BRCA1 Insufficiency Induces a Hypersialylated Acidic Tumor Microenvironment That Promotes Metastasis and Immunotherapy Resistance

Tumor initiation, progression, and metastasis are influenced heavily by the interactions of malignant cells and the tumor microenvironment (TME; ref. 1). Acidity, which is one of the major features of many types of cancers, results from the accumulation of metabolic waste products caused by cancer cells with high metabolic activity and insufficient blood perfusion (2). During progression and metastasis processes, the acidic tumor microenvironment (ATME; pH 5.6–7.0; refs. 3, 4) is believed to be caused by lactate/H⁺ from glycolysis, bicarbonate, and proton from the hydronation of CO₂ (5) and the removal of H⁺ from the cytosol to the interstitial space (6). The acidic microenvironment not only eases cancer cell migration through the extracellular matrix (ECM; ref. 7) toward blood vessels that have a relatively higher pH (7.35–7.45; ref. 8) but also hampers the infiltration of active immune cells into tumor tissues by reducing the activity of T (9) and natural killer cells (10). To date, it is unclear whether acidic tumor-permissive microenvironment (ATPME) conditions can also be caused by different mechanisms in premalignant and tumor tissues. Thus, understating the initiation of the ATPME is critical for the precision prevention of cancers.

Breast cancer occurs at the highest frequency among all cancers with 2.26 million new breast cancer cases and approximately 685,000 patients with cancer died in 2020 (11). Nearly all breast cancers are carcinomas originating from mammary epithelial cells, which line the lobules and terminal ducts as bilayers. The inner layer contains the secretory luminal epithelial cells, and the outer layer is comprised of contractile myoepithelial cells (or basal cells) that are surrounded by the basement membranes (BM). The integrity of the epithelial bilayers is important for maintaining the normal development of the mammary gland and preventing the malignant transformation of the epithelial cells (12–14). In addition, the dissemination of malignant cells to the secondary organs depends on their ability to break through the basal layer and BMs that mainly contain the laminin and collagen IV (13–15) and invade the ECM, followed by intravasation into the circulation, extravasation from blood vessels, and outgrowth at distant organs (16).

Although the majority of breast cancers occur sporadically without a family history, approximately 5%–10% of breast cancers are inheritable (17). Breast cancer–associated gene 1 (BRCA1) is a tumor-suppressor gene, the germline mutation of which causes familial breast cancer (18). BRCA1 has many important functions, including transcriptional regulation, DNA damage repair, centrosome duplication, cell-cycle checkpoint, chromatin remodeling, and protein ubiquitination (19–25). Consistent with its tumor-suppressor function, approximately 25% of mice carrying a mammary tissue–specific disruption of Brca1 mediated by MMTV-Cre (Brca1^Co/Co; MMTV-Cre, or Brca1^MKO) developed mammary tumors by 1.5 years (26) and about 20% of these mice developed metastasis in distant organs. Further studies have revealed that Brca1 deficiency causes numerous abnormalities in mammary gland development characterized by blunted ductal morphogenesis, dysregulation of gene expression genome-wide, the enhanced firing of DNA replication fork (26, 27), and increased apoptosis that can be partially suppressed by the loss of function of P53, ATM, CHK2, and 53PB1 (28–31) or the activation of some oncogenes (32–34). Patients with breast cancer with BRCA1 mutations also experienced a high frequency of metastasis in multiple organs (35). However, it is not clear whether acidic conditions exist in premalignant mammary gland tissues and breast cancer tissues in patients carrying BRCA1 mutation. It is also unknown that acidity if exists, would have an impact on mammary gland integrity and cancer metastasis in those patients.

In this study, the ATPME was detected in premalignant mammary glands and tumor tissues of Brca1^MKO mice and BRCA1ness human breast cancer tissues. The ATPME is contributed by enriched sialic acid ligand polysialic acid (PSA) on mammary epithelial cell surface caused by elevated Vegfa/Il6–TGFβ–St8sia4 signaling. We demonstrated that an increased level of PSA mediated by elevated sialyltransferase St8sia4 induces a metastatic niche and tumor immunosuppressive microenvironment (TIME) that facilitates cancer metastasis. Neutralization of ATPME with sialyltransferase inhibitor (STi), 3FaxP-Neu5Ac, sensitizes the mammary tumors to αPD1 treatment.

MDA-MB-231, HCC1937, and MCF10A cell lines were acquired from the ATCC. The MCF10A cell line containing a heterozygous BRCA1 185delAG mutation and MCF10A wild-type (WT) isogenic cell lines were kindly provided by Dr. Ben Ho Park (The Sidney Kimmel Comprehensive Cancer Center, The Johns Hopkins University School of Medicine, Baltimore, MD; ref. 36). 95 human breast cancer samples in tissue microarray (TMA) were kindly provided by the Third Affiliated Hospital of Sun Yat-sen University. 30 human paraffin breast cancer samples for immunofluorescence staining were kindly provided by the First Affiliated Hospital of Sun Yat-sen University.

For virus production, HEK293T cells were cultured in DMEM supplemented with 10% FBS and 1% penicillin/streptomycin. For the 10-cm-dish, 9 μg psPAX2, 3 μg pMD2.G, and 12 μg transfer plasmid were mixed well in 500 μL serum-free medium, 72 μL of 1 mg/mL PEI was added to another tube with 500 μL serum-free medium (μg DNA: μg PEI is 1:3). The diluted PEI was gently added to the diluted DNA, mixed well and incubated for 15–20 minutes at room temperature. Afterward, the transfection mix was carefully transferred dropwise onto the HEK293T packaging cells to avoid dislodging them. The cells were incubated for 18 hours, and the medium was carefully replaced with a 15 mL complete medium. The supernatant containing viruses was harvested 48 hours post-transfection and centrifuged at 500 × g for 5 minutes to remove the packaging cells. The viral supernatant was filtered through a 0.45-μm PES filter, aliquoted, and stored at −80°C as soon as possible. One to 2 mL fresh viruses (or frozen) were added carefully to the pre-seeded cells in 6-well plates and incubated for at least 48 hours, followed by puromycin selection to obtain a stable expression cell line.

All mouse experiments performed in this study were approved by the University of Macau Animal Ethics Committee. The Brca1-conditional knockout mouse model (Brca1^MKO) has been well-established in our laboratory. To establish a tumor allograft model, constructed Brca1 WT (B477) mammary epithelial cells (2 × 10⁵) and Brca1-MT (G600) mammary epithelial cells (2 × 10⁵) were orthotopically implanted into the right 4th mammary fat pad of nude mice ages 5–6 weeks. To establish a tumor metastasis model, constructed 545 (Brca1 mutant low metastatic), 628W (Brca1 WT high metastatic), and EMT6 cells (5 × 10⁵) were orthotopically injected into the right 4th mammary fat pad of nude and Balb/c mice or directly injected through the tail vein (5 × 10⁴ cells). Tumor volume was measured and calculated according to the formula: V = 0.5 × L ×W², where L represents the longer diameter whereas W represents the shorter diameter. For drug treatment experiments, different drugs were administered when tumor volume reached approximately 50 mm³, 3Fax-P-Neu5Ac (20 mg/kg) was intraperitoneally injected for 7 consecutive days and Stattic (10 mg/kg) was intraperitoneally injected three times a week. For nanoparticle delivery, Stattic (10 mg/kg) and 3Fax-P-Neu5Ac (20 mg/kg) were intravenously injected every three days three times. The ratiometric fluorescent probe was kindly provided by Prof. Xuanjun Zhang from the University of Macau and nanoparticles packed with Stattic and 3Fax-P-Neu5Ac were kindly provided by Prof. Yunlu Dai from the University of Macau. For T-cell depletion, 300 μg anti-CD8a mAb was intraperitoneally administered 2 days before and 1 day after tumor cell injection, and maintained once per week during the whole treatment process.

Five hundred μL of culture medium (10% FBS as a chemoattractant) was added into a 24-well plate, then the Transwell inserts were put into each well. Four hundred μL cell suspension (the serum-free, same number in each group) was added onto the upper chamber and incubated for 24 hours in 37°C and 5% CO₂. The medium was removed and 4% paraformaldehyde was used to fix the cells, and gently washed twice with PBS. Afterward, cells were stained using Crystal violet, and cells in the upper chamber were removed gently by a cotton swab until it turns white. The Transwell inserts were put inversed and images were acquired by Leica M165FC microscope.

Bone marrow, blood, spleen, peritoneal cells, and mammary tissues collected from 10 months of mice and breast tumors collected from Brca1^MKO tumor mice were dissociated in TRizol, total RNAs were subjected to mRNA isolation and library construction followed by RNA sequencing (RNA-seq; Hiseq, pair-end, 6GB raw data per sample). Data were analyzed using HISAT, StringTie, and Ballgown to obtain differentially expressed genes. FPKM values were picked up for each interested gene to generate a matrix and heat maps were drawn for different important genes in different organs by RStudio. The immune cell composition was calculated in the mammary gland and breast tumor by ImmuCC and plotted by heat map. Gene expression analysis of St8sia4 was performed by TNMplot differential gene expression analysis in normal and tumor tissues analyzer (https://tnmplot.com/analysis/). To predict the immune cell abundance correlation with St8sia4 expression, human datasets of BRCA samples were acquired from TISIDB (http://cis.hku.hk/TISIDB/index.php). Data of correlation in either transcription or protein level of two interested genes were from patients with breast cancer in the TCGA database whose plots were obtained from TIMER 2.0 (http://timer.cistrome.org/) and Linkedomics (http://www.linkedomics.org/login.php). Survival curves with high or low expression of genes in patients with breast cancer were obtained using a Kaplan–Meier plotter (https://kmplot.com/analysis/).

All human and mouse paraffin slides were deparaffinized and rehydrated according to standard protocol. The slides were washed with PBS and then heated in R-Buffer-A (10 mL in 90 mL of water). After processing completion of the program (or overnight incubation), the slides were washed with PBS and then treated with 0.5% Triton X-100 and 0.5 mg/mL of sodium borohydride (in PBS) at room temperature for 10 minutes, respectively. Then slides were incubated with blocking solution (50% 3% BSA and 50% Animal-Free Blocker) overnight or at least 1 hour at room temperature and incubated with primary antibodies with different sources at 4°C overnight. Secondary antibodies and DAPI were incubated for 1 hour at room temperature and anti-fade reagents were covered onto the tissues and a coverslip was placed on each slide, images were scanned by Carl Zeiss LSM 880 super-resolution microscopes.

Cells were seeded in a 4-well chamber, washed with PBS 2 times, and fixed with 4% formaldehyde for 15 minutes. Then, the cells were washed with PBS thoroughly and treated with 0.5% Triton X-100 for 10 minutes, followed by incubation with blocking solution for at least 1 hour at room temperature. Primary antibodies were incubated overnight at 4°C and secondary antibodies were incubated for 1 hour at room temperature, images were scanned using Carl Zeiss LSM 880 super-resolution microscopes.

All human and mouse paraffin slides were deparaffinized and rehydrated according to standard protocol. The slides were washed with PBS and then cooked in R-Buffer-A (10 mL in 90 mL of water). After completion of the program (or overnight), slides were washed with PBS followed by incubation with quenching solution (10 mL of 30% H₂O₂ to 90 mL of absolute methanol, mix well) for 20 minutes. Sections were incubated with blocking solution (50% 3% BSA and 50% Animal-Free Blocker, filtered) for at least 1 hour at room temperature and incubated with primary antibody at 4°C overnight. Sections were incubated with Biotinylated goat anti-mouse and rabbit (Abcam) for 10 minutes at room temperature after wash. Then Streptavidin peroxidase was added to incubate for 10 minutes at room temperature. DAB solution (DAB substrate kit peroxidase, Vector) was prepared before use (1 drop reagent 1, 2 drops reagent 2, and 1 drop reagent 3 in 2.5 mL ddH₂O) and added to the sections. Once the color was changed, sections were washed with tap water immediately for 5 minutes and counterstained in hematoxylin for 1 minute and washed with tap water for 5 minutes. Slides then were put into 70%, 95% ethanol for 1 minute, 100% ethanol for 2 minutes, and xylene for 5 minutes each. Slides were placed to dry and DPX mounting was covered onto the tissues and a coverslip was placed on each slide.

Serum from 6-month-old mice and supernatant from different cell lines were diluted using 1× blocking buffer according to the manufacturer's instructions (ab169820, Abcam). Diluted samples were added onto the membranes in a 4-well tray, and were pre-incubated with 1× blocking buffer at room temperature for 30 minutes and incubated overnight at 4°C. The membranes were placed into clean containers and washed with 20 to 30 mL of 1× Wash Buffer I per membrane for 30 to 45 minutes. Then they were returned to the 4-well tray and washed with 2 mL Wash Buffer II twice at room temperature. 2 mL of 1× Biotin-Conjugated Anti-Cytokines was pipetted into each well, and the membranes were incubated overnight at 4°C. 2 mL of 1× HRP-Conjugated Streptavidin was added into each well and membranes were incubated overnight at 4°C. Membranes were all washed and processed according to previous descriptions on the second day. Detection Buffer C and D were mixed in equal volumes (1:1) and covered the membranes in the tray. Following 2 minutes of incubation at room temperature, membranes were exposed immediately, and images were captured using a Bio-Rad ChemiDoc Touch imaging system.

Fresh tissues were collected and put into a holder filled with low-melting agarose gel and waiting for solidification at 4°C. Then they were cut into 250-μm slices using a vibratome, Leica VT1200S (Leica Biosystems) and stained with a ratiometric fluorescent probe (40 μmol/L) for 1 hour at 37°C and then washed with PBS. Images were acquired using a Leica M165FC stereomicroscope. pH values were determined according to the formula: (y = 31.8403–4.4898x, R = 0.9928).

Tissues or cells were lysed with RIPA buffer (150 mmol/L NaCl, 50 mmol/L Tris-HCl, pH 7.4, 10% glycerol, 1% NP-40, 0.5% sodium deoxycholate, 0.1% SDS, and 1 mmol/L EDTA, supplemented with protease inhibitor and phosphatase inhibitor cocktail (Thermo Fisher Scientific). Membranes with different target proteins were incubated with primary antibodies overnight at 4°C and with secondary antibodies for 1 hour at room temperature. Images were acquired by an ODYSSEY CLx system.

RNA was isolated from the fresh mammary gland, tumor, and spleen tissues with TRIzol, and reverse transcription was conducted using the PrimeScript RT Reagent Kit with gDNA Eraser (Takara #RR047Q). Real-time qPCR was performed with SYBR Green ER Master Mix (Roche, 24759100) and data were acquired using a QuantStudio 7 Flex Real-Time PCR System (Thermo Fisher Scientific).

Single cells from mouse MG and tumor tissues were collected and digested with a digestion medium containing 5% FBS, 5 μg/mL insulin (I-1882; Sigma-Aldrich), 500 ng/mL hydrocortisone (H0888; Sigma-Aldrich), 300 U/mL collagenase III (S4M7602S, Worthington), and 100 U/mL hyaluronidase (H3506; Sigma-Aldrich) for 2 hours at 37°C. Samples were then digested in a medium containing 5 mg/mL dispase II (10295825001; Roche Diagnostics) and deoxyribonuclease (58C10349; Worthington) for 5 minutes at 37°C after one wash with HPSS (14170–112; Life Technologies). Red blood cells were depleted by 3 minutes lysis in 1x RBC lysis buffer at room temperature and quenched by HBSS. The suspensions were then filtered through a 40-μm cell strainer and washed with PBS. Cells were resuspended to a density of 2 × 10⁷ cells/mL in a prewarmed serum-free medium, and an equal volume of 10 μmol/L cisplatin working solution was added to the cell suspension (final concentration of cisplatin: 5 mmol/L). Cells were resuspended and incubated at room temperature for 5 minutes. Then, 5× volume of the Maxpar Cell Staining Buffer was added to quench the cisplatin staining, and the cells were centrifuged at 300 × g for 5 minutes. Cells were then washed once with 1 mL of Maxpar Cell Staining Buffer, and 3 million cells were suspended in 80 μL of Maxpar Cell Staining Buffer. The suspended cells were added to 20 μL of Fc-blocking solution. The cells were incubated for 10 minutes at 4°C and resuspended in 50 μL of Maxpar Cell Staining Buffer. Fifty μL of the antibody cocktail was added to each tube, and the cells were incubated for 30 minutes at room temperature following a gentle vortex. The cells were then washed twice with 1 mL of Maxpar Cell Staining Buffer, and 1 mL of a cell intercalation solution was added (Cell-ID Intercalator-Ir diluted with Maxpar Fix and Perm Buffer). The samples were mixed well and left overnight at 4°C. The next day, the cells were washed twice with 1 mL of Maxpar Cell Staining Buffer and 1 mL of Maxpar water. The cell concentration was adjusted to 2.5–5 × 10⁵/mL with positive control bead buffer (EQTM Four Element Calibration, Cat. #201078), and then data were acquired on a CyTOF instrument (Helios, Fluidigm).

Single cells were collected from nude mice with breast tumors, bone marrow, blood, spleen, and peritoneum followed by treatment with RBC lysis buffer and washed with PBS. Cells were resuspended in 100 μL PBS and incubated with the conjugated antibody for 30 minutes on ice. Then cells were washed twice and resuspended in PBS, data were acquired using BD FACSCalibur and analyzed using Flowjo 10.0 software.

Materials used to encapsulate drugs were provided by Prof. Yunlu Dai's laboratory and synthesized according to their protocol. The synthesis of polymerized monomers is based on the reaction between tert-butyl (2-aminoethyl) carbamate and methacryloyl chloride, the subsequent product is purified by column chromatography (eluent: CH2Cl2/ CH3OH, 1/1 v/v). Afterward, according to the reported literature, PEG5000 and 4-Cyano 4-phenylcarbonothioylthio pentanoic acids undergo an esterification reaction to form the chain transfer agent (PEG-CTA). Then reversible addition-fragmentation chain transfer polymerization was employed for the synthesis of PEG-b-NHBoc. When the polymerization reaction is over, using trifluoroacetic acid removes the protective group and reacts with 3,4-dihydroxy benzaldehyde to form the PEG-b-Pho. PEGs and drugs were dissolved in methanol, and the mixture was added to the solution (deionized water/THF) for 5 minutes sonication. Following the removal of the organic phase by vacuum suction, the unencapsulated component was filtered using a 220-nm filtration membrane, the remaining solution was concentrated and stored at 4°C in a dark bottle. The morphology of nanoparticles was investigated by TEM and UV-vis absorption spectra were measured by SHIMADZU VU-1800 spectrophotometer. 200 μL control PEGs labeled with RFP were injected into nude mice bearing tumor through the tail vein and enrichment of nanoparticles in the tumor site was observed using an IVIS Illumina live system. PEGs with drugs were also injected through the tail vein three times once the average tumor size reached approximately 50 mm³, and mice were sacrificed on day 26.

Statistical analysis was performed by GraphPad Prism 8.0 (GraphPad Software). An unpaired Student t test was used to compare differences between two groups and one-way ANOVA was used for comparisons between multiple groups. Correlations analyses were performed by Pearson-Correlation test. Data are presented as the means ± SD from at least three independent experiments. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001 were considered to be statistically significant compared with the control group.

The RNA-seq data generated in this study have been deposited in the National Center for Biotechnology Information Sequence Read Archive database under accession numbers: PRJNA875667, PRJNA876112, and PRJNA876204. Information on antibodies, drugs, primers, sgRNAs, and vectors are provided in Supplementary Tables S8 and S9. Human BRCA sample datasets were acquired from TISIDB (http://cis.hku.hk/TISIDB/index.php) and TCGA data was obtained from TIMER 2.0 (http://timer.cistrome.org/) and Linkedomics (http://www.linkedomics.org/). All other raw data generated in this study are available upon request from the corresponding author.

BRCA1 mutation carriers frequently experience distant lymph node and lung metastasis (35) and multiorgan metastasis was also observed in our Brca1^MKO;mT/mG mouse models (Supplementary Fig. S1A). To study potential interplays involved in tumorigenesis and metastasis, we first used immunofluorescence (IF) staining with markers for luminal epithelial cells (CK18), basal or myoepithelial cells (CK14), and BM (collagen IV) in both human patients with breast cancer without or with BRCA1 mutation and Brca1-MKO mouse model to examine the bilayer mammary structures that are composed of luminal and basal layers and surrounded by the BM. Our analysis revealed the severe loss of the basal layers (Fig. 1A–D), disrupted BMs (Fig. 1E–H), and high levels of vimentin proteins (Supplementary Fig. S1B–S1E) in both tumor-adjacent mammary tissues and tumor tissues of BRCA1 mutation carrier patients compared with those of non-BRCA1 mutation patients (Supplementary Table S1). These observations suggest that tumorigenesis associated with BRCA1 deficiency has a stronger impact than sporadic tumorigenesis on the integrity of mammary structures, and impaired mammary gland structure may pave the path for premalignant or cancer cells to escape from the original tumor initiation site. To investigate this, we next examined the impact of Brca1 deficiency on the integrity of mammary architecture using 8- and 10-months-old WT and Brca1^MKO mice, which are at stages before or onset of the tumor formation. We observed that disrupted basal layers and BMs of the mammary gland were increased in 8-month-old Brca1^MKO mice (MTMG) compared with that in age-matched WT mice (WTMG; Fig. 1I and K; Supplementary Fig. S1F–S1H). These phenotypes were exacerbated by severe loss of basal layers, disrupted BMs, and randomly localized cells with undetermined fate in mammary glands when the mice ages 10-months-old (Fig. 1J and K and Supplementary Fig. S1G and S1H). Consistent with a high level of vimentin level in mammary tissues carrying BRCA1 mutations, the undetermined cell fate was also observed with double positive of CK18 and CK14 mammary epithelial cells in both Brca1^MKO mouse model (Fig. 1L and M) and patients with breast cancer (Fig. 1N and O). These data indicate that the mammary epithelial cells in Brca1-deficient mice suffered abnormality in their architecture and differentiation before and during tumorigenesis.

This notion was further supported by analysis of the gene expression profile of RNA-seq analysis by comparing age-matched WTMG, MTMG, and Brca1 mutant primary tumors. The data revealed that regulators of epithelial–mesenchymal transition (EMT) markers, including Ddr2 and Itga5, and well-known EMT markers, including Fn1, MMP14, Twist1, Snai1, Snai2, Zeb1, and Zeb2, are increased significantly from MTMG to breast tumor tissues, and the markers for mammary epithelial differentiation, including Krt19, Krt18, Krt5, and Krt14, were decreased dramatically in breast tumor tissues of Brca1^MKO mice. These findings indicate that the mammary epithelial cells in Brca1^MKO mice were experiencing atypical fate determination (Supplementary Fig. S1I, Supplementary Table S2).

To identify the factors that might be responsible for impaired mammary gland structure associated with BRCA1 deficiency, we conducted a cytokine array screening by using the serum of WT and Brca1-MT mice at 6 months of age before the tumor formation. We detected 34 differentially presented (20 increased and 14 decreased) cytokines in the blood of Brca1^MKO mice compared with those in WT mice. Of note, the sialic acid-binding receptors, including soluble E-selectin (Sele), L-selectin (Sell), Vcam1, and Madcam1, which may affect tumor progression and metastasis (37), were all dramatically increased in the blood of Brca1^MKO mice (Fig. 1P; Supplementary Fig. S1J, Supplementary Table S3), suggesting that the increased interaction between the sialic acid ligands on the tumor cells with the receptors on the endothelial and immune cells. Sialyltransferases are the enzymes that are responsible for transferring the sialic acid to glycans on the cell membrane (38). Next, we examined the expression of 24 enzymes that are involved in the sialic acid transferring or hydrolysis process in G600 (Brca1-MT) and B477 (Brca1-WT) mammary epithelial cell lines. We found that 8 of them were increased dramatically at mRNA level in G600 cells compared with B477 cells (Fig. 1Q; Supplementary Table S4). Increased mRNA levels of Sele, Sell, and St8sia4 were found in the mammary tissues in Brca1^MKO mice compared with those in Brca1-WT mice (Supplementary Fig. S1K and S1L), and increased level of St8sia4 was also found in Brca1-MT G600 mammary epithelial cells compared with B477 cell line (Supplementary Fig. S1L). Consistent with increases at mRNA level, the protein levels of these genes were also elevated in G600 cells, Brca1^MKO mammary tissues, MCF10A185, and HCC1937 human BRCA1 deficient cell lines by Western blot (Fig. 1R–U) and statistical analysis (Supplementary Fig. S1M and S1N), respectively. TCGA datasets analysis showed that St8sia4 expressed at much higher levels in patients with breast cancer than in normal patients (Fig. 1V), breast cancer with either expression of BRCA1^low (Fig. 1W), and its expression is also positively correlated with SELE level in patients with basal-like breast cancer (Fig. 1X), and in general patients with breast cancer (Fig. 1Y). These data indicate that the elevated sialyltransferases St8sia4 that are required for the synthesis of PSA might be responsible for the impairment of mammary gland structure.

Because the expression of St8sia4 was increased in mammary tissues of Brca1-MKO mice, we examined the St8sia4 mRNA level in B477 cells expressing shBrca1 at different concentrations and overexpression (OE) of mBrca1 in G600 cells to clarify how St8sia4 was regulated by Brca1. The data revealed that the St8sia4 expression was negatively regulated by Brca1 as its mRNA and protein levels are increased in both mouse B477 and human MCF10A cell lines with the expression of shBrca1 (Fig. 2A; Supplementary Fig. S2A and S2B), respectively, but decreased in Brca1-MT G600 cells expressing mBrca1 cDNA at both mRNA and protein levels, respectively (Supplementary Fig. S2C and S2D).

In humans, it is reported that St8sia4 is responsible for the production of PSA (39). PSA is a highly negatively charged polymer molecule (40), which could decrease local pH by charging attraction in the presence of polyanions on the cell membranes (41). We hypothesized that an acidic microenvironment might exist in mammary tissues in Brca1-deficient mice. Thus, we examined the level of PSA in mammary epithelial cells of both mice and humans without or with BRCA1 deficiency by PSA antibody. The data revealed that the PSA level was elevated in both mice (Fig. 2B and C) and humans (Fig. 2D and E) with BRCA1 deficiency, and this was further supported by elevated PSA levels in Brca1^MKO/FGFR2 double mutant mice, but not in mammary tumor tissues of MMTV-cNeu and FGFR2 mouse models (Supplementary Fig. S2E; ref. 42), showing that hypersialylation occurred inside these cells and on the cell membranes. Because negatively charged PSA in/on the cell surface could attract H+ (39, 41), we then examined the pH of the mammary tissues of 8-month-old Brca1^MKO and age-matched WT mice, as well as tumors from Brca1^MKO and WT mice using the ratiometric fluorescent probe with the ratio of red and green intensities (43). The data revealed an average pH value of 6.6 and 6.25 in the mammary tissues and tumors of Brca1^MKO mice, respectively, compared with the pH value of 7.4 in the control mice (Fig. 2F and G).

Although germline mutation of BRCA1 (BRCA1^no) is only detected in about 3%–4% of human breast cancer, roughly 1/3 of breast cancers exhibit a low level of BRCA1 (BRCA1^low), many of which might be insufficient for DNA damage repair and are defined as BRCA1ness (44, 45). To study whether there is a correlation between BRCA1 and PSA, we conducted a TMA analysis of 95 human breast cancer samples and the data showed that 45 of them exhibited a high level of PSA (more than 50% positive signal, Supplementary Table S5) and positively correlated with ST8SIA4 (Supplementary Fig. S2F). Further analysis revealed that BRCA1 expression was negatively correlated with PSA and ST8SIA4 expression (Fig. 2H and I; Supplementary Fig. S2G), whereas such a correlation was not detected between BRCA1 and Lactate dehydrogenase A (LDHA; Supplementary Fig. S2H and SI), as well as PSA and LDHA (Supplementary Fig. S2J), indicating that acidic tumor microenvironment induced by no or low expression of BRCA1 was not caused by overproduction of lactate.

Because the highly negatively charged PSA polymer could attract local H+, we next studied whether increased expression of St8sia4 could drive the dynamic change of pH in local mammary tissues. We implanted B477 control cells with OE-St8sia4 or G600 cells with expression of sgSt8sia4 into mammary fat pat and examined the changes of pH in vivo. The data revealed that tumors with OE-St8sia4 had bigger volumes (Fig. 2J) and lower interstitial tissue pH (Fig. 2K and L) with increased PSA levels compared with the control tumor (Supplementary Fig. S2K and S2L). Conversely, the tumor volume was smaller (Fig. 2M), and the pH of interstitial mammary tissues was increased from pH 6.5 to pH 7.0 (Fig. 2N and O) with decreased PSA in G600 cells expressing sgSt8sia4 (Supplementary Fig. S2K and S2L), showing that elevated St8sia4 could contribute to acidosis in mammary tissues. Further data analysis revealed higher expression of St8sia4 in human patients with breast cancer correlated with poor survival outcomes in patients with breast cancer (Supplementary Fig. S2K).

Altogether, these data demonstrate that besides previously known mechanisms for inducing acidic conditions primarily by metabolic activities of cancer cells, hypersialylation associated with a high level of St8sia4 in mammary tissues could build up an ATPME in premalignant and mammary tumor tissues with deficiency and insufficiency of BRCA1.

Given the observation that the Brca1-mutant mammary tissues have a high level of PSA and low interstitial tissue pH, we hypothesized that an acidic environment might impair mammary gland structure, which might provide an easy path for the tumor cells to migrate out from the original location. To test this hypothesis, we used the EMT6 and 545 cell lines with the overexpressing St8sia4 (OE-St8sia4), which have an intermediate metastatic ability with a low expression level of St8sia4 (Fig. 3A), sgSt8sia4 in the 628W cell line, which is highly metastatic and expresses St8sia4 at a higher level than 545 cells (Supplementary Fig. S3A and S3B). We first implanted both the parental EMT6 (EMT6-Ctr) and EMT6-OE-St8sia4 cells into the mammary fat pad of BALB/c mice. We then compared the profiles for tumor progression and its impact on the integrity of the tumor-adjacent mammary tissues with or without the treatment of 3Fax-P-Neu5Ac (STi), a pan-inhibitor for sialyltransferases. Our data indicated that tumors became visible on day 7 after the implantation of cells, and tumors generated by EMT6-OE-St8sia4 cells were significantly bigger than those in 3Fax-P-Neu5Ac treated and EMT6-ctr groups. Tumors formed by EMT6-OE-St8sia4 maintained the fastest growth at day 26; however, the growth was reversed to about the control level by the STi (Supplementary Fig. S3C). Our analysis of the mammary gland revealed that disrupted basal layers in tumor-adjacent mammary gland were observed on day 14, and much more severe damage of the basal layer was observed on day 26 in mice implanted with EMT6-OE-St8sia4 cells than with EMT6-Ctr cells by co-staining with CK14 and CK18 antibodies, and this impaired mammary gland structure could be recovered after the treatment with STi (Fig. 3B).

After screening several mammary tumor cell lines derived from Brca1-mutant mammary tumors, we found one cell line, 545, which expresses St8sia4 at a low level, and we also overexpressed St8sia4 into it to generate an isogenic pair of cell lines (545-Ctr and 545-OE-St8sia4; Fig. 3A). Using this pair of cells, we observed the same phenotype of impaired tumor adjacent mammary gland at day 26 (Fig. 3C), elevated PSA level, disrupted BMs (Fig. 3D), and EMT status in mammary tissues with OE-St8sia4 (Fig. 3E), which were observed in EMT6-Ctr and EMT-OE-St8sia4 pair of cells. Altogether, these data indicated that hypersialylated malignant cells induced by elevated St8sia4 are responsible for the disruption of mammary architecture.

To investigate the effects of elevated PSA induced by St8sia4 in the processes of cancer metastasis, we hypothesized that displacement of sialic acids on the cell membranes might generate metastatic niches to enhance the dissemination of tumor cells and/or induce cellular morphological changes that facilitate cancer metastasis. To this aim, we first examined the morphology of EMT6-Ctrl and EMT6-OE-St8sia4 cell pair and found that EMT6-OE-St8sia4 cells exhibited mesenchymal-like structure (Supplementary Fig. S3D) and significantly more migrative revealed by the migration assay (Fig. 3F and G). Enhanced migration ability was also observed from the 545-OE-St8sia4 cells compared with 545-Ctrl cells (Fig. 3H and I) and reduced migration ability in MDA-MB-231 cell with the expression of sgST8SIA4 (Supplementary Fig. S3E and S3F). To further investigate whether elevated St8sia4 could affect the behavior of the cells in 3-dimensional culture, we examined the invasiveness of EMT6-Ctr and EMT6-OE-St8sia4 cell pairs and B477-Ctr and B477-OE-St8sia4 cell pairs in Matrigel culture and the data revealed that both EMT6-OE-St8sia4 and B477-OE-St8sia4 cells acquired invasive features (Supplementary Fig. S3G and S3I).

Next, we studied the effects of St8sia4 overexpression in vivo using allograft tumors formed by mammary fat pad implantation experiments. After implanting 545-OE-St8sia4 and 545-Ctr cells into the fat pad of nude mice, we found that all the mice carrying tumors formed by 545-OE-St8sia4 cells developed bigger tumors than control mice and extensive lung metastatic foci, whereas all the mice carrying tumors formed by 545-Ctr cells only had tumors in much smaller sizes without lung metastasis (Fig. 3J–L; Supplementary Fig. S3J). To further investigate this, we identified another cancer cell line, 628W, which is highly metastatic and expresses St8sia4 at a higher level than 545 cells (Fig. 3M) and generated an isogenic pair of 628W-Ctr and 628W-sgSt8sia4 (Fig. 3N). Our data showed that all mice carrying tumors formed by 628W-Ctr cells formed lung metastasis with bigger tumors in size, and sgSt8sia4 in 628W cells blocked the metastasis with smaller tumors in size (Fig. 3O–Q; Supplementary Fig. S3J). Altogether, these data demonstrate that accumulation of PSA induced by elevated St8sia4 could impair the mammary gland architecture, acquire invasiveness for mammary epithelial cells in the primary tumor site, and eases the malignant cells out for metastasis.

To identify the oncogenic factors that are responsible for the upregulation of St8sia4 in Brca1^mut mammary epithelial cells, we paid attention to the high levels of Vegfa and Il6 in the serum of Brca1^MKO mice (Fig. 1P; Supplementary Table S3), which could induce the transcriptional response involved in tumorigenesis, cancer progression, and metastasis (46). Histological analysis of tumor tissues derived from the allograft mouse model initiated with Brca1-MT cells revealed extensive angiogenesis in the tumor tissues by staining with antibodies against CD31 and Endomucin and quantification of double-positive area for these two markers (Supplementary Fig. S4A). This observation was further supported by increased mRNA levels in the luminal subpopulation sorted by Cd24 and Cd29 from GFP-positive cells (Supplementary Fig. S4B and S4C) and protein levels (Supplementary Fig. S4D) of Vegfa and Il6 from premalignant mammary tissues in Brca1^MKO mice, suggesting that Brca1 deficiency might trigger increased expression of these oncogenes in mutant cells before tumor formation. To investigate whether Brca1 could negatively regulate Vegfa and Il6, we examined the expression of Vegfa and Il6 in shBrca1-B477 cells and OE-Brca1-G600 cells. The data revealed that the mRNA level of both Vegfa and Il6 was increased coordinately with the amount of shBrca1 transfected into the B477 WT cells (Fig. 4A) and decreased in Brca1-MT G600 cells with the OE-Brca1 cDNA (Fig. 4B). In support of this notion, higher promotor activities of Vegfa and Il6 were observed in G600 cells compared with B477 cells, and their activities were suppressed when mBrca1 was expressed in both types of cells (Fig. 4C and D), demonstrating that Brca1 negatively regulates the expression of both Vegfa and Il6. Because Vegfa/Il6 and St8sia4 are both negatively regulated by Brca1, we wanted to further investigate whether the Vegfa/Il6 could positively regulate St8sia4. We examined the protein level of St8sia4 in B477 cells expressing shBrca1, shBrca1/sgVegfa, and shBrca1/sgIl6, and in G600 cells with OE-mBrca1, OE-Brca1/OE-Vegfa, and OE-Brca1/OE-Il6. The data showed that the protein level of St8sia4 was increased in B477 cells expressing shBrca1 alone, but the increase was reversed by coexpressing either sgVegfa or sgIl6 (Fig. 4E). On the other hand, the suppression of St8sia4 by OE-Brca1 in G600 cells was partially released by the expression of either OE-Vegfa or OE-Il6 (Fig. 4F), demonstrating that Brca1 negatively regulates St8sia4 through Vegfa/Il6.

Both Vegfa and Il6 are transcriptional regulators involved in angiogenesis and tumor cell invasion (46, 47) and it also has been reported that Vegfa signaling can stimulate TGFβ production in A549 cells (48). In our experiments, we found that the increased protein levels of Vegfa, and Il6 were accompanied by increased TGFβ1, pSmad3, and pStat3 in both premalignant mammary glands (Fig. 4G) and tumor tissues (Fig. 4H) Brca1-MKO mice. To explore whether the TGFβ is involved in hypersialyation in mammary tumor formation, we examined St8sia4 expression at mRNA and protein levels in both epithelial cell lines and the RAW264.7 macrophage cell line upon TGFβ and its inhibitor treatment. The data revealed that TGFβ induced the expression of St8sia4 in mammary epithelial cells (Fig. 4I; Supplementary Table S6), and mRNA and protein levels of St8sia4, Smad2, and vimentin in the macrophage cell line (Fig. 4I; Supplementary Fig. S4E and S4F and Supplementary Table S6). The protein levels of St8sia4, pStat3, and pSmad3 were decreased upon the treatment with the TGFβ inhibitor in both 628W and G600 cell lines (Fig. 4J). Further analysis revealed that OE-Vegfa or OE-Il6 caused increased protein levels of St8sia4, Stat3/pStat3, and TGFβ/pSMAD3, whereas KO-Vegfa or KO-Il6 reduced them (Fig. 4K). Similar results were obtained from EMT6 and MDA-MB-231 cells, respectively (Supplementary Fig. S4G and S4H). These data indicated that TGFβ increases these proteins and OE-Vegfa/Il6 could also increase them and TGFβ; however, it remains unclear whether the induction of OE-Vegfa/Il6 to St8sia4 and Stat3/pStat3 is mediated by TGFβ. To investigate this, we treated the cells with OE-Vegfa or OE-Il6 with a TGFβ inhibitor (LY2109761). The data indicated that the TGFβ inhibitor could completely block the induction of St8sia4 and pStat3/Stat3 by OE-Vegfa/Il6 (Fig. 4L; Supplementary Fig. S4I). Conversely, the levels of these proteins were recovered by TGFβ treatment in cells expressing either sgVegfa or sgIl6 (Fig. 4M; Supplementary Fig. S4J), demonstrating that TGFβ works downstream of Vegfa/Il6 signal and mediates their effects in the regulation of St8sia4 and pStat3/Stat3.

Next, we investigate the effects of TGFβ and its inhibitor LY2109761 on PSA displacement on the cell membrane, accumulation of PSA inside the cell, and distribution of the E-selectin on the endothelial cell membrane. The data revealed that TGFβ could expand the accumulation of PSA (Fig. 4N and O) and E-selectin (Supplementary Fig. S4M and S4N) on the cell membrane, respectively, and this effect could be reduced by LY2109761 treatment, expression of Brca1 cDNA (Fig. 4N and O), or expression of either sgVegfa or sgIl6 (Supplementary Fig. S4K and S4L). In support of our notion, the expression of St8sia4 and TGFβ is positively correlated in patients with breast cancer from the TCGA database (Fig. 4P; Supplementary Table S7) and disruption of either Vegfa or Il6 or both minimized tumor growth (Supplementary Fig. S4O and S4P) and lung metastasis (Supplementary Fig. S4Q and S4R). Collectively, our data demonstrate that oncogenic Vegfa/Il6 signals caused by Brca1 deficiency elevate St8sia4 through TGFβ, leading to the increased accumulation of PSA, and enhancing the ATPME (Fig. 4Q).

Dissemination of cancer cells needs support from the TIME (49, 50) in primary and distant metastatic organs. To further explore whether the hypersialylation triggers TIME, we performed CyTOF analysis with antibodies against CD45, CD11b, Ly6G, Ly6C, CD3, CD4, and CD8 for various immune cell populations from mammary gland tissues in BALB/c mice, mammary tumor tissues implanted with EMT6 parental cell, OE-St8sia4 cell, OE-St8sia4/sgSt8sia4 cells at days 14 and 26 after implantation (Fig. 5A and B). The data revealed that the parental tumors accumulated more PMN-MDSCs (polymorphonuclear myeloid-derived suppressor cells) and M-MDSCs (mononuclear myeloid-derived suppressor cells), and reduced CD4⁺ and CD8⁺ cells than normal MG. Overexpression of St8sia4 in the primary tumors significantly enhanced this phenotype, which was reversed by sgSt8sia4 at days 14 and 26 (Fig. 5A and B).

To further define the status of MDSCs and macrophages in tumor tissues, we carried out IHC and IF staining in the spleen and mammary tumor tissues from EMT6-OE-St8sia4 and 545-OE-St8sia4 allograft mouse models with antibodies of Gr1 (Ly6G and Ly6C) for MDSCs, F4/80 and CD86 for M1-like, F4/80 and CD206 for M2-like macrophages. The data showed that MDSCs were increased in the spleen of Brca1^MKO mice (Supplementary Fig. S5A–S5C). M1-like macrophages were decreased, and M2-like macrophages were increased significantly in tumor tissues from both EMT6-OE-St8sia4 and 545-OE-St8sia4 mouse models (Fig. 5C–E; Supplementary Fig. S5D–S5F). The above phenotype was also observed at different ages and different organs in our Brca1^MKO mice compared with the control mice by flow cytometry analysis (Supplementary Fig. S5G–S5I) because Brca1 negatively regulates St8sia4 expression.

Because St8sia4 was positively regulated by Vegfa/Il6, next, we examined the MDSC populations by IHC with S100a9 antibody and by flow cytometry analysis with Cd11b and Ly6G in spleen and tumor tissues expressing either sgVegfa or sgIl6. The data revealed the decreased MDSCs in tumor tissues expressing either sgVegfa or sgIl6 (Fig. 5F and G), increased MDSCs in the spleen of 545-OE-St8sia4 tumor-bearing mice, and decreased MDSCs in the spleen of 628W tumor-bearing mice expressing sgSt8sia4 (Fig. 5H and I). This notion was further supported by reduced MDSCs in both spleen and mammary tumor tissues expressing either sgVegfa or sgIl6 by flow cytometry analysis in B477 and G600 mouse models (Supplementary Fig. S5J–S5L).

To provide evidence of molecular actions in TIME, we analyzed Bulk RNA-seq from the bone, blood, spleen, and peritoneum of WT mice, Brca1^MKO mice without tumors, and Brca1^MKO mice bearing mammary tumors followed by comparing the expression pattern with that acquired from TCGA database. Consistent with the above notions, B, T, and dendritic cells started to decrease, whereas immune cells from myeloid cell linage started to increase in the blood and spleen of Brca1^MKO mice without tumor, and further increased in Brca1^MKO tumor-bearing mice (Supplementary Fig. S5M). These observations were further illustrated by the gene expression of immune cells (Supplementary Fig. S5N). The featured gene expression patterns in the immune cells exhibited highly expressed immunosuppressive genes, including Tgfβ1, Smad3, Mmp9, Arg1, Arg2, Il10, Il1β, and PD-L1 in the blood, spleen, and peritoneum (Supplementary Fig. S5O), suggesting stronger immunosuppressive environment in those organs caused by Brca1 deficiency. The elevated protein levels of TGFβ signaling by Western blot further supported the above gene expression analysis (Fig. 5J). Furthermore, correlation analysis between the expression of St8sia4 and the immune cell population in the TCGA database showed that St8sia4 expression is positively correlated with the abundance of MDSCs, macrophages, and Treg cells (Fig. 5K–M), and the patients with breast cancer with high expression of St8sia4 and low CD8⁺ T-cell or CD8⁺ central memory T-cell infiltration have poor survival outcome (Fig. 5N; Supplementary Fig. S5P). These data demonstrate that elevated St8sia4 reshapes the tumor microenvironment and supports tumor cell growth and metastasis.

Because ATPME induced by hypersialylation could impair the architecture of the mammary bilayer, which facilitates the malignant cells to move out, next, we selected two drugs, 3Fax-P-Neu5Ac (STi) and Stattic, to test whether neutralization of the ATPME and inhibition of TGFβ signal could inhibit tumor growth and metastasis in our mouse models. 3Fax-P-Neu5Ac is a sialyltransferase pan-inhibitor (STi; refs. 51, 52) that we have tested earlier for inhibition of St8sia4 activity (Fig. 3B), whereas Stattic is an inhibitor that can disrupt TGFβ signaling induced by the activation of pSmad3 (53). We first conducted the mono and combinatory treatments with these two drugs delivered by intraperioneal injection in allografted tumors produced by 628W cells (Fig. 6A). The data revealed that mono treatment with either STi (20 mg/kg) or Stattic (10 mg/kg) inhibited the primary tumor growth (Fig. 6B and C) and GFP metastatic signals in the lung (Supplementary Fig. S6A and S6B) to some extent with the reduced lung weight (Supplementary Fig. S6C), and the combinatory treatment with these two drugs further reduced tumor outgrowth (Fig. 6B and C) and the metastatic signal in the lungs (Supplementary Fig. S6A and S6B) with reduced lung weight (Supplementary Fig. S6C), showing that neutralization of the acidic environment and TGFβ signal inhibition could synergistically retard the primary tumor growth and block the metastasis with reversed splenomegaly phenotype (Supplementary Fig. S6D and S6E). No obvious cytotoxic effects in the lung, kidney, liver, and spleen were observed by hematoxylin and eosin (H&E) staining (Supplementary Fig. S6F).

To further elucidate the action of these two drugs on mammary gland architecture in malignant mammary tissues, we examined tumor and tumor-adjacent mammary glands treated with STi, Stattic, or both drugs together with antibodies against CK18/collagen IV or CK18/CK14, respectively. The data revealed that much more collagen bundles (above 90%) were formed in tumor tissues detected by CK18/collagen IV staining that may prevent the escape of tumor cells from the original location compared with the parental control tumor tissues (Fig. 6D and E) in mono or combinatory drug treatment, and more mammary gland ducts with intact basal cell layers (above 90%) were observed after treatment with either single or two drugs together by double staining of CK14 and CK18 (Fig. 6F and G), demonstrating that effects of the acidic condition in mammary tissues caused by hypersialylation could be overcome by these two drugs.

It was reported that although an effective dose of 3Fax-P-Neu5Ac can inhibit myeloma tumor growth, at the same time, it also induces edema in the peritoneal cavity of mice and toxicity in the liver and kidney (54, 55). To avoid the potential side effects of these anticancer drugs, we packed drugs into nanoparticles with amphiphilic polymer and delivered them through intravenous injection every 3 days three times when the average tumor size reached approximately 50 mm³ after implantation. We designed a new treatment plan in which the two drugs were delivered via nanoparticles (Fig. 6H). The data showed that the nanoparticles could be enriched in mammary tissues after 48 hours via the tail vein injection (Fig. 6I). The primary tumor growth and GFP metastatic signals were significantly reduced by nanoparticles packed with either Stattic or STi, respectively (Fig. 6J, K, L, and M).

Finally, we measured tumor tissue pH with the ratiometric fluorescent probe in the same cohort of mice and the ratio of green color versus red color was much bigger in controls (PBS and PEG treatment groups) compared with the drug treatment group (Fig. 6N) and the pH in mammary tissues was recovered from approximately 6.5 to 6.8 to above 7.0 after treatment with Stattic or STi (Fig. 6O). These findings demonstrate that neutralization of APTME by STi and inhibition of TGFβ signaling via nanoparticles could significantly reduce tumor outgrowth and minimize lung metastasis caused by hypersialylation in mammary tissues.

Immunotherapy resistance is a very challenging problem during the treatment of patients with cancer (56). Our data showed that elevated expression of St8sia4 reduces effector T cells in mammary tissues, we then hypothesized that ATPME induced by St8sia4 might contribute to the resistance to αPD1 treatment. To test the hypothesis, we examined expression levels of PD-L1 and PD1 in mammary tissues, tumors, and spleen from BALB/c mice implanted with EMT6-Ctr, EMT6-OE-St8sia4, and EMT6-OE-St8sia4/sgSt8sia4 (knockout of St8sia4) cells. We found that increased PD-L1 from the mice with implantation of EMT6-Ctr or EMT6-OE-St8sia4 cells and the increased PD-L1 cell population could be suppressed in tumors formed by EMT6-OE-St8sia4/sgSt8sia4 cells by IF and flow cytometry analysis (Fig. 7A and B; Supplementary Fig. S7A). Next, we examined the expression of PD1 and CD3 by IF staining. We detected increased PD1-positive cells in the tumors formed by EMT6-OE-St8sia4 cells compared with that by EMT6-Ctr cells and this increase can be reversed in mice implanted EMT6 cells expressing St8sia4/sgSt8sia4 (Supplementary Fig. S7B and S7C). To find the status of the PD1 expression cells, we examined the expression of PD1 and CD3 on the immune cells in the spleen and the data indicated that PD1⁺ cells and the percentage of PD1⁺/CD3⁺ cells were all increased significantly in tumor-bearing mice formed by implantation of EMT6 cells, and such increases were further enhanced in EMT6-OE-St8sia4 cells (Fig. 7C–E). These data suggest that the increased PSA induced by St8sia4 could not only change the status of epithelial cells but also shape the immunosuppressive tumor microenvironment by affecting functional T cells.

Because PSA is a highly negatively charged polymer molecule, we hypothesized that removing the PSA from the tumor cell surface and neutralizing the local pH might sensitize the tumor to αPD1 treatment. Next, we treated the tumor-bearing mice overexpressing St8sia4 with αPD1, STi NPs, and STi NPs in combination with αPD1 (Fig. 7F). The data revealed that OE St8sia4 enhanced the tumor growth, and treatment of αPD1 did not change the tumor outgrowth at all; however, tumor outgrowth was reduced significantly by STi treatment, and the tumor growth was further suppressed by combinatory treatment of αPD1 and STi compared with treatment with STi only (Fig. 7G and H). The splenomegaly phenotype and metastasis phenotype in the lung was also reversed significantly in STi and combination treatment groups, respectively (Supplementary Fig. S7D, S7E, S7H, and S7I). To further examine the population of MDSCs, CD4⁺, and CD8⁺ T cells before and after treatment in tumor tissues with OE-St8sia4, we performed CyTOF analysis (Fig. 7I and J). In support notion above, the data revealed that (i) OE-St8sia4 decreased CD4⁺ and CD8⁺ T-cell populations and increased PMN-MDSCs and M-MDSCs; (ii) the immune populations had no significant change with αPD1 mono-treatment compared with the control groups; (iii) the CD4⁺ and CD8⁺ T-cell populations were recovered and the populations of PMN-MDSCs and M-MDSCs were reduced by mono-treatment STi, and (iv) The CD4⁺ and CD8⁺ T-cell populations kept increase and PMN-MDSCs and M-MDSCs populations decreased much more in the group with the combinatory treatment of αPD1 and STi than the mono-treatment group with either αPD1 or STi (Fig. 7I and J), demonstrating that STi greatly facilities the killing effect of αPD1 on the tumor cells.

Consistent with CyTOF analysis, the CD3-positive cells, and caspase-3 protein level were increased (Supplementary Fig. S7J; Fig. 7K and L), and the PSA level was decreased significantly in both STi and STi plus αPD1 treatment groups (Fig. 7M and N), showing that the pan-inhibitor of sialyltransferases (STi) could eliminate PSA induced by OE-St8sia4. To further investigate whether the synergy effect of STi and αPD1 is mediated by T cells, we depleted CD8 cells in mice implanted with EMT6-St8sia4 cells and with a combination treatment of STi and αPD1. The data revealed that combination treatment reduced tumor volume and splenomegaly, but the tumor outgrowth and splenomegaly come back again if we depleted CD8 T cells in Balb/c mice (Fig. 7O–Q; Supplementary Fig. S7F–S7G), showing that T-cell indeed mediated the synergy effect of STi and αPD1. The H&E staining also showed recovered lung morphology from metastatic lungs in both STi and STi plus αPD1 treatment groups and no obvious toxic effects were observed in the liver and spleen before and after the treatment (Supplementary Fig. S7K). Altogether, these data demonstrate that ATPME caused by OE-St8sia4 is associated with immunotherapy resistance, and removing PSA by STi sensitizes this type of tumor to αPD1 treatment.

By studying premalignant mammary tissues and tumors of both Brca1^MKO mice and human breast cancer samples, we have made several findings. (i) BRCA1-deficiency and BRCA1-low induce ATPME characterized by hypersialylation mediated by St8sia4 in premalignant mammary tissues and most human BRCA1ness breast cancers; (ii) The hypersialylation in mammary tissues is directed by the activation of Vegfa/Il6 mediated by TGFβ signals; (iii) The ATPME creates malignant niches that facilitate mammary tumor formation and metastasis; and (iv) Inhibition of ATPME blocks breast tumor metastasis in experimental models with hypersialylated mammary tissues.

It is generally believed that acidic pockets in tumor tissues are built up by the accumulation of overproduced acidic metabolic wastes by H⁺ mainly from fermentative glycolysis, CO₂, lack of oxygen, or hypoxia (57, 58). Cancer cells adapt to the acidic environment for survival and proliferation via changes of enhanced glycolysis, increased activities of the transporter system on the cell surface, and dysregulated glutamine and lipid metabolism (2). Mapping of acidic regions with pH low insertion peptide (pHLIP), it was shown that invasive fronts at the tumor–stroma interface are acidic in vivo at the cellular level and acidity modulates the splicing of specific types of RNA-binding proteins (59). Affected by acidity, the melanoma cancer cells display more aggressive invasive behavior when they are acclimated to low pH growth conditions in vitro (60) and experiments have also revealed that cells at mildly acidic pH culture conditions become more invasive (61). An increase of peritumoral pH through oral administration of sodium bicarbonate could reduce tumor growth and local invasion in a mouse model (62). Although the majority of the above studies focused on the acidity of the interstitial space TME contributed by metabolic acidic waste using in vitro experiments, emerging evidence suggests that PSA may also play important role in breast cancer formation and progression. It was shown that enhanced expression of PSA is correlated with malignant phenotype in breast cancer cell lines and clinical tissue samples (63). The ST8SIA4 expression level was significantly higher in breast tumor tissues compared with adjacent tissues, and at a higher level in highly metastatic MDA-MB-231 cells than in MCF-7 cells with lower ability of metastasis (64), and knocking down ST8SIA4 in MDA-MB-231 cells significantly inhibited their malignant behaviors, including cell proliferation and invasion. Further analysis identified ST8SIA4 as one of the miR-26a/26b-targeted genes (64). Of note, higher expression of ST8SIA4 in tumor cells was associated with poor patient outcomes whereas ST8SIA4 expression in infiltrating stromal cells was associated with good patient outcomes, thus the prognostic significance of ST8Sia4 expression seems dependent on the cell type within the tumor (65).

Our study provides a significant advance in further understanding the potential role of ST8SIA4/PSA in breast cancer from a new angle. We demonstrated that ST8SIA4 is negatively regulated by BRCA1 and the hypersialylation on mammary epithelial cell surface induced by accumulation and overproduction of PSA sialic acid ligand in response to elevated sialyltransferase St8sia4 in both human and mouse mammary tissues harboring BRCA1 mutations and breast cancers with “no” or “low” levels of BRCA1. In human breast cancers that contain low or no expression of BRCA1 or BRCA2 are referred to BRCAness, which is later extended to cancers that are defective in BRCA functions, including DNA homologous recombination repair, replication fork protection, and synthetic lethality with PARP inhibitors (44, 45). It is worthwhile to note that although the germline mutations of BRCA1 are only found in a small fraction (<4%) of breast cancers, up to 25% of cancers exhibit hypermethylation in the BRCA1 promoter, leading to its low expression (45). Thus, some of these BRCA1-low expression cancers might not be counted as BRCA1ness cancers, which are mainly classified by lacking several major functions of the BRCA (44). We demonstrated that hypersialylation in mammary tissues builds up the ATPME (pH 6.6) before tumor formation in Brca1^MKO mice and the acidity impaired mammary gland integrity, promoting tumor metastasis. Surprisingly, hypersialylation, as reflected by increased expression of PSA caused by overexpression of sialyltransferase is quite common in breast cancers, occurring in about 50% of breast cancers, and the level of PSA is negatively correlated with BRCA1^low but not with their lactate/H⁺ levels. We further demonstrated that the persistent acidic condition in BRCA1-deficient mammary tissues may be one of the major causes of impaired mammary gland structure, leading to metastasis.

The metastasis process includes many steps, including the invasion of tumor cells from their primary sites into ECM, intravasation into the blood vessel, traveling and survival from the blood vessel, extravasation out from the blood vessel, and outgrowth at distant organs (66). It has been reported that acidity induces increased expression of EMT markers and invasiveness of melanoma cells in vitro but not in vivo (58). The factors responsible for ECM degradation, such as MMPs and cathepsins, are activated under acidic conditions (58) and facilitate the degradation of the ECM (67). But very little is known about how microenvironmental acidity affects the specific steps of metastasis processes in vivo (2). Although the acidity caused by BRCA1 insufficiency is different from previously known mechanisms, the low pH might generate similar impacts on tumor growth and cancer metastasis. Here, we show that the increased St8sia4 in the premalignant mammary gland creates acidic metastatic niches and activates oncogenic signaling, which impairs mammary bilayer structure and induces EMT of BRCA1-MT cells, facilitating the circulation of a tumor cell to the distant lung. Accumulation of PSA, which presents as stacks of sialic acids generated by St8sia4 onto membrane glycans (68), impairs the integrity of the basal myoepithelial layer and BM, facilitating the loss of the cell–cell interactions (69). Thus, our studies advance the understanding of how acidity could enhance tumorigenesis and benefit metastasis in vivo.

The oncogenic actions of Vegfa and IL6 lead to phosphorylation of Stat3, and pStat3, which, in turn, upregulates the oncogenic expressions of Vegfa/Il6 and MMP14 (70–72), leading to the aberrant activation of the endothelium. Under the acidic condition, secreted MMPs and cathepsins from malignant cells and endothelial cells are activated, resulting in the degradation of ECM and release the growth factors such as TGFβ (73, 74). The loss of function of Brca1, which is a transcriptional regulator (19), dysregulates expressions of many genes, especially in the fast proliferation cells (26). In this study, we further illustrated that Brca1 negatively regulates Vegfa/Il6, which, in turn, elevates St8sia4, leading to an accumulation of sialic acid ligands on the epithelial cell surface. In addition, the oncogenic actions of Vegfa/Il6 can be mediated by TGFβ, which can enhance the accumulation of the sialic ligand on mammary epithelial cell membranes. Because the binding of pStat3 to Smad3 enhances the TGFβ action (53), we selected 3FaxP-Neu5Ac and Stattic to target acidity and TGFβ signaling, and the data showed it not only minimized primary tumor growth but also diminished metastatic signals in the lung. These data significantly advance our understanding of the contribution of sialic acid ligands to ATPME and provide a selective strategy for precision prevention and tailored medicine for patients with breast cancer.

Immunotherapy resistance is commonly observed in virtually all cancers, especially in solid cancers (56). Our data indicated that the tumor growth in our animal model was markedly blocked by the combinatory treatment of αPD1 and 3Fax-P-Neu5Ac. These data suggest that the neutralization of ATPME plays a critical role in enhancing the efficiency of αPD1 blockade in suppressing mammary tumor growth. It is known that multiple factors might be involved in the efficacy of immune checkpoint blockade therapy mediated by PD-1/PD-L1, including levels of PD-1/PD-L1 expression, various cellular activities, tumor microenvironment, tumor immunogenicity, etc. (56, 75). Our data indicated that OE-St8sia4 generated profound changes in the mammary epithelial bilayer structure by inducing an acidic environment with an accumulation of PSA. These changes might not only yield a strong impact on TIME but also hinder the effectiveness of αPD1. Indeed, we found that OE-St8sia4 increased M-MDSCs and PMN-MDSCs populations, decreased CD4⁺ and CD8⁺ cell populations as well as increased expression levels of PD-L1 and PD1 in cancer cells and immune cells, respectively, all of which should promote tumor growth. Thus, we believe that the treatment of 3Fax-P-Neu5Ac to suppress St8sia4 could not only reduce the expression of PD-L1 and PD1 but also decrease the population of MDSCs and increase the population of the effector T cells. Furthermore, the neutralization of ATPME might also help the binding of antibodies to PD1 to release the ICB in tumor-bearing mice. The combinatory treatment of 3Fax-P-Neu5Ac and αPD1 reduces tumor growth much more than the single treatment.

In sum, our study, for the first time, revealed that hypersialylation occurs in a high fraction of human breast cancers, with a preference for BRCA1ness cancers. The correlation is not 100%, as it is subjected to a complex regulation with BRCA1 as the main regulator, together with some other, yet to be identified, factors. We demonstrated how BRCA1 insufficiency induces an acidic condition characterized by hypersialylation mediated by St8sia4/PSA with the activation of Vegfa/Il6 and enhanced by TGFβ signals. The ATPME impairs the mammary epithelial bilayer structure and allows the formation of malignant niches that facilitate mammary tumor formation and metastasis. Using the combinatory treatment with STi, 3Fax-Peracetyl Neu5Ac together with Stattic or 3Fax-Peracetyl Neu5Ac with αPD1, we developed strategies to inhibit the formation of ATPME, which minimizes primary tumor growth and lung metastasis.

No disclosures were reported.

X. Shu: Conceptualization, software, formal analysis, validation, investigation, methodology, writing–review and editing. J. Li: Resources, investigation. U.I. Chan: Methodology. S.M. Su: Methodology. C. Shi: Resources, investigation. X. Zhang: Methodology. T. An: Resources, methodology. J. Xu: Methodology. L. Mo: Investigation. J. Liu: Investigation. Y. Wang: Investigation. X. Li: Investigation. M. Deng: Formal analysis, methodology. J. Lei: Resources, methodology. C. Wang: Resources, methodology. H. Tian: Resources, methodology. S. Heng: Methodology. J.S. Shim: Resources, methodology. X. Zhang: Resources, methodology. Y. Dai: Resources, methodology. Z. Yao: Resources. X. Kuang: Resources. Y. Lin: Resources. C.-X. Deng: Conceptualization, supervision, funding acquisition, writing–review and editing. X. Xu: Conceptualization, resources, supervision, funding acquisition, writing–original draft, writing–review and editing.

This work was supported by a multi-year research grant (MYRG2016–00138-FHS, MYRG2017–0008-FHS, MYRG2019–0064-FHS) by the University of Macau, Macau SAR, China; The Science and Technology Development Fund (FDCT) grants (027/2015/A1, 029/2017/A1, 0101/2018/A3, 0011/2019/AKP, 0045/2021/AFJ, 0097/2021/A2, 0004–2021-AKP and 0092/2020/AMJ) of Macau SAR, China, and National Natural Science Foundation of China key grant (82030094). The authors thank the members of the Xu laboratory for helpful advice and discussion; the Animal Research Core for providing the animal housing, the Proteomics, Metabolomics, and Drug Development Core, and the Bioimaging and Stem Cell Core.

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.