Abstract
Immunotherapy has shown promising efficacy for breast cancer (BC) patients. Yet the predictive biomarkers for immunotherapy response remain lacking. Based on two GEO datasets, 53 differentially expressed genes associated with durvalumab treatment response were identified. Using least absolute shrinkage and selection operator (LASSO) and univariate Cox regression, four genes (COL12A1, TNN, SCUBE2, and FDCSP) revealed prognostic value in the TCGA BC cohort. COL12A1 outperformed the others, without overlap in its survival curve. Survival analysis by Kaplan–Meier plotter demonstrated that COL12A1 was negatively associated with BC patients’ prognosis. A COL12A1-based nomogram was further developed to predict the overall survival in BC patients. The calibration plot revealed an optimal agreement between nomogram prediction and actual observation. Moreover, COL12A1 expression was significantly up-regulated in BC tissues and COL12A1 knockdown impaired the proliferation of MDA-MB-231 and BT549 cells. Gene Ontology, Kyoto Encyclopedia of Genes and Genomes, and Gene Set Enrichment analysis pathway indicated that the function of COL12A1 was related to immunity-related pathways. Immunological analyses illustrated that COL12A1 was correlated with M2 macrophage infiltration and M2 macrophage markers (transforming growth factor beta 1 (TGFB1), interleukin-10, colony stimulating factor 1 receptor (CSF1R) and CD163) in BC. Immunohistochemistry staining further revealed a highly positive relationship of COL12A1 with TGF-β1. The co-incubated models of BC cells and M2 macrophges showed COL12A1 knockdown suppressed M2 macrophage infiltration. Additionally, silencing COL12A1 suppressed TGF-B1 protein expression, and treating with TGFB1 could reverse the inhibitory effects on M2 macrophage infiltration by COL12A1 knockdown. Using immunotherapy datasets, we also found elevated expression of COL12A1 predicted poor response to anti-PD-1/PD-L1 therapy. These results reinforce the current understanding of COL12A1’s roles in tumorigenesis and immunotherapy response in BC.
Introduction
Breast cancer (BC), a complex disease related to high incidence and fatality rate, is the most common malignancy threatening women’s health worldwide (Aghamiri et al. 2021, Cui et al. 2021). It is estimated that 2,261,419 new cases of BC and 684,996 deaths from BC have been reported globally in 2020 (Sung et al. 2021). Conventional treatments for BC encompass surgery, radiotherapy, chemotherapy, or combinations thereof, but these choices are often related to serious side effects and may even cause late recurrence. In this scenario, cancer immunotherapies with higher efficacy and more targeted capability have positioned themselves as promising methods to win combat against BC (Zeng et al. 2021). Immunotherapies applying immune checkpoint inhibitors (ICIs) that block immune checkpoint-related molecules, such as cytotoxic T lymphocyte-associated antigen-4, programmed death-1 (PD-1), its ligand (PD-L1), and lymphocyte activating-3, have brought about unprecedented paradigm shifts for the treatment of several hematological and solid tumors. These agents could restrain inhibitory pathways in cytotoxic T cells, in turn resulting in boosted anti-tumor immune responses (Cai et al. 2021, Rizzo et al. 2022a ). Single-agent ICI appears to be effective in merely a small fraction of metastatic triple-negative breast cancer (TNBC) patients. In this scenario, multiple ICI-based combination therapies encompassing combinations of ICIs with chemotherapy, antibody–drug conjugates, and other treatment strategies have been assessed in TNBC to ameliorate the efficacy of ICIs, thus enhancing overall responses and clinical outcomes (Rizzo et al. 2022b ). Furthermore, ICIs have been implicated in the development of early death, which could be mitigated via ICI-based combination therapies (Wang et al. 2021, Viscardi et al. 2022). Despite these encouraging benefits, only a few BC patients have achieved a durable response to ICIs in clinical practice. Biomarkers predicting response may aid in selecting patients who will optimally benefit from ICIs. It has been evidenced that several emerging biomarkers are predictive of BC response to ICI, including PD-L1 expression status, tumor-infiltrating lymphocytes, and tumor mutational burden (Rizzo & Ricci 2022). Unfortunately, current biomarkers still possess many disadvantages; for instance, many biomarkers displayed intra/intratumor heterogeneity, the treatment cost is exceptionally high, and the predictive capability is far from ideal, hindering the clinical translation of these biomarkers (Long et al. 2021). Thus, discovering novel and more reliable predictive biomarkers is desperately needed.
Collagen type XII α1 chain (COL12A1), located on human chromosome 6q12-q13, belongs to the fibril-associated collagen family with interrupted triple helices (Gerecke et al. 1997). COL12A1 serves as a mediator in the interplay between fibrils and the surrounding stroma, and its mutations are correlated with myopathy (Hicks et al. 2014). Recently, the vital role of COL12A1 in cancers has attracted increasing attention. COL12A1 was up-regulated in various types of cancers, encompassing gastric cancer (Duan et al. 2018), colorectal cancer (Wu and Xu 2020), and BC (Xu et al. 2020), and this elevation led to tumor migration, invasiveness, and metastasis (Karagiannis et al. 2012, Verghese et al. 2013, Xiang et al. 2019). Other studies also detected an increase in COL12A1 levels in cisplatin and doxorubicin-resistant ovarian cancer cell lines (Januchowski et al. 2016). Additionally, COL12A1 expression was tightly associated with anti-tumor immunity/immunotherapy. For example, Wu et al. identified that COL12A1 was positively correlated with the abundance of M0 and M2 macrophages in more than six tumor types and exhibited significant differential expression between responding and non-responding patients receiving immunotherapy (Wu et al. 2021). However, there is little comprehensive and systematic evidence regarding the role of COL12A1 on BC immunity.
The purpose of the current analysis was to explore the detailed role of COL12A1 in BC immunotherapy response. By analyzing the data deposited in several public databases, COL12A1 was found to influence the therapeutic responses toward durvalumab, an anti-PD-L1 inhibitory (Bachelot et al. 2021) and predict poor prognosis in BC. Moreover, COL12A1 was highly expressed in BC tissues, estrogen receptor (ER)-positive, human epidermal growth factor receptor-2 (HER-2)-positive, progesterone receptor (PR)-positive, and non-basal-like breast tumors. Functional enrichment analysis revealed that COL12A1’s function was closely linked with immune response. Immunology analysis showed that COL12A1 was positively related to M2 macrophage infiltration and M2 markers. Further in vitro experiments evidenced that COL12A1 could increase TGFB1 expression and subsequently enhance M2 macrophage infiltration. Additionally, COL12A1 high expression was associated with poor immunotherapy response and predicted dismal survival of patients receiving immunotherapy. Our study indicated the predictive role of COL12A1 in BC immunotherapy response.
Materials and methods
Identification of differentially expressed genes with prognostic value
Datasets suitable for the study were mined from Gene Expression Omnibus (GEO, http://www.ncbi.nlm.nih.gov/geo) database (Clough & Barrett 2016) using the keywords ‘durvalumab’ and ‘breast cancer’. Two datasets associated with durvalumab treatment in BC were incorporated, including GSE139050 and GSE173839. The GSE139050 dataset contains three pairs of tissues from durvalumab responding and non-responding BC patients. GSE173839 contains 71 tissue specimens from BC patients following durvalumab and olaparib treatment and 34 control BC tissue specimens. Based on the setting cut-off criteria (|log2 fold change| > 0.6), the differential expressed genes (DEGs) associated with durvalumab therapeutic response were identified. Then, RNA-seq data and detailed clinicopathological information in the TCGA-BRCA cohort were collected from the TCGA portal (https://portal.gdc.cancer.gov/repository) (Olness 1989). To screen genes from the communal DEGs that are associated with overall survival (OS) of BC patients in the TCGA-BRCA cohort, we implemented the least absolute shrinkage and selection operator (LASSO) Cox regression via the ‘glmnet’ package (Friedman et al. 2010). Subsequently, univariate Cox regression and Kaplan–Meier curves were completed to further validate the prognostic values of each gene in the TCGA-BRCA cohort.
Prognosis analysis of COL12A1
For survival analysis, patients were stratified into high and low expression groups based on the optimal cut-off value of COL12A1 expression. Kaplan–Meier plotter (kmplot.com/analysis) was applied to investigate the prognostic significance of COL12A1 mRNA and protein levels on BC (Györffy et al. 2010). The prognostic index includes OS and distant metastasis-free survival (DMFS). Subsequently, univariate and multivariate Cox regression analyses were formulated in the TCGA-BRCA cohort to assess the prognostic effects of COL12A1 and other potential risk factors. To guide clinical decision-making, a quantitative model with clinical applicability is required to predict the OS of BC patients. Hence, we introduced a nomogram that incorporated COL12A1 and other independent prognostic factors attained by univariate analysis such as age, pathological stage, T, and N stage. To confirm the practical implication of this model, a calibration curve was plotted to verify.
COL12A1 expression analysis
The expression of COL12A1 in different types of cancer tissues including BC tissues was analyzed through Human Protein Atlas (HPA, https://www.proteinatlas.org/) (Thul & Lindskog 2018). Then, a pan-cancer expression analysis was done by tumor immune estimation resource, version 2 (TIMER2, http://timer.cistrome.org/) (Li et al. 2020) and we mainly focus on the expression difference of COL12A1 between BC tissues vs normal tissues. Also, the Xiantao tool and UALCAN portal (http://ualcan.path.uab.edu) (Chandrashekar et al. 2017) were further employed to verify the expression difference. Additionally, using the Xiantao tool, we also explored expression differences of COL12A1 among groups of BC patients, based on distinct clinical indicators, containing ER, PR, HER-2 status, as well as tumor subtypes. The DNA methylation level of COL12A1 promoter in BC was also analyzed through UALCAN database.
COL12A1-related enrichment analyses
The top 100 genes which were similar to COL12A1 in BC were attained by exploiting the ‘similar genes detection’ module in Gene Expression Profiling Interactive Analysis 2 (GEPIA2, http://gepia2.cancer-pku.cn/#index) (Tang et al. 2019). A correlation heatmap was created to display the expression correlation between COL12A1 and the top 20 similar genes. Based on the 100 similar genes, Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses were conducted by Xiantao platform to investigate the functional roles of COL12A1-related genes in biological process (BP), cellular components (CC), molecular functions (MF), and KEGG pathways. Only pathways of interest were displayed. Additionally, to explore COL12A1’s specific role in durvalumab response in-depth, Gene Set Enrichment Analysis (GSEA) by Xiantao tool was accomplished based on the DEGs in GSE139050. Significantly enriched hallmarks were determined when adjusted P < 0.05 and false discovery rate < 0.25. The top 10 hallmarks-related pathways were shown.
Immunological analysis
The relationship between COL12A1 expression and immune cells was initially determined using ssGSEA analysis by Xiantao tool, with results shown as a bubble plot. Biomarker Exploration of Solid Tumors (BEST, https://rookieutopia.com/app_direct/BEST/) portal was used for validation. Then, correlation analyses between COL12A1 expression and the infiltration degree of different macrophage subtypes were performed using distinct algorithms by TIMER2. Besides, Xiantao tool, TIMER2, and GEPIA2 databases were applied for the cross-validation of markers for different macrophage subtypes that were associated with COL12A1. Finally, the association between COL12A1 and immunomodulators, immunotherapeutic response, as well as immunotherapy prognosis, was analyzed by BEST application.
Cell culture, siRNA transfections, TGF-β1 treatment, and macrophage polarization
BC cell lines (MDA-MB-231 and BT549) were provided by Cancer Research Institute, Central South University. Human monocytic cell line (THP-1) were purchased from Chinese Academy of Sciences. BC cells and THP-1 cells were grown in RPMI-1640 medium (BasalMedia, Shanghai, China) supplemented with 10% fetal bovine serum (FBS, Procell, Wuhan, China) and 1% penicillin-streptomycin (Gibco) at 37°C with 5% CO2. Small-interfering RNAs (siRNAs) targeting COL12A1 (si-COL12A1-1 and si-COL12A1-2) and non-targeting control siRNAs were synthesized by Ribo Life Science Company (Guangzhou, China). As previously depicted (Chen et al. 2021), siRNAs were transfected into MDA-MB-231 and BT549 cells for 48 h utilizing Lipofectamine 3000 (Thermo Fisher Scientific). Recombinant human TGF-β1 was bought from R&D. Forty-eight hours after siCOL12A1 transfection, MDA-MB-231 and BT549 cells were treated with TGF-β1 (5 ng/mL) for 24 h. To polarize M0 macrophages into M1, 2.5 × 105 THP-1 cells were seeded in a 6-well plate and stimulated with 320 nM phorbol-12-myristate-13-acetate (MedChemExpress, Monmouth Junction, NJ, USA) at 37°C for 6 h. For M2 polarization, cells were stimulated with M2-polarizing reagents IL-4 (20 ng/mL, MedChemExpress) plus IL-13 (20 ng/mL, Peprotech, Rocky Hill, NJ, USA) and then incubated at 37°C for 72 h.
Western blot detection
Cells were collected and lysed in a RIPA buffer containing protease inhibitor cocktail. The resulting protein sample was quantified using a BCA protein assay kit, separated via SDS-PAGE gels electrophoresis, electro-transferred onto PVDF membranes, and blocked using 5% skim milk. Subsequently, membranes were incubated with anti-COL12A1 primary antibodies (abcam, Shanghai, China) and TGF-β1 (Proteintech, Rosemont, IL, USA) with a dilution of 1:1000 (v/v). Afterward, the membranes were rinsed in PBST thrice for 3 min each, treated with HRP-conjugated secondary antibodies (Proteintech) for 1 h and final band intensities of protein were determined via the Image Lab software (Bio-Rad). β-actin (1:1000; Santa Cruz) served as a normalization control.
Colony formation assay
The MDA-MB-231 and BT549 cell lines were transfected as earlier. Afterward, approximately 1000 cells were inoculated into 6-well-culture plates and incubated for around 15 days. Cell colonies were counted after crystal violet staining.
Immunohistochemistry
The BC tissue array (Cat no: HBre-Duc060CS-04) was obtained from Shanghai Outdo Biotech (Shanghai, China). Immunohistochemical (IHC) staining was accomplished as depicted previously (Yan et al. 2019, Xu et al. 2022). In brief, IHC was conducted utilizing Histomouse SP Kit (Invitrogen). Tissue sections were immunostained employing a streptavidin-peroxidase method following antigen retrieval by microwave heating. Signal detection was done using a 3,3′-diaminobenzidine reagent. The antibodies against COL12A1 (Cat no: ab121304) and TGF-β1 (Cat no: 21898-1-AP) were both purchased from Proteintech (USA). Two experienced pathologists independently examined and quantified the image of sections. The stained intensity was classified into four levels: 0 (negative), 1 (weak), 2 (moderate), or 3 (strong). The score for staining extent was scaled as 0 (≤10%), 1 (11–25%), 2 (26–50%), 3 (51–75%), or 4 (>75%). The final score was designated as the multiplication of the above two scores and was categorized as 1–3 (weakly positive), 4–6 (positive), or 7–12 (strongly positive).
M2 macrophage infiltration assay
After the polarization of THP-1 cells, the M2 macrophage infiltration assay was completed as described previously (Xu et al. 2022). Briefly, 2.5 × 105 M2 macrophage cells were added without serum for 12 h into the upper compartment of a transwell plate (5-mm pore size, Corning), and 2.5 × 105 MDA-MB-231 and BT549 cells were cultured with 10% FBS in the lower compartment. After 24 h of co-incubation at 37°C, the cells in the upper compartment were fixated in 4% formalin and stained using 0.3% crystal violet. For infiltrated M2 macrophage cell counting, three fields from each membrane were randomly chosen and counted.
Statistical analysis
Pairwise comparisons were performed applying the Student's t-test, while multi-group comparisons were conducted applying ANOVA. Kaplan–Meier methods were adopted to build survival curves, and the log-rank test was done to estimate survival differences between groups. All experiments were repeated at least thrice, with data reported as mean ± s.d. Data analyses were performed using GraphPad Prism 6 and SPSS 23.0 (IBM). Significant differences were accepted when P < 0.05.
Results
Identification of differentially expressed durvalumab response-related genes
DEGs were screened utilizing gene expression profiles of two GEO datasets regarding durvalumab therapy, namely, GSE139050 and GSE173839. Based on the screening criteria of |log2FC| > 0.6, we identified 17,206 DEGs in GSE139050, and 87 DEGs in GSE173839. After analysis with Venn diagram, 53 co-DEGs were identified for subsequent analyses (Fig. 1A). Combining the prognostic information of the TCGA BC cohort, LASSO Cox regression model was exploited to determine the most useful prognostic factors out of the co-DEGs and identified TNM, SCUBE2, KRT5, FDCSP, and COL12A1 (Fig. 1B and C). Among these five genes (Supplementary Table 1, see section on supplementary materials given at the end of this article), COL12A1 showed a positive coefficient and appeared as a risk factor, as its high expression indicated high risk. The other four genes with negative coefficients appeared as protective factors, as their high expression suggested low risks. Univariate analysis and Kaplan–Meier curves were further employed to evaluate the prognostic value of each gene. Univariate analysis showed that apart from KRT5, other genes were significantly related to BC patients’ OS (Fig. 1D). Kaplan–Meier curves indicated that highly expressed COL12A1 was significantly related to shorter OS and its survival curves revealed few overlaps, but survival curves of the other three genes were visually overlapping (Fig. 1E, F, G, H and I). Thus, COL12A1 was selected for an in-depth study. To determine the effect of COL12A1 on the therapeutic response of BC patients, we checked COL12A1 levels in GSE139050 and GSE173839. According to a previous study, COL12A1 expression was induced by exposing drug-sensitive ovarian cell lines to antineoplastic drugs, and COL12A1 could block the penetration of antineoplastic drugs into the cancer tissues and enhance the apoptosis resistance, thus increasing the drug resistance (Januchowski et al. 2016). Analysis of GSE173839 revealed an enhanced expression of COL12A1 following treatment of durvalumab and olaparib (Fig. 1I), which might also correlate with enhanced resistance toward durvalumab and olaparib in BC. Also in GSE139050, COL12A1 in durvalumab non-responders exhibited a trend toward increased expression, although the P value failed to reach significance owing to the abnormally high expression of COL12A1 in one of the responders (Supplementary Fig. 1). These findings indicated that COL12A1 might be a negative biomarker for patients’ responses to durvalumab.
Identification of durvalumab response-related genes with prognostic implication in breast cancer (BC) patients. (A) Venn diagram of common differential expression genes (co-DEGs) from the GEO database. (B) Acquirement of the optimal penalty parameter (λ) in the LASSO model by a minimum combination of five genes. The two dotted vertical lines represented the optimal values using minimum criteria and 1 s.e. criteria by 10-fold cross-validation. (C) LASSO coefficient profiles of 53 co-DEGs. (D) Forest plot of five candidate prognostic genes. (E–H) Kaplan–Meier survival analysis was adopted to assess the prognostic value of these candidate genes, respectively. (I) The expression of COL12A1 was compared between durvalumab/olaparib-treated and control BC tissues in the GSE173839 dataset. A full color version of this figure is available at https://doi.org/10.1530/ERC-23-0012.
Citation: Endocrine-Related Cancer 30, 5; 10.1530/ERC-23-0012
COL12A1 predicts dismal prognosis of breast cancer patients
Next, the prognostic implication of COL12A1 was validated using Kaplan–Meier database. Of note, high COL12A1 protein level was significantly associated with bad OS of BC patients in Liu_2014 (hazard ratio (HR) = 2.2, P = 0.011) (Fig. 2A) and Tang_2018 cohorts (HR = 2.46, P = 0.025) (Fig. 2B). As well, patients with up-regulated COL12A1 protein were marginally related to inferior DMFS in Liu_2014 dataset (HR = 1.75, P = 0.055) (Fig. 2C). Consistently, high mRNA expression of COL12A1 demonstrated poor DMFS in BC patients with basal-like breast tumors (HR = 1.6, P = 0.049) (Fig. 2D). Then, univariate and multivariate Cox regression analyses were completed in the TCGA-BRCA dataset to determine whether COL12A1 could independently predict prognosis of BC patients. As shown in Table 1, the univariate analysis identified COL12A1 (HR = 1.463, P = 0.021), age (HR = 2.020, P < 0.001), pathologic stage (HR = 2.962 for stage III, P < 0.001; HR = 11.607 for stage IV, P < 0.001), T stage (HR = 3.755 for T4, P < 0.001), as well as N stage (HR = 1.956 for N2, P < 0.001; HR = 2.519 for N3, P < 0.001; HR = 4.188 for N4, P < 0.001) as an independent undesirable prognostic variable for OS in patients afflicted by BC. Moreover, multivariate analyses uncovered that COL12A1 (HR = 1.479, P = 0.031) and age (HR = 2.314, P < 0.001) were still independent prognostic factors in BC. Then, the above independent prognostic factors for OS were incorporated to construct a COL12A1-based nomogram to predict 10-year survival probability in the TCGA-BRCA cohort (Fig. 2E). The nomogram can easily be applied by providers to assess BC patients’ prognosis, and the only clinical details required to use the nomogram are age, pathological stage, T stage, N stage, and COL12A1 levels. The calibration curve also illustrated satisfactory concordance between the predicted and actual 10-year survival probability (Fig. 2F).
Prognostic value of COL12A1 in BC. Kaplan–Meier survival analyses of COL12A1 protein expression on overall survival (OS) based on the the Liu_2014 cohort (A) and Tang_2018 cohort (B). (C) Kaplan–Meier survival analyses of COL12A1 protein expression on distant metastasis-free survival (DMFS) based on the Liu_2014 cohort. (D) Kaplan–Meier survival analyses of COL12A1 mRNA expression on DMFS of basal-like BC patients. (E) A nomogram integrating the COL12A1 expression with the clinicopathological characteristics (age, pathological stage, T stage, and N stage) in the TCGA cohort. (F) Calibration curve of the nomogram for predicting 10-year OS in the TCGA cohort. A full color version of this figure is available at https://doi.org/10.1530/ERC-23-0012.
Citation: Endocrine-Related Cancer 30, 5; 10.1530/ERC-23-0012
Univariate and multivariate analyses of risk factors and OS in TCGA.
| Characteristics | Total (n) | Univariate analysis | Multivariate analysis | ||
|---|---|---|---|---|---|
| Hazard ratio (95% CI) | P value | Hazard ratio (95% CI) | P value | ||
| Race | 993 | ||||
| Asian | 60 | Reference | |||
| Black or African American | 180 | 1.525 (0.463–5.024) | 0.488 | ||
| White | 753 | 1.325 (0.420–4.186) | 0.631 | ||
| Age | 1082 | ||||
| ≤60 | 601 | Reference | |||
| >60 | 481 | 2.020 (1.465–2.784) | <0.001 | 2.314 (1.617–3.311) | <0.001 |
| Pathologic stage | 1059 | ||||
| Stage I | 180 | Reference | |||
| Stage II | 619 | 1.697 (0.985–2.922) | 0.057 | 1.301 (0.540–3.132) | 0.557 |
| Stage III | 242 | 2.962 (1.664–5.273) | <0.001 | 1.953 (0.586–6.503) | 0.276 |
| Stage IV | 18 | 11.607 (5.569–24.190) | <0.001 | 3.646 (0.941–14.129) | 0.061 |
| T stage | 1079 | ||||
| T1 | 276 | Reference | |||
| T2 | 629 | 1.334 (0.889–2.002) | 0.164 | 1.197 (0.629–2.279) | 0.583 |
| T3 | 139 | 1.572 (0.933–2.649) | 0.089 | 1.012 (0.447–2.288) | 0.978 |
| T4 | 35 | 3.755 (1.957–7.205) | <0.001 | 1.722 (0.639–4.638) | 0.283 |
| N stage | 1063 | ||||
| N0 | 514 | Reference | |||
| N1 | 357 | 1.956 (1.329–2.879) | <0.001 | 1.565 (0.955–2.563) | 0.075 |
| N2 | 116 | 2.519 (1.482–4.281) | <0.001 | 1.512 (0.610–3.748) | 0.372 |
| N3 | 76 | 4.188 (2.316–7.574) | <0.001 | 1.891 (0.796–4.491) | 0.149 |
| PR status | 1029 | ||||
| Positive | 687 | Reference | |||
| Negative | 342 | 1.367 (0.977–1.912) | 0.068 | ||
| ER status | 1032 | ||||
| Positive | 792 | Reference | |||
| Negative | 240 | 1.405 (0.977–2.019) | 0.066 | ||
| HER2 status | 715 | ||||
| Positive | 157 | Reference | |||
| Negative | 558 | 0.628 (0.383–1.028) | 0.064 | ||
| Histological type | 977 | ||||
| Infiltrating ductal carcinoma | 772 | Reference | |||
| Infiltrating lobular carcinoma | 205 | 0.827 (0.526–1.299) | 0.410 | ||
| COL12A1 | 1082 | ||||
| Low | 540 | Reference | |||
| High | 542 | 1.463 (1.059–2.021) | 0.021 | 1.479 (1.037–2.110) | 0.031 |
P < 0.05 was used for multi-Cox regression.
ER, estrogen receptor; HER2, human epidermal growth factor receptor-2; OS, overall survival; PR, progesterone receptor.
High expression of COL12A1 in breast cancer and its relation with clinical characteristics
To further characterize COL12A1’s roles in the malignant progression of BC, we investigated its expression profile using various bioinformatics databases. HPA portal revealed that mRNA expression of COL12A1 was relatively higher in BC than in other cancer types (Supplementary Fig. 1A). Pan-cancer expression analysis by TIMER2 database further revealed significantly up-regulated COL12A1 expression in BC tissues relative to normal tissues (P < 0.001) (Supplementary Fig. 1B). Consistent observations were also gained from Xiantao tool and UALCAN. Based on Xiantao tool, COL12A1 expression was obviously increased in BC, regardless of unpaired (Fig. 3A) or paired tissues (both P < 0.001) (Fig. 3B). Using Ualcan, we also detected highly expressed COL12A1 in BC at both transcript (Fig. 3C) and protein levels (both P < 0.001) (Fig. 3D). To further elucidate the biological meaning of COL12A1 in BC, we knocked down COL12A1 with two siRNAs in MDA-MB-231 and BT549 cell lines (Fig. 3E and F). Clone formation assays indicated that cell proliferation was significantly repressed after transfection of siCOL12A1 in MDA-MB-231 and BT549 cells (all P < 0.01) (Fig. 3G, H and I). Additionally, we probed the association between COL12A1 expression and the clinical characteristics of BC patients using Xiantao tool. BC patients with positive ER, PR, or HER2 status showed elevated levels of COL12A1 than those with negative ER, PR, or HER2 status (all P < 0.01) (Fig. 3J, K and L). Likewise, COL12A1 was significantly decreased in the basal-like subtype compared with other subtypes (all P < 0.001) (Fig. 3M). Methylation analysis showed that BC tissues had higher levels of methylation compared with normal tissues (P < 0.001) (Fig. 3N). These results indicated that COL12A1 expression may represent a potential diagnostic indicator in BC.
The expression of COL12A1 in BC and its correlation with clinical indicators. Expression comparison of COL12A1 between BC and normal tissues in unpaired TCGA-BRCA (A) and paired TCGA-BRCA (B) by Xiantao tool. Expression comparisons of COL12A1 expression between BC and normal tissues at transcript (C) and protein levels (D) by UALCAN portal. (E and F) Protein expression of COL12A1 transfected with siCOL12A1 or siNC in MDA-MB-231 and BT549 cells. Quantitation of COL12A1 protein was shown in supplementary figure S4A and B. (G–I) COL12A1 knockdown suppressed the cell clone formation rate of MDA-MB-231 and BT549 cells. Box plot visualizing the relationship between COL12A1 expression and different clinical indicators including ER status (J), HER2 status (K), PR status (L), and tumor subtypes (M) by Xiantao tool. (N) Promoter methylation levels of COL12A1 in BC and normal tissues by the UALCAN. **P < 0.01, ***P < 0.001. A full color version of this figure is available at https://doi.org/10.1530/ERC-23-0012.
Citation: Endocrine-Related Cancer 30, 5; 10.1530/ERC-23-0012
Functional enrichment analysis of COL12A1-related genes
We acquired 100 genes showing similar expression patterns with COL12A1 using the ‘Similar Genes Detection’ module of GEPIA2, further investigated the biological significance of these genes, and performed GO and KEGG analyses using Xiantao tool to annotate those significantly enriched functions. Figure 4A visualized the heatmap of the top 20 similar genes of COL12A1. Results of GO and KEGG analyses showed that COL12A1 and its similar genes were mainly enriched in the following information, cellular response to transforming growth factor beta (TGFB) stimulus, response to TGFB and extracellular matrix (ECM) organization in BP, wnt signalosome, ECM component and collagen-containing ECM in CC, growth factor binding, ECM structural constituent conferring compression resistance and platelet-derived growth factor binding in MF, and TGFB signaling pathway, PI3K-Akt signaling pathway, and ECM–receptor interaction in KEGG pathways (Fig. 4B). Furthermore, GSEA analysis of DEGs in GSE139050 was performed to further interrogate COL12A1 in durvalumab response, and the top 10 pathways are determined as negative regulation, as shown in Fig. 4C. Among them, immune-associated signaling pathways included interferon-gamma (IFNγ) response, inflammatory response, and interferon alpha (IFNα) response (Fig. 4D). These findings suggested that COL12A1 might take part in immune-related functional networks in BC.
Functional enrichment analysis. (A) Heatmap of expression correlation between COL12A1 and the top 20 similar genes in BC. (B) GO and KEGG analysis based on functionally similar genes of COL12A1. (C) GSEA analysis of DEGs in GSE139050. (D) Three remarkably enriched immune-related HALLMARK pathways by GSEA analysis. A full colour version of this figure is available at https://doi.org/10.1530/ERC-23-0012.
Citation: Endocrine-Related Cancer 30, 5; 10.1530/ERC-23-0012
The effect of COL12A1 in M2 macrophage infiltration
As the enrichment analyses above suggested that immune-related pathways were significantly enriched, we wondered how COL12A1 influenced the immune microenvironment of BC. The association between COL12A1 and diverse immune infiltration cells was explored by employing Xiantao tool with ssGSEA algorithm. As shown in Fig. 5A, various immune cells were found to be positively associated with COL12A1 expression, while COL12A1 had the highest correlation with macrophages (Fig. 5A). Using BEST portal, macrophages also showed a consistently positive correlation with the expression of COL12A1 in TCGA-BRCA, GSE21653, GSE22219, and GSE97342 datasets, although the opposite result was obtained in GSE9893 (Fig. 5B). Through TIMER2, we further evaluated the relationship between COL12A1 expression and different macrophage states. As illustrated by Fig. 5C, COL12A1 exhibited a significantly positive correlation with macrophages in BC and lumA BC subtype, M0 macrophages in basal-like BC and lumA BC subtypes, as well as M2 macrophages in BC, basal-like BC, and lumA BC subtypes, yet, fail to obtain consistent correlation with M1 macrophages using different algorithms (Fig. 5C). Further analysis of the typical phenotypic markers for M (AIF1), M1 (IL12A, TNF, NOS2, and PTGS2), and M2 (IL10, CD163, TGFB1, and CSF1R) based on TCGA-BRCA cohort revealed that COL12A1 had a stronger relationship with markers for M (AIF1) and M2 (IL10, CD163, TGFB1, and CSF1R) while the inconsistent relationship with M1 markers (IL12A, TNF, NOS2, and PTGS2) was found (Fig. 5D). In parallel, correlation analyses by TIMER2 and GEPIA2 databases attested that COL12A1 had no consistent relationship with four M1 markers (Supplementary Fig. 3A and B), instead, exhibited a consistent positive correlation with four M2 markers, among which TGFB1 born the strongest positive correlation (Fig. 6A and B). Then, we examined expression levels of TGFB1 in clinical specimens by IHC staining and conducted a correlation analysis of COL12A1 and TGFB1. Results indicated a stronger staining intensity of COL12A1 and TGFB1 in BC than normal (Fig. 6C and D) and a positive correlation between the staining intensity of TGFB1 and COL12A1 (Pearson’s r = 0.5914, P < 0.001) (Fig. 6E). To explore the effect of COL12A1 on M2 macrophage migrative activity, THP-1 cells were differentiated into M2 macrophages and MDA-MB-231 and BT549 cell lines with COL12A1 knockdown were established. Results showed that COL12A1 knockdown in BT549 and T47D cells diminished the infiltration of M2 macrophages (Fig. 6F). A previous study has demonstrated that TGFB1 signaling served a critical role in macrophage migration (Wu et al. 2022). We hypothesized that TGFB1 may participate in COL12A1-mediated M2 macrophage migration. To confirm the influence of COL12A1 expression on TGFB1, the expression of TGFB1 in COL12A1-silenced MDA-MB-231 and BT549 BC cell lines was measured via Western blot assay. As expected, COL12A1 knockdown reduced the protein expression of TGFB1 (Fig. 6G). Exogenous TGFB1 treatment in the knockdown cells rescued this reduction (Fig. 6H). Moreover, TGFB1 treatment weakened the suppressive effects of siCOL12A1 on the migration of M2 macrophages (Fig. 6I).
The correlation between COL12A1 with immune cell infiltration. (A) Lollipop of correlation between COL12A1 expression and immune cells using ssGSEA algorithms by Xiantao tool. (B) Heatmap of the link between COL12A1 expression and immune cells based on BEST portal. (C) Heatmap of association between COL12A1 expression and different macrophage subsets based on TIMER2. (D) Pearson correlation matrix of correlation between COL12A1 expression and classical phenotype markers of M, M1, and M2 macrophages. A full color version of this figure is available at https://doi.org/10.1530/ERC-23-0012.
Citation: Endocrine-Related Cancer 30, 5; 10.1530/ERC-23-0012
The effect of COL12A1 on M2 macrophage infiltration. (A) The correlation between COL12A1 expression and M2 macrophage markers (IL10, CD163, TGFB1, and CSFR1) based on TIMER2. (B) The correlation between COL12A1 expression and M2 markers (TGFB1, IL10, and CSFR1) based on GEPIA2. (C) Representative images of COL12A1 and TGFB1 protein expression in BC and normal tissues. (D) COL12A1 was up-regulated in BC tissues relative to normal tissues. (E) The relationship between COL12A1 and TGFB1 in all BC tissues. (F) Infiltration of M2 macrophages in MDA-MB-231 and BT549 cells following transfection with siCOL12A1 or siNC. (G) Assessment of TGFB1 expression in siCOL12A1-transfected MDA-MB-231 and BT549 cells using Western blot. Quantitation of COL12A1 and TGFB1 protein was shown in Supplementary Fig. 4C and D. (H) siCOL12A1-transfected MDA-MB-231 and BT549 cells were treated with TGFB1 (5 ng/mL) or without TGFB1, and protein expression of TGFB1 was determined via Western blot. Quantitation of COL12A1 and TGFB1 protein was shown in Supplementary Fig. S4E and F. (I) Infiltration of M2 macrophages in siCOL12A1-transfected MDA-MB-231 and BT549 cells following treatment with TGFB1 (5 ng/mL) or without TGFB1. A full color version of this figure is available at https://doi.org/10.1530/ERC-23-0012.
Citation: Endocrine-Related Cancer 30, 5; 10.1530/ERC-23-0012
The role of COL12A1 in immunotherapy response and prognosis
To further investigate the effects of COL12A1 on immune responses, we analyzed the correlation between COL12A1 and immunomodulators, including antigen presentations, immunoinhibitors, immunostimulators, chemokines, and chemokine receptors. As shown in Fig. 7A, COL12A1 expression showed a strongly positive correlation with immunostimulators, including TNFSF4, NT5E, and CD276, negative association with antigen presentation, HLA-DOB among all five datasets, but had no significant or inconsistent relationship with other immunomodulators. Moreover, we want to see whether aberrant COL12A1 expression could impact the immunotherapeutic response of BC. As shown in Fig. 7B, the expression of COL12A1 was elevated in anti-PD-1/PD-L1 nonresponders in the Kim cohort and Wolf cohort (both P = 0.056). Kaplan–Meier curve demonstrated that higher expression of COL12A1 was associated with inferior PFS of patients receiving anti-PD-1/PD-L1 treatment in the Kim cohort (P = 0.042) (Fig. 7C). The area under receiver operating characteristic curve (AUC) values for Kim and Wolf cohorts were 0.730 and 0.605, respectively, indicating a good capacity of COL12A1 to distinguish anti-PD-1/PD-L1 responders and nonresponders (Fig. 7D).
The association of COL12A1 and immunotherapy response. (A) The correlation between COL12A1 expression and immunomodulators. (B) Expression comparison of COL12A1 in anti-PD-1/PD-L1 responders and nonresponders based on the Kim and Wolf cohorts (C) Kaplan–Meier curves representative of the correlation between COL12A1 and PFS of BC patients receiving anti-PD-1/PD-L1 based on Kim cohort. (D) ROC curve of COL12A1 for patients in Kim cohort and Wolf cohort. A full color version of this figure is available at https://doi.org/10.1530/ERC-23-0012.
Citation: Endocrine-Related Cancer 30, 5; 10.1530/ERC-23-0012
Discussion
Cancer immunotherapy has evolved into a highly promising modality (Calmeiro et al. 2021, Hadadi & Acloque 2021, Moujaber et al. 2021). Numerous immunotherapy approaches have been investigated for BC, covering tumor-targeting antibodies, adoptive T-cell therapy, vaccines, and ICIs such as anti-PD-1/PD-L1. Among these treatments, ICIs stand out because of the substantial clinical benefits for patients with BC (Barzaman et al. 2021). Unfortunately, merely a portion of BC patients will truly respond to ICIs (Isaacs et al. 2021). Thus, developing biomarkers for predicting responsiveness and resistance to ICIs has been a focal point in the research on BC. In our work, based on two GEO datasets, 53 co-DEGs associated with anti-PD-L1 durvalumab therapeutic response were identified. Accumulative studies have reported that LASSO Cox regression could ameliorate the accuracy of bioinformatics analyses and allow for the simultaneous interpretation of every independent variable to screen out the most valuable parameters (Meng et al. 2020). Due to the existence of various confusing features, powerful feature selection and shrinkage were still demanded to avoid overfitting and improve interpretation. To solve this problem, we leveraged the LASSO regression model to select the prognostic genes out of the co-DEGs, and five genes including TNM, SCUBE2, KRT5, FDCSP, and COL12A1 were identified to have prognostic values. Then, using univariate analysis and Kaplan–Meier curves, four genes among them were ultimately correlated with BC patients’ prognosis, but only COL12A1 had non-overlapping survival curves, bearing a significant correlation with poor prognosis in BC patients. According to a prior study, higher expression of COL12A1 was correlated with enhanced chemotherapeutic resistance (Januchowski et al. 2016). Our study found that treatment of durvalumab and olaparib significantly up-regulated COL12A1 expression, implying an increased drug resistance. Later using immunotherapy datasets, up-regulated COL12A1 was correlated with poor response to anti-PD-L1 therapy. Consequently, COL12A1 was a negative biomarker for BC patients’ responses to anti-PD-L1 durvalumab.
COL12A1 encodes the α-chain of collagen XII, and its tumor-promoting role has been extensively investigated. Li et al. found that enhanced expression of miR-1180-3p could suppress COL12A1 levels and subsequently decrease the proliferation and mobility of colorectal cancer (Li et al. 2021a ). Besides, COL12A1 overexpression could counteract the inhibitor effects of methyltransferase-like 3 on the proliferation and metastasis of esophageal squamous cell carcinoma (Li et al. 2022). Consistently, Verghese and colleagues evidenced that COL12A1 was up-regulated in BC stroma and remarkably related to tumor recurrence (Verghese et al. 2013). Plentiful studies also reported that high COL12A1 expression is frequently related to the bad prognosis of several cancers, containing BC (Xu et al. 2020), gastric cancer (Jiang et al. 2019), and pancreatic adenocarcinoma (Li et al. 2021 b). Our results confirmed the findings of previous studies, suggesting that COL12A1 had oncogenetic activity and increased expression of COL12A1 indicated unfavorable prognoses in BC. Additionally, it is worth emphasizing that a novel COL12A1-based predictive tool, which integrated COL12A1 levels and other clinical covariates such as age, pathological stage, T, and N stage, aimed to achieve an adequate survival assessment for BC, was first developed by our team. Impressively, the nomogram yields ideal calibration in BC prognosis prediction.
GO and KEGG enrichment analyses showed that the similar genes of COL12A1 were mainly enriched in several biological pathways, including the TGFB signaling pathway, ECM organization, and PI3K-Akt signaling pathways. These pathways are primarily associated with cancer and immunity. TGF-β is well-established as an immune-suppression mediator within the tumor microenvironment (TME). Activation of TGFB signaling contributes to tumor immune evasion and poor cancer immunotherapy response (Batlle & Massagué 2019). ECM can drive proliferation, migration and invasiveness, neo-angiogenesis, and immune evasion of tumor cells via interacting with tumor and stromal cells (Jiang et al. 2022). PI3K/AKT signaling mediates the chemoresistance of tumor cells through shielding against immune responses and activating pro-survival signaling pathways in multiple human cancers (Kaboli et al. 2021). Further GSEA analysis of DEGs associated with durvalumab responsiveness revealed negative enrichment for hallmarks including IFNγ response, inflammatory response, and IFNα response. IFN signaling covers type-I IFNs (IFNα and IFNβ) as well as type-II IFN (IFNγ). IFNγ signaling is a critical driver of ICIs-induced antitumor immunity, and dysfunction of IFNγ signaling in cancer cells is a mechanism underlying immunotherapy resistance (Du et al. 2022). IFNα is an immunostimulatory molecule and synergistic immunotherapy with IFNα may ameliorate the effects of immunotherapy (Vidal 2020). Therapeutic approaches to augment inflammatory signaling contributes to sensitizing cancers to ICIs (Chen et al. 2022). The functional enrichment results offered novel mechanistic insights into BC development and immunotherapy responsiveness. However, further experimental evidence is still required to confirm these biological mechanisms.
Immunotherapy using ICIs, such as anti-PD-1/PD-L1 antibodies, has shown impressive treatment efficacy for BC in several clinical trials. Nevertheless, the response occurs in merely a minority of patients and a better appreciation of the mechanisms underlying ICI resistance in BC may help ameliorate the patterns and efficacy of immunotherapy. One hurdle impacting the ICI efficacy is the immunosuppressive grid constructed particularly by tumor-associated macrophages (TAMs). TAMs are among the most plentiful immune cell populations within the tumor environment and can be roughly categorized into two contrasting groups: pro-inflammatory M1 macrophages that boost tumoricidal immunity and immunosuppressive M2 macrophages that are tumor-supporting (Xiang et al. 2021). Among them, M2-like TAMs are greatly linked with immunity down-regulation and can drive resistance toward anti-PD-1/PD-L1 antibodies (Gao et al. 2022). Clinically, the presence of M2 macrophages is often connected with an inferior prognosis in BC patients (Tiainen et al. 2015). To reinforce the efficacy and conquer resistance toward ICIs, TAM-modulating strategies in the setting of ICIs are currently under development. These therapeutic approaches are being designed to either diminish TAM recruitment or reprogram them to be more immunogenic (Chamseddine et al. 2022). CSF-1/CSF-1R pathways are implicated in TAMs recruitment and M2 polarization, and its blockade could boost the efficacy of a wide variety of immunotherapeutic modalities like anti-PD-1/PD-L1 antibodies (Magkouta et al. 2021). In syngeneic murine models, the combination of a CSF-1R inhibitor (BLZ945) and an anti-PD-L1 antibody contributed to a synergistic effect against mesothelioma as compared with single-agent regimens, abrogating TAM infiltration, favoring their polarization toward M1 phenotype, stimulating CD8+ T-cell activation (Magkouta et al. 2021). Moreover, a phase I trial is currently underway combining anti-PD-L1 antibody (durvalumab) with CSF-1R tyrosine kinase inhibitor (pexidartinib) in metastatic pancreatic or colorectal cancers (Breakstone 2021). In different tumor types, TGFB was primarily produced by M2 macrophages, moreover, it promotes M2 polarization and drives the EMT, thereupon causing excellent immunosuppressive functions, motility, and invasion (Chamseddine et al. 2022). SHR-1701 represents a bifunctional fusion protein, which consisted of an anti-PD-L1 monoclonal antibody fused with TGF-β receptor II extracellular domain (a TGF-β ‘trap’). In a clinical-expansion phase I study, SHR-1701 demonstrated superior efficacy among the gastric cancer cohort, yielding an objective response rate of 25.7% and a 12-month OS rate of 54.5%, better than historical data for PD-L1 monotherapy (Liu et al. 2022). In our study, COL12A1 was found to be a negative marker of anti-PD-L1 inhibitor therapy. Correlation analyses between COL12A1 and immune-infiltrating cells in BC demonstrated that COL12A1 expression had the strongest positive correlation with the infiltration of macrophages. Further analyses revealed that COL12A1 was positively related to M2 macrophage infiltration and M2 markers, among which TGFB1 had the strongest positive relationship. COL12A1-mediated M2 macrophage migrates toward BC cells via TGFB1 signaling. These results collectively indicated that COL12A1 promoted macrophage recruitment and M2 polarization, M2 macrophages migration toward BC cells and consequently establish an immunosuppressive environment, which was one of the pivotal causes for anti-PD-L1 therapy resistance in BC. We speculated that COL12A1 blockade might be clinically feasible to overcome the primary hypo-responsiveness of anti-PD-L1 therapy in a broad spectrum of BC patients. Huge efforts have been made to target TAMs employing siRNAs-loaded nanoparticles to reduce their recruitment to the TME and to reprogram them to fight cancers (Shobaki et al. 2020). Nevertheless, the development of TAM-modulating therapies build on nanoparticles still confronts us with tremendous challenges, such as how to achieve the preferential delivery to immunosuppressive M2 macrophages or how to obtain durable and sufficient anti-tumor responses. Fortunately, engineering of novel nanomedicines offers novel opportunities through (1) leveraging nanoparticles decorated with ligands that could identify the specific markers of M2 macrophages to realize target delivery and (2) re-educating TAMs in a durable fashion using carriers with drug-controlled release properties (Xiang et al. 2021). Our study also provides a theoretical grounding for the design of siRNA-loaded nanoparticles with the objective of targeting and manipulating TAMs to boost cancer immunotherapy.
We also investigated the association between COL12A1 and immunomodulators in BC. COL12A1 expression had a positive correlation with immunostimulators (TNFSF4, NT5E, and CD276), yet, a negative correlation with antigen presentation, HLA-DOB in BC. There existed plentiful reports regarding these modules and BC. CD73 (NT5E) enzyme within the tumor and immune cells cause the occurrence of high adenosine concentration in the BC microenvironment, thereupon suppressing the tumoricidal immune responses provoked by anti-PD-1 monoclonal antibody therapy (Liu et al. 2021). CD276 (B7-H3) could potentially contribute to the aggressiveness of BC through the excessive secretion of IL-10, causing a strong immunosuppressive effect (Liu et al. 2013). Our findings provided evidence for a close correlation between COL12A1 and these immune molecules in BC, suggesting the immunosuppressive role of COL12A1.
Important limitations of our study must be considered. Although our study included analyses regarding human samples and in vitro experiments, additional animal models injected COL12A1 knockdown BC cell lines or using COL12A1 knockout mice are demanded to further explore the intrinsic mechanisms of COL12A1 in macrophage behaviors of BC. Besides, whether combining COL12A1 inhibition with anti-PD-L1 therapy could improve its efficacy in vitro and in vivo is required to be validated. Through analyses of BC cells, the TME, and the complex crosstalk between the two, many scholars have elucidated some of the mechanisms underlying anti-PD-L1 resistance. Nevertheless, these findings constitute a small fraction of the potential mechanisms, and deeper knowledge of the complexity of tumoral plasticity and heterogeneity, and the distinct mechanisms of action and resistance of anti-PD-L1 agents is crucial. Multiple combination therapy modalities are available for different mechanisms underlying anti-PD-L1 resistance in BC, and some of these modalities have been proven to be effective in preclinical studies. Unfortunately, the mechanisms of anti-PD-L1 resistance in BC await further investigation and verification in clinical trials. To sum up, better in-depth studies on immune resistance mechanisms of existing ICIs and investigation of more effective targets will be the focus of future works, and more opportunities and challenges may appear.
Conclusion
In conclusion, this is the first report that COL12A1 is associated with durvalumab responsiveness of BC and showed a significant correlation with bad prognosis. Moreover, a COL12A1-based nomogram was developed to assess the survival outcome of patients with BC, which revealed good predictive accuracy. Additionally, COL12A1 expression is related to M2 macrophage infiltration,and immunomodulators, thus shaping an immunosuppressive microenvironment and promoting poor immunotherapy responsiveness. Therefore, our study reveals that COL12A1 potentially serves a critical function in tumor immunity and represents a promising biomarker for prognosis and immunotherapy response prediction of BC patients.
Supplementary materials
This is linked to the online version of the paper at https://doi.org/10.1530/ERC-23-0012.
Declaration of interest
The authors declare that the study was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
Funding
This work was supported by the National Natural Science Foundation of China (grant number 82272659), and the Science and Technology Innovation Program of Hunan Province (grant numbers 2021RC3029, 2022RC1210).
Data availability statement
The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found in the article/Supplementary Material.
Author contribution statement
YY and QL: Acquisition of data. ZX and YY: Analysis and interpretation of data. ZX: Conception and design. YL: Data curation. YL and SZ: Development of methodology. YY and QL: Writing and/or revising the manuscript. All authors contributed to the article and approved the submitted version.
References
Aghamiri S, Zandsalimi F, Raee P, Abdollahifar MA, Tan SC, Low TY, Najafi S, Ashrafizadeh M, Zarrabi A, Ghanbarian H, et al.2021 Antimicrobial peptides as potential therapeutics for breast cancer. Pharmacological Research 171 105777. (https://doi.org/10.1016/j.phrs.2021.105777)
Bachelot T, Filleron T, Bieche I, Arnedos M, Campone M, Dalenc F, Coussy F, Sablin MP, Debled M, Lefeuvre-Plesse C, et al.2021 Durvalumab compared to maintenance chemotherapy in metastatic breast cancer: the randomized phase II SAFIR02-BREAST immuno trial. Nature Medicine 27 250–255. (https://doi.org/10.1038/s41591-020-01189-2)
Barzaman K, Moradi-Kalbolandi S, Hosseinzadeh A, Kazemi MH, Khorramdelazad H, Safari E & & Farahmand L 2021 Breast cancer immunotherapy: current and novel approaches. International Immunopharmacology 98 107886. (https://doi.org/10.1016/j.intimp.2021.107886)
Batlle E & & Massagué J 2019 Transforming growth factor-β signaling in immunity and cancer. Immunity 50 924–940. (https://doi.org/10.1016/j.immuni.2019.03.024)
Breakstone R 2021 Colon cancer and immunotherapy-can we go beyond microsatellite instability? Translational Gastroenterology and Hepatology 6 12. (https://doi.org/10.21037/tgh.2020.03.08)
Cai T, Liu H, Zhang S, Hu J & & Zhang L 2021 Delivery of nanovaccine towards lymphoid organs: recent strategies in enhancing cancer immunotherapy. Journal of Nanobiotechnology 19 389. (https://doi.org/10.1186/s12951-021-01146-2)
Calmeiro J, Carrascal MA, Tavares AR, Ferreira DA, Gomes C, Cruz MT, Falcão A & & Neves BM 2021 Pharmacological combination of nivolumab with dendritic cell vaccines in cancer immunotherapy: an overview. Pharmacological Research 164 105309. (https://doi.org/10.1016/j.phrs.2020.105309)
Chamseddine AN, Assi T, Mir O & & Chouaib S 2022 Modulating tumor-associated macrophages to enhance the efficacy of immune checkpoint inhibitors: a TAM-pting approach. Pharmacology and Therapeutics 231 107986. (https://doi.org/10.1016/j.pharmthera.2021.107986)
Chandrashekar DS, Bashel B, Balasubramanya SAH, Creighton CJ, Ponce-Rodriguez I, Chakravarthi BVSK & & Varambally S 2017 UALCAN: a portal for facilitating tumor subgroup gene expression and survival analyses. Neoplasia 19 649–658. (https://doi.org/10.1016/j.neo.2017.05.002)
Chen M, Linstra R & & van Vugt MATM 2022 Genomic instability, inflammatory signaling and response to cancer immunotherapy. Biochimica et Biophysica Acta. Reviews on Cancer 1877 188661. (https://doi.org/10.1016/j.bbcan.2021.188661)
Chen X, Yan YL, Zeng SS, Gong ZC & & Xu ZJ 2021 Caveolin-1 promotes radioresistance via IRGM-regulated autophagy in lung cancer. Annals of Translational Medicine 9 47. (https://doi.org/10.21037/atm-20-3293)
Clough E & & Barrett T 2016 The gene expression omnibus database. Methods in Molecular Biology 1418 93–110. (https://doi.org/10.1007/978-1-4939-3578-9_5)
Cui G, Wu J, Lin J, Liu W, Chen P, Yu M, Zhou D & & Yao G 2021 Graphene-based nanomaterials for breast cancer treatment: promising therapeutic strategies. Journal of Nanobiotechnology 19 211. (https://doi.org/10.1186/s12951-021-00902-8)
Du W, Frankel TL, Green M & & Zou W 2022 IFNγ signaling integrity in colorectal cancer immunity and immunotherapy. Cellular and Molecular Immunology 19 23–32. (https://doi.org/10.1038/s41423-021-00735-3)
Duan S, Gong B, Wang P, Huang H, Luo L & & Liu F 2018 Novel prognostic biomarkers of gastric cancer based on gene expression microarray: COL12A1, GSTA3, FGA and FGG. Molecular Medicine Reports 18 3727–3736. (https://doi.org/10.3892/mmr.2018.9368)
Friedman J, Hastie T & & Tibshirani R 2010 Regularization paths for generalized linear models via coordinate descent. Journal of Statistical Software 33 1–22. (https://doi.org/10.18637/jss.v033.i01)
Gao J, Liang Y & & Wang L 2022 Shaping polarization of tumor-associated macrophages in cancer immunotherapy. Frontiers in Immunology 13 888713. (https://doi.org/10.3389/fimmu.2022.888713)
Gerecke DR, Olson PF, Koch M, Knoll JH, Taylor R, Hudson DL, Champliaud MF, Olsen BR & & Burgeson RE 1997 Alpha 1(IX). Genomics 41 236–242. (https://doi.org/10.1006/geno.1997.4638)
Györffy B, Lanczky A, Eklund AC, Denkert C, Budczies J, Li Q & & Szallasi Z 2010 An online survival analysis tool to rapidly assess the effect of 22,277 genes on breast cancer prognosis using microarray data of 1,809 patients. Breast Cancer Research and Treatment 123 725–731. (https://doi.org/10.1007/s10549-009-0674-9)
Hadadi E & & Acloque H 2021 Role of circadian rhythm disorders on EMT and tumour-immune interactions in endocrine-related cancers. Endocrine-Related Cancer 28 R67–R80. (https://doi.org/10.1530/ERC-20-0390)
Hicks D, Farsani GT, Laval S, Collins J, Sarkozy A, Martoni E, Shah A, Zou Y, Koch M, Bönnemann CG, et al.2014 Mutations in the collagen XII gene define a new form of extracellular matrix-related myopathy. Human Molecular Genetics 23 2353–2363. (https://doi.org/10.1093/hmg/ddt637)
Isaacs J, Anders C, McArthur H & & Force J 2021 Biomarkers of immune checkpoint blockade response in triple-negative breast cancer. Current Treatment Options in Oncology 22 38. (https://doi.org/10.1007/s11864-021-00833-4)
Januchowski R, Świerczewska M, Sterzyńska K, Wojtowicz K, Nowicki M & & Zabel M 2016 Increased expression of several collagen genes is associated with drug resistance in ovarian cancer cell lines. Journal of Cancer 7 1295–1310. (https://doi.org/10.7150/jca.15371)
Jiang X, Wu M, Xu X, Zhang L, Huang Y, Xu Z, He K, Wang H, Wang H & & Teng L 2019 COL12A1, a novel potential prognostic factor and therapeutic target in gastric cancer. Molecular Medicine Reports 20 3103–3112. (https://doi.org/10.3892/mmr.2019.10548)
Jiang Y, Zhang H, Wang J, Liu Y, Luo T & & Hua H 2022 Targeting extracellular matrix stiffness and mechanotransducers to improve cancer therapy. Journal of Hematology and Oncology 15 34. (https://doi.org/10.1186/s13045-022-01252-0)
Kaboli PJ, Imani S, Jomhori M & & Ling KH 2021 Chemoresistance in breast cancer: PI3K/Akt pathway inhibitors vs the current chemotherapy. American Journal of Cancer Research 11 5155–5183.
Karagiannis GS, Petraki C, Prassas I, Saraon P, Musrap N, Dimitromanolakis A & & Diamandis EP 2012 Proteomic signatures of the desmoplastic invasion front reveal collagen type XII as a marker of myofibroblastic differentiation during colorectal cancer metastasis. Oncotarget 3 267–285. (https://doi.org/10.18632/oncotarget.451)
Li C, Jin W, Zhang D & & Tian S 2021a Clinical significance of microRNA-1180-3p for colorectal cancer and effect of its alteration on cell function. Bioengineered 12 10491–10500. (https://doi.org/10.1080/21655979.2021.1997694)
Li J, Li Z, Xu Y, Huang C & & Shan B 2022 METTL3 facilitates tumor progression by COL12A1/MAPK signaling pathway in esophageal squamous cell carcinoma. Journal of Cancer 13 1972–1984. (https://doi.org/10.7150/jca.66830)
Li T, Fu J, Zeng Z, Cohen D, Li J, Chen Q, Li B & & Liu XS 2020 TIMER2.0 for analysis of tumor-infiltrating immune cells. Nucleic Acids Research 48 W509–W514. (https://doi.org/10.1093/nar/gkaa407)
Li Y, Su Z, Wei B, Qin M & & Liang Z 2021 bBioinformatics analysis identified MMP14 and COL12A1 as immune-related biomarkers associated with pancreatic adenocarcinoma prognosis. Mathematical Biosciences and Engineering 18 5921–5942. (https://doi.org/10.3934/mbe.2021296)
Liu C, Liu J, Wang J, Liu Y, Zhang F, Lin W, Gao A, Sun M, Wang Y & & Sun Y 2013 B7-H3 expression in ductal and lobular breast cancer and its association with IL-10. Molecular Medicine Reports 7 134–138. (https://doi.org/10.3892/mmr.2012.1158)
Liu D, Zhou J, Wang Y, Li M, Jiang H, Liu Y, Yin X, Ge M, Xiang X, Ying J, et al.2022 Bifunctional anti-PD-L1/TGF-βRII agent SHR-1701 in advanced solid tumors: a dose-escalation, dose-expansion, and clinical-expansion phase 1 trial. BMC Medicine 20 408. (https://doi.org/10.1186/s12916-022-02605-9)
Liu S, Li D, Liu J, Wang H, Horecny I, Shen R, Zhang R, Wu H, Hu Q, Zhao P, et al.2021 A novel CD73 inhibitor SHR170008 suppresses adenosine in tumor and enhances anti-tumor activity with PD-1 blockade in a mouse model of breast cancer. OncoTargets and Therapy 14 4561–4574. (https://doi.org/10.2147/OTT.S326178)
Long J, Wang D, Yang X, Wang A, Lin Y, Zheng M, Zhang H, Sang X, Wang H, Hu K, et al.2021 Identification of NOTCH4 mutation as a response biomarker for immune checkpoint inhibitor therapy. BMC Medicine 19 154. (https://doi.org/10.1186/s12916-021-02031-3)
Magkouta SF, Vaitsi PC, Pappas AG, Iliopoulou M, Kosti CN, Psarra K & & Kalomenidis IT 2021 CSF1/CSF1R axis blockade limits mesothelioma and enhances efficiency of anti-PDL1 immunotherapy. Cancers (Basel) 13 2546. (https://doi.org/10.3390/cancers13112546)
Meng Z, Ren D, Zhang K, Zhao J, Jin X & & Wu H 2020 Using ESTIMATE algorithm to establish an 8-mRNA signature prognosis prediction system and identify immunocyte infiltration-related genes in Pancreatic adenocarcinoma. Aging (Albany NY) 12 5048–5070. (https://doi.org/10.18632/aging.102931)
Moujaber T, Balleine RL, Gao B, Madsen I, Harnett PR & & DeFazio A 2021 New therapeutic opportunities for women with low-grade serous ovarian cancer. Endocrine-Related Cancer 29 R1–R16. (https://doi.org/10.1530/ERC-21-0191)
Olness K 1989 Hypnotherapy: a cyberphysiologic strategy in pain management. Pediatric Clinics of North America 36 873–884. (https://doi.org/10.1016/s0031-3955(1636726-8)
Rizzo A, Cusmai A, Acquafredda S, Rinaldi L & & Palmiotti G 2022a Ladiratuzumab vedotin for metastatic triple negative cancer: preliminary results, key challenges, and clinical potential. Expert Opinion on Investigational Drugs 31 495–498. (https://doi.org/10.1080/13543784.2022.2042252)
Rizzo A & & Ricci AD 2022 Biomarkers for breast cancer immunotherapy: PD-L1, TILs, and beyond. Expert Opinion on Investigational Drugs 31 549–555. (https://doi.org/10.1080/13543784.2022.2008354)
Rizzo A, Ricci AD, Lanotte L, Lombardi L, Di Federico A, Brandi G & & Gadaleta-Caldarola G 2022b Immune-based combinations for metastatic triple negative breast cancer in clinical trials: current knowledge and therapeutic prospects. Expert Opinion on Investigational Drugs 31 557–565. (https://doi.org/10.1080/13543784.2022.2009456)
Shobaki N, Sato Y, Suzuki Y, Okabe N & & Harashima H 2020 Manipulating the function of tumor-associated macrophages by siRNA-loaded lipid nanoparticles for cancer immunotherapy. Journal of Controlled Release 325 235–248. (https://doi.org/10.1016/j.jconrel.2020.07.001)
Sung H, Ferlay J, Siegel RL, Laversanne M, Soerjomataram I, Jemal A & & Bray F 2021 Global cancer statistics 2020: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA: A Cancer Journal for Clinicians 71 209–249. (https://doi.org/10.3322/caac.21660)
Tang Z, Kang B, Li C, Chen T & & Zhang Z 2019 GEPIA2: an enhanced web server for large-scale expression profiling and interactive analysis. Nucleic Acids Research 47 W556–W560. (https://doi.org/10.1093/nar/gkz430)
Thul PJ & & Lindskog C 2018 The human protein atlas: a spatial map of the human proteome. Protein Science 27 233–244. (https://doi.org/10.1002/pro.3307)
Tiainen S, Tumelius R, Rilla K, Hämäläinen K, Tammi M, Tammi R, Kosma VM, Oikari S & & Auvinen P 2015 High numbers of macrophages, especially M2-like (CD163-positive), correlate with hyaluronan accumulation and poor outcome in breast cancer. Histopathology 66 873–883. (https://doi.org/10.1111/his.12607)
Verghese ET, Drury R, Green CA, Holliday DL, Lu X, Nash C, Speirs V, Thorne JL, Thygesen HH, Zougman A, et al.2013 MiR-26b is down-regulated in carcinoma-associated fibroblasts from ER-positive breast cancers leading to enhanced cell migration and invasion. Journal of Pathology 231 388–399. (https://doi.org/10.1002/path.4248)
Vidal P 2020 Interferon α in cancer immunoediting: from elimination to escape. Scandinavian Journal of Immunology 91 e12863. (https://doi.org/10.1111/sji.12863)
Viscardi G, Tralongo AC, Massari F, Lambertini M, Mollica V, Rizzo A, Comito F, Di Liello R, Alfieri S, Imbimbo M, et al.2022 Comparative assessment of early mortality risk upon immune checkpoint inhibitors alone or in combination with other agents across solid malignancies: a systematic review and meta-analysis. European Journal of Cancer 177 175–185. (https://doi.org/10.1016/j.ejca.2022.09.031)
Wang Z, Chen L, Ma Y, Li X, Hu A, Wang H, Wang W, Li X, Tian B & & Dong J 2021 Peptide vaccine-conjugated mesoporous carriers synergize with immunogenic cell death and PD-L1 blockade for amplified immunotherapy of metastatic spinal. Journal of Nanobiotechnology 19 243. (https://doi.org/10.1186/s12951-021-00975-5)
Wu H, Wang H, Jiang Z & & Chen Y 2021 Identification of three core secretome genes associated with immune infiltration in high tumor mutation burden across 14 major solid tumors. International Journal of General Medicine 14 6755–6767. (https://doi.org/10.2147/IJGM.S333141)
Wu T, Wang W, Shi G, Hao M, Wang Y, Yao M, Huang Y, Du L, Zhang X, Ye D, et al.2022 Targeting HIC1/TGF-β axis-shaped prostate cancer microenvironment restrains its progression. Cell Death and Disease 13 624. (https://doi.org/10.1038/s41419-022-05086-z)
Wu Y & & Xu Y 2020 Integrated bioinformatics analysis of expression and gene regulation network of COL12A1 in colorectal cancer. Cancer Medicine 9 4743–4755. (https://doi.org/10.1002/cam4.2899)
Xiang X, Wang J, Lu D & & Xu X 2021 Targeting tumor-associated macrophages to synergize tumor immunotherapy. Signal Transduction and Targeted Therapy 6 75. (https://doi.org/10.1038/s41392-021-00484-9)
Xiang Z, Li J, Song S, Wang J, Cai W, Hu W, Ji J, Zhu Z, Zang L, Yan R, et al.2019 A positive feedback between IDO1 metabolite and COL12A1 via MAPK pathway to promote gastric cancer metastasis. Journal of Experimental and Clinical Cancer Research 38 314. (https://doi.org/10.1186/s13046-019-1318-5)
Xu YH, Deng JL, Wang LP, Zhang HB, Tang L, Huang Y, Tang J, Wang SM & & Wang G 2020 Identification of candidate genes associated with breast cancer prognosis. DNA and Cell Biology 39 1205–1227. (https://doi.org/10.1089/dna.2020.5482)
Xu Z, Chen X, Song L, Yuan F & & Yan Y 2022 Matrix remodeling-associated Protein 8 as a novel indicator contributing to glioma immune response by regulating ferroptosis. Frontiers in Immunology 13 834595. (https://doi.org/10.3389/fimmu.2022.834595)
Yan Y, Xu Z, Qian L, Zeng S, Zhou Y, Chen X, Wei J & & Gong Z 2019 Identification of CAV1 and DCN as potential predictive biomarkers for lung adenocarcinoma. American Journal of Physiology. Lung Cellular and Molecular Physiology 316 L630–L643. (https://doi.org/10.1152/ajplung.00364.2018)
Zeng X, Liu C, Yao J, Wan H, Wan G, Li Y & & Chen N 2021 Breast cancer stem cells, heterogeneity, targeting therapies and therapeutic implications. Pharmacological Research 163 105320. (https://doi.org/10.1016/j.phrs.2020.105320)
