunosuppressive properties through the expression of immune checkpoint ligands and secretion of immunosuppressive cytokines. The sustained crosstalk between cancer cells and TAMs is central to orchestrating tumor progression within the TME.³³ In bladder cancer, tumor-derived CCL2 recruits TAMs and endows them with M2-like functional properties. These M2-polarized TAMs subsequently secrete factors including IL-10 and EGF, which further facilitate tumor progression and malignant development.³⁴ Therefore, deciphering intercellular communication within the TME of CRC is instrumental in developing new therapeutic strategies.

Our study identified HOXD13 in cancer cells as a pivotal mediator of TAMs-induced immunosuppression, underscoring its key role in orchestrating the immunosuppressive microenvironment within CRC. Ectopic expression of HOXD13 was observed in CRC, notably in highly metastatic cell lines and in metastatic cases. Moreover, elevated HOXD13 protein expression was positively correlated with the abundance of M2 macrophages and negatively correlated with CD8⁺ T cells infiltration within CRC tissues. Mechanistically, HOXD13 transcriptionally activated downstream target genes AREG and PILRA, thereby suppressing antitumor immunity and promoting CRC progression. Clinically, patients with high HOXD13 expression exhibited poor prognosis, and HOXD13 could serve as an independent risk factor for unfavorable outcomes in patients with CRC. Survival analysis also revealed that patients with CRC with dual positivity for HOXD13/AREG and HOXD13/PILRA had the worst prognosis among all patient groups.

PILRA belongs to the paired immunoglobin like type 2 receptor family and is expressed as a monomeric transmembrane protein. PILRA contains two immunoreceptor tyrosine-based inhibitory motifs, which mediate inhibitory signaling through the recruitment of SHP-1 and SHP-2.³⁵ PILRA, which is predominantly expressed on cancer cells and myeloid immune cells—including monocytes, dendritic cells, and macrophages—is consistently associated with the suppression of immune responses. PILRA knockout mice demonstrate heightened production of pathogenic cytokines during inflammatory responses and exhibit increased susceptibility to autoimmune arthritis.³⁶ Nevertheless, the impact of PILRA on CRC progression and its underlying mechanisms has yet to be elucidated. Our study demonstrated that HOXD13 transcriptionally activated PILRA expression in CRC cells. Patients with CRC with high PILRA expression were associated with a poor prognosis and a more aggressive tumor phenotype. Knockdown of PILRA in CRC cells marked suppressed tumor progression and liver metastasis. Furthermore, PILRA was found to inhibit both the proliferation and effector functions of CD8⁺ T cells. A recent study, whose findings are consistent with and may provide a mechanistic explanation for our own work, revealed that tumor cells expressed PILRA that binds to CD99 on T cells and suppresses T cell-mediated antitumor immunity via the ZAP70/NFAT/IL-2/JAK/STAT signaling pathway.²⁷

AREG, a ligand of the EGFR, activates key intracellular signaling pathways through its interaction with the receptor.³⁷ Beyond its essential functions in the development and maturation of mammary glands, bone tissue, and oocytes, AREG is also critically involved in multiple oncogenic processes. These effects include modulating tumor-intrinsic phenotypes—such as promoting cancer cell proliferation, angiogenesis, and invasion—as well as fostering an immunosuppressive TME. AREG promotes PD-L1 expression on cancer cells, fostering an immunosuppressive TME that enables immune evasion.³⁸ Furthermore, in BRCA1-related breast cancer, tumor-derived AREG drives tumor growth by promoting proangiogenic programming in TAMs.³⁹ A recent study has demonstrated that AREG expression is upregulated in tumor tissues following radiotherapy. This elevation in AREG promotes distant metastatic growth through two key mechanisms: inducing an immunosuppressive state in EGFR⁺ mononuclear phagocytes and enhancing tumor cell resistance to phagocytosis via upregulation of the CD47 signaling pathway.⁴⁰ Evidence from previous studies indicates that AREG expression is elevated in CRC, with particularly prominent upregulation observed in CRC-derived liver metastases.⁴¹ Nevertheless, the precise mechanism underlying its overexpression and its functional contributions to CRC pathogenesis remains poorly defined. Our study demonstrated that HOXD13 transactivated AREG expression in CRC cells and that cancer cells derived AREG promoted M2 polarization of TAMs via the EGFR/PI3K/AKT signaling pathway. Moreover, targeting AREG potentiated antitumor immunity in the TME, bolstered the response to anti-PD-L1 therapy, and curtailed HOXD13-mediated CRC progression and metastasis.

The TGF-β signaling pathway is a central regulator of a broad spectrum of biological processes, spanning normal cellular homeostasis to pathological states such as fibrosis and carcinogenesis. TGF-β is produced by and acts on a wide variety of cells, including both parenchymal elements (eg, epithelial and tumor cells) and stromal components (eg, immune cells and fibroblasts), thereby forming a complex and dynamic signaling network.⁴² This network evolves over time and plays a context-dependent role in disease, underpinning the dual role of TGF-β as both an inhibitory factor and a promotive force in tumor progression. In the progression of human CRC, TGF-β1 shifts from acting as a suppressor of tumor growth to facilitating cancer cells proliferation and invasion. Metastatic colon carcinoma cells exhibit proliferative responses to TGF-β, whereas moderately to well-differentiated primary colon carcinomas are typically growth-inhibited by TGF-β.⁴³ Clinically, patients exhibiting elevated TGF-β expression in primary CRC demonstrate a significantly increased risk of disease recurrence compared with those with low TGF-β levels.⁴⁴ TGF-β upregulates PD-1 expression via Smad3-dependent transcriptional activation, leading to suppression of T cells function. Furthermore, TGF-β inhibits the infiltration of CD8⁺ T cells into tumors by downregulating the chemokine receptor CXCR3 in a Smad2/ALK5-dependent manner.^{45 46} This study revealed that TGF-β1 upregulated HOXD13 expression in CRC cells through Smad2/3 signaling, which in turn promoted M2 polarization of TAMs via the AREG/EGFR/PI3K/AKT axis. Meanwhile, TGF-β1 derived from M2-type macrophages formed a positive feedback loop with CRC cells, continuously driving tumor progression. Moreover, disrupting HOXD13 expression in CRC cells effectively suppressed TGF-β1-induced cancer progression. Collectively, these data established a pivotal role for the HOXD13-dependent feedback loop in mediating the tumor-promotive effects of TGF-β1.

Taken together, our results demonstrated that HOXD13-mediated crosstalk between CRC cells and TAMs suppressed antitumor immunity and promoted CRC progression. TGF-β1 induced HOXD13 overexpression in CRC cells through the Smad2/3 complex. HOXD13 then transcriptionally upregulated the expression of AREG and PILRA in CRC cells, leading to M2 polarization of TAMs and inhibition of CD8⁺ T cells mediated antitumor responses. Notably, TGF-β1 secreted by infiltrated M2-type macrophages further enhanced HOXD13 expression in CRC cells, establishing a positive feedback loop between cancer cells and TAMs. Concurrent blockade of AREG and PD-L1 effectively disrupted this interaction, restoring antitumor immunity and suppressing tumor progression.

Data availability statement

Data are available upon reasonable request.