Public datasets retrieval

The Cancer Genome Atlas (TCGA) data: The pan-cancer normalized RNA-seq datasets, copy number variant (CNV) data processed by GISTIC algorithm, 450 K methylation data, mutation profiles, and clinical information were obtained from UCSC Xena data portal (https://xenabrowser.net/datapages/). The somatic mutation data were obtained from the TCGA (http://cancergenome.nih.gov/) database and then used to calculate the tumor mutation burden (TMB) by R package “maftools”. The abbreviations for TCGA cancer types were shown in Table S1.

The METABRIC data: The normalized RNA-seq dataset, CNV data processed by GISTIC algorithm, mutation profiles, and clinical data in the METABRIC cohort were downloaded from the cBioPortal data portal (http://www.cbioportal.org/datasets) [13]. The sample size for each dataset was shown in Table S2.

Pan-cancer analysis of the correlation between ACE2 and immunological features

To evaluate the pan-cancer immunological correlation of ACE2, we first collected expression levels of 122 immunomodulators including major histocompatibility complex (MHC), receptors, chemokines, and immunostimulators from the study of Charoentong et al.  [14]. Then, the correlations between ACE2 and immune checkpoints were also assessed. The TISIDB [15] tool was used to estimate the abundance of immune cells infiltration, and the correlations between ACE2 and the infiltration levels of immune cells were next evaluated.

Linked Omics database analysis

The Linked Omics database (http://www.linkedomics.org/login.php) is a web-based tool to analyze multi-dimensional datasets [16]. The functional roles of ACE2 in BC were predicted using the Linked Omics tool in term of Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways by the gene set enrichment analysis (GSEA). Default options were used for all parameters.

Assessment of the immunological features in TME of BC

The immunological features of TME in BC contained immunomodulators, the activities of the cancer immunity cycle, infiltration levels of TIICs, and the expression of inhibitory immune checkpoints.

The expressions of 122 immunomodulators were first included in this part. Considering the cancer immunity cycle which contains seven stages reflects the anti-cancer immune response and the activities of each step decide the fate of tumor cells, we subsequently calculated the activation scores of each step by the single-sample gene set enrichment analysis (ssGSEA) according to the expression level of specific signatures of each step [17]. The ssGSEA algorithm, implemented by extending GSEA method, allows the definition of an enrichment score that represents the absolute enrichment of gene sets in each sample within a given dataset. The gene expression profile of the given dataset was sorted and normalized firstly. Then, the enrichment scores were assessed based on the empirical cumulative distribution function of signatures in the gene sets and remaining genes. Given of the biological and technical batch effects among datasets came from different researches and the algorithm schedule, ssGSEA algorithm was applied to calculate the enrichment scores of samples for each dataset respectively, rather than for samples in the integrated dataset. Meanwhile, all bioinformatic analysis was performed for each dataset, respectively. Moreover, in order to avoid calculation errors resulting from various algorithms which were developed to explore the relative abundance of TIICs in TME, we comprehensively estimated the infiltration levels of TIICs using five independent algorithms: TIMER [18], EPIC [19], MCP-counter [20], quanTIseq [21] and TISIDB [15]. The ESTIMATE algorithm was also performed to calculate Tumor Purity, ESTIMATE Score, Immune Score and Stromal Score [22]. Furthermore, we also collected several well-known effector genes of TIICs, and computed the T cell inflamed score according to the linear combination of the expression levels and weighting coefficient of 18 genes reported by Ayer et al.  [23].

To verify the role of ACE2 in mediating cancer immunity in BC, we grouped the patients into the high ACE2 and the low ACE2 group with the 50% cutoff based on the median expression levels of ACE2, and then analyzed the difference of the immunological features of TME concerning the above aspects between the high and the low ACE2 groups.

Immunophenoscore analysis

As previously reported, a patient’s immunophenoscore (IPS) can be calculated without bias using machine learning by consideration of the 4 major categories of components that measure immunogenicity: effector cells, immunosuppressive cells, MHC molecules, and immunomodulators [14]. The IPS values of BC patients were obtained from the Cancer Immunome Atlas (TCIA) (https://tcia.at/home).

Calculation of the enrichment scores of various gene signatures

According to previous research [24], we collected several gene-sets positively associated with therapeutic responses to immunotherapy, targeted therapy and radiotherapy and specific markers of biological processes correlated with anti-tumor immunity. The enrichment scores of these signatures were obtained using the “GSVA” R package [25]. Detailed information on immunotherapy-related gene signatures was shown in Table S3.

Prediction of therapeutic response

The role of ACE2 in predicting the response to chemotherapy was also evaluated. First, BC-related drug-target genes were screened by using the Drugbank database (https://go.drugbank.com/). The R package “pRRophetic” where the samples’ half-maximal inhibitory concentration (IC50) was calculated by ridge regression were utilized to predict the response to anti-cancer therapy for patients in different cohots. The statistical models were firstly built based on the gene expression and drug sensitivity data in a considerably large panel of cancer cell lines obtained from the Cancer Genome Project (CGP) database (https://www.sciencedirect.com/topics/neuroscience/cancer-genome-project). Then, these models were combinated the gene expression profiles from tumor biopsies of different cohorts to predict the clinical drug response, respectively. Meanwhile, the tenfold cross validation on the test set was performed to estimate the accuracy of the phenotype prediction.Default options were used for all parameters [26].

Clinical samples

Two tissue microarrays (TMAs, HBreD050Bc01 and HBreD090Bc03) were obtained from Outdo Biotech (Shanghai, China). The TMAs were embedded in paraffin, and the thickness of the TMAs was 4 μm. The HBreD050Bc01 microarray contained 40 BC and 10 adjacent samples. The HBreD090Bc03 microarray contained 85 BC and 5 adjacent samples. Thus, a total of 125 BC samples and 15 adjacent samples were involved in the current research. Ethical approval (YBM-0502, 2020–03) for this part was granted by the Clinical Research Ethics Committee, Outdo Biotech.

In addition, 30 plasma samples from BC patients receiving the Docetaxel-based neoadjuvant chemotherapy were obtained before and after they received chemotherapy. The detailed information for these patients was shown in Table S4. These BC patients were recruited by the First Affiliated Hospital with Nanjing Medical University. After 8 cycles of neoadjuvant chemotherapy, they received a surgical operation. The response to neoadjuvant chemotherapy was evaluated according to the Miller-Payne criterion. Ethical approval (NJMU-2020–93, 2020–03) for this part was granted by the Clinical Research Ethics Committee, Nanjing Medical University.

Immunohistochemistry and semi-quantitative evaluation

Next, Immunohistochemistry (IHC) staining was conducted on these tissue slides. The primary antibodies used in the research were as follows: anti-ACE2 (1:3000 dilution, Cat. ab15348, Abcam, Cambridge, UK), anti-CD8 (Ready-to-use, Cat. PA067, Abcarta, Suzhou, China), and anti-PD-L1 (Ready-to-use, Cat. GT2280, GeneTech, Shanghai, China). Antibody staining was visualized with DAB and hematoxylin counterstain, and stained sections were scanned using Aperio Digital Pathology Slide Scanner. All stained sections were independently evaluated by two independent pathologists. For semi-quantitative evaluation of ACE2 and PD-L1 staining, the immunoreactivity score (IRS) was applied as previously described [27]. For CD8 staining, infiltration level was assessed by estimating the percentage of cells with strong intensity of membrane staining in the stroma cells. In addition, tumors were demarcated into three phenotypes based on the spatial distribution of CD8+ T cells, including the inflamed, the excluded, and the deserted phenotypes [28].

Enzyme-linked immunosorbent assay (ELISA)

The levels of plasma Ang-1–7 in BC patients were detected by the ELISA assay. The ELISA kit for Ang-1–7 (Cat. E-EL-H5518) was obtained from Elabscience (Wuhan, China). All samples were assayed in duplicates according to the manufacturer’s protocol, and the average values were reported as pg/mL. The association between plasma Ang-1–7 levels and the response to neoadjuvant chemotherapy was next assessed.

Cell culture and plasmid transfection

MDA-MB-231 (RRID: CVCL_0062) and 4T1 (RRID: CVCL_0125) cell lines were purchased from KeyGEN BioTECH (Nanjing, China). MDA-MB-231 cells were maintained in Leibovitz’s L-15 medium (KeyGEN BioTECH, Nanjing, China) supplemented with 10% fetal bovine serum (FBS) (Hyclone, Thermo Scientific, Waltham, MA) at 37 °C with 5% CO2. 4T1 cells were maintained in DMEM medium (KeyGEN BioTECH, Nanjing, China) supplemented with 10% FBS at 37 °C with 5% CO2. All experiments were performed with mycoplasma-free cells. MDA-MB-231 cell lines have been authenticated using short tandem repeat profiling. The ACE2 overexpression plasmid were synthesized by KeyGEN BioTECH (Nanjing, China). The cloning vector for ACE2 overexpression plasmid was pcDNA3.1( +)-EGFP. For subsequent assays, MDA-MB-231 cells were transfected with ACE2 plasmid (2.5 μg) using Lipofectamine 3000 Reagent (Invitrogen) according to the manufacturer’s instructions.

Quantitative real‑time PCR (qRT-PCR)

Total RNA of BC cells was extracted using TRIzol reagent (Invitrogen). The primers for ACE2, PD-L1 and GAPDH mRNA reverse transcription were synthesized in KeyGEN BioTECH (Nanjing, China). qRT-PCR was conducted using the One Step TB Green™ PrimeScript™ RT-PCR Kit II (SYBR Green, TaKaRa). Primers used for gene amplification as following: ACE2: 5’ GCTCTTCCTGGCTCCTTCTCAG 3’ (forward), 5’ AGGTCTTCGGCTTCGTGGTTAA 3’ (reverse); PD-L1: 5’ GCCGAAGTCATCTGGACAAGC 3’ (forward), 5’ TGATTCTCAGTGTGCTGGTCAC 3’ (reverse); GAPDH: 5’ AGATCATCAGCAATGCCTCCT 3’ (forward), 5’ TGAGTCCTTCCACGATACCAA 3’ (reverse). The 2−ΔΔCt method was used for mRNA expression analysis.

Western blotting analysis

Cells were plated in 35-mm dishes (6 × 105 cells/dish). 48 h after transfection, all cells were harvested the proteins with lysis buffer. SDS–polyacrylamide gel electrophoresis and Western blotting analysis were performed as standard protocols. For nuclear and cytoplasmic protein extraction, nuclear and cytoplasmic Protein Extraction Kit were used (KeyGEN BioTECH, Nanjing, China). The primary antibodies for ACE2 (1:1000 dilution, Cat. ab15348, Abcam), PD-L1 (1:1000 dilution, Cat. 17,952–1-AP, ProteinTech) and Tubulin (1:2000 dilution, Cat. 10,094–1-AP, ProteinTech) were used. Expression levels of proteins were normalized to Tubulin for each sample.

Cell Counting Kit-8 (CCK-8) assay

In general, BC cells were grouped into three groups: blank control group, negative control group (transfected with control plasmid) and ACE2 overexpression group. Twenty-four hours after transfection, cells were digested using 0.25% trypsin for 1 min and resuspended. Suspended cells were seeded on a 96-well plate with the cell density adjusted to 2 × 104 cells/ml (100 μl/well) and fostered at 37 °C in a constant-temperature incubator with 5% CO2 for 24, 48, 72 and 96 h, respectively. To each well, 10 μl CCK-8 was added, after which the plate was put in the incubator for 2 h. The OD value of each well was measured at 450 nm by a microplate reader. Each experiment was repeated three times.

Mouse BC model and pharmacodynamic evaluation

For subcutaneous tumor models, female BALB/c mice purchased from Shanghai SLAC Laboratory Animal Co.,Ltd were used in in vivo analysis. The mice were housed and maintained in laminar flow cabinets under specific pathogen-free conditions. All experiments were approved by the Laboratory Animal Ethics Committee at Nanjing Medical University (IACUC-1902016, 2019–02). To establish BC tumor xenografts, 4T1 mouse tumor cells were subcutaneously injected (1 × 107 cells) into the flanks of 5- to 6-week-old female mice (~ 20 g on average). Tumors were monitored and regularly measured with calipers every two to three days. When tumors reached about 100 mm3 in volume, mice were randomized into six different groups: vehicle group, Ang-1–7 group, Docetaxel group, Ang-1–7 + Docetaxel group, anti-PD-1 and Ang-1–7 + anti-PD-1 group. Ang-1–7 (Cat. HY-12403, MedChemExpress) and Docetaxel (Cat. D107320, Aladdin) were dissolved in PBS. Ang-1–7 was administered to mice through subcutaneous injection at 1 mg/kg daily. Docetaxel was administered to mice through intraperitoneal injection at 10 mg/kg three times a week. Anti-PD-1 antibody (BioXCell, Cat. BE0273) was dissolved in PBS, and 200 μg was administered intraperitoneally three times a week. The vehicle solution consisted of PBS only. Treatment was continued until tumors reached 20 mm along the long axis or until 25 days after treatment initiation.

To evaluate the infiltrating levels of immune cells in control and Ang-1–7 treated tumors, IHC staining was used. Tumors were fixed in 4% paraformaldehyde for 20 min and permeabilized with 0.5% Triton X-100 in PBS for 10 min. Tumor sections were blocked with BSA 5% for 30 min. The sections were incubated with primary antibodies against CD8 (1:100 dilution, Cat. A11856, ABclonal, Wuhan, China) and CD3 (1:1000 dilution, Cat. 17,617–1-AP, ProteinTech, Wuhan, China) at 4 °C overnight and the corresponding secondary antibody. After washing, DAPI was used for the nuclei staining. Images were scanned using OlyVIA Slide Scanner.

To detect the expression of effective and suppressive cytokines in control and Ang-1–7 treated tumors, ELISA test was used. The ELISA kit for IFN-γ (Cat. E-EL-M0048), CXCL9 (Cat. E-EL-M0020), CXCL10 (Cat. E-EL-M0021), TGF-β1 (Cat. E-EL-0162), IL-10 (Cat. E-EL-M0046) were obtained from Elabscience (Wuhan, China). The values were standardized by protein concentration of each sample and were reported as pg/mg or ng/mg.