DUSP22 dephosphorylates LGALS1 to enhance T cell-driven antitumor immunity

Abstract

Background

Insufficient infiltration of CD8⁺ T cells in the tumor microenvironment (TME) critically restricts antitumor immunity and cancer immunotherapy efficacy. The purpose of this study was to identify novel tumor cell-intrinsic regulators of T-cell infiltration and to elucidate their mechanisms of action.

Methods

We performed a genome-wide Sleeping Beauty transposon mutagenesis screen in murine breast cancer models. Protein–protein interactions were identified by mass spectrometry and validated by co-immunoprecipitation. Gene and protein expression levels were assessed by reverse transcription and quantitative PCR and western blotting. T-cell infiltration and function were evaluated using flow cytometry, immunohistochemistry (IHC), multiplex IHC, and by analyzing bulk and single-cell RNA sequencing data complemented by bioinformatic analysis. The specific dephosphorylation sites on LGALS1 were confirmed through phosphomimetic mutant experiments. T-cell infiltration was further validated using an in vitro T-cell transendothelial migration assay and in vivo mouse models.

Results

Our screening identified 39 candidate genes, with tumor cell-intrinsic dual-specificity phosphatase 22 (DUSP22) expression correlating with enhanced CD8⁺ T-cell accumulation and suppressed tumor progression. Overexpression of DUSP22 resulted in increased CD8⁺ T-cell infiltration and enhanced T-cell function. Mechanistically, DUSP22 binds to LGALS1 and dephosphorylates it at the Ser8 and Thr58 residues, leading to LGALS1 degradation and subsequent alleviation of LGALS1-mediated immunosuppression. In human breast cancer samples, LGALS1 expression was negatively correlated with both DUSP22 levels and CD8⁺ T-cell infiltration. Therapeutic targeting of the DUSP22-LGALS1 axis significantly enhanced CD8⁺ T-cell infiltration and synergized with anti-programmed cell death protein-1 therapy to boost antitumor responses.

Conclusions

Our findings unveil a novel phosphorylation-dependent DUSP22-LGALS1 axis that reprograms the immunosuppressive TME. This work thus proposes a promising therapeutic strategy to overcome immune checkpoint blockade resistance in breast cancer.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Insufficient infiltration of CD8⁺ T cells into the tumor microenvironment (TME) is a major limitation for effective antitumor immunity and is a key reason for the poor efficacy of current cancer immunotherapies. While the immunosuppressive nature of the TME is recognized, the specific tumor cell-intrinsic genetic regulators that control T-cell infiltration were not fully understood.

WHAT THIS STUDY ADDS

• This study identifies the dual-specificity phosphatase 22 (DUSP22)-LGALS1 axis as a novel tumor cell-intrinsic pathway that governs T-cell influx. We mechanistically demonstrate that DUSP22 binds to and dephosphorylates LGALS1, leading to its degradation and the subsequent alleviation of immunosuppression. Furthermore, we provide translational evidence by showing the clinical relevance of this axis in human breast cancer and, most importantly, demonstrate that its therapeutic targeting potently enhances CD8⁺ T-cell infiltration and synergizes with anti-programmed cell death protein-1 therapy to bolster antitumor responses.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• This discovery redefines our understanding of the phosphorylatory control of the immunosuppressive TME and proposes a promising new therapeutic strategy. It is poised to direct future research towards exploring this pathway and similar mechanisms, with the potential to influence clinical practice by providing a novel combination strategy to overcome immune checkpoint inhibitor resistance in patients with breast cancer.

Background

The immunosuppressive tumor microenvironment (TME) remains a major challenge in treating breast cancer, particularly aggressive subtypes like triple-negative breast cancer (TNBC).¹ CD8⁺ T cells, pivotal mediators of antitumor immunity and cancer immunotherapy efficacy, are strongly associated with favorable patient prognosis.2,5 However, TME of breast cancer frequently impedes T-cell recruitment and infiltration through secretion of immunosuppressive factors,^{6 7} recruitment of regulatory immune cells,^{8 9} and upregulation of immune checkpoint molecules like programmed cell death 1 ligand 1 (PD-L1).^{10 11} These barriers underscore the urgent need to identify molecular targets that enable reprogramming the TME to facilitate T-cell infiltration and sustained antitumor activity.

Dual-specificity phosphatase 22 (DUSP22, also known as JKAP or JSP-1), a member of the dual-specificity phosphatase (DUSP) family, is recognized as a tumor suppressor frequently downregulated in malignancies. Its canonical function involves dephosphorylating and modulating MAP kinases (eg, JNK, p38).¹² Emerging evidence reveals that DUSP22 exhibits multifaceted regulatory roles across diverse signaling networks. A notable example is observed in lung adenocarcinoma, where DUSP22 attenuates malignant progression by downregulating the epidermal growth factor receptor/c-MET axis and its downstream effectors, thereby suppressing tumor proliferation and metastatic potential.¹³ In hepatocellular carcinoma associated with non-alcoholic steatohepatitis, DUSP22 physically associates with focal adhesion kinase to impede its phosphorylation at Tyr397 and catalytic phosphorylation at Tyr576/577, which subsequently mitigates hyperactivation of the ERK1/2-MAPK cascade and NF-κB-driven inflammatory signaling.¹⁴ Furthermore, recent investigations highlight DUSP22’s capacity to destabilize non-kinase substrates via phosphatase activity-dependent mechanisms. For instance, DUSP22 destabilizes ubiquitin protein ligase E3 component n-recognin 2 in T cells through dephosphorylating its Ser1694 and Tyr1697, promoting ubiquitin-mediated degradation.¹⁵ Despite these advances in understanding DUSP22’s tumor-intrinsic functions, its role in sculpting the tumor-immune microenvironment remains largely unknown. Specifically, whether and how DUSP22 influences immune cell recruitment and function in breast cancer, particularly in immunologically cold subtypes like TNBC, constitutes a critical knowledge gap. This gap is especially intriguing given DUSP22’s established roles in T-cell signaling and its potential to regulate key immunomodulatory pathways.

Galectin 1 (Gal1, encoded by LGALS1), a secreted β-galactoside-binding lectin, serves as an immunosuppressive factor in TME by inducing apoptosis of activated T cell.¹⁶ However, its pathophysiological relevance has been debated due to discrepancies between apoptotic concentrations and physiological TME levels.^{17 18} Recent evidence highlights LGALS1’s role in reinforcing the endothelial barrier^{19 20} through upregulating PD-L1 on vascular endothelial cells, physically restricting CD8⁺ T-cell infiltration.^{21 22} Clinically, elevated LGALS1 expression is significantly associated with reduced overall survival in breast cancer23,25 and multiple solid tumors,425,27 positioning it as a compelling therapeutic target to disrupt TME-mediated immune evasion.

While the immunosuppressive role of LGALS1 in hindering T-cell infiltration is becoming clear, the upstream regulatory mechanisms controlling its expression and stability in the TME are poorly understood. Identifying such regulators is crucial for developing novel strategies to disrupt the LGALS1-mediated immune barrier. Intriguingly, given DUSP22’s emerging role as a destabilizer of protein substrates through both phosphatase-dependent and independent mechanisms, we hypothesized that DUSP22 might serve as a previously unrecognized negative regulator of LGALS1 in breast cancer. We postulated that loss of DUSP22 function, as frequently occurs in malignancies, could lead to LGALS1 stabilization, thereby contributing to the exclusion of CD8⁺ T cells and fostering an immunosuppressive TME.

To uncover novel regulators of antitumor immunity, we employed an unbiased in vivo Sleeping Beauty (SB) transposon mutagenesis screen.^{28 29} This approach led us to identify DUSP22 as a potent novel modulator of tumor-infiltrating lymphocytes. We demonstrate that DUSP22, frequently downregulated in breast cancer, acts as a critical immune-sensitizing agent by binding to, dephosphorylating, and promoting the degradation of the key immunosuppressive factor LGALS1. This novel DUSP22-LGALS1 axis represents a central mechanism controlling CD8⁺ T-cell infiltration into breast tumors. Furthermore, based on this mechanistic insight, we devised a synergistic therapeutic strategy combining programmed cell death protein-1 (PD-1) blockade with LGALS1 neutralization, which effectively reversed T-cell exclusion and unleashed potent antitumor immunity. Our work not only elucidates a previously unknown regulatory pathway in breast cancer immunity but also provides a compelling rationale for targeting the LGALS1 pathway to overcome resistance to current immunotherapies.

Results

Sleeping Beauty transposon mutagenesis identifies DUSP22 as a key regulator of tumor-infiltrating lymphocytes

The SB DNA transposon system facilitates genome-wide functional genomic screening by randomly cutting and pasting into different genome loci to induce gain-of-function or loss-of-function alterations, thereby tagging candidate cancer driver genes.³⁰ To identify genes governing T-cell infiltration, we performed immunohistochemical (IHC) staining with pan-T cell marker CD3 to stratify tumors into low and high T cell-enriched subgroups. Further correlation analysis with driver genes identified through SB transposon sequence revealed that the driver genes potently govern T-cell infiltration (figure 1A,B). Mice bearing high T-cell enriched tumors demonstrated significantly prolonged survival compared with those with low T-cell infiltration (online supplemental figure 1A). Consistent with clinical observations, CD3⁺ T lymphocyte abundance was associated with improved patient outcomes.³¹

Figure 1. Sleeping Beauty transposon mutagenesis identifies DUSP22 as a key regulator of tumor-infiltrating lymphocytes. (A) Schematic diagram of the screening process for genes associated with T-cell infiltration in SB mouse tumors. (B) Representative figures of IHC staining with an anti-CD3 antibody in SB mice tumors (top) and their quantitative results (bottom). Cell numbers of at least five figures were randomly picked in each sample for counting, and average cell counts were plotted. Scale bars, 100 µm. (C) The oncoplot illustrates insertion mutations in candidate genes that regulate T-cell infiltration. (D) Scatter plot illustrates the Spearman correlation between the fractions of CD3/CD8 T cells and the expression levels of candidate genes across the indicated dataset. The fractions of CD3, CD4, and CD8 T cells were computationally deconvoluted using the CIBERSORT algorithm in the TCGA-BRCA dataset, with total CD3 T-cell fractions defined as the combined fractions of CD4 and CD8 T-cell fractions. (E) Representative images and results from multiplex IHC staining of patients with breast cancer show that DUSP22 expression is positively correlated with both CD3⁺ and CD8⁺ T-cell proportions. Scale bar, 50 µm. Data are represented as mean±SEM. Statistics were analyzed by two-tailed Student’s t-test in B. BRCA, breast cancer; DUSP22, dual-specificity phosphatase 22; IHC, immunohistochemical; SB, Sleeping Beauty; TCGA, the Cancer Genome Atlas.

Comparative genomic analysis revealed 356 candidate genes specific to low T-cell enriched tumors and 678 genes associated with high T-cell infiltration (figure 1C; online supplemental figure 1B; online supplemental table S2). Functional enrichment analysis highlighted distinct biological signatures: low T cell-enriched genes were enriched in adherens junction, aldosterone signaling, and glycosaminoglycan biosynthesis pathways (online supplemental figure 1C), whereas high T-cell subgroup genes implicated focal adhesion, axon guidance, and glutamatergic synapse regulation (online supplemental figure 1D). By cross-comparing the insertion frequencies in low and high T cell-enriched tumors, we identified 39 genes that are strongly associated with T-cell infiltration in tumors. Notably, previously reported T-cell regulators including DOCK2³² and CYLD³³ were among the candidates (figure 1C,D; online supplemental table S3), validating our screening strategy. Gene Ontology analysis further unscored the enrichment of candidates related to leukocyte cell–cell adhesion, T-cell activation, and type II immune responses (online supplemental figure 1E), reinforcing the role of these genes in shaping a T cell-permissive TME.

To prioritize clinically relevant candidates, we integrated our findings with a human breast cancer database. First, deconvolution of the Cancer Genome Atlas - Breast Cancer (TCGA-BRCA) RNA sequencing (RNA-seq) data using CIBERSORT³⁴ (online supplemental figure 2A) revealed a striking positive correlation between DUSP22 expression and both CD3⁺ and CD8⁺ T-cell infiltration (figure 1D; online supplemental figure 2B,C). Subsequently, clinical validation using multiplex IHC (mIHC) staining in human breast tumors confirmed that high DUSP22 protein levels are associated with elevated densities of both CD3⁺ and CD8⁺ T cells (figure 1E). Consistently, mouse SB tumors harbored Dusp22 insertions that displayed enhanced CD3⁺ T-cell infiltration compared with other tumors (online supplemental figure 2D), with transposon insertion patterns suggesting Dusp22 activation during tumorigenesis (online supplemental figure 2E). Collectively, these multimodal analyses establish DUSP22 as a tumor intrinsic driver of T-cell infiltration in breast cancer.

DUSP22 suppresses tumor growth in a T cell-dependent manner

Given the clinical association between DUSP22 downregulation in tumors (online supplemental figure 3A) and its putative role in T-cell infiltration, we overexpressed (OE) Dusp22 in murine breast cancer cell line 4T1 and EMT6, which was validated by reverse transcription and quantitative PCR (online supplemental figure 3B) and western blotting (figure 2A,B). To profile its tumor-intrinsic effects, we first assessed cell proliferation using a 96-hour AlamarBlue assay. The results showed that Dusp22 OE did not confer a growth advantage (online supplemental figure 3C). Subsequently, orthotopic implantation of Dusp22 OE cells into immunodeficient hosts (nude and non-obese diabetic/severe combined immunodeficiency (NOD/SCID) mice) showed no difference in tumor growth compared with vector controls (figure 2C,D; online supplemental figure 3D,E). However, Dusp22 OE tumors in immunocompetent mice exhibited profound growth suppression in both 4T1 (figure 2E) and EMT6 models (figure 2F). Survival ratio analysis also demonstrated significant benefits in Dusp22 OE groups with immunocompetent hosts (figure 2G,H). We next tested the immunomodulatory potency of DUSP22 under a moderate tumor burden by inoculating 50,000 cells per mouse. DUSP22 overexpression significantly delayed tumor progression in the 4T1 model (online supplemental figure 3F-H) and, strikingly, induced regression in three out of five EMT6 tumors (online supplemental figure 3I,K), demonstrating a consistent antitumor effect at different conditions. These data unequivocally demonstrate that DUSP22-mediated tumor suppression is contingent on T cells.

Figure 2. DUSP22 suppresses tumor growth in a T cell-dependent manner. (A and B) Western blotting of DUSP22 expression in 4T1 (A) and EMT6 (B) Dusp22 OE and vector cell lines. (C and D) Tumor growth curve (left) and weight (right) of orthotopic tumors derived from 4T1 (C) or EMT6 (D) cells in immunodeficient mice (n=4–6 mice per group). (E and F) Tumor growth curve and weight of orthotopic tumors derived from 4T1 (E) or EMT6 (F) cells in immunocompetent mice (n=4–6 mice per group). (G and H) Survival curve of mice bearing 4T1 (G) or EMT6 (H) Dusp22 OE and vector tumor (n=10 per group for 4T1, n=15 per group for EMT6). (I) Schematic representation of treatment of 4T1 Dusp22 OE or vector tumor-bearing mice with IgG or anti-CD3 antibodies on the indicated days (n=5 per group). Mice were administered anti-CD3 or IgG (10 µg/dose) intraperitoneally on the indicated days. (J and K) Tumor growth curve (J) and tumor weight (K) in mice bearing Dusp22 OE or vector tumors treated with IgG or anti-CD3 antibodies. All data are represented as mean±SEM. Statistics were analyzed by two-way ANOVA in C (left), D (left), E (left), and F (left), by two-way ANOVA with multiple comparisons in J, by two-tailed Student’s t-test in C (right), D (right), E (right), F (right) and K or by log-rank test in G and H. ANOVA, analysis of variance; DUSP22, dual-specificity phosphatase 22; ns, not significant; OE, overexpressed.

To directly test the dependency on T cells, we employed an in vivo depletion strategy (figure 2I). Specifically, 4T1 Dusp22 OE and vector control cells were orthotopically implanted into the mammary fat pads of immunocompetent BALB/c mice. Subsequently, T-cell depletion was achieved through the administration of anti-CD3 antibodies. Flow cytometric analysis confirmed effective T-cell elimination in peripheral blood (online supplemental figure 3L). Notably, T-cell ablation fully rescued tumor growth in Dusp22 OE bearing mice, with tumor volume and weight demonstrating comparable values to those of vector controls (figure 2J,K). This functional rescue conclusively establishes that DUSP22 exerts its antitumor effects exclusively through T cell-mediated mechanisms.

DUSP22 drives cytotoxic CD8⁺ T-cell infiltration into tumors

To investigate whether Dusp22 OE-mediated antitumor effects stem from enhanced T-cell infiltration, we first quantified tumor-infiltrating lymphocytes via flow cytometry (online supplemental figure 4A). Dusp22 overexpression significantly increased the proportions of CD3⁺ and CD8⁺ T cells in both 4T1 and EMT6 tumors compared with controls, while CD4⁺ T-cell infiltration remained unchanged (figure 3A,B; online supplemental figure 4B,C,F,G; online supplemental figure 5A). IHC validation further confirmed these findings, demonstrating robust enhanced CD8⁺ T-cell infiltration in Dusp22 OE tumors (figure 3C,D; online supplemental figure 4D,E,H,I).

Figure 3. DUSP22 drives cytotoxic CD8⁺ T-cell infiltration in tumors. (A and B) Flow cytometry analysis showing the percentage of CD3⁺ and CD8⁺ T cells in 4T1 (A) and EMT6 (B) Dusp22 OE and vector tumors. (C and D) Representative images (left) and results (right) of IHC staining for CD8 in 4T1 (C) and EMT6 (D) Dusp22 OE and vector tumors. Cell numbers of at least five figures were randomly picked in each sample for counting, and average cell counts were plotted. Scale bar, 100 µm. (E) UMAP plot of 4T1 Dusp22 OE and vector tumor cells. (F) GSEA enrichment plots of intratumoral CD8⁺ T cells from 4T1 Dusp22 OE versus vector tumors. (G) UMAP plot shows secondary clusters of intratumoral CD8⁺ T cells in 4T1 tumors. (H) Bar plot shows the percentage of different CD8⁺ T-cell subclusters from 4T1 Dusp22 OE and vector tumors. (I and J) Flow cytometry analysis showing the percentage of IFNγ⁺, GZMB⁺ and Ki67⁺ cells among CD8⁺ T cell in 4T1 (I) and EMT6 (J) Dusp22 OE and vector tumors. (K) Schematic diagram of T-cell transendothelial migration assay. CD8⁺ T cells isolated from BALB/c mouse spleens or peripheral blood mononuclear cells were activated using plate-bound CD3/CD28 antibodies (5 µg/mL each) for 48 hours, then co-cultured with Dusp22 OE or vector cancer cells. (L) Quantification of T-cell passing through the membrane of Transwell in 4T1 (left) and EMT6 (right) Dusp22 OE versus vector cells. (M) Validation of DUSP22 overexpression by western blot in the indicated human TNBC cell lines. α-tubulin was used as a control. (N) CD8⁺ T-cell migration from the indicated TNBC cells was assessed using a transwell system. DUSP22 overexpression significantly promoted T-cell migration. All data are represented as mean±SEM. Statistics were analyzed by two-tailed Student’s t-test. DC, dendritic cell; DUSP22, dual-specificity phosphatase 22; GSEA, gene set enrichment analysis; GZMB, granzyme B; IHC, immunohistochemical; IFN, interferon; NK, natural killer cell; NKT, nature killer T cell; OE, overexpressed; T_cm, central memory T cells; T_eff, effector T cells; T_em, effector memory T cells; T_ex, exhausted T cells; Treg, regulatory T cell; TNBC, triple-negative breast cancer; UMAP, Uniform Manifold Approximation and Projection.

We further performed bulk RNA-seq on Dusp22 OE vs vector tumors and identified 3,810 significantly differentially expressed genes (DEGs). Notably, cytotoxic effector molecules—including Gzma, Gzmb, Gzmk, and Ifng—were markedly upregulated in Dusp22 OE tumors (online supplemental figure 5B). Functional enrichment analysis revealed strong association of DEGs with immune response pathways, particularly lymphocyte activation, T-cell proliferation, differentiation, and migration (online supplemental figure 5C,D). Consistently, using various deconvolution methods,35,37 it was speculated that the infiltration fraction of CD3⁺ and CD8⁺ T cells in tumors of Dusp22 OE mice was significantly higher than that in the control group (online supplemental figure 5E). These data collectively demonstrate that Dusp22 overexpression remodels the tumor immune microenvironment by enhancing cytotoxic T-cell infiltration and killing effect, thereby enabling T cell-mediated tumoricidal activity to restrict cancer progression.

To delineate how cancer cell-intrinsic DUSP22 regulates the TME, particularly in regulating T-cell infiltration, we conducted single-cell RNA sequencing (scRNA-seq) on 4T1 tumors derived from Dusp22 OE and vector cells. The analysis captured 14,046 single cells, which were resolved into 10 major cellular populations based on canonical cell type-specific markers, including epithelial cells, fibroblasts, CD8 T cells, CD4 T cells, and myeloid cells (figure 3E; online supplemental figure 6A-C). Consistently, transcriptional analysis in CD8⁺ T cells revealed enhanced effector programs, including lymphocyte migration, lymphocyte-mediated immunity, inflammatory response, and interferon-gamma (IFN-γ) signaling in Dusp22 OE tumors (figure 3F; online supplemental figure 6D). Subclustering delineated five CD8⁺ T-cell subsets: naïve T cells, CD8 central memory T (T_cm) cells, CD8 effector memory T (T_em) cells, CD8 effector T (T_eff) cells, and CD8 exhausted T cells (figure 3G; online supplemental figure 6E-G). Notably, Dusp22 OE tumors exhibited expanded proportions of effector populations, including CD8 T_eff, CD8 T_cm and CD8 T_em cells, and reduced exhaustion T-cell populations (figure 3H), suggesting that DUSP22 enhances both CD8⁺ T-cell infiltration and functional activation. Flow cytometry further validated those findings with demonstration of increased IFNγ and granzyme B (GZMB) expression in CD8⁺ T cells after Dusp22 overexpression (figure 3I,J), with no significant changes observed in CD4⁺ T cells (online supplemental figure 5F,G). Ki67 staining confirmed that increased T-cell populations were driven by recruitment, not proliferation (figure 3I,J; online supplemental figure 5F,G). Given DUSP22’s involvement in the JAK/STAT3 pathway, we assessed its overexpression in tumor cells. Consistent with its role as a phosphatase and potential tumor suppressor, we observed that DUSP22 expression led to a cell-intrinsic reduction in phosphorylation of STAT3 at Y705 (online supplemental figure 5H). However, transcriptomic analyses revealed no broad JAK/STAT pathway activation (online supplemental figure 5I,J).

To directly test DUSP22’s role in CD8⁺ T-cell migration, we developed an in vitro transendothelial migration assay (figure 3K). Co-culture with Dusp22 OE tumor cells significantly enhanced CD8⁺ T-cell migration compared with co-culture with control cells (figure 3L). Furthermore, overexpression of DUSP22 in human TNBC cell lines (MDA-MB-231 and HCC1143; figure 3M) significantly promoted CD8⁺ T-cell migration (figure 3N), demonstrating its conserved role in human models. Taken together, these data establish that cancer cell-intrinsic DUSP22 orchestrates CD8⁺ T cells infiltration, thereby suppressing tumorigenesis.

DUSP22 modulates T-cell infiltration by interacting with LGALS1

To unravel how tumor intrinsic DUSP22 controls T-cell recruitment, we employed tetracycline-inducible Dusp22 overexpression system in 4T1 cells (online supplemental figure 7A). Immunoprecipitation coupled with liquid chromatography-mass spectrometry (IP-LC/MS) identified 202 DUSP22-interacting proteins, among which LGALS1 was reported as a secreted protein that regulates T-cell functions,^{16 21 22} emerged as a top candidate (figure 4A; online supplemental figure 7B,C). To validate it, we conducted co-immunoprecipitation (Co-IP) assays with HA-tagged DUSP22 and Flag-tagged LGALS1, and the results confirm interaction between DUSP22 and LGALS1 (figure 4B).

Figure 4. DUSP22 interacts with LGALS1 and decreases its expression. (A) Strategy to identify DUSP22-interacting proteins. We first constructed tetracycline-inducible HA-tagged Dusp22 overexpressing 4T1 cells. These cells were treated with or without Dox (1 µg/mL) for 12 hours. The cell lysates were then co-incubated with anti-HA magnetic beads. Finally, the immunoprecipitated products were analyzed by LC-MS to identify interacting proteins. (B) Results by Co-IP assays in HEK293T cells transfected with Flag-tagged LGALS1 and HA-tagged DUSP22. Anti-Flag and anti-HA magnetic beads were used as immunoblotting probes. (C and D) Western blotting analysis of LGALS1 expression in 4T1 (C) and EMT6 (D) Dusp22 OE and vector cells. (E) Western blotting analysis of LGALS1 expression in 4T1 tetracycline-induced Dusp22 OE cells at the indicated time points. (F and G) Western blotting analysis of DUSP22 and LGALS1 expression in 4T1 (F) and EMT6 (G) Dusp22 OE and vector tumors. (H and I) ELISA analysis of the levels of LGALS1 protein secreted from 4T1 tetracycline-induced Dusp22 OE or vector cells (H) and EMT6 Dusp22 OE and vector cells (I). All data are represented as mean±SEM. (J) Quantification of secreted LGALS1 in conditioned media from control and DUSP22-overexpressing TNBC cells by ELISA. (K) Systemic levels of LGALS1 were quantified by ELISA from the serum of mice bearing 4T1 or EMT6 tumors expressing either Dusp22 OE or vector control. Serum samples were analyzed at a 1:50 dilution. All data are represented as mean±SEM. Statistics were analyzed by two-tailed Student’s t-test. Co-IP, co-immunoprecipitation; Dox, doxycycline; DUSP22, dual-specificity phosphatase 22; LC-MS, liquid chromatography-mass spectrometry; OE, overexpressed; TNBC, triple-negative breast cancer.

Concomitantly, we observed that Dusp22 OE cells demonstrated lower expression level of LGALS1 protein (figure 4C–E), as well as in tumor tissues (figure 4F,G). Given LGALS1’s role as a secreted immunomodulator, we further quantified its extracellular secretion levels by ELISA. DUSP22 overexpression consistently downregulated LGALS1 secretion in both murine (4T1, EMT6; figure 4H,I) and human (MAD-MB-231, HCC1143; figure 4J) TNBC cell lines, suggesting a negative correlation between DUSP22 and LGALS1. To determine whether DUSP22 also influences systemic LGALS1 levels, we measured serum LGALS1 using ELISA in 4T1 and EMT6 tumor-bearing mice. Consistent with the TME data, DUSP22-OE reduced serum LGALS1 levels in tumor-bearing mice (figure 4K). Consistently, TCGA-BRCA dataset analysis showed an inverse correlation between LGALS1 expression and CD3⁺ T-cell infiltration (online supplemental figure 8A). Additionally, an examination of breast cancer proteomic dataset demonstrated that LGALS1 protein levels were negatively associated with CD8A levels (online supplemental figure 8B). To further validate these correlations, an mIHC analysis was conducted on 90 primary human breast cancer samples. This analysis focused on the expression patterns of LGALS1, DUSP22, pan-cytokeratin⁺ tumor epithelial cells, and CD8⁺ tumor-infiltrating T cells. Notably, the results showed high LGALS1 expression correlated with low DUSP22 protein level and low CD8⁺ T-cell densities (figure 5A–C), suggesting LGALS1 may negatively regulate T-cell infiltration in tumor.

Figure 5. LGALS1 blunts the immunomodulatory effects of DUSP22. (A) Representative images of multiplex IHC staining in human breast cancer tissues (n=90). Scale bar, 50 µm. (B and C) The results of multiplex IHC staining in A demonstrate that DUSP22 expression was negatively correlated with both LGALS1 expression and the proportion of CD8⁺ T cells. Scale bar, 50 µm. (D) Schematic illustration of CD8⁺ T-cell transendothelial migration. Naïve CD8⁺ T cells isolated from BALB/c mouse spleens were activated with plate-bound CD3/CD28 antibodies (5 µg/mL each) for 48 hours and then co-cultured with cancer cells treated with rGal1 protein (400 ng/mL) or a neutralizing anti-LGALS1 (anti-Gal1) antibody (300 ng/mL). (E) Migration efficacy of activated CD8⁺ T cells co-cultured with 4T1 Dusp22 OE or vector cells under indicated treatments. (F) Schematic representation of mice bearing 4T1 Dusp22 OE or vector tumors treated with PBS or OTX008 on the indicated days. Mice were administered OTX008 (0.1 mg/dose) or PBS intraperitoneally on the indicated days. (G and H) Tumor growth curve (G) and weight (H) of orthotopic tumors treated with PBS or OTX008 (n=5 mice per group). (I and J) Results of IHC staining for CD3 (I) and CD8 (J) in 4T1 Dusp22 OE and vector tumors treated with PBS or OTX008. Cell numbers of at least five figures were randomly picked in each sample for counting, and average cell counts were plotted. (K) Western blotting shows the protein expression of LGALS1 and DUSP22 in 4T1 cells stably transduced with indicated lentivirus. (L to N) Tumor growth curve (L), weight (M), and percentage of CD8⁺ T cell (N) of orthotopic tumors derived from established cells in K. All data are represented as mean±SEM. Statistics were analyzed by two-tailed Student’s t-test in E, H, I, J, M, and N, or by two-way ANOVA with multiple comparisons in G and L. ANOVA, analysis of variance; anti-Gal1, anti-LGALS1 antibody; DUSP22, dual-specificity phosphatase 22; IHC, immunohistochemical; PBS, phosphate-buffered saline; OE, overexpressed; rGal1, recombinant LGALS1 protein; TMA, tissue microarray.

To validate the functional impact of LGALS1 on CD8⁺ T-cell infiltration, we performed in vitro migration assays by co-culturing CD8⁺ T cells with recombinant LGALS1 protein (rGal1) (figure 5D). As results shown, rGal1 significantly attenuated the pro-migratory effect induced by Dusp22 OE. Notably, this inhibitory effect was reversed by LGALS1-neutralizing antibodies (anti-Gal1) (figure 5E; online supplemental figure 8C), demonstrating the critical role of LGALS1 in DUSP22-mediated T-cell migration modulating.

To further validate whether DUSP22 regulates T-cell infiltration in solid tumors by reducing LGALS1, we knocked out (KO) Lgals1 in 4T1 and EMT6 breast cancer cell lines (online supplemental figure 8D). As predicted, Lgals1 KO recapitulated the phenotypes of Dusp22 overexpression, namely, suppressed tumor growth (online supplemental figure 8E,F) and enhanced CD8⁺ T-cell infiltration in 4T1 tumors (online supplemental figure 8G). Similar phenotypes have been observed in the EMT6 model (online supplemental figure 8H–J). Consistent with DUSP22 overexpression effects, LGALS1 pharmacological inhibition via OTX008³⁸ (figure 5F) recapitulated tumor suppression in 4T1, EMT6, and AT-3 models (figure 5G,H; onlinesupplemental figures 8K,L and 9AD). This antitumor effect was associated with the expansion of CD3⁺ and CD8⁺ T-cell populations within the TME (figure 5I,J; onlinesupplemental figures 8M 9E,F), while splenic T-cell populations remained unaffected (online supplemental figure 9G,H). Together, these results indicate that DUSP22 orchestrates T-cell infiltration and antitumor immunity may be through LGALS1-dependent mechanisms.

To further determine whether DUSP22’s antitumor effects are achieved through LGALS1 regulation, we OE both Lgals1 and Dusp22 in 4T1 and EMT6 cells (figure 5K; online supplemental figure 10A) followed by orthotopic implantation. Strikingly, Lgals1 overexpression abolished the tumor-suppressive phenotype induced by Dusp22 (figure 5L,M; online supplemental figure 10B,C). Subsequent IHC analysis demonstrated that overexpression of Lgals1 significantly abolished Dusp22-stimulated CD8⁺ T-cell infiltration (figure 5N; online supplemental figure 10D), confirming that DUSP22-dependent tumor suppression strictly requires LGALS1 downregulation to enable cytotoxic T-cell recruitment.

DUSP22 induces LGALS1 degradation via site-specific dephosphorylation at Ser8/Thr58 residues

DUSP22, a dual-specificity phosphatase, has been reported to facilitate the degradation of downstream proteins via dephosphorylation. To validate whether DUSP22 destabilizes LGALS1, we first performed cycloheximide chase assays, and the results revealed that overexpression of Dusp22 significantly reduced the half-life of LGALS1 (online supplemental figure 11A). To further explore whether DUSP22 modulates LGALS1 degradation through phosphorylation, we pulled down LGALS1 and then investigated the phosphorylation level. The results show that both serine (Ser) and threonine (Thr) phosphorylation levels were reduced in the presence of DUSP22. However, the phosphatase-dead DUSP22 (C88S) mutant¹⁵ could not alter LGALS1 phosphorylation levels (figure 6A), suggesting that DUSP22 relies on its phosphatase activity to degrade LGALS1.

Figure 6. DUSP22 induces LGALS1 degradation via site-specific dephosphorylation at Ser8/Thr58 residues. (A) Western blot analysis was performed to detect Ser, Thr, and Tyr phosphorylation of immunoprecipitated Flag-LGALS1 in HEK293T cells transfected with the indicated plasmids. (B) The lollipop plot illustrates the distribution and number of LGALS1 phosphorylation sites in mice. (C) Western blotting was performed to assess the expression levels of Flag-tagged LGALS1 and GFP-tagged DUSP22 in HEK293T cells transfected with the indicated plasmids. (D) Western blotting analysis shows the protein expression of LGALS1 and DUSP22 in 4T1 cells transduced with the indicated lentivirus. (E) Migration efficacy of activated CD8⁺ T cells co-cultured with 4T1 cells transduced with the indicated lentivirus. (F to I) Tumor growth curve (F), weight (G), percentage of CD8⁺ T cell (H) and representative CD8 IHC staining images (I) of 4T1 orthotopic tumors derived from established cells in D. Cell numbers of at least five figures were randomly picked in each sample for counting, and average cell counts were plotted. Scale bar, 50 µm. All data are represented as mean±SEM. Statistics were analyzed by two-tailed Student’s t-test in E, G, and H, or by two-way ANOVA with multiple comparisons in F. ANOVA, analysis of variance; DUSP22, dual-specificity phosphatase 22; IHC, immunohistochemical; Ser, serine; Thr, threonine; Tyr, tyrosine.

To further identify specific phosphorylation site(s) targeted by DUSP22, all reported LGALS1 phosphorylation residues (S8, S26, S30, S39, T58, T63, T71, and Y120), based on PhosphoSitePlus database annotations³⁹ (figure 6B), were mutated to aspartic acid (D) residue to mimic its phosphorylated states. Co-transfection of Dusp22 with a variety of Lgals1 mutants demonstrated that DUSP22 degraded wild-type LGALS1 and several mutants, including S26D, S30D, T71D, and Y120D. However, the S8D, S39D, T58D, and T63D mutants retained high levels of LGALS1, suggesting that these residues are critical for DUSP22-mediated dephosphorylation and regulation (figure 6C, online supplemental figure 11B). To further identify the potent target for regulating T-cell infiltration, we transfected these four phosphomimetic mutants (Lgals1-S8D/S39D/T58D/T63D) into EMT6 Dusp22 OE cells to estimate the T-cell migration. The results demonstrated that Lgals1-S8D/T58D mutants significantly reduced the migration ability of CD8⁺ T cells (online supplemental figure 11C) after DUSP22 overexpression, indicating immune regulation function of DUSP22 may be through dephosphorylation of Lgals1-S8/T58 residues. Furthermore, we stably OE the Lgals1-WT/S8D/T58D mutant in the 4T1 Dusp22 OE cells to identify its functional mutation site(s) (figure 6D). T-cell migration assay showed Lgals1-S8D/T58D mutants significantly reduce the migration ability of CD8⁺ T cells (figure 6E). Further in vivo orthotopic tumor models showed that the S8D/T58D mutants restricted DUSP22-mediated tumor inhibition (figure 6F,G) by reducing CD8⁺ T-cell infiltration (figure 6H,I).

Collectively, DUSP22 directly dephosphorylates LGALS1 at the Ser8/Thr58 residues, thereby decreasing its stability, suppressing LGALS1 secretion, and unleashing CD8⁺ T cell-mediated antitumor immunity. This molecular mechanism likely underpins DUSP22-mediated remodeling of the TME, ultimately driving cytotoxic CD8⁺ T-cell infiltration and antitumor immunity.

Targeting LGALS1 facilitates T-cell infiltration and improves the efficacy of immunotherapy

Our findings establish LGALS1 as a tumor-intrinsic immunosuppressive checkpoint whose degradation promotes CD8⁺ T-cell infiltration, supporting the therapeutic relevance of LGALS1 inhibition. To validate its synergy with immune checkpoint blockade, we developed an orthotopic 4T1 mammary tumor model in BALB/c mice and tested the combination of LGALS1 targeting and anti-PD-1 therapy (figure 7A). As anticipated, 4T1 tumors—which typically display a T cell-excluded phenotype—demonstrated resistance to anti-PD-1 monotherapy. Strikingly, the combination of OTX008 (a selective LGALS1 inhibitor) with anti-PD-1 achieved superior tumor control compared with either agent alone, with significant reductions in tumor volume observed throughout the treatment course (figure 7B,C).

Figure 7. Targeting LGALS1 facilitates T-cell infiltration and improves the efficacy of immunotherapy. (A) Schematic diagram showing the treatment regimen for mice bearing 4T1 tumors. Mice were treated with OTX008 or PBS in combination with anti-PD-1 or isotype IgG on the indicated days (n=5 mice per group). Mice received intraperitoneal injections of OTX008 (5 mg/kg) and/or anti-PD-1/IgG (0.2 mg per dose) on the indicated days (n=5 mice per group). (B and C) Tumor growth curve (B) and weight (C) of 4T1 orthotopic tumors treated with OTX008 or PBS in combination with anti-PD-1 or isotype IgG. (D) Schematic diagram of the treatment regimen for mice bearing 4T1 tumors. Mice received intraperitoneal injections of anti-Gal1/anti-pGal1 (0.15 mg per dose) and/or anti-PD-1/IgG (0.2 mg per dose) on the indicated days, with both types of antibodies administered via the same route (n=5 mice per group). (E and F) Tumor growth volume (E) and weight (F) of 4T1 orthotopic tumors treated with anti-Gal1 or anti-pGal1 antibodies in combination with anti-PD-1 or isotype IgG. (G) Representative images (left) and results (right) of IHC staining for CD8, GZMB and for cleaved caspase-3 in 4T1 tumors. Cell numbers of at least five figures were randomly picked in each sample for counting, and average cell counts were plotted. Scale bar, 50 µm. All data are represented as mean±SEM. Statistics were analyzed by two-tailed Student’s t-test (C, E, F and G) and two-way ANOVA with multiple comparisons (B). ANOVA, analysis of variance; anti-Gal1, anti-LGALS1 antibody; GZMB, granzyme B; IHC, immunohistochemical; PBS, phosphate-buffered saline; PD-1, programmed cell death protein-1; pGal1, phosphorylated LGALS1.

To further validate LGALS1’s role as a therapeutic target, we employed antibody-based targeting strategies against both native and phosphorylated LGALS1 (anti-Gal1 and anti-pGal1, respectively) in the same model system (figure 7D). While anti-Gal1 monotherapy showed limited efficacy, its combination with anti-PD-1 resulted in synergistic tumor growth inhibition (figure 7E,F).

Mechanistic analyses revealed that combination therapy produced profound immunomodulatory effects, significantly increasing intratumoral CD8⁺ T-cell density and expanding GZMB-positive cytotoxic lymphocyte populations (figure 7G; online supplemental figure 12A). This enhanced antitumor immunity correlated with elevated tumor cell apoptosis, as evidenced by increased cleaved caspase-3 staining (figure 7G; online supplemental figure 12A). To evaluate treatment safety, we performed longitudinal monitoring of systemic toxicity markers. While combination therapy induced modest splenic atrophy (reduced spleen-to-body weight ratio), mice maintained stable body weights throughout the study, suggesting acceptable treatment tolerability (online supplemental figure 12B,C). The observed splenic changes may reflect either treatment-related immune activation or targeted effects on LGALS1-expressing stromal components, warranting further investigation.

Collectively, our findings establish LGALS1 as an important regulator of the immunosuppressive TME, whose therapeutic targeting disrupts the immune-evasive niche through complementary mechanisms. The synergistic cooperation between LGALS1 blockade (achieved via antibodies or pharmacological inhibitors) and anti-PD-1 therapy demonstrates remarkable capacity to reverse adaptive immune resistance in poorly infiltrated tumors. This work positions LGALS1 as an immune checkpoint with dual therapeutic value: (1) enhancing T-cell trafficking into immunologically “cold” tumors, and (2) amplifying cytotoxic effector functions within the tumor bed. The translational relevance is underscored by the compatibility of LGALS1-targeting strategies with existing checkpoint inhibitors, offering a clinically actionable approach to expand immunotherapy efficacy across multiple cancer types. Future investigations should prioritize biomarker development to identify LGALS1-driven immunosuppression and optimize combination regimens for clinical translation.

Discussion

Our study delineates a novel, phosphorylation-dependent mechanism of cancer immune evasion centered on the DUSP22-LGALS1 axis. We position DUSP22 not merely as a tumor suppressor but as a crucial tumor-intrinsic immune sensitizer, which exerts its function by targeting and destabilizing the immunosuppressive lectin LGALS1. This dephosphorylation-driven checkpoint operates upstream of established barriers, controlling the abundance of a multifunctional T-cell suppressor and thereby governing the fundamental capacity of CD8⁺ T cells to infiltrate tumors. Specifically, we show that DUSP22 mediates site-specific dephosphorylation of LGALS1 at Ser8/Thr58 residues, leading to its destabilization. This previously unrecognized post-translational regulation mechanistically links elevated LGALS1 levels—clinically associated with poor T-cell infiltration and immunotherapy resistance to tumor immune evasion.

Beyond this mechanistic insight, our findings significantly expand the functional repertoire of DUSP family proteins beyond their canonical roles in regulating MAPK signaling. While DUSPs are well-known for intracellular signal modulation, we establish DUSP22 as a pivotal mediator of tumor-immune crosstalk through its direct action on an extracellular immunosuppressive factor. This raises the intriguing possibility that other DUSP family members may similarly regulate key components of the TME, opening a new frontier in understanding phosphatases as arbiters of antitumor immunity.

The DUSP22-LGALS1 axis unveils a previously unappreciated layer of immunoregulation: post-translational control of a soluble immunosuppressive factor. While the roles of LGALS1 in inducing T-cell apoptosis and reinforcing endothelial barriers were known, the upstream events dictating its stability and hence its functional output in the TME were elusive. Our discovery that DUSP22 dephosphorylates LGALS1 at specific residues to trigger its degradation provides a mechanistic explanation for the frequent downregulation of DUSP22 in cancers. It suggests that tumors hijack this regulatory node not only for growth advantage but also to establish immune privilege by stabilizing a key immunosuppressant. This paradigm distinguishes it from canonical cell-surface checkpoints like PD-L1, which are often regulated at the transcriptional level, and highlights the therapeutic potential of targeting protein stability within the TME.

Therapeutic targeting of this axis offers dual mechanistic advantages. Pharmacological LGALS1 inhibition (via OTX008) synergizes with PD-1 blockade to overcome resistance in 4T1 “cold” tumors, while phosphosite-specific anti-pGal1 antibodies precisely neutralize LGALS1’s immunosuppressive activity. Unlike PD-L1 inhibitors that act through blockade of the PD-1/PD-L1 pathway to reverse T-cell exhaustion,^{40 41} LGALS1 modulation concurrently mitigates endothelial barrier dysfunction and T-cell apoptosis,^{16 21 22} addressing two major bottlenecks in lymphocyte infiltration. This phosphorylation-centric strategy minimizes off-target risks by exerting activity exclusively in disease contexts characterized by LGALS1 hyperphosphorylation (eg, TMEs), mirroring the precision of modern degradation platforms such as PROTACs that similarly leverage target-specific molecular signatures.

Bulk and scRNA-seq analyses reveal that DUSP22-mediated LGALS1 degradation remodels the TME: upregulating cytotoxic effectors (GZMB and IFN-γ) while expanding effector memory T-cell populations. Crucially, the resistance of phosphomimetic S8D/T58D LGALS1 mutants to degradation confirms these residues as critical regulatory sites. Evolutionary conservation of these phosphosites enhances translational relevance, suggesting broad applicability across human malignancies.

Notably, the DUSP22-LGALS1 axis represents a metabolic-immune interface distinct from classical receptor-ligand checkpoints. Tumor cells exploit DUSP22 downregulation to maintain LGALS1-mediated immune privilege—a mechanism corroborated by genetic rescue experiments showing reversal of DUSP22 overexpression phenotypes through LGALS1 restoration. This linear regulatory hierarchy contrasts with pleiotropic checkpoints like PD-L1, positioning LGALS1 as a tractable therapeutic target.

A persistent challenge in T cell-based therapies for solid tumors lies in overcoming immunosuppressive barriers that restrict T-cell infiltration and function. Our study demonstrates that targeting the DUSP22-LGALS1 axis directly addresses the bottleneck: converts immunologically cold tumors into T cell-inflamed microenvironments. These findings suggest that disrupting LGALS1 could precondition the TME to synergize with adoptive T-cell therapies, such as chimeric antigen receptor T-cell (CAR-T), by facilitating T-cell entry and persistence. The evolutionary conservation of LGALS1 phosphorylation sites further supports its broad applicability across various malignancies.

Our work thus provides a compelling rationale for clinical translation targeting the DUSP22-LGALS1 axis. The synergistic effect of LGALS1 inhibition with anti-PD-1 therapy offers an immediate combinatorial strategy for immunologically cold tumors. Looking forward, our study also outlines a precision medicine framework: tumors with low DUSP22 expression, high total LGALS1, or specifically, high phosphorylated LGALS1 (pGal1) could be prioritized for LGALS1-targeted therapy. The development of clinical-grade pGal1 IHC assays or other biomarkers to identify these patient populations will be crucial. Furthermore, exploring strategies to reactivate or deliver DUSP22 function represents an alternative therapeutic avenue. Ultimately, targeting this axis may bridge the persistent gap between immunotherapy promise and the challenge of solid tumor resistance.

Methods

Animal models

Surgical procedures were strictly followed by the guidelines of the Animal Care and Use Committee of the Faculty of Health at the University of Macau. BALB/c mice, NOD/SCID and BALB/c nude mice were obtained from the animal facility center of the Faculty of Health and Sciences. Female mice aged 6–8 weeks were randomly assigned to experimental groups and housed in groups of 4–6 per cage. For orthotopic tumor implantation, 4T1 or EMT6 cells (5,000 or 1×10⁶) and AT-3 cells (2.5×10⁵) were resuspended in 100 µL of phosphate-buffered saline (PBS) and injected into the mammary fat pads of recipient mice. Tumor volume was measured two times per week using the formula: 0.5×width²×height. Mice with tumors reaching 2.0 cm in diameter were euthanized as endpoints.

In vivo treatments

For in vivo LGALS1 inhibition, OTX008 (5 mg/kg, HY-19756, MCE) or PBS was administered intraperitoneally every 2 days, starting 4 days post tumor cell inoculation. For anti-PD-1 immunotherapy experiments, mice were intraperitoneally injected with 200 µg anti-mouse PD-1 antibody (Bio X Cell, BE0273) or isotype control (Bio X Cell, BE0089) every 4 days after tumor volume reached ~100 mm³. For neutralization of LGALS1/p-LGALS1, mice were treated with 150 µg anti-LGALS1 (anti-Gal1) or phosphor-specific anti-pLGALS1 (T58) antibodies custom-synthesized by GenScript (Singapore) or isotype control via intraperitoneal injection. For in vivo T-cell depletion, mice were injected intraperitoneally with 10 µg of Ultra-LEAF Purified anti-mouse CD3ε antibody (100340, BioLegend) and its Ultra-LEAF Purified Armenian Hamster IgG Isotype control (400940, BioLegend) antibody 2 days before cancer cell inoculation, and every 4 days thereafter.

Cell lines

HEK293T (CRL-3216), Lenti-X 293T (Takara, #632180), MDA-MB-231 (HTB-26), and mouse breast cancer AT-3 (Cellverse, #iCell-m107) cells, and HUVEC (CRL-1730) cell lines were cultured in Dulbecco's Modified Eagle Medium (DMEM) supplemented with 10% fetal bovine serum (FBS) (Gibco). Murine endothelial cells SVEC4-10 (CRL-2181), mouse breast cancer cells 4T1 (CRL-2539) and EMT6 (CRL-2755), and human TNBC cell line HCC1143 (CRL-2321) were cultured in Roswell Park Memorial Institute (RPMI) 1640 medium with 10% FBS (Gibco). All media contained 100 IU mL⁻¹ of penicillin/streptomycin. Cells were incubated at 37°C under a humidified atmosphere containing 5% CO₂ and routinely tested for Mycoplasma.

Plasmid construction, transfection and infection

GFP-tagged Dusp22, and C88S mutation (Cys → Ser substitution) Dusp22 were cloned into pcDNA3 vector (30124, Addgene, Cambridge, Massachusetts, USA). 3×Flag-tagged Lgals1 was cloned into pcDNA3 vector (30125, Addgene, Cambridge, Massachusetts, USA). Plasmids were validated by Sanger sequencing and transfected into HEK293T cells using polyethylenimine (Polysciences, 23966). HA-tagged Dusp22 and 3×Flag-tagged Lgals1 were cloned into pLJM1 vector (91980, Addgene, Cambridge, Massachusetts, USA), and also HA-tagged Dusp22 was cloned into a tetracycline-inducible vector (41393, Addgene, Cambridge, Massachusetts, USA). Human DUSP22 was cloned into pCDH-CMV vector (72265, Addgene, Cambridge, Massachusetts, USA). Lgals1 knockout (sgLgals1) and non-targeting (sgCtrl) sgRNA oligonucleotides (online supplemental table S1) were cloned into lentiCRISPR v2 (52961, Addgene, Cambridge, Massachusetts, USA). Lentivirus was produced in Lenti-X 293T cells using constructed vectors with packaging plasmid delta 8.2 and pMD2.G using polyethylenimine. The supernatant containing virus particles was harvested at 60 hours and filtered through a 0.45 µm polyethersulfone filter, followed by infection of 4T1 and EMT6 cells, then selected using the appropriate antibiotics (InvivoGen) corresponding to the vector resistance.

Cell proliferation

Cell proliferation was assessed using the Alamar Blue assay according to the manufacturer’s protocol. Briefly, medium containing 10% v/v Alamar Blue (Sigma, R7017) was added to 96-well cultured cells and incubated at 37°C for 2 hours, and absorbance was measured at 590 nm using an excitation wavelength of 530–560 nm.

Immunohistochemistry

Tissue microarrays were made after tissues were fixed in 4% paraformaldehyde and embedded in paraffin. Paraffin sections with a thickness of 5 µm were used for staining. Primary antibodies used in this study included anti-CD3E (R10037, Zenbio), anti-CD4 (ab183685, Abcam), anti‐CD8a (98941S, Cell Signaling Technology), anti-cleaved caspase 3 (25128–1-AP, Proteintech), anti-GZMB (17215S, Cell Signaling Technology). Secondary antibodies used in this experiment included horseradish peroxidase (HRP) goat anti-rabbit IgG antibody (MP-7451–15, Vector Laboratories) or HRP goat anti-mouse IgG antibody (MP-7452–15, Vector Laboratories). After sectioning and staining, slides were scanned at 40×using Hamamatsu Digital Slide Scanner NanoZoomer S60 (Hamamatsu Photonics K.K., Hamamatsu, Japan), and then analysis with Visiopharm Integrator System (Visiopharm A/S, Hoersholm, Denmark).

Clinical samples and multiplex IHC staining

Breast cancer tissue microarrays (HBreD090Bc01, 90 cases; Outdo Biotech, Shanghai, China) were subjected to mIHC using the TSA 7-Color Kit (abs50037-100T; Absin Bioscience, Shanghai, China). Sequential staining was performed according to standardized protocols, employing the following primary antibodies: anti-CD3 (ab135372, Abcam), anti-CD8α (85336S, Cell Signaling Technology), anti-DUSP22 (NBP1-83078, Novus Biologicals), anti-LGALS1 (GenScript, Singapore), and anti-pan Cytokeratin (ab7753, Abcam). Nuclear counterstaining was achieved with 4',6-diamidino-2-phenylindole (DAPI). Whole-slide imaging was performed at 40×magnification using the PhenoImager Fusion system (Akoya Biosciences, Massachusetts, USA), followed by quantitative analysis with inForm V.3.0 software (Akoya Biosciences).

ELISA

LGALS1 secretion was quantified in supernatant of 4T1 and EMT6 cells and in the serum of mice bearing 4T1 and EMT6 tumors was measured with Mouse Galectin-1 ELISA Kit (EK2382, Multi Science Biotech), and corresponding measurements for human MDA-MB-231 and HCC1143 cell supernatants were performed using a Human Galectin-1 ELISA Kit (SEKH-0526–48T, Solarbio). Absorbance was measured at 450 nm with a reference wavelength of 630 nm.

Western blotting analysis

Protein lysates were extracted using radio-immunoprecipitation assay buffer (RIPA) lysis buffer supplemented with proteinase and phosphatase inhibitors (Roche Diagnostics, 4693159001). 30 µg protein, quantified by bicinchoninic acid (BCA) assay (Thermo Fisher Scientific) was separated on 10% sodium dodecyl sulfate – polyacrylamide gel electrophoresis (SDS-PAGE) gel, transferred to PVDF membrane (Millipore), and probed with the following primary antibodies overnight at 4°C: DUSP22 (PA5-118836, Invitrogen), LGALS1 (R26784, Zenbio), phosphor-Threonine (9386, Cell Signaling Technology), phosphor-Tyrosine (9411, Cell Signaling Technology), phosphor-Serine (E0AB-21334, Elabscience), HA (KMTSM1315, AlpaLifeBio), Flag (KMTSM1318, AlpaLifeBio), α-tubulin (11 224–1-AP, Proteintech) and β‐actin (A5316, Sigma). Subsequently, the membranes were incubated with HRP‐conjugated secondary antibodies: goat anti‐rabbit (7074, Cell Signaling Technology) or anti‐mouse (7076, Cell Signaling Technology) at room temperature for 1 hour. Membranes were washed with tris-buffered saline with Tween-20 (TBS-T) buffer and developed using enhanced chemiluminescence reagent with a digital imaging system (Bio-Rad Laboratories).

Co-immunoprecipitation

HEK293T cells were transfected with 5 µg of overexpression plasmids (HA-Dusp22 and 3×Flag-Lgals1) in 10 cm dishes for 48 hours and then washed three times in pre-chilled PBS before being harvested and lysed 30 min with IP buffer supplemented with phosphatase and protease inhibitor cocktail (Roche). The supernatants were spun down at 13,500 rpm for 30 min then quantified with BCA. Proteins were immunoprecipitated using anti-HA magnetic beads (HY-K0201A, MCE) and anti-Flag magnetic beads (M8823, Sigma) at 4°C overnight. The obtained magnetic beads were washed three times and boiled in loading buffer at 95°C for 10 min. Western blot analysis was then performed using the corresponding primary and secondary antibodies.

Mass spectrometry

Immunoprecipitated proteins were digested with trypsin, desalted over C18 resin and then loaded onto an LTQ linear ion trap mass spectrometer (Thermo Finnigan) for LC-MS/MS analyses. MS/MS spectra were searched using SEQUEST against a target-decoy database of tryptic peptides, and candidate proteins were screened with an online tool (https://www.uniprot.org/).

Bulk RNA sequencing and analysis

The RNA-seq libraries were sequenced on Illumina platform (150 bp paired-end). Reads were aligned to the mouse reference genome (mm10) using HISAT2⁴² and counted using featureCounts.⁴³ DEGs were identified using edgeR,⁴⁴ and the threshold was set as fold changes >4 with an adjusted p value<0.05. Functional enrichment analysis was performed using R package clusterProfiler.⁴⁵

Quantitative RT-PCR

Total RNA was extracted using TRIzol reagent (Thermo). Complementary DNA was synthesized using Qiagen OneStep RT-PCR kit (Qiagen), and quantitative reverse transcription-PCR was performed using FastStart SYBR Green Master Mix (Roche). The 2^−ΔΔCT method was used and housekeeping gene Gapdh was used to normalize gene expression. The primer sequences were listed in online supplemental table S1.

T-cell transendothelial migration assay

SVEC4-10 or HUVEC cells were seeded on a Transwell insert, and 5×10⁴ tumor cells were added to the lower chamber of each Transwell. The cells were then co-cultured for 24 hours to allow the SVEC4-10 or HUVEC cells to form a compact monolayer. CD8⁺ T cells isolated from mouse spleen by EasySep Mouse CD8⁺ T Cell Isolation Kit (19853, STEMCELL) or peripheral blood mononuclear cells by EasySep Human CD8⁺ T Cell Isolation Kit (17953, STEMCELL), and then activated with mouse CD3 (16-0032-85, Invitrogen) and CD28 antibodies (16-0281-85, Invitrogen) with recombinant murine interleukin-2 (CK24, Novoprotein, Shanghai, China), or human CD3 (317302, BioLegend) and CD28 antibodies (302902, BioLegend) with recombinant human interleukin-2 (C013, Novoprotein, Shanghai, China), for a duration of 2–3 days. T cells were added into the upper layer of the Transwell. After a 2–4 hours incubation at 37°C, the migrated T cells were harvested from the lower chamber and quantified by flow cytometry following staining with the appropriate antibodies.

Flow cytometry

For cell surface staining, cells isolated from tumors or blood were incubated with antibodies for 30 min on ice in the dark. For intracellular staining, cells were fixed and permeabilized using Cytofix/Cytoperm kit (426803, BioLegend) and stained. For in vitro analysis of IFN-γ and GZMB production by tumor-infiltrating T cells, cells were stimulated with a phorbol 12-myristate 13-acetate (PMA)/ionomycin cocktail containing Brefeldin A at 1/400 dilution (423303, BioLegend) and incubated for 4 hours at 37°C. The following antibodies and reagents were used: CD45-PerCP (103130, BioLegend), CD3-APC (100236, BioLegend), CD3-BV510 (100234, BioLegend), CD4-Alexa Fluor 700 (100430, BioLegend), CD8a-PE-Cyanine7 (100722, BioLegend), CD8a-BV605 (100744, BioLegend), GZMB-PE (372208, BioLegend), Ki67-PE (350504, BioLegend), IFN-γ-PE (505808, BioLegend) and Fixable Viability dye A780 (65-0865-14, Invitrogen).

scRNA-seq analysis

Dusp22 OE or vector 4T1 cells were injected into BALB/c mice. 16 days later, tumors were collected and digested by collagenase II, IV (0.25%), and DNase I (0.05%) in Hanks’ Balanced Salt Solution (HBSS) for 30 min at 37°C. scRNA-seq libraries were prepared using 10x Genomics kit (Chromium Next GEM Single Cell 3ʹ Reagent Kits V.3.1) and then sequenced using an Illumina NextSeq sequencer. Raw data were mapped to mouse genome mm10 with CellRanger V.3.1.0. Downstream analyses were performed in R (V.4.1.1) using Seurat (V.4.4.0). For differential expression analysis, the Wilcoxon rank-sum test was applied, using |log2FC|>0.25, and adjusted p value<0.05 as the cut-off.

Proteomic mass-spectrometry data analysis

Proteomic mass-spectrometry data from the Breast Invasive Carcinoma (TCGA, PanCancer Atlas) was obtained through cBioPortal (https://www.cbioportal.org).⁴⁶ Spearman’s rank correlation analysis was employed to evaluate statistical associations between the expression levels of indicated proteins and CD8A across these datasets.

Identification of T-cell infiltration-associated genes

To identify “T cell infiltration–associated genes”, we analyzed 88 spontaneous tumors from Brca1Co;WAP-Cre;SB; T2Onc3 transgenic mice (one tumor per mouse). Tumor progression was quantified by the time-to-endpoint, with tumors harvested on reaching a diameter of 1–2 cm or when the mice reached a predetermined humane endpoint. Subsequently, all tumors underwent transposon sequencing and IHC staining. Based on the CD3⁺ T-cell proportion, the tumors were classified into either low or high T-cell–enriched groups for subsequent analysis. The number of tumors harboring insertional mutations in each candidate cancer driver gene was compared between low and high T-cell–enriched tumors using Fisher’s exact test, with statistical significance defined as p<0.05.

Data availability statement

Data are available upon reasonable request.