BC patients (N = 67) were recruited in the Harbin Medical University Cancer Hospital from 2016 to 2020, and the cancerous and non-cancerous tissues were collected by surgery, were judged by two experienced pathologists in our hospital, and were immediately stored at -80°C conditions for further utilization. According to their therapy treatment history, the BC patients were divided into two groups with (N = 34, considered as PR-BC patients) or without (N = 33, considered as PS-BC patients) paclitaxel treatment history. Specifically, the patients who had paclitaxel treatment history were considered as PR-BC patients, whereas the patients did not receive any chemotherapy were deemed as PS-BC patients. The clinical characteristics of the involved patients were listed in Table 1 . The clinical experiments were approved by the Ethics Committee Affiliated to Harbin Medical University Cancer Hospital, and the informed consent forms were also signed by all the participates.

The human TNBC cell lines (MDA-MB-453 and MCF7) were purchased from BeNa Culture Collection (Beijing, China), All cells were regularly authenticated using short tandem repeat (STR) DNA profiling and tested to confirm that they were mycoplasma contamination-free. The above cells were maintained in the DMEM medium (Gibco, USA) containing 10% fetal bovine serum (FBS, Gibco) under the conditions with 5% CO₂ atmosphere at 37°C. In addition, according to the experimental protocols recorded in our previous work (42), the paclitaxel-resistant MDA-MB-453/PTX and MCF7/PTX cell lines was established by treating the parental paclitaxel-sensitive BC cells with continuous paclitaxel treatment in a step-wise manner. The PR-BC cells were maintained in the DMEM medium (Gibco, USA) containing 10% fetal bovine serum, and the PR-BC cells at passage 2 to 4 with paclitaxel-resistant properties were used in this study. Moreover, the vectors for gp96 overexpression/downregulation and p53 overexpression were synthesized by Sangon Biotech (Shanghai, China), which were delivered into the TNBC cells by using the Lipofectamine 2000 transfection reagent (Invitrogen, USA). Moreover, the PS-BC cells were exposed to the hypoxic environment (0.2% oxygen, 5% CO₂, and 37°C) for three days to establish the hypoxic PS-BC cell models as previously described (43).

The EXO Quick-Exosome Isolation kit (Biosciences System, USA) was purchased to obtain the exosomes from patients’ plasma and cells’ supernatants in keeping with the manufacturer’s protocols. Specifically, for the plasma exosomes, the patient blood was collected and were subjected to centrifugation (3000g, 15min) to acquire the supernatants, which were regarded as plasma. Then, 300 μl of the plasma were mixed with the exosome precipitation solution of the kit, and the plasma exosomes were further collected. For the cell-derived exosomes, the BC cells were cultured in vitro and their supernatants were collected through sequential centrifugation, including 300g for 10 min, 2000g for 15min, 12,000g for 30 min to remove the cells and debris, which were further filtered by 0.22μm filters (Millipore, USA). The filtered supernatants were mixed with the working solution of the kit, and the exosomes were isolated and purified according to the kit’s instructions.

To observe the morphologies of the exosomes, we performed the TEM assay based on the previous protocols. The purified exosomes were fixed with 1% glutaraldehyde for 10 min at room temperature, and the exosomes were further placed onto a formvar carbon-coated 300-mesh copper electron microscopy grids (Agar Scientific Ltd., USA) for 5 min, and the exosomes were stained with 2% uranyl oxalate for 2 min. The grids were washed with PBS and a TEM machine (Jeol, Japan) was employed to observe and photograph the exosomes.

The CD8⁺ T cells were isolated and cultured according to the previous literature (35). Briefly, the volunteers were recruited and the blood was obtained, which were further subjected to the Ficoll kit (Sigma-Aldrich, USA) to obtain the peripheral mononuclear cells (PBMCs) in keeping with the reagent’s protocols. A flow cytometer (BD Bioscience, USA) was further used to acquire the CD8⁺ T subgroup from the PBMCs as previously described (44).

Total RNA was extracted from BC tissues and cells through using the TRIzol reagent (Beyotime, Shanghai, China), and further Real-Time qPCR was performed to quantify genes expressions. Specifically, the cDNA synthesis kit (ThermoFisher Scientific, CA, USA) was used to reversely transcribed the mRNA into complementary DNA (cDNA), through using the Bio-Rad Real-Time qPCR detection system (Bio-Rad, CA, USA), we performed Real-Time qPCR to determine the mRNA levels of HSP gp96, p53, IL-1β, IL-18, ZEB1, Vimentin, Twist, Slug and Snail, and the relevant expression levels of the above genes were normalized by the GAPDH. The primers were designed and constructed by Sangon Biotech (Shanghai, China), and their detailed sequences were shown in Table 2 .

RIPA lysis buffer (Biotime, Hangzhou, China) was used to obtain total proteins from BC tissues and cells, and the quality of the protein was determined by BCA kit (Pierce, USA). Proteins were separated by 10% SDS-PAGE, and were subsequentially transferred onto the PVDF membranes (Millipore, MA, USA). The membranes were blocked by 5% non-fat milk, and incubated with the primary antibodies against gp96 (1:1000, #ab227293, Abcam, UK), TSG101 (1:1000, #ab125011, Abcam, UK), p53 (1:2000, #ab1101, Abcam, UK), GAPDH (1:2000, #ab8245, Abcam, UK), CD63 (1:1500, #ab271286, Abcam, UK) and β-actin (1:2000, #ab179467, Abcam, UK) at 4°C overnight. In the second day, the membranes were probed with the secondary antibodies for 1h at room temperature, and an ECL system (Bio-Rad, CA, USA) was employed to visualize the protein bands. The expression levels of the proteins were normalized by GAPDH.

MTT assay was performed to evaluate cell viability as previously reported. In brief, the BC cells were cultured in the 96-well plates in the standard culture conditions, which were further added with MTT reaction solution at the density of 20 μl per well for 4h. Then, the supernatants were carefully discarded, and the formazan was diluted by adding 150 μl DMSO. The plates were fully vortexed, and a microplate reader (Molecular Devices, CA, USA) was employed to determine cell viability at the wavelength of 450 nm.

The cytokine secretion (IL-1β and IL-18) in BC cells’ supernatants were determined by performing ELISA through using their corresponding ELISA kit purchased from RAPIDBIO (CA, USA). The cell supernatants were collected and incubated with the reaction solution, and a microplate reader (Molecular Devices, CA, USA) was further employed to measure the optical density (OD) values at the wavelength of 450 nm, and the relative expression levels of the two cytokines were calculated according to the OD values.

The BC cells with differential treatments and vector transfection were sequentially stained with Annexin V-FITC and PI in the commercial apoptosis detection kit (Solarbio, Beijing, China), and a FCM (BD Bioscience, USA) was employed to examine the apoptotic cell ratio, which were defined as Annexin V-FITC or PI-positive cells. The experiments were conducted in accordance with the manufacturer’s instruction.

The EdU kit (Solarbio, Beijing, China) was purchased to examine cell proliferation in keeping with the manufacturer’s protocol. Specifically, the CD8⁺ T cells with or without gp96-containing exosomes co-incubation were incubated with EdU reaction solution for 2h, and the cells were further washed by PBS buffer for three times, fixed by 4% paraformaldehyde, neutralized by glycine, and reacted with fluorescence dyes. Then, the cells were stained with DAPI to visualize their nucleus, and a fluorescence microscope (Olympus, Japan) was employed to observe and photograph the EdU-positive cells, which were regarded as CD8⁺ T cells with good proliferative abilities.

The BC cells were cultured in the 25 cm³ flasks and were fixed by incubating cells with 4% paraformaldehyde for 20 min at 37°C. Then, the fixed cells were washed by PBS buffer for three times to fully remove the residual paraformaldehyde, and were permeabilized by using the PBST buffer (PBS buffer with 0.05% Triton X-100 supplementation). Then, the cells were sequentially incubated with the primary antibodies (Gasdermin D, 1:200) and secondary antibodies, and the expression levels and cellular localization of the proteins were examined by performing a confocal microscope (Zeiss, Germany).

The MCF7 cells were subcutaneously injected into the dorsal flank of the six-week-old female C57BL/6J mice, and the cell concentration for injection was 1 × 10⁶ cells per mouse, and tumor volume was monitored every 3 days. Until the tumor volume reached about 100 mm³, the gp96-containing or deficient exosomes were injected into the mice tumors, and the tumor volume were further monitored for 6 days with 2 days intervals. The mice were divided into three groups with 6 mice each group, including negative control (NC), gp-96 containing exosomes, and gp-96-deficient exosomes. Then, the mice were anesthetized and sacrificed, and the tumors were obtained and weighed. Our animal experiments were conducted in accordance with the guidelines of the Ethics Committee affiliated to Harbin Medical University Cancer Hospital.

SPSS 18.0 software was employed to analyze the data in the present study, which had been shown as Means ± Standard Deviation (SD), and the graphs were generated by using the GraphPad Prism 8.0 software. One-way ANOVA analyzed the differences among multiple groups (>2), while the differences in two groups were compared by performing the student’s t-test. Individual experiment was repeated at least 3 times. *P < 0.05, **P < 0.01 and ***P < 0.001 were considered as statistical significance.

For the clinical experiments, the cancerous and adjacent non-cancerous tissues were collected from BC patients (N = 67), and Real-Time qPCR results showed that the mRNA levels of gp96 were significantly increased in BC tissues comparing to the corresponding normal tissues (Fold changes: 2.12, Figure 1A ). Also, the correlation of gp96 expressions with patients’ clinical features were analyzed, and our findings suggested that BC patients with higher clinical stage (III-IV), positive lymph node metastasis and higher tumor size (≥ 5cm) showed significant high-expressed gp96, but its expression status had nothing to do with patients’ age ( Table 1 ). Interestingly, the BC patients were divided into two groups with (N = 34, considered as PR-BC patients) or without (N = 33, considered as PS-BC patients) paclitaxel treatment history, and we surprisingly found that gp96 mRNA levels were apparently increased in the cancer tissues collected from BC patients with paclitaxel-resistant properties in contrast with the PS-BC patients (Fold changes: 1.92, Figure 1B ). In addition, the plasma exosomes were isolated and purified from the above BC patients’ blood samples, the morphology of the exosomes was photographed by performing the TEM assay ( Figure S1A ), and further Western Blot analysis verified that the exosome markers, including CD63 and TSG101, were expressed, but the cellular indicator β-actin was absent in the isolated exosomes ( Figure S1B ), implying that the quality of the exosomes were guaranteed. Data in Figures 1C, D showed that gp96 protein was upregulated in the exosomes isolated from PR-BC patients comparing to the PS-BC patients (Fold changes: 4.83). For the cellular experiments, the parental PS-BC cells (MDA-MB-453 and MCF7) were subjected to continuous low-dose paclitaxel exposure to induce the descent PR-BC cells (MDA-MB-453/PTX and MCF7/PTX) as previously reported, and the experimental procedures were shown in Figure 1E . The MTT assay results indicated that PR-BC cells were much more resistant to paclitaxel treatment, in contrast with the PS-BC cells ( Figures 1F–I ). Of note, comparing to the PS-BC cells, we found that low-dose paclitaxel increased gp96 expression levels in PR-BC cells at both mRNA and protein levels, as it indicated by Real-Time qPCR ( Figure 1J ) and Western Blot analysis ( Figures 1K, L ) results. Consistently, the exosomes were isolated from BC cells’ supernatants ( Figures S1A, B ), and our data supported that gp96 protein tended to be enriched in the PR-BC-exo but not in the PS-BC-exo (Fold changes: 2.83, Figures 1M, N ). The above data suggested that intracellular and extracellular gp96 was closely associated with BC malignancy and drug-resistance.

The following experiments were conducted to determine the role of gp96 in regulating paclitaxel-resistance in BC cells, and to achieve this, we transfected the gp96 overexpression vectors into PS-BC cells ( Figure S2A ), and the gp96 downregulation vectors were delivered into the PR-BC cells ( Figure S2B ), respectively. The vector transfection efficiency was determined by Real-Time qPCR analysis ( Figures S2A, B ). Data in Figures 2A–C showed that upregulation of gp96 rescued cell viability and decreased the ratio of Annexin V-FITC or PI-positive apoptotic cells in the PS-BC cells subjecting to high-dose paclitaxel stimulation, as determined by MTT assay ( Figures 2A, B ) and FCM assay ( Figure 2C and Figures S3A, B ), implying that gp96 overexpression increased paclitaxel-resistance in PS-BC cells. Conversely, we evidenced that knockdown of gp96 suppressed cell viability ( Figures 2D, E ) and promoted cell apoptosis ( Figure 2F , Figures S3C, D ) in paclitaxel-treated PR-BC cells, hinting that silence of gp96 increased paclitaxel-sensitivity in PR-BC cells. The potential underlying mechanisms were also preliminarily uncovered, and previous literatures report that p53 is the hub gene for regulating paclitaxel chemo-resistance (45, 46), and p53 can be targeted and degraded by gp96 (47), which encouraged us to assume that gp96 might regulate BC cell death through modulating p53. To investigate the effects of gp96 on p53, we conducted Real-Time qPCR and Western Blot analysis, and our results hinted that gp96 merely suppressed p53 at protein levels (Fold changes: 0.63, Figures 2H, I ) but not in the mRNA levels (Fold changes: 1.03, Figure 2G ) in PS-BC cells. Thus, p53 was overexpressed in PS-BC cells ( Figures S2C–E ), and the functional experiments evidenced that upregulated p53 enhanced the inhibiting effects of paclitaxel on cell viability in the PS-BC cells ( Figure 2J, K ). Also, p53 overexpression abrogated the regulating effects of gp96 overexpression on paclitaxel-resistance in PS-BC cells ( Figures 2J, K ), suggesting that intracellular gp96 controlled paclitaxel-resistance in BC cells in a p53-dependent manner.

Cancer cells derived exosomes contribute to chemo-resistance in multiple cancers, including BC (6, 7), and we next investigated whether PR-BC-exo affected paclitaxel-resistance in PS-BC cells. To explore this issue, the exosomes secreted by PR-BC cells with or without gp96-deficiency were incubated with PS-BC cells, and our data showed that PKH67-labelled exosomes translocated into the PS-BC cells ( Figure 3A ). The following Western Blot analysis evidenced that gp96 protein was successfully delivered into the PS-BC cells by PR-BC-exo (Fold changes: 3.62 and 4.32, Figures 3B, C ), and as expected, Real-Time qPCR analysis evidenced that PR-BC-exo did not affect gp96 mRNA levels in the PS-BC cells ( Figure 3D ). Interestingly, our MTT assay data showed that PR-BC-exo with gp96 expressions was capable of increasing cell viability in paclitaxel-treated PS-BC cells ( Figures 3E, F ). Consistent with this, the FCM assay results supported that paclitaxel-induced apoptotic cell death in PS-BC cells were also reversed by co-incubating cells with PR-BC-exo ( Figure 3G and Figures S4A, B ). However, the PR-BC-exo with gp96-deficiency did not influence paclitaxel-resistance in PS-BC cells ( Figures 3E–G and Figures S4A, B ), suggesting that PR-BC-exo transmitted gp96 to increase paclitaxel-resistance in PS-BC cells. The in vivo experiments supported that PR-BC-exo also reversed the suppressing effects of paclitaxel on the tumorigenesis of MCF7 cells in tumor-bearing mice models ( Figures 3H, I and Figures S4C, D ). Those data suggested that PR-BC cells increased paclitaxel-resistance of its surrounding PS-BC cells via delivering gp96-containing exosomes.

As previously reported, gp96 participates in the regulation of immune evasion and surveillance during cancer progression (9), and CD8⁺ T is considered as the main immune cells to eliminate malignant cells by recognizing tumor-associated antigens in tumor cells (35, 36). However, malignant cancer cells were able to shed or hide their antigens, and a subgroup of cancer cells communicate with CD8⁺ T cells by secreting information molecules, resulting in the death of those immune cells and facilitate immune evasion (35, 36). Thus, we considered exosomal gp96 as the communicating molecule which might mediate CD8⁺ T cell death. Hence, the pBMSCs-derived CD8⁺ T cells were isolated and were co-cultured with PR-BC-exo with or without gp96 deficiency. The EdU assay ( Figures 4A, B ) and MTT assay ( Figure 4C ) were performed, and the results showed that PR-BC-exo with gp96 expression significantly suppressed cell viability and proliferation in the CD8+ T cells, while PR-BC-exo with gp96 deficiency had little effects on the above cellular functions ( Figures 4A–C ). In addition, we noticed that PR-BC-exo induced upregulation of cellular cytokines, including IL-1β and IL-18, in both CD8⁺ T cells (Fold changes: 3.45 and 4.32, Figure 5A ) and its supernatants (Fold changes: 5.22 and 4.71, Figure 5B ), as determined by Real-Time qPCR and ELISA. Since IL-1β and IL-18 are classical biomarkers for pyroptotic cell death, and gp96 has been proved to activate NLRP3 inflammasome in murine APCs (48), we then performed the following experiments to investigate whether exosomal gp96 triggered pyroptotic cell death in CD8⁺ T cells. As shown in Figure 5C , the immunofluorescent staining assay results suggested that PR-BC-exo promoted Gasdermin D expressions in CD8⁺ T cells through delivering gp96. The above results suggested that PR-BC-exo delivered gp96 to restrain cell viability in CD8⁺ T cells, which was associated with cell pyroptosis.

Hypoxia is common phenomena in solid tumors when tumor volume reaches about 2 mm, and hypoxic pressure brings significant changes to the cellular functions of both tumor and stromal cells in the tumor microenvironment, which further facilitates angiogenesis and cancer metastasis (13–15). Of note, recent data indicate that hypoxia increases gp96 expressions through upregulating HIF-1 (9), which were verified in our study. Specifically, the PS-BC cells were subjected to HIF-1 knockdown vectors ( Figure S6 ) and hypoxic treatments, and the cancer cells and their exosomes in the supernatants were collected. By performing Real-Time qPCR ( Figure 6A ) and Western Blot analysis ( Figures 6B, C ), we verified that hypoxia promoted gp96 generation (Fold changes: 3.22 and 4.01) and secretion (Fold changes: 2.34 and 2.91) in the PS-BC cells and supernatants, which were reversed by silencing HIF-1, suggesting that hypoxia promoted intracellular and extracellular gp96 expressions through inducing HIF-1 expressions. Then, we explored the paclitaxel-resistant characteristics of the PS-BC cells under normal or hypoxic environment ( Figures 6D–H and Figures S5A, B ). To achieve this, the mentioned PS-BC cells were subjected to high-dose paclitaxel treatment, and the MTT assay supported that paclitaxel especially suppressed cell viability in PS-BC cells without hypoxic stimulation ( Figures 6D, E ). Also, the FCM results evidenced that hypoxic PS-BC cells were much more resistant to paclitaxel-induced cell apoptosis than the normal PS-BC cells ( Figures 6F, G and Figures S5A, B ), suggesting that hypoxia increased paclitaxel-resistant properties in PS-BC cells. Moreover, we profiled the EMT associated makers in the PS-BC cells, and the Real-Time qPCR analysis results validated that the mRNA levels of the mesenchymal biomarkers, including ZEB1, vimentin, twist, slug, snail and fibronectin, were all elevated by hypoxic stimulations in PS-BC cells ( Figures 6I, J ). Taken together those data, we concluded that Hypoxia affected gp96 expressions and secretion in BC cells via inducing HIF-1, and hypoxic condition was capable of increasing paclitaxel-resistance and promoting EMT process in the PS-BC cells.