Breast cancer bone metastases are common and incurable. Tumoral integrin β3 (β3) expression is induced through interaction with the bone microenvironment. Although β3 is known to promote bone colonization, its functional role during therapy of established bone metastases is not known. We found increased numbers of β3+ tumor cells in murine bone metastases after docetaxel chemotherapy. β3+ tumor cells were present in 97% of post-neoadjuvant chemotherapy triple-negative breast cancer patient samples (n = 38). High tumoral β3 expression was associated with worse outcomes in both pre- and postchemotherapy triple-negative breast cancer groups. Genetic deletion of tumoral β3 had minimal effect in vitro, but significantly enhanced in vivo docetaxel activity, particularly in the bone. Rescue experiments confirmed that this effect required intact β3 signaling. Ultrastructural, transcriptomic, and functional analyses revealed an alternative metabolic response to chemotherapy in β3-expressing cells characterized by enhanced oxygen consumption, reactive oxygen species generation, and protein production. We identified mTORC1 as a candidate for therapeutic targeting of this β3-mediated, chemotherapy-induced metabolic response. mTORC1 inhibition in combination with docetaxel synergistically attenuated murine bone metastases. Furthermore, micelle nanoparticle delivery of mTORC1 inhibitor to cells expressing activated αvβ3 integrins enhanced docetaxel efficacy in bone metastases. Taken together, we show that β3 integrin induction by the bone microenvironment promotes resistance to chemotherapy through an altered metabolic response that can be defused by combination with αvβ3-targeted mTORC1 inhibitor nanotherapy. Our work demonstrates the importance of the metastatic microenvironment when designing treatments and presents new, bone-specific strategies for enhancing chemotherapeutic efficacy.
Bone metastases remain a significant, unmet challenge in the treatment of breast cancer. The majority of patients with metastatic breast cancer will develop bone involvement (1), with predominantly osteolytic lesions accompanied by refractory pain, increased fracture risk, and decreased survival (2). Bone-targeted therapies such as bisphosphonates and the anti-RANKL mAb denosumab have substantially improved quality of life, reducing fracture incidence and impeding bone metastatic progression. Unfortunately, these agents are associated with a survival benefit in only a subset of patients (3), and resistance to chemotherapy and radiation is common (4).
Interaction between the tumor microenvironment and cancer cells has been recognized as an important mechanism driving chemoresistance (5, 6), confounding studies that focus on in vitro treatment data. The bone represents a distinct metastatic niche, comprised of unique cell types, extracellular matrix (ECM) components, and soluble factors compared with visceral metastatic sites. Moreover, the progression from single, disseminated tumor cells on a quiescent bone surface to floridly outgrowing osteolytic lesions is a highly dynamic process, with the importance of individual microenvironmental factors likely varying over time (4, 7). Some critical factors have been identified at different stages of bone metastatic progression, but more targets are needed to enhance the efficacy of available therapies against clinically detectable lesions.
Integrins are heterodimeric transmembrane receptors that bind ligand moieties in the ECM, initiating signaling events with broad consequences for cell survival, proliferation, and migration (8). Integrin β3 (β3, as part of αvβ3 and αIIbβ3 heterodimers) can be a marker of tumor aggressiveness and is expressed on cells in the bone tumor microenvironment, including activated endothelium, osteoclasts, platelets, and immune cells (8–13). We recently showed that β3 is upregulated on breast cancer cells through TGFβ signaling in the bone microenvironment and can be exploited for bone-specific nanoparticle drug delivery (14). β3 has been identified as an important factor for bone colonization by breast cancer cells (9, 15), and has also been shown to promote resistance to EGFR inhibition across multiple cancer types (16). Although studies have previously linked β3 signaling and chemotherapy resistance (17), its role in vivo, and particularly during therapy of established bone metastases, is poorly characterized.
In this study, we provide evidence for β3 as an important promoter of resistance to taxane chemotherapy in breast cancer bone metastases. We show that β3 expression is associated with an alternative metabolic response to taxanes in vitro and in vivo, and that β3-mediated resistance can be defused by combination therapy with mTORC1 inhibitors. Our work demonstrates the importance of the metastatic microenvironment when designing treatments and presents new, bone-specific strategies for enhancing chemotherapeutic efficacy.
Materials and Methods
All animal studies were performed according to Washington University Institutional Animal Care and Use Committee (WU IACUC, protocol No. 20190104) guidelines. Female C57BL/6J (Jax, RRID:IMSR_JAX:000664) and BALB/c (Jax, RRID:IMSR_JAX:000651) mice were obtained from The Jackson Laboratory and injected at 6–7 weeks of age. All mice were housed under pathogen-free conditions according to the WU IACUC.
Cell lines and constructs
The C57BL/6 background PyMT-BO1-GFP-Luc murine breast tumor cell line was developed and validated as described previously (10). The BALB/c background 4T1-FL-GFP murine breast tumor cell line (derived from 4T1, RRID:CVCL_0125) was originally from Dr. David Piwnica-Worms (The University of Texas, Houston, TX) as described previously (18). All cell lines were cultured at low passage (used within 1–3 passages after thaw) and tested regularly for Mycoplasma-specific DNA by PCR amplification of cell or supernatant samples (last test January of 2020). For in vitro experiments involving coated culture dishes, nontissue culture-treated plates were coated prior to cell seeding with either poly-l-lysine (Sigma: P4707) or Vitronectin XF (STEMCELL Technologies: 07180) according to manufacturer's recommendations. In vitro drug treatments were performed using docetaxel (LC Laboratories: RP 56976), everolimus (RAD001; Selleck Chemicals: cat No. S1120), and doxorubicin hydrochloride (Sigma: PHR1789) at the stated concentrations.
CRISPR knockout of the Itgb3 gene in the PyMT-BO1 line was previously described (19). pMx, pMx-Δβ3, and pMx-hβ3 retroviral vectors used for rescue of β3 expression in the clone No. 1 β3KO PyMT-BO1 (β3KO1-BO1) line were a gracious gift from Steven Teitelbaum (Washington University School of Medicine, St. Louis, MO). Virus was packaged along with the pCMV-VSVG plasmid using the Plat-E cell line (RRID:CVCL_B488; ref. 20). Tumor cell lines were transduced with viral supernatant for 12 hours at 37°C in 6-well tissue culture plates. Transduced cells were selected in 2 μg/mL blasticidin (Sigma: 203350), and stable protein expression of the wild-type (hβ3) and signaling mutant (Δβ3) integrin constructs was validated by Western blot analysis (below).
CRISPR knockout of Itgb3 in the 4T1 cell line was achieved by stable transduction of the Cas9 gene and the following gRNAs: 5′-CACCGCCGGGATAACCTCGTTGTTG-3′; 5′-AAACCAACAACGAGGTTATCCCGGC-3′, using the lentiCRISPR v2-Puro vector system (Addgene No. 98290). 293T cells (ATCC, RRID:CVCL_0063) were used for viral packaging with the pCMV-DR8.2 and pCMV-VSVG plasmids. Tumor cell lines were transduced with viral supernatant for 12 hours at 37°C in 6-well tissue culture plates. Transduced cells were selected in 10 μg/mL puromycin (Sigma: P8833) and further purified by serial FACS sorting of TGF-β1 (2 ng/mL, R&D Systems: 7666-MB-005) stimulated cells based on β3 expression. Itgb3 knockout cell lines were validated by sequencing and FACS.
Cell viability assays
Viability assays were performed as described previously (21, 22), with minor modifications (Supplementary Information).
For apoptosis assays, caspase-3/7 activity was determined using the Caspase-Glo 3/7 Assay System (Promega: G8090) and normalized to cell viability as reported by incubation with CellTiter-Blue (Promega: G8080). Colorimetric and luminescent readouts were measured with a SpectraMax i3 plate reader (Molecular Devices).
In vivo modeling of metastasis and therapy
Distant metastases were established in mice by intracardiac (i.c.) inoculation of PyMT-BO1 or 4T1 cells into the left ventricle as described previously (14). Tumor burden was monitored by in vivo bioluminescence imaging (BLI). Mice were assigned randomly by cage to treatment groups; mouse weight and hindlimb tumor burden were compared to ensure no statistical differences between groups prior to treatment initiation. Docetaxel (5 mg/kg, LC Laboratories: RP 56976) or equivalent vehicle was freshly prepared and administered by tail vein injection (Supplementary Data). Freshly prepared working solution of rapamycin (Sigma: R0395) or equivalent vehicle was administered 2 mg/kg by intraperitoneal injection (Supplementary Data). An equimolar equivalent of nanoparticle-encapsulated rapamycin or cargo-free nanoparticle control was administered by tail vein injection for nanoparticle experiments.
Bioluminescence imaging and radiography
Bioluminescence imaging (BLI) and X-ray assessment of tumor burden and bone lesion area were performed as described previously (14), with minor modifications (Supplementary Data).
All slides were stained in parallel, using identical staining conditions, with anti-integrin β3 (clone: D7 × 3P, 1:200, Cell Signaling Technology, RRID:AB_2798136) using previously described protocols (14). Images were acquired on a NanoZoomer (Hamamatsu Photonics).
Postchemotherapy biopsies from patients with triple-negative breast cancer
Primary breast cancer specimens were obtained from M0 patients with localized, triple-negative disease at time of surgical resection and subsequently banked, curated, and assembled into a tissue microarray by the St. Louis Breast Tissue Registry. Clinical data were obtained in accordance with the Washington University Institutional Review Board (IRB No. 201102394) and WAIVER of Elements of Consent per 45 CFR 46.116 (d) and deidentified prior to investigator access. IRB-directed human research activities were guided by principles set forth in the Belmont Report.
Areas of invasive tumor and tumor cell β3 expression by DAB staining were confirmed in consultation with a board-certified pathologist. β3 expression was scored by a group of investigators, all blinded to clinical annotation, using a bimodal classification system focused exclusively on positive staining in tumor cells. For each specimen, the percentage of identifiable tumor cells with β3 staining was jointly determined by the scoring group. On the basis of the range of tumoral β3 staining observed in the cohort as a whole, a cutoff of 10% was determined, with samples below this threshold assigned to the “low” expressing group and those above assigned to the “high” expressing group. Samples for which the scoring group was not unanimous were referred to consulting pathologist for final resolution.
Survival analysis in human patients
Recurrence-free survival (RFS, defined as date of diagnosis to date of first local or distant recurrence, otherwise censored at last known recurrence-free date) of β3 low versus high triple-negative breast cancer (TNBC) core samples after neoadjuvant chemotherapy was determined by Kaplan–Meier analysis using Cox proportional hazards and log rank test in consultation with a statistician in the Siteman Biostatistics Shared Resource. Relevant patient demographic data and tumor characteristics were compared between groups using Wilcoxon rank-sum and Fisher exact tests. RFS in publicly available data was determined by Kaplan–Meier analysis (KM-Plotter, database version 2017, accessed 05/26/2020) (23), comparing patients with TNBC receiving any chemotherapy in the lowest quartile of β3 expression with those in the three upper quartiles.
Transmission electron microscopy of murine bone metastases
Mice bearing 4T1 bone metastases were established, treated, and monitored as described above and in Fig. 2C. After tissue processing (Supplementary Data), X-ray microscopy (XRM Versa 520, Zeiss) was performed to identify tumor regions in hindlimb bone samples for thin sectioning. Thin sections (70 nm) were prepared on grids, stained with 2% aqueous uranyl acetate followed by Reynold's lead citrate, and imaged on a TEM (JEOL JEM-1400 Plus) at 120 KeV. Ultrastructural parameters were quantified using ImageJ (NIH, Bethesda, MD; RRID:SCR_003070).
RNA sequencing and analysis
RNA-sequencing (RNA-seq) was performed with the Genome Technology Access Center at Washington University School of Medicine (Supplementary Data). Data are accessible through NCBI GEO (GSE166315, GSE166536).
Flow cytometric analysis
After preparation as single-cell suspensions, ex vivo and in vitro samples were stained and analyzed as described previously (14), with assay-specific modifications (Supplementary Data).
Western blot analysis
Western blot was performed as described previously (10) using anti-integrin β3 (D7 × 3P, RRID:AB_2798136), phospho-S6 ribosomal protein (Ser240/244; D68F8, RRID:AB_10694233), phospho-S6 ribosomal protein (Ser235/236) (D57.2.2E, RRID:AB_916156), or β-actin loading control (13E5, RRID:AB_2223172), followed by horseradish peroxidase–conjugated anti-rabbit secondary antibody (all from Cell Signaling Technology) according to manufacturer's protocols. Bands were developed via enhanced chemiluminescence and analyzed by densitometry in ImageJ (NIH, Bethesda, MD, RRID:SCR_003070). Blots in Supplementary Fig. S6A and S6C were run from samples harvested after 24 and 48 hours of treatment (one blot for each time point). After transfer, blots were cut in half. Top halves were blotted for integrin β3. Bottom halves were blotted sequentially for pS6 (S240/244), pS6 (S235/236), and β-actin, stripping between each with Restore Western Blot Stripping Buffer (Thermo Fisher Scientific: 21059). Because of this, the β-actin loading control for 24-hour panels in these two figures is the same.
Oxygen consumption analyses
Cells were seeded and treated in 6-well plates as indicated, then lifted with trypsin and reseeded onto Seahorse XF96 V3 PS Cell Culture Microplates (Agilent: 101085–004) overnight at experimentally optimized density. Extracellular flux analysis of oxygen consumption rate (OCR) was performed on the Seahorse Biosciences XF96 Flux Analyzer (Agilent) at baseline and after serial injection of oligomycin (1.5 μmol/L), FCCP (0.5 or 1 μmol/L), and antimycin A/rotenone (0.5 μmol/L; Seahorse XF Cell Mito Stress Test Kit, Agilent: 103015–100) according to manufacturer's recommendations. After analysis, tumor cell sample luciferase activity was determined for normalization using the Steady-Glo Luciferase Assay System (Promega: E2510) and a SpectraMax i3 plate reader (Molecular Devices). Data normalization, analysis, and calculation of maximum OCR were performed using Wave Desktop v2.6 (Agilent, RRID:SCR_014526).
Galuminox imaging of radical oxygen species
Live cell fluorescence imaging was performed as described previously (24), with minor modifications (Supplementary Data).
Synthesis of αvβ3-RAPA nanoparticles
Phospholipid/polysorbate 80 micelle nanoparticles (NPs) were prepared as a microfluidized suspension of 20% (v/v) combining polysorbate 80 (NOF America) with a 2.0% (w/v) commixture and 1.7% (w/v) glycerin in pH 6.5 carbonate buffer. The commixture included 2 mole% rapamycin, 0.15 mole% αvβ3-PEG2000-PE (Supplementary Data), and high-purity phosphatidylcholine (Lipoid). Rapamycin was excluded from commixture for targeted, drug-free NPs. The lipid commixtures were combined with the polysorbate, buffer, and glycerin and homogenized at 20,000 psi for 4 minutes at 4°C with a microfluidics homogenizer (M110s or LV1, Microfluidics, Inc). NPs were sterile filtered and preserved under inert gas in sterile sealed vials until use. Dynamic light scattering (Zeta Plus, BrookHaven) showed nominal particle size of 23.9 nm, with polydispersity of 0.258 and an average electrophoretic zeta potential of –1.61mv for αvβ3-RAPA-NPs, which were closely similar to αvβ3-CF-NP control.
Serum chemistry analysis
Blood was obtained by submandibular venous puncture and collected in Microtainer serum separator tubes (BD Biosciences: BD365967) for serum chemistry analysis using the Liasys 330 AMS Diagnostic liquid chemistry analyzer. Investigators were blinded to treatment groups during analysis.
All sample sizes reported in the study are the minimum number of samples. For animal studies, sample sizes were decided on the basis of our previous work in these models. Statistical differences were analyzed using either one- or two-tailed unpaired t test with Welch correction, ANOVA with Tukey or Sidak test for post hoc multiple comparisons, or ANOVA with test for linear trend using Prism 8 (GraphPad Software Inc., RRID:SCR_002798). Results were considered to reach significance at P ≤ 0.05, indicated with asterisks (*, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001). Data are presented as mean values; error bars represent ±SD.
Integrin β3 expression is increased in breast cancer cells after chemotherapy
Dysregulated expression of integrin β3 (β3) is associated with increased aggressiveness and drug resistance in cancer (16). We first asked whether exposure to the chemotherapeutic agent docetaxel (DTX) alters the proportion of β3-expressing cells in tumor populations. To test this, the 4T1 and PyMT-BO1 murine breast cancer cell lines (modeling triple-negative and luminal B disease, respectively) were administered docetaxel in vitro and cell surface β3 expression was assessed by flow cytometry. We observed an increase in the percentage of β3+ cells in both cell lines after docetaxel treatment (Fig. 1A; Supplementary Fig. S1A), with a stronger dose-dependent response in the PyMT-BO1 line. In vitro exposure to doxorubicin, another chemotherapeutic agent frequently administered in breast cancer, yielded similar increases in the percentage of β3+ cells (Supplementary Fig. S1B).
We next evaluated integrin β3 expression after docetaxel in the bone metastatic environment. We had previously demonstrated increased tumoral β3 in bone metastases compared with primary breast tumors, both in human patients and in 4T1 and PyMT-BO1 preclinical models (14). We found that docetaxel failed to attenuate osteolytic lesions generated by 4T1 and PyMT-BO1 (Fig. 1B), indicating that both cell lines were fairly chemoresistant. This prompted us to measure β3 expression in the resistant tumor cells that remained. Given their greater β3 response to docetaxel in vitro, we focused on PyMT-BO1 bone metastases, harvesting live, GFP+ tumor cells for assessment of β3 expression by ex vivo flow cytometry. We found a significant increase in the proportion of GFP+β3+ tumor cells in bone metastasis samples from mice receiving docetaxel compared with those from mice receiving vehicle (vehicle: 33% β3+; DTX: 54% β3+, P < 0.0001; Fig. 1C; Supplementary Fig. S1C).
To gauge translational relevance, we assessed tumoral β3 expression in a tissue microarray (TMA) of high-risk, post-chemotherapy clinical specimens taken from 38 patients with localized TNBC who did not achieve pathologic complete response (pCR) after neoadjuvant chemotherapy (Fig. 1D). We evaluated tumor cell-specific β3 expression by IHC, designating samples as either low or high based on staining intensity and β3+ cell frequency (see Materials and Methods). Ninety-seven percent of patient specimens had positive tumoral β3 staining after chemotherapy. 27 (71%) were characterized as low tumoral β3 expression, whereas 11 (29%) were high (Fig. 1D and E). As expected, we observed a consistent vascular pattern of strong β3 expression on tumor neoangiogenic endothelium, serving as a positive staining control (Supplementary Fig. S1D; ref. 25). Kaplan–Meier analysis of differences in recurrence-free survival (RFS) between patients with β3 low and high postchemotherapy residual tumors revealed a trend toward increased risk of recurrence in the high group, particularly after the first 1.5 years after diagnosis (note curve cross in Fig. 1F), although this was not statistically significant in our relatively small sample (HR, 1.75; 0.66–4.74; P = 0.254; Fig. 1F). To validate this finding in a larger cohort, we used publicly available microarray data to perform a separate RFS analysis in 315 patients with TNBC who had received chemotherapy (23). In this dataset, patients with increased tumoral β3 expression (high, upper three quartiles) were twice as likely to experience recurrence compared with patients in the lowest quartile of expression (low; HR = 2.01; logrank P < 0.0095; Fig. 1G). Together, these data suggest that tumoral β3 expression is increased and associated with worse outcomes after chemotherapeutic challenge.
Integrin β3 promotes docetaxel resistance in bone metastases
We next considered functional differences in β3+ tumor cells that might drive poor outcomes after treatment, measuring proliferation changes in cells with high and low β3 expression after chemotherapy in vitro. PyMT-BO1 cells were exposed to docetaxel in vitro and BrdU incorporation was assessed by flow cytometry in β3hi (high) and β3lo (low) populations. The β3lo population of docetaxel-treated PyMT-BO1 cells exhibited significantly reduced BrdU incorporation compared with β3lo cells receiving vehicle (48% reduction, P < 0.0001). Interestingly, BrdU incorporation by β3hi cells present in the same cultures was unchanged between docetaxel- and vehicle-treated samples (Fig. 2A), suggesting that β3-expressing cells respond differently to docetaxel. To address this, we employed CRISPR/Cas9 technology to generate Itgb3 knockout (β3KO) derivatives of the 4T1 (Supplementary Fig. S2A) and PyMT-BO1 (Supplementary Fig. S2C; ref. 19) murine breast cancer cell lines.
Cell viability of β3KO derivatives was measured by MTT assay. Both β3WT and β3KO 4T1 cells were sensitive to docetaxel administration in vitro, with β3WT cells showing only modestly higher viability (Fig. 2B). Given that integrins enhance cell adhesion (8), and that tumor cell coculture with bone marrow stromal cells (BMSCs) increases chemoresistance (26), we next assessed the docetaxel viability of β3WT and β3KO 4T1 derivatives in BMSC coculture. In these conditions, β3WT 4T1 cells exhibited enhanced resistance to docetaxel compared with single culture, while BMSC-cocultured β3KO 4T1 derivatives remained sensitive (Fig. 2B). We found similar in vitro docetaxel viability trends in β3WT and β3KO PyMT-BO1 derivatives (Supplementary Fig. S2D).
Given the contribution of the tumor microenvironment to therapeutic responses (6), we next interrogated the role of β3 for chemoresistance in murine bone metastases. Mice bearing disseminated β3WT or β3KO 4T1 cells were administered either docetaxel or vehicle and assessed for organ tumor burden by ex vivo BLI. Across all organs analyzed (kidneys, lung, liver, and hindlimb bones) in β3WT tumors, there was no significant difference in bioluminescence between vehicle and docetaxel-treated groups (Fig. 2C; Supplementary Fig. S2B). In contrast, hindlimb bones from docetaxel-treated mice bearing β3KO cells exhibited significantly (50.6-fold) reduced tumor burden compared with vehicle, with visceral organs also exhibiting trends toward decrease (kidneys: 5.3-fold; lung: 1.2-fold; liver: 5.7-fold; Fig. 2C; Supplementary Fig. S2B). Parallel experiments using β3WT and β3KO PyMT-BO1 cells revealed similar findings, with β3KO bone metastases showing the greatest decrease after docetaxel (Supplementary Fig. S2E and S2F). Taken together, these results suggest that tumoral β3 plays a functional role in the chemoresistant phenotype.
Rescue of integrin β3 expression restores chemoresistance in a signaling-dependent manner
Having established that β3 deletion sensitizes bone metastases to docetaxel, we next asked whether β3 rescue in β3KO breast cancer cells was sufficient to restore docetaxel resistance. To do this, clone No. 1 β3KO PyMT-BO1 cells (β3KO1-BO1) were retrovirally engineered to express either an empty vector (pMx), a functional human integrin β3 construct (hβ3), or the DiYF integrin β3 mutant (Δβ3), which can bind ligand but is incapable of downstream signaling (Fig. 3A; ref. 27). In vitro, hβ3-expressing cells were significantly more viable than pMx-expressing cells by MTT after DTX exposure (∼3.6 nmol/L vs. ∼1.6 nmol/L IC50, P < 0.0001). Rescue with signaling-deficient Δβ3 mutant, in contrast, had no effect on viability (∼1.2 nmol/L vs. ∼1.6 nmol/L IC50, P = 0.3678; Fig. 3B). We also observed diminished apoptosis and enhanced proliferation in hβ3-expressing cells compared with both empty vector and Δβ3-rescue after docetaxel exposure (Fig. 3C and D). In BMSC coculture, hβ3 rescue significantly increased DTX resistance compared with wild-type PyMT-BO1 cells (Supplementary Fig. S3A), which have lower β3 expression at baseline.
To confirm this in vivo, we established β3KO1-BO1 derivative metastases in mice and administered docetaxel or vehicle. Empty vector β3KO metastases of the kidneys, lungs, and hindlimbs were sensitive to docetaxel (4.8-fold, 2.5-fold, and 6.3-fold decrease from vehicle, respectively). Organs harboring hβ3-expressing tumors, meanwhile, exhibited no significant difference in BLI between docetaxel and vehicle-receiving mice. Importantly, signaling-deficient Δβ3-rescued tumors were notably sensitive to docetaxel, with statistically similar fold decreases to empty vector groups (Fig. 3E). Taken together, these results suggest that β3 expression is sufficient to promote increased resistance to docetaxel in vitro and in vivo, and that this phenotype requires intact integrin signaling.
Integrin β3 mediates an alternative metabolic response to docetaxel
Our results suggested that β3-mediated chemoresistance was most evident in the bone metastatic microenvironment. To evaluate the role of β3 expression in the docetaxel response of individual tumor cells in this context, we analyzed docetaxel-treated β3WT and β3KO 4T1 murine bone metastases by transmission electron microscopy (TEM; Fig. 4A). Vehicle-treated bone metastases were grossly similar, with no evident β3-dependent differences in organelle morphology or ECM composition. In docetaxel-treated β3KO tumors, many breast cancer cells exhibited membrane blebbing and fragmentation, consistent with a higher level of cell death. Individual mitochondrial area was increased compared with vehicle in β3KO (Supplementary Fig. S4A), but the ratio of neither total mitochondrial area nor rough endoplasmic reticulum (ER) area to cytosolic area was altered (Fig. 4B). In docetaxel-treated β3WT bone metastases, breast cancer cells remained largely intact, and were notably embedded in abundant fibrillar ECM not observed in vehicle-treated tumors. In contrast to β3KO, rough ER area was markedly increased from vehicle, with pronounced cisternae clearly visible (WT vehicle: 4.5% rough-ER-to-cytosol; WT DTX: 14.9% rough-ER-to-cytosol, P < 0.0001; Fig. 4B). Similar to docetaxel-treated β3KO tumors, individual mitochondrial area was increased in docetaxel-treated β3WT (Supplementary Fig. S4A), while total mitochondrial area was unchanged. Together, these results suggest that docetaxel administration elicits tumoral β3-dependent changes in the cellular and microenvironmental ultrastructure of bone metastases.
To identify mechanistic links between β3 expression and chemoresistance, we generated RNA-seq transcriptomic profiles of β3WT versus β3KO 4T1 cells and hβ3-rescued versus empty vector β3KO1-BO1 cells after docetaxel or DMSO exposure in vitro. Gene set enrichment analysis (GSEA) of biological process and cellular compartment gene ontology (GO) terms in 4T1 profiles revealed β3-dependent enrichment of genes associated with ER, the unfolded protein response, and collagen-containing ECM after docetaxel administration (Fig. 4C; Supplementary Fig. S4B). Using hallmark GSEA (28), we next isolated functional pathways of interest, focusing on those where β3-expressing and β3KO docetaxel responses were most different. A group of metabolism-related pathways was the most enriched during the β3-mediated docetaxel response in hβ3-rescued β3KO1-BO1 cells, many of which were also positive in the 4T1 analysis (Fig. 5A, see dashed line box). Interestingly, the hallmark pathway with the greatest positive difference in both 4T1 and PyMT-BO1 was OXPHOS (4T1: +3.2 net NES; BO1: +7.3 net NES; Fig. 5A).
To functionally validate OXPHOS enrichment during the β3WT docetaxel response, we performed in vitro extracellular flux analysis of oxygen consumption rate (OCR) in both 4T1 and PyMT-BO1 lines. We found significantly increased maximum OCR after docetaxel in hβ3-rescued β3KO1-BO1 cells, while empty vector (pMx) exhibited no or minimal increase compared with vehicle treatment (Fig. 5B). Likewise, in 4T1, we found significant increases in maximum OCR after docetaxel in β3WT not seen in β3KO cells (Fig. 5B). Interestingly, the differences in OCR between 4T1 β3WT and β3KO were observed on plates coated with vitronectin (a ligand recognized by activated αvβ3 integrin) but not on regular tissue culture–treated plates (Supplementary Fig. S5A and S5B).
These differences in bulk oxygen handling after chemotherapy prompted investigation of reactive oxygen species (ROS), another pathway identified in our hallmark analysis (Fig. 5A). Galuminox, a novel fluorescent metalloprobe (24), allowed us to directly image hydrogen peroxide and superoxide in live 4T1 cells by confocal microscopy. These studies revealed a nearly fivefold increase in β3WT ROS after docetaxel, while ROS after docetaxel in β3KO cells was not significantly different (Fig. 5C). Taken together, our results suggest that β3 mediates an alternative metabolic response to docetaxel treatment in breast cancer cells.
mTORC1 inhibition reverses β3-mediated chemoresistance
We searched our hallmark GSEA for metabolically relevant signaling pathways that could be targeted in combination with docetaxel to overcome β3-mediated resistance. We found mTORC1 activity and its target E2F, both established regulators of mitochondrial metabolism and protein synthesis (29), to be among the most significantly enriched signaling pathways in hβ3-expressing cells exposed to docetaxel (mTORC1 NES 3.7, q < 0.0001; E2F NES 4.1, q < 0.0001; Fig. 6A). To functionally validate the importance of mTORC1 activity in β3WT 4T1 cells without chemotherapy, we assessed viability after exposure to the mTORC1 inhibitor everolimus (mTORCi). Although everolimus abrogated phosphorylation of ribosomal protein S6 in both β3WT and β3KO 4T1 cells (Supplementary Fig. S6A), β3WT 4T1 cells exhibited significant viability reduction compared with DMSO control, while β3KO viability was unaffected (Fig. 6B). We next asked if, similar to our docetaxel experiments, β3 expression is increased after everolimus treatment. Western blot analysis of 4T1 cells after in vitro exposure to everolimus demonstrated significantly higher total β3 protein expression compared with cells receiving vehicle (Fig. 6C). Flow cytometry analysis showed that the proportion of β3+ PyMT-BO1 cells was likewise increased after in vitro everolimus treatment (Supplementary Fig. S6B).
Our in vivo TEM images revealed that β3WT cells undergo rough ER expansion after exposure to docetaxel, possibly as part of an unfolded protein stress response (Fig. 4). To determine the effect of combination DTX and mTORCi on this phenotype in vitro, we assessed de novo protein production in β3WT and β3KO 4T1 cells exposed to either docetaxel, mTORCi, or both. At baseline, β3WT cells incorporated almost 65% more HPG-methionine than β3KO. Docetaxel reduced HPG-methionine incorporation by 25% in β3WT cells, but did not affect de novo protein production in β3KO. Importantly, while mTORCi alone had no effect on HPG-methionine incorporation by β3WT cells, combination with docetaxel resulted in a 60% reduction compared with vehicle, almost twice the reduction observed in β3KO (Fig. 6D). Given previous demonstration of a link between β3 signaling and mTORC1 activity (30, 31), in addition to the clinically approved use of mTORC1 inhibitors in patients with breast cancer, we decided to pursue it as a candidate for combination therapy with docetaxel in breast cancer bone metastases.
To evaluate this combination strategy, we established β3WT PyMT-BO1 bone metastases by intracardiac injection. Mice were randomized to receive either vehicle, docetaxel alone, the mTORC1 inhibitor rapamycin alone (RAPA), or combined treatment (COMBO). Although ex vivo BLI bone tumor burden in groups receiving docetaxel or RAPA alone was not significantly different from vehicle, combination therapy synergistically attenuated bone metastases (5.5-fold reduction compared with vehicle, P < 0.01; Fig. 6E). This effect was not observed in visceral metastases (Supplementary Fig. S6C).
αvβ3-targeted NPs loaded with mTOR inhibitor enhance docetaxel efficacy in bone metastases
To confirm β3-dependent synergy and provide proof of principle for this strategy in a precision medicine setting, we modified our αvβ3-targeted micelle nanoparticle (previously described in ref. 14) with a rapamycin cargo (αvβ3-RAPA-NP) to specifically deliver rapamycin to cells expressing activated αvβ3 integrin heterodimers (Fig. 6F; Supplementary Fig. S6D). Mice bearing β3WT PyMT-BO1 bone metastases were randomized to receive either cargo-free control NPs (αvβ3-CF-NP), combination αvβ3-CF-NP and free docetaxel, or combination αvβ3-RAPA-NP and free docetaxel. By ex vivo BLI and X-ray analysis, we found that combination αvβ3-RAPA-NP and free docetaxel was significantly more effective to decrease bone tumor burden and tumor-induced bone loss (osteolysis) than cargo-free NPs and free docetaxel (Fig. 6G). As before, the effect on tumor burden was not significant in visceral metastases (Supplementary Fig. S6E). In addition, rapamycin loading did not increase serum markers of therapy-induced toxicity compared with cargo-free particles in combination with docetaxel (Supplementary Fig. S6F). Taken together, these data suggest mTORC1 inhibition as a strategy for enhancing response to taxane therapy in αvβ3-expressing bone metastases.
Bone metastases are common in breast cancer (32), manifesting as osteolytic lesions with a fundamentally different biology than the primary tumor or visceral metastatic sites (4). Bone-targeted agents such as bisphosphonates and denosumab have improved patient quality of life, but these therapies are not curative and largely spare the tumor itself (3).
Exposure to the bone microenvironment modulates tumor cell phenotype (5, 33). In previous studies, we found that bone-induced TGFβ signaling upregulates β3 expression in breast cancer cells (14). This study expands on this finding, demonstrating that tumoral β3 expression itself promotes chemoresistance characterized by an alternative metabolic response to docetaxel. We further showed that combination rapamycin and docetaxel overcomes β3-mediated resistance. Finally, administration of rapamycin-loaded, αvβ3-targeted NPs specifically improved docetaxel response in murine bone metastases, providing proof of principle for an effective strategy that might circumvent possible toxicities associated with combination therapy.
β3+ murine breast cancer cells were increased after in vitro chemotherapy, corroborating results in human cells reported by Vellon and colleagues (34). Docetaxel in vitro failed to reduce proliferation in the β3hi population, suggesting that DTX selects for resistant cells with higher β3 expression. Our in vivo findings supported this, with β3+ tumor cells enriched in bone metastases remaining after systemic docetaxel treatment. In human patients, incomplete response to neoadjuvant chemotherapy is associated with significantly worse outcomes (35), likely driven by selection for and reprogramming toward resistance in the cells that survive (36). Using publicly available data, we found that β3 expression at diagnosis was associated with a higher recurrence risk in patients with TNBC receiving any chemotherapy. To our knowledge, this is the first report of a correlation between αvβ3 expression and recurrence-free survival in patients with TNBC receiving chemotherapy. We would expect this effect to be most significant in patients with bone metastases, where β3 expression is known to be increased. Unfortunately, large databases of bone metastases are not readily available, possibly contributing to this association in primary breast tumors, where absolute β3 expression is low, going unreported until now. We found populations of β3+ residual tumor cells in 97% of postchemotherapy primary tumor specimens we analyzed from high-risk patients with localized TNBC who failed to achieve a pCR after neoadjuvant chemotherapy. Although survival significance in this TMA cohort was limited by sample size, we found a trend toward increased risk of recurrence in patients with high β3 expression. This difference was especially pronounced after a curve crossing event approximately 1.5 years after diagnosis, raising the possibility that β3 expression might be more relevant to recurrence later in the course of TNBC. Studies are planned to analyze a larger cohort of patient samples with longer follow-up times.
β3 is an important promoter of bone metastasis (15), is upregulated in bone metastases compared with the primary and visceral sites (14), is important to protumor microenvironment cells such as osteoclasts (9) and tumor blood vessels (11, 12), and has previously been implicated in resistance to therapies across several cancer types (17, 37, 38). Despite this, direct pharmacologic blockade of αvβ3 has not shown significant activity in clinical trials of aggressive and advanced cancers (39). Furthermore, the addition of the RGD-mimetic integrin αvβ3/αvβ5-inhibitor cilengitide to temozolomide chemotherapy was not associated with clinical benefit in a phase III trial of patients with glioblastoma (40). Trial design and advanced patient stage may have contributed to this lack of efficacy; however, off-target effects of pharmacologic αvβ3 blockade on host cells, such as enhancing protumor neoangiogenesis (41) and promoting immunosuppression (10), could also be at play. Interestingly, a preclinical study in lung and pancreatic cancer models demonstrated that low doses of cilengitide could enhance the activity of gemcitabine chemotherapy in vivo by simultaneously increasing tumor cell–extrinsic delivery and tumor cell–intrinsic drug uptake (42), raising the possibility that the effects of these agents on the tumor microenvironment might still be exploited for synergistic antitumor effects when used in the right context.
β3 studies in breast cancer have hinged primarily on in vitro characterization with pharmacologic blockade or on its role in promoting metastasis (15, 17). Using CRISPR/Cas9 technology, we abrogated tumoral β3 expression in the setting of an intact immune system. Because manipulation of tumoral integrin β3 could also affect tumor growth (15, 43), we used each genetic line as its own control, only comparing response to chemotherapy across genotypes. We show here that bone metastases lacking integrin β3 were significantly more sensitive to docetaxel than wild-type metastases.
In vitro, β3WT and β3KO breast cancer cells were both highly sensitive to docetaxel, but in vivo β3WT bone metastases were relatively resistant. In vitro survival increased when β3WT cells were plated on BMSCs, suggesting ligand availability as a potential factor in αvβ3 integrin–mediated resistance. Integrin activation, which induces a conformational change that exposes the ligand binding domain, is required for ligand binding and signaling (44), and it is possible that β3 activation in the bone microenvironment results in easier access to ECM ligands. The bone metastatic microenvironment also has a lower pH and oxygen concentration, higher stress modulus, and distinct nutrient and chemical milieu (4), all of which can influence cancer cell reliance on αvβ3 signaling (31, 45, 46).
Although BLI, histology, and X-ray indicated that docetaxel had minimal effect on β3WT bone metastases, TEM analysis uncovered a dramatic metabolic response to chemotherapy in these resistant cells characterized by profound increases in protein production and ECM deposition. Experiments are underway to profile these ECM proteins, determine the mechanism through which β3 ligand binding and signaling are involved, and more specifically exploit this metabolic vulnerability to enhance therapeutic efficacy.
We identified protein synthesis, ECM enrichment, and OXPHOS as potential downstream targets of β3 signaling in response to chemotherapy. Although no β3-mediated changes in mitochondrial ultrastructure were evident by TEM, our in vitro studies showed that DTX consistently increased OCR in β3WT compared with β3KO cells. Live-cell imaging of 4T1 cells further demonstrated robust β3-mediated increases in docetaxel-induced ROS generation, suggesting an alternative metabolic response that drives therapeutic resistance (47, 48).
Consistent with increased ER observed in vivo, we found that in vitro protein production was higher at baseline and more responsive to docetaxel in β3WT cells. Recent evidence indicates that tumoral ER stress is a common feature of breast cancer bone metastases (49), and others have shown that integrin signaling bolsters in vitro protein production during hypoxia (31). These data suggest that diminished tolerance for ER stress (50) could drive the large chemosensitizing effect we observe with β3KO in the bone compared with other metastatic sites, where tumoral β3 expression is not as high. Future studies, including single-cell RNA and ribosomal sequencing of tumor cells collected directly from bone metastases, are planned to more specifically evaluate ER stress and unfolded protein response during β3-mediated chemoresistance in vivo.
mTORC1 pathway genes were enriched after docetaxel treatment of β3WT cells and are demonstrated targets of αvβ3 signaling in breast cancer (29–31). Administration of either the mTORC1 inhibitor rapamycin or docetaxel alone had little effect on PyMT-BO1 bone tumor burden, while rapamycin and docetaxel together significantly attenuated bone metastases. Notably, this synergistic effect was exclusive to bone, where tumor expression of β3 is much higher than in visceral metastases (14). mTORC1 inhibition alone has been shown to restrict tumor growth in bone micrometastases (51), suggesting that combination with docetaxel may provide additional benefit for adjuvant metastasis prevention (52). Although single-agent mTOR inhibition does not attenuate tumor growth in bone macrometastases (51, 53), it has been shown to protect against osteolytic bone loss, which might contribute indirectly to our findings (53).
To test the specific effect of rapamycin inhibition in αvβ3-expressing cells, we coadministered free docetaxel with rapamycin-loaded, αvβ3-targeted NPs (14). Combination of mTOR inhibitors and taxane chemotherapy is clinically challenging due to toxicity (54), but αvβ3-NPs can reduce drug availability in the circulation by influencing release kinetics (14, 55). In this study, combination therapy with docetaxel and αvβ3-RAPA-NP was more effective than docetaxel alone against PyMT-BO1 bone metastases, demonstrating that the efficacy of mTORC1 inhibition is, in part, mediated by its specific activity in cells expressing activated αvβ3 integrin. The combination of αvβ3-RAPA-NP with docetaxel did not significantly affect serum toxicity markers in mice, which holds potential for clinical translation.
The bone specificity of rapamycin/docetaxel combination therapy was a striking finding from our study. We have previously shown that tumoral integrin β3 expression is increased in bone metastases compared to the primary or visceral metastatic sites, facilitating increased drug delivery to the bone by αvβ3-targeted NPs (14). Beyond enhancement of drug delivery, our transcriptomic and functional studies suggest that mTOR activity is also more important for β3-expressing cells, particularly in the setting of docetaxel treatment. Consistent with this, increased tumoral β3 expression could render bone metastases more dependent on mTOR activity for proliferation after exposure to docetaxel. The effect of mTOR inhibition on visceral metastases is likely less pronounced because β3-expressing tumor cells represent a significantly lower proportion of the overall tumor burden than what we observe in the bone. In our aggressive preclinical metastatic models, experimental duration is likely not long enough for mTOR-dependent β3+ cells to sufficiently accumulate. Overexpression of β3 in β3KO1-BO1 cells resulted in increased chemoresistance in visceral metastases. Experiments are underway to determine the role of mTOR for chemoresistance in β3-expressing cells at these visceral sites. In addition to what we observe in bone metastases, we propose that patients with visceral metastases who have been previously treated with chemotherapy and would be predicted to have increased β3+ cells could benefit from combination mTORC1 inhibition and docetaxel.
Finally, although it is of high experimental utility, intracardiac injection exhibits limitations as a model of breast cancer bone metastasis. It overlooks selection processes that occur at the primary site and during intravasation into the circulatory system, and has lower latency rates between dissemination and frank macrometastatic outgrowth. These differences all have the potential to impact therapeutic response, but currently, no genetic mouse models of spontaneous breast cancer bone metastasis exist. Studies are in progress to evaluate the efficacy of our combination strategy on bone metastasis following resection of mammary fat pad tumors and monitoring for spontaneous metastases. Nevertheless, our findings here highlight the need for therapeutic strategies that consider the microenvironmental context of the tumor when targeting metastatic cells.
No disclosures were reported.
G.C. Fox: Conceptualization, data curation, formal analysis, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. X. Su: Supervision, investigation, methodology, writing–review and editing. J.L. Davis: Data curation, formal analysis, validation, investigation, visualization, writing–review and editing. Y. Xu: Investigation. K.A. Kwakwa: Conceptualization, data curation, validation, investigation, visualization, methodology. M.H. Ross: Conceptualization, data curation, validation, investigation, visualization, methodology. F. Fontana: Data curation, validation, investigation, visualization, methodology, writing–review and editing. J. Xiang: Data curation, investigation, writing–review and editing. A.K. Esser: Data curation, validation, investigation. E. Cordell: Validation, investigation. K. Pagliai: Data curation, formal analysis, validation, investigation, visualization, methodology. H.X. Dang: Resources, data curation, software, formal analysis, validation, investigation, visualization, methodology. J. Sivapackiam: Conceptualization, formal analysis, funding acquisition, validation, investigation, visualization, methodology. S.A. Stewart: Conceptualization, resources, data curation, software, formal analysis, funding acquisition, visualization, methodology. C.A. Maher: Conceptualization, resources, data curation, software, formal analysis, visualization, methodology, writing–original draft, writing–review and editing. S.J. Bakewell: Resources, data curation, formal analysis, funding acquisition, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. V. Sharma: Resources, data curation, formal analysis, funding acquisition, validation, investigation, visualization, methodology, project administration. S. Achilefu: Conceptualization, resources, data curation, formal analysis, funding acquisition, investigation, visualization, methodology. D.J. Veis: Conceptualization, resources, data curation, formal analysis, supervision, funding acquisition, investigation, visualization, methodology, project administration, writing–review and editing. G.M. Lanza: Conceptualization, resources, data curation, supervision, funding acquisition, visualization, methodology, writing–original draft, project administration, writing–review and editing. K.N. Weilbaecher: Conceptualization, resources, data curation, supervision, funding acquisition, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing.
This study was supported in whole or in part by the following grants: NCI R01 CA216840 (to G.C. Fox, X. Su, J.L. Davis, Y. Xu, K.A. Kwakwa, M.H. Ross, F. Fontana, J. Xiang, A.K. Esser, G.M. Lanza, K.N. Weilbaecher), NCI P01 CA100730 (to G.C. Fox, X. Su, Y. Xu, M.H. Ross, F. Fontana, J. Xiang, A.K. Esser, K.N. Weilbaecher), NIH U54CA199092 (S. Achilefu, K.N. Weilbaecher), DoD BCRP W81XWH-16-1-0286 (S. Achilefu, X. Su, K.N. Weilbaecher), NIAMS R21 AR073507 (to D.J. Veis), NIAMS R01 AR070030 (to D.J. Veis), NHLBI R01 HL111163 (to J. Sivapackiam, V. Sharma), NHLBI R01 HL142297 (to J. Sivapackiam, V. Sharma), NIBIB P41 EB025815 (to J. Sivapackiam, V. Sharma), and training grants NIAMS T32AR060719 (to G.C. Fox and M.H. Ross) and NIGMS GM07200 (to G.C. Fox). Additional support was provided by the Genome Technology Access Center for sequencing (NIDDK P30-CA91842, ICTS/CTSA UL1TR002345); the Washington University Center for Cellular Imaging (WUCCI) for contributions to preparation, acquisition, and interpretation of electron microscopy data (CDI-CORE-2015–505, CDI-CORE-2019–813, Foundation for Barnes-Jewish Hospital 3770, NCI P30-CA091842, NIH ORIP OD021694); the Musculoskeletal Research Center for histology and radiography (NIAMS P30-AR057235); the Molecular Imaging Center at Washington University for bioluminescence imaging (NCI S10 OD02742, to S. Achilefu); the Alvin J. Siteman Cancer Center Biostatistics Shared Resource for statistical analysis of patient clinical data (NCI P30 CA091842); the Pat Burkhart Breast Cancer Fund, the Barnes-Jewish Foundation, the St. Louis Men'`s Group Against Cancer (to K.Weilbaecher). The authors also acknowledge Vecteezy and BioRender for graphics, and the Hope Center Alafi Neuroimaging Lab (NIH Shared Instrumentation Grant S10 RR027552). The authors thank Drs. Jason Mills, Kareem Azab, David Ornitz, Roberta Faccio, and Vivek Arora for their incisive suggestions and criticism. We gratefully acknowledge Crystal Idleburg, Lynne Collins, Julie Prior, Laura Luecking, Tom Walsh, Dr. Rosy Luo, Dr. Kathryn Tormos, Craig Smith, Dr. Erica Lantelme, Dorjan Brinja, Max Fisher, Dr. Sanja Sviben, Dr. Greg Strout, and Dr. Peter Bayguinov for their invaluable technical assistance and expertise.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.