2.1 Prognostic role of STAT4 in independent cohort

Demonstrating the prognostic role of STAT4 in our collected cohort of 86 TNBC samples (Table S1), STAT4 protein was highly expressed in tumor tissues compared to that in paired normal tissues (Figure 1A–C), which was consistent with previous report.²³ Survival analysis revealed high STAT4 expression was significantly related to better prognostic outcomes. Patients with high STAT4 expression showed longer survival time (Figure 1D). The univariate Cox proportional hazard assay implied clinical features age and tumor size were considered as risk factors of prognosis (hazard ratio > 1, p < 0.05), while the STAT4 score was a protective factor of prognosis (Figure 1E). Furthermore, significant univariate Cox characteristics were introduced into multivariate COX analysis, which identified STAT4 as an independent prognostic factor (Figure 1F). The Cancer Genome Atlas (TCGA) and METABRIC datasets were analyzed to further verify the prognostic value of STAT4 in a large cohort. As shown in Figure 1G,H, STAT4 was significantly associated with overall survival, and high STAT4 subgroups, particularly in TNBC, showed better prognostic survival.

2.2 STAT4 was specifically relevant with T cell in tumor microenvironment

To further decipher the role of STAT4 in TME, single-cell transcriptomics datasets of nine breast cancer cohorts were analyzed. As presented in Figure 2A, a distinct expression pattern was discovered among various datasets, with different expression level and distribution in cell clusters were observed, unveiling the expression heterogeneity of STAT4 among patient cohort. STAT4 was predominantly expressed in immune cells but not in malignant cells and stromal cells. Detailed cell annotations suggested STAT4 was high expression in CD4+ T cells, proliferating T cells, and CD8+ T cells in more than half the data sets (Figure 2A). The single-cell sequencing dataset of clinical cohort GSE176078 (a comprehensive dataset comprising 26 primary breast cancer tumors) was selected to further explore STAT4 in TME. In breast cancer subtypes, STAT4 was higher in HER2+ samples than in ER+ and TNBC patients (Figure 2B). Moreover, the expression heterogeneity of STAT4 in patients was reveled in Figure 2C. Importantly, the elevated and specific expression of STAT4 was observed in T cells (Figure 2D–F), and further minor annotations implied that STAT4 was expressed in multiple anti-tumor T-cell subtype: CD4+ T cells, CD8+ T cells, NK T cells, NK cells, and cycling T cells (Figure 2G,H). The significant and positive correlation (Cor = 0.91, p-value < 0.001) between T-cell ratio and STAT4-positive cell ratio in the 46 primary tumor samples was shown (Figure 2I). More detailed annotations suggested STAT4+ cells were strongly correlated (Cor > 0.8, p-value < 0.001) with CD4+ T cells, CD8+ T cells, and NK cells (Figure 2J), having implications for understanding favorable prognosis in the STAT4-high cohort. Briefly, high-STAT4 cohort exhibited higher T-cell level, resulting in better prognosis status.

2.3 STAT4 as a tumor driver in vitro

To unveil gene function of STAT4 in breast cancer cell lines in vitro, we first detected the mRNA and protein expression of STAT4 in several common breast cancer cell lines. Almost no STAT4 expression is detected in three of breast cancer cell lines, but MCF-10A cells display high levels of STAT4 expression (Figure 3A,B). STAT4 gene was overexpressed by plasmid delivery into BT549 cells. Overexpression of STAT4 promoted BT549 cell proliferation revealed by cell counting kit-8 (CCK-8) and cell clone assays (Figure 3C,D). Wound healing and Transwell experiments implied that STAT4 facilitated migration and invasion (Figure 3E,F). Simultaneously, STAT4 overexpression led to a shortening of the G1 cell cycle and a decrease in the proportion of apoptotic cells (Figure 3G,H).

2.4 STAT4 upregulated PD-L1 via IL-12R/JAK2–STAT3–STAT4/PD-L1 axis

TNBC cell lines expressed PD-L1,²⁴ and immune checkpoint inhibitors have shown great treatment potential in TNBC.²⁵ Interestingly, the co-transcription between STAT3 and STAT4 has been reported in colorectal carcinoma,²⁶ but whether STAT4 indirectly transactivates CD274 (coding PD-L1) through synergistic effect with STAT3 remains unknown. Therefore, we explored the interaction between STAT4 and JAK2/STAT3/PD-L1 axis. A strong correlation between STAT4 and CD274 was unmasked in clinical single-cell dataset GSE176078 and TCGA–breast invasive carcinoma (BRCA) cohort, and in breast cancer cell lines (Figure 4A–C). Herein, STAT4 overexpression promoted PD-L1 expression and JAK2/STAT3 overactivation (Figure 4D), whereas the possibility that STAT4 transcriptionally induced PD-L1 expression with STAT3 dependent was not excluded.^{27, 28} Previous research reported the close correlation between IL12RB1 and the activation of STAT3 and STAT4, while IL12RB2 specifically activated STAT3.²⁹ In this study, co-immunoprecipitation assay verified the direct binding between STAT3 and STAT4 (Figure 4E). Both STAT3 and STAT4 were responsible for the transcription of IL12RB1 and IL12RB2 in breast cancer cell (Figure 4F,G). Silencing STAT3 reduced CD274 level induced by STAT4 overexpression, indicating that STAT4 regulated PD-L1 expression dependent on STAT3 (Figure 4H). The above results implied that STAT4 promoted PD-L1 expression through the JAK2/STAT3/IL12RB axis, thus we proposed the potential feedback loop: STAT4/IL12RB/JAK2–STAT3–STAT4/PD-L1 pathway (Figure 4I). Interestingly, in xenograft nude mice model, overexpressed STAT4 decreased tumor growth (Figure 4J). Overexpressed STAT4 elevated PD-L1 expression in protein and mRNA level in vivo, instead of STAT3 (Figure 4K,L). These results suggested that STAT4 enhanced cell proliferation in a STAT3-dependent manner in vitro, while STAT4 inhibited tumor growth without STAT3 dependent in vivo.

2.5 STAT4-related pathway score predicted anti-PD-1 immunotherapy response

Given that STAT4 elevated PD-L1 expression via the feedback loop: STAT4/IL-12R/JAK2–STAT3–STAT4, in order to further apply above signaling axis for predicting immunotherapy response, the feedback loop was defined as the STAT4-related pathway. The STAT4-related pathway score (Srps) was calculated using its gene symbols (JAK2, STAT3, STAT4, CD274, IL12RB1, IL12RB2) in GSE194040 cohort. The higher expression of components of STAT4-related pathway and Srps in the responder cohort than in non-responder cases was observed, respectively, with a consistent trend among seven immune activation biomarkers³⁰ (Figure 5A), suggesting high-Srps cases had more “hot” tumor environment than patients with low Srps levels.³¹ As shown in Figure 5B,C, the Srps was significantly elevated after anti-PD1 treatment, and responders were more frequent in the high-Srps subgroup. The predictive accuracy of Srps (area under the curve [AUC] = 0.83) toward treatment response was better than that of seven biomarkers (Figure 5D). In single-cell cohort of GSE176078, STAT4 and Srps were mainly expressed in T cells, comprising CD4+ T cells, CD8+ T cells, NK cells, and NK T cells, and Srps was strongly associated with T-cell levels (Figure 5E–G). These findings highlighted Srps as a T-cell-related signature for immunostratified therapy, with potential to predict anti-PD-1 treatment response.

2.6 The immune landscape of STAT4-related pathway in anti-PD1 immunotherapy

To further investigate the role of STAT4-related pathway in tumor environment following anti-PD1 treatment, we analyzed the single-cell dataset of breast cancers treated with anti-PD-1 immunotherapy (cases = 29). As present in Figure 6A,B, STAT4 was high expression in T cells. Of importance, the STAT4-related pathway members and Srps were increased following anti-PD1 treatment (Figure 6C). Notably, STAT4 and Srps were enriched in T cells, and exhibited strong correlation with T-cell (Figure 6D,E), verifying their robust expression pattern in T cell with or without anti-PD1 treatment, and strong relationship with T cell (Figure 6E–G). Moreover, STAT4-high subgroup had significantly greater T-cell infiltrating level than STAT4-low subgroup (Figure 6F). Furthermore, the expression level and expression percent of STAT4 and Srps were significantly increased in T-cell expansion patients (Figure 6G–K). It is worthy to note that Srps level in pre-treatment samples could predict T-cell status (expansion vs. non-expansion) after anti-PD-1 with available accuracy (Figure 6L). Collectively, Srps signatures showed predictive potential for T-cell expansion following anti-PD1 treatment. We also applied the Srps on large-scale dataset with anti-PD-1/PD-L1/CTLA4 treatment and found that Srps performed certain accuracy to predict treatment response on eight of total 25 cohorts (AUC > 0.7), which demonstrated the potential utility of Srps in pan-cancer immunotherapy response (Figure S1).

4.1 Data source and processing

The mRNA sequencing data of BRCA with fragments per kilobase per million (FPKM) files were retrieved from TCGA (https://portal.gdc.cancer.gov/) and normalized by log2(FPKM +1) for further analysis (113 normal and 1109 tumor cases) in R software (version 4.0). The gene expression and clinical data of Molecular Taxonomy of Breast Cancer International Consortium (METABRIC, 1903 cases) cohort were retrieved from cbioportal (http://www.cbioportal.org/). The gene expression matrix and annotations of 1019 cell lines (more than 20 cancer types, including 59 breast cancer cell lines) were download from the Cancer Cell Line Encyclopedia (https://sites.broadinstitute.org/ccle) and then normalized by above log2 method. Duplicate samples and those without prognostic information were excluded from further analysis. The 86 breast cancer samples with clinicopathological data were collected for discovering protein expression and prognosis in breast cancer (Table S1). The STAT4 expression in nine single-cell transcriptomic datasets of human breast cancer was deciphered in Tumor Immune Single-Cell Hub 2 online server (http://tisch.comp-genomics.org/). Single-cell dataset of 26 primary breast tumors (including 11 ER+, five HER2+, and 10 TNBC samples) was download from Gene Expression Omnibus (GEO, https://www.ncbi.nlm.nih.gov/geo/) under the accession number GSE176078 and processed following previous study.⁴³ The bulk transcriptomic dataset of PD1 immunotherapy trail (pembrolizumab, 69 cases) was retrieved and refined from GEO database (GSE194040). The single-cell dataset with anti-PD1 treatment (pembrolizumab, 29 cases) in breast cancer was retrieved from previous study.⁴⁴ The data of immunotherapy on pan-cancer were analyzed in TIDE (http://tide.dfci.harvard.edu).⁴⁵ The STAT4-related pathway score (gene set: JAK2, STAT3, STAT4, CD274, IL12RB1, IL12RB2) was computed using ssGSEA method of GSEA package in R software (version 4.3.0).⁴⁶ The mean expression in Z-score was shown in heat map, and dot plot was used to exhibit gene expression at single-cell level.

4.2 Expression and prognosis of STAT4 in BRCA

The median value of STAT4 expression level was used as the cut-off to distinguish between the high and low expression groups. The significance of the Kaplan–Meier survival curve was obtained from log-rank test. The tissues microarrays (TMA) ship of 86 TNBC samples and 45 paired tumor adjacent tissues were collected from National Human Genetic Resources Sharing Service Platform (NHGRP; ID: 2005DKA21300). The subsequent experiments were performed at Shanghai Outdo Biotech Co., Ltd. and approved by the company's ethics committee (no. HBreD139Su01; HBre-Duc090Sur-01). TMAs were constructed and fixed with formalin. The staining of immunohistochemistry (IHC) was completed by auto-stainer link 48 instruments (Dako) with STAT4 antibody (1:500, Abcam, ab68156) overnight at 4°C. IHC results were scanned by Ape-rio XT (Leica Microsystems Inc.) and evaluated by two pathologists independently via modified immunoreactive score (IRS). The staining intensity of negative, weak, moderate and strong were scored as 0, 1, 2 and 3, respectively. The percentage scores of positive cells were described as 0, 0%; 1, 1%–25%; 2, 26%–50%; 3, 51%–75% and 4, 76%–100%. The collected clinical cohort was divided into STAT4-high and STAT4-low subgroups according to the median of IHC scores, and table was constructed by R package TableOne with the default significance test method for categorical variables: Fisher's exact test.

4.3 Cell culture and cell transfection

MCF-7, MCF-10A, MDA-MB-231, and BT549 cells were purchased from the Cell Bank of the Chinese Academy of Sciences. With the exception of MCF-10A, cell lines were cultured in RPMI 1640 medium (Gibco) supplemented with 10% fetal bovine serum (FBS, Gibco; Thermo Fisher Scientific) and 1% penicillin and streptomycin solution (HyClone, Cytiva), while the MCF-10A cells were incubated with DMEM-F12 (Gibco) with 10% FBS, 0.3 g/L L-glutamine, 20 ng/mL Epidermal growth factor (EGF), 10 μg/mL insulin, 500 μg/mL hydrocortisone, and 40 mg/L gentamicin 23. All cells were maintained under environmental conditions of 37°C and 5% CO₂. The amplified cDNA of human STAT4 was cloned into the pLenti-CMV-EGFP lentiviral vector to construct the STAT4 overexpression vector (Gene Pharma). After 6 h of transfection with polyethylenimine agent for si-STAT3 (sense: 5′- GGCCAGCAAAGAAUCACAUTT-3′, antisense: 5′-AUGUGAUUCUUUGCUGGCCTT-3′), the medium was transferred to complete medium.

4.4 Western blotting and co-immunoprecipitation assay

For western blotting, proteins were extracted from the cells using RIPA lysis buffer (Beyotime, P0013). The protein concentration was measured using bicinchoninic acid protein assay kit (Beyotime, P0009). Denatured total protein was loaded onto 10% SDS-PAGE gel and then transferred to the polyvinylidene fluoride (PVDF) membrane. The blots were incubated with monoclonal antibodies (STAT4, 2653, CST; p-STAT4, 4134, CST; PD-L1, 13684, CST; STAT3, 4904, CST; p-STAT3, 9145, CST; JAK2, 3230, CST; β-actin, AC026, ABclonal), and then secondary horseradish peroxidase (HRP)-conjugated anti-rabbit antibody (ab6721, Abcam) was added. After washing, membranes were subjected to chemiluminescence, and the relative optical density was analyzed using ImageJ software (NIH).

Immunoprecipitation: after transfection, cells were lysed in Pierce IP Lysis Buffer (Thermo Fisher Scientific) and the supernatants were cultured overnight with anti-Flag (ab205606, Abcam) or anti-HA (ab236632, Abcam) antibody. Immunoprecipitation reaction mixtures were incubated with protein A/G beads (Santa Cruz Biotechnology). The beads were washed three times to elute the proteins before performing western blotting as described above.

4.5 Reverse transcription-quantitative polymerase chain reaction and chromatin immunoprecipitation

Reverse transcription-quantitative polymerase chain reaction (RT-qPCR): total RNA was extracted using TRIzol reagent. After 24 h of cell inoculation, reverse transcription was performed to synthesize cDNA through the Primescript RT Reagent Kit (Takara Bio Inc., RR047A). The qPCR analysis was performed using quantum studio 3 (Thermo Fisher Scientific) based on SYBR kit TB green TM premix ex taqtm II (Takara Bio Inc., RR820A). Thermal cycling conditions were set according to the manufacturer's instructions, and the primer sequences were as follows: ACTB—forward primer: CGATCCGCGCGTCCCACA and reverse primer: ACGCAGCTCATTGTAGAAGGGTGGTG; STAT4—forward primer: AGCCTTGCGAAGTTTCAAGA and reverse primer: ACACCGCATACACACTTGGA; CD274—forward primer: GCCTCCACTCAATGCCTCAATTTG and reverse primer: TTCACAACCACACTCACATGACAAG. Relative mRNA expression levels were measured via the 2^{− ΔΔCT} method.

Chromatin immunoprecipitation (ChIP) analysis was carried out using the Pierce Agarose ChIP Kit (Thermo Fisher Scientific, 26156), according to the manufacturer's instructions. In brief, cross-linked protein–DNA complexes were immunoprecipitated by following antibodies: anti-STAT3 (Santa Cruz Biotechnology, sc-8019), anti-STAT4 (Santa Cruz Biotechnology, sc-398228), or negative control IgG (Abcam, ab172730). PCR reactions (Takara, RR900Q) electrophoresis were used to locate the target gene promoter sequences in the immunoprecipitated DNA. The sequences of the PCR primers for IL12RB promoter are as follows: IL12RB1—forward primer: GCTGGAGTGTGGTGGCATGATC and reverse primer: AGGTGGGTGGATCACTTAGGGTTAG; IL12RB2—forward primer: GTAATGCCCAGAAGCGTGGT and reverse primer: CTGCCAGGTAAACAATAACAATGC.

4.6 Cell proliferation assay

After transfection of the three groups (blank control, empty vector, and STAT4 overexpression), cells were harvested and seeded into 96-well plates at a density of 2000 cells/well. After 24, 48, 96, and 120 h, 10 μL of CCK-8 (CK04, Dojindo) reagent was added to each well according to the manufacturer's instructions and the cells were incubated at 37°C for 1 h. The absorbance at 450 nm was obtained using microplate reader (Thermo Fisher Scientific).

4.7 Wound healing and Transwell analysis

BT-549 cells were seeded in six-well plates and cultured to approximately 100% confluence in each well. Linear scratch was generated via 10-μL pipette tips. Photographs of scratches were taken after 24 and 48 h for generating statistics through ImageJ software. In Transwell assay, 1 × 10⁵ transfected cells were suspended in 200 μL of serum-free medium and were transplanted to the upper Transwell chamber (Corning, Inc., 0.8 μm pore) covered with Matrigel (Corning, Inc.).⁴⁷ Complete medium of RPMI 1640 (500 μL) was added to the lower chamber. After 24 h, the cells were washed with phosphate-buffered saline buffer, fixed with 4% paraformaldehyde, stained with 0.1% crystal violet and counted via ImageJ.

4.8 Flow cytometry assay

BT-549 cells (2 × 10⁵ cells/well) were seeded into six-well plates for cell cycle analysis and apoptosis assay. Transfected cells were fixed by 70% ethanol overnight at −20°C before staining with propidium iodide (PI) cycle kit (Nanjing KeyGen Biotech. Co. Ltd.) for cell cycle analysis. For apoptosis analysis, transfected cells were collected and stained with annexin V/PI apoptosis detection kit (KeyGen Biotech). The flow cytometry analysis was exerted using flow cytometer (CytoFlex, Beckman Coulter), and the MODIFIT LT 5.0 (BD Biosciences) was used for data analysis of the cell cycle assay.

4.9 Animal experiment

BALB/c nude mice were purchased from SPF Biotechnology Co., Ltd.. Animal experiments were performed and approved at the Animal Central of Chengdu University of Traditional Chinese Medicine. Female mice were grouped randomly and subcutaneously injected with wild-type BT549 cells or STAT4-overexpressed BT549 cells (5 × 10⁶ cells/mouse) to establish a cell-derived xenograft model. Animals were caged in a specific pathogen-free laboratory with free access to food and water for 25 days. After the animals were euthanized by cervical dislocation, tumor weight, and tumor volume were measured. Tumor tissues were used to perform qPCR and western blot assays.

4.10 Statistical analysis

All statistical analyses of public database were processed and visualized in R software. Statistical significance was calculated using Wilcoxon signed rank test and Kruskal–Wallis test in two or more groups, respectively. All correlation were assessed by Pearson's method. The experiments were repeated thrice in parallel, and GraphPad PRISM software (version 7.0) was applied to calculate the significance of all experiments. The p-value of less than 0.05 was considered statistically significant (ns, p ≥ 0.05; ^*p < 0.05; ^**p < 0.01; ^***p < 0.001).