Hormonal Regulation of Semaphorin 7a in ER⁺ Breast Cancer Drives Therapeutic Resistance

Breast cancer is the most commonly diagnosed cancer and the second leading cause of cancer-associated death in women in the United States. Estrogen receptor–positive (ER⁺) breast cancers comprise approximately 70% of all breast cancer cases. The development of effective antiestrogen-based endocrine therapies, including aromatase inhibitors (AI) and selective ER modulators (e.g., tamoxifen) or degraders (e.g., fulvestrant/Faslodex), has facilitated the successful treatment of ER⁺ primary tumors. However, depending on nodal involvement, 22% to 52% of women with ER⁺ breast cancer will experience disease recurrence within 20 years (1), frequently in a distant metastatic site, such as lung, liver, bone, or brain. These recurrent ER⁺ breast cancers uniformly develop resistance to endocrine therapies, rendering most therapeutic interventions in the clinical setting ineffective at preventing metastatic outgrowth. Therefore, novel molecular targets are needed to predict for and treat recurrent and/or metastatic ER⁺ breast cancers.

Semaphorin 7a (SEMA7A) is a glycosylphosphatidylinositol-anchored member of the semaphorin family of signaling proteins. SEMA7A has well-described roles in the promotion of axon guidance, immune modulation, and cellular migration (2–6). In adults, SEMA7A is expressed in the brain, bone marrow, lung, and skin, but is minimally expressed in many adult tissues, including the mammary gland (information via https://www.proteinatlas.org). Interestingly, in neuronal cells of the hypothalamus, SEMA7A expression is regulated by steroid hormones (7), whereas in pulmonary fibrosis models, SEMA7A is regulated by TGFβ (8). In its normal physiologic role, SEMA7A binds β1-integrin to activate downstream signaling cascades, including proinvasive MAPK/ERK and prosurvival PI3K/AKT pathways (2, 5, 8). Our group and others have described tumor-promotional roles for SEMA7A in mammary cancers, including tumor and host derived mechanisms of increased tumor growth, cellular migration, epithelial-to-mesenchymal transition, immune cell infiltration, remodeling of the blood and lymphatic vasculature, and metastatic spread (9–13). More recently, it has been reported that SEMA7A expression in lung cancer confers resistance to EGFR tyrosine kinase inhibition (14).

Studies to date have investigated the role of SEMA7A in breast cancer using preclinical models of triple-negative breast cancer (TNBC), which lack expression of ER (ER⁻), progesterone receptor (PR⁻), and amplification of HER2 (HER2⁻). Specifically, our studies in TNBC reveal that SEMA7A promotes TNBC cell growth, invasion, lymphangiogenesis, lymph vessel invasion, and metastasis (10, 15). However, our previous genomics analysis in patient data sets demonstrated that SEMA7A expression correlates with poor prognosis in patients with both ER⁻ and ER⁺ breast cancer and that SEMA7A expression levels were highest in luminal A tumors, which are generally ER⁺ and/or PR⁺ and HER2⁻ (9). Thus, understanding the role of SEMA7A in ER⁺ breast cancer is critical for a complete understanding of its role and impact on breast cancer progression. Our results presented herein show that SEMA7A mRNA expression is highest in ER⁺ breast cancers and that SEMA7A induces pleiotropic tumor-promotional effects in ER⁺ breast cancer cells. We also describe a novel relationship between hormone signaling and expression of SEMA7A in breast cancer. Additionally, we report for the first time that expression of SEMA7A confers resistance to standard-of-care endocrine therapy and to CDK 4/6 inhibition in preclinical ER⁺ models. Finally, using preclinical models, we identify a therapeutic vulnerability of SEMA7A-expressing tumors with the BCL2 inhibitor venetoclax (ABT-199/GDC-0199). These data provide promising evidence that SEMA7A⁺ER⁺ breast cancers that do not respond to endocrine therapies may respond to novel treatment regimens in the clinic.

MCF7 cells were obtained from K. Horwitz (University of Colorado Anschutz Medical Campus). T47D and BT474 cells were obtained from V. Borges and M. Borakove (University of Colorado Anschutz Medical Campus). MCF10DCIS cells were obtained from K. Polyak and A. Marusyk (Harvard University). MDA-MB-231 cells were obtained from P. Schedin (Oregon Health and Sciences University, Portland, OR). Tamoxifen-resistant T47D cells were obtained from A. Ward and C. Sartorius (University of Colorado Anschutz Medical Campus). Cell lines were validated by fingerprinting analysis at the University of Colorado Anschutz Medical Campus Molecular Biology Core. Cells were routinely tested for Mycoplasma contamination throughout the studies described herein with either Lonza MycoAlert PLUS or Mycoflour Mycoplasma Detection Kit (Molecular Probe), last tested September 4, 2019. Cells were cultured at 37°C and 5% CO₂. MCF7 and BT474 cells were grown in MEM with 5% heat-inactivated fetal bovine serum (FBS; Atlanta Biologicals) and 1% penicillin/streptomycin (P/S; Thermo Fisher Scientific). T47D and E0771 cells were grown in RPMI-1640, with 10% heat-inactivated FBS and 1% P/S. Experiments were performed with cells under passage 30. Forced suspension cultures were performed by coating plates with 12 mg/mL poly-2-hydroxyethyl methacrylate (Sigma). For exogenous SEMA7A treatments, SEMA7A was purified from SEMA7A-Fc–expressing MDA-MB-231 cells in collaboration with the CU Anschutz Medical Campus Protein Purification/MoAB/Tissue Culture Core.

For knockdown studies, SEMA7A-targeting shRNA plasmids (shSEMA7A KD1 and KD2) or a scrambled control (Scr; SA Biosciences, and the Functional Genomics Facility at the CU Anschutz Medical Campus) was transfected using X-tremeGENE transfection reagent (Millipore Sigma) as described by the manufacturer. SEMA7A overexpression in MCF7 cells was achieved using a SEMA7A-Fc plasmid (generous gift from R. Medzhitov, Yale University, New Haven, CT). A pcDNA3.1 empty vector control plasmid was obtained from H. Ford (CU Anschutz Medical Campus). All other plasmids were obtained from the Functional Genomics Core at the University of Colorado Anschutz Medical Campus. Knockdown and overexpression were confirmed via qPCR and Western blot analysis.

The reporter construct was generated by cloning a −820 to +182 base pair fragment of the SEMA7A promoter into the pGL3-Basic Luciferase reporter (Invitrogen). The four mutations for the ERE1/2 sites were as follows: TGACCC →TGTTCC, GGGTCC → GCATCC, GGGTCA → GCATCA, GGGCGG → TTACAG. Three SP1 sites were mutated: GGGCGG → TTACAG. Cells (50,000) were seeded in a 12-well plate and incubated for 5 minutes with 50 μL Opti-MEM (Invitrogen) and 3 μL GeneJuice (Millipore Sigma) per well. DNA (0.5–1 μg) was added to GeneJuice Transfection Reagent/serum-free medium mixture. Cells were harvested 48 hours after transfection using Reporter Lysis Buffer (Promega), unless treated with 10 nmol/L E2. If treated with E2, cells were serum starved for 48 hours in serum-free medium. Cells were then treated with 10 nmol/L E2 for 48 hours and harvested as stated above. Luciferase activity was normalized to total protein.

Fulvestrant, 4-hydroxytamoxifen, and 17-β estradiol (E2) were purchased from Millipore Sigma. For in vitro hormone treatments, cells were cultured in charcoal-stripped serum (Thermo Scientific) containing phenol red-free media for 3 to 5 days, then treated with 10 nmol/L E2 as determined by our dose response and the literature (16–18). For clonogenic and growth assays, fulvestrant and 4-hydroxytamoxifen treatments were at a final concentration of 7.5 nmol/L and 5 μmol/L, respectively. Palbociclib and SU6656 were acquired from MedChemExpress and treatments were at a final concentration of 2.5 and 5 μmol/L, respectively, in DMSO. LY294002 was purchased from Cell Signaling Technologies and in vitro treatments were at a final concentration of 2.5 μmol/L LY294002 in DMSO. In vitro venetoclax (AbbVie) was at a final concentration of 300 nmol/L in DMSO.

MCF7 cells (100,000) were seeded per well of a 6-well dish to assess growth in vitro. Cells were fixed with 10% NBF and stained with crystal violet at the indicated time points. 1,000 MCF7 cells or 5,000 T47D cells were seeded per well of a 6-well dish and assessed 7 to 14 days later for clonogenic assays. 3D matrigel and collagen cultures were performed as previously described (9, 19). Cleaved caspase-7 was measured using a Caspase Glo assay (Promega) as described by the manufacturer. Exogenous SEMA7A treatments were at a final concentration of 5 μg/mL.

RNA was isolated using the commercially available Quick-RNA kit (Zymo Research). RNA concentration and purity were assessed using a Nanodrop machine via absorbance at 280 nm, 260 nm, and 230 nm. RNA (1 μg) was used to synthesize cDNA with the Bio-Rad iScript cDNA Synthesis kit with the following conditions: 25°C for 5 minutes, 42°C for 30 minutes, and 85°C for 5 minutes.

cDNA was quantified via qPCR using Bio-Rad iTaq Universal SYBR Green Supermix and primers for SEMA7A (Bio-Rad PrimePCR) using the following conditions: 95°C for 3 minutes, then 40 cycles of 95°C for 15 sec and 60°C for 1 minute, 95°C for 30 seconds, and 55°C for 1 minute. Primer fidelity was assessed for each qPCR experiment using a 95°C melt curve. SEMA7A mRNA was detected with Bio-Rad primers (forward: CTCAGCCGTCTGTGTGTATT, reverse: CTCAGGTAGTAGCGAGAGTTTG). Primers for ESR1 (forward: 5-GAACCGAGATGATGTAGCCA, reverse: GTTTGCTCCTAACTTGCTCTTG) and PGR (forward: ATTTCATCCACGTGCCTATCC, reverse: CCTTCCTCCTCCTCCTTTATCT) were purchased from IDP. mRNA quantification was normalized using the geometric average of two reference genes: GAPDH (forward: CAAGAGCACAAGAGGAAGAGAG, reverse: CTACATGGCAACTGTGAGGAG) and RPS18 (forward: GCGAGTACTCAACACCAACA, reverse: GCTAGGACCTGGCTGTATTT). Data were analyzed using the relative standard curve method.

Western blots were performed by loading equal amounts of protein lysates as determined by Bradford assay. Samples were separated on a 10% tris-glycine SDS polyacrylamide gel and electroblotted onto a polyvinylidene fluoride membrane. Antibody information is provided in Supplementary Table S1. Blots were then exposed and developed on film. Color images were converted to black and white for presentation, and, if applied, color corrections were applied uniformly to all lanes of the blot to ease viewing of bands. No nonlinear intensity or color adjustments to specific features were performed. Multiple replicates were performed for each blot, and representative blots with corresponding quantifications normalized to loading controls are shown where appropriate.

2,500,000 SEMA7A overexpressing or empty vector control MCF7 cells were injected into the four mammary fat pads of 6- to 7-week-old female, estrogen-supplemented (20), NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ (NSG) mice (BCBC cat. # 4142, RRID:BCBC_4142; The Jackson Laboratory) or NOD-Prkdcem26Cd52Il2rgem26Cd22/NjuCrl (NCG) mice (IMSR cat. # CRL:572, RRID:IMSR_CRL:572; Charles River). Power calculations based on pilot studies were utilized to determine the number of animals per group. Tumors were measured using digital calipers every 3 to 4 days with the observer blinded to treatment group, and tumor volume was calculated as (L × W × W × 0.5). For estrogen-independent studies, animals received sham beeswax pellets. For drug treatment studies, fulvestrant dosing began when average tumor volume reached ∼200 mm² with the following dosing regimen: 2.5 mg fulvestrant in 10% ethanol and 90% sesame oil, injected subcutaneously every 5 days. This dose is equivalent to approximately 83 mg/kg per mouse and, using the FDA Dosing Guidance for Industry: Conversion of Animal Doses to Human Equivalent Doses, equivalent to 6.7 mg/kg in humans. For venetoclax studies, animals were randomized to treatment groups. Venetoclax was dissolved in phosal-50G:polyethylene glycol-400:ethanol at 6:3:1 to a final concentration of 12 mg/mL and administered via oral gavage. Tumors were harvested at designated study endpoints. Mammary glands with intact tumor and lungs were harvested, formalin-fixed, and paraffin embedded for IHC analysis. All animal studies were performed under the approval of the University of Colorado Anschutz Medical Campus Institutional Animal Care and Use Committee.

Tissues were formalin-fixed and paraffin-embedded as previously described (21). Antibody information is provided in Supplementary Table S1. For staining quantification, slides were scanned into the Aperio Image Analysis software (Leica). The total tumor area was quantified by removing necrotic and stromal areas. Custom algorithms for specific stains were used to quantify stain intensity or percent-positive stain as assessed by percent weak, medium, and strong determined using the color-deconvolution tool.

Kaplan–Meir plotter (22) was used to assess relapse-free survival (RFS) data for 5,143 breast cancer patients with a mean follow-up of 69 months. SEMA7A was queried and stratified into high and low expression groups using the autoselect best cutoff level function. The generated data were exported, graphed, and analyzed in GraphPad Prism to calculate hazard ratio (HR) with 95% confidence intervals and log-rank P values. The Gene Expression–Based Outcome for Breast Cancer Online platform (http://co.bmc.lu.se/gobo/) was used to query SEMA7A expression in 1,881 available sample tumors and a panel of 51 breast cancer cell lines. Data were stratified as high and low SEMA7A based on median expression and reported by molecular subtype. Outcomes data were reported only in patients with ER⁺ breast cancer. The Oncomine Platform (https://www.oncomine.org) was used to query SEMA7A in available breast cancer data sets. The P value threshold was set to 0.05. All fold changes and gene ranks were accepted. Data are reported from the molecular subtype analysis, correlation with stage and grade, and drug-sensitivity screens.

Five thousand SEMA7A overexpressing or empty vector control MCF7 cells were seeded per well in a 96-well plate and allowed to adhere overnight. Media were replaced with fresh drug- or vehicle-treated media the next morning. Drug treatments included 7.5 nmol/L fulvestrant, 5 μmol/L 4-hydroxytamoxifen, 2.5 μmol/L palbociclib, 10 μmol/L LY294002, 5 μmol/L Su6656, and growth was assessed in an IncuCyte live-cell analysis system at the University of Colorado Cancer Center Tissue Culture Core. Confluence was reported every 4 hours.

Fulvestrant-resistant cell lines were generated by injecting 2,500,000 SEMA7A overexpressing or empty vector control MCF7 cells into contralateral fourth mammary fat pads of a 6-week-old female, estrogen-supplemented (20) NCG mouse (IMSR cat. # CRL:572, RRID:IMSR_CRL:572; Charles River). Fulvestrant treatments began 7 days later with the following dosing regimen: 2.5 mg fulvestrant in 10% ethanol and 90% sesame oil, injected subcutaneously every 5 days. Forty-eight days after implantation, tumors were harvested and minced in digestion mixture: MCF7 media with 500 units/mL collagenase type II and IV, 20 μg/mL DNase (Worthington Biochemical), and 1× antibiotic–antimycotic (Thermo Fisher). Disassociation mixture was incubated at 37°C for 1 hour. Dissociated cells were centrifuged at 1,500 RPM for 7 minutes, then resuspended and plated in standard culture media with 1× antibiotic–antimycotic and returned to the incubator overnight. The next day, 7.5 nmol/L fulvestrant was added to the media, and cells were maintained in these culture conditions. Ex vivo fulvestrant resistance was confirmed with clonogenic assays as described above.

Long-term estrogen-deprived (LTED) cells were cultured in phenol red-free media (MEM for MCF-7 and RPMI-1640 for T47D) supplemented with 5% dextran charcoal-stripped FBS for 6 months.

Four thousand SEMA7A overexpressing or empty vector control MCF7 cells were seeded per well in a 96-well plate and allowed to adhere for 6 hours. Vehicle or increasing concentrations of 4-hydroxytamoxifen or fulvestrant were added. Growth was assessed in BioSpa live-cell imaging system (BioTek). Confluence was reported every 4 hours.

Estradiol ELISA kit (Cayman Chemical) was used to measure serum estradiol levels from estrogen-supplemented or sham beeswax pellet control mice according to the manufacturer's instructions.

MCF7 cells were enzymatically dissociated in Accutase cell detachment solution (STEMCELL Technology: cat. # 07920) for approximately 10 minutes at 37°C. Single cells were stained with anti-mouse/human CD44-AF647 (BioLegend: cat. # 103018) and anti-human CD24-FITC (BioLegend: cat. # 311104) for 30 minutes at 4°C in the dark. Cells were analyzed on the YETI-ZE5 flow cytometer, and data were analyzed using Kaluza software. Cells were delineated from debris by SSC-A/FSC-A, and singlets were identified by FSC-H/FSC-A. Single cells were subsequently assessed for CD44/CD24 expression.

All cell line and xenograft experiments were performed in biological duplicate or triplicate. Data are expressed as means ± SEM for animal studies and ± SD for in vitro studies. All statistical analyses were performed on raw data. An unpaired two-tailed Student t test was performed for analyses comparing two groups, whereas a one-way ANOVA followed by a Tukey–Kramer test was used for multiple comparisons of groups of three or more. For tumor volume data and IncuCyte analyses, a repeated-measures one-way ANOVA was used. For Kaplan–Meier analyses, the P value was determined using a log-rank test. All P values less than or equal to 0.05 were considered statistically significant, and all statistical analyses were performed in GraphPad Prism 7.01.

To determine if SEMA7A is expressed in ER⁺ breast cancer, we assessed SEMA7A mRNA expression using the Gene-Expression–Based Outcome for Breast Cancer Online (GOBO) tool. We observed that 67% of luminal subtype cell lines, which are typically ER⁺, exhibit increased SEMA7A expression versus 36% to 50% of the basal subtypes (which are typically ER⁻; Fig. 1A). Average expression was higher when ER⁺ cell lines were compared with TNBC cell lines and when luminal cell lines were compared with basal subtype (Fig. 1B). SEMA7A mRNA expression is also increased in ER⁺ patient tumors (Fig. 1C), and we confirmed that SEMA7A was overexpressed and/or had a higher copy-number rank in ER⁺ or PR⁺ disease in seven additional data sets using Oncomine (Supplementary Table S2). Additionally, SEMA7A was highest in stage III and IV, grade 2 and 3, and lymph node–positive (LN⁺) disease in multiple data sets (Supplementary Table S2). We next examined a role for SEMA7A status in predicting outcomes for patients with ER⁺ breast cancer; patients with high tumor expression of SEMA7A had decreased overall and distant metastasis-free survival (Fig. 1D and E; Supplementary Fig. S1A and S1B). Additionally, using KM plotter, we further stratified RFS by LN positivity in the ER⁺ (HER2^+/−) subset and observed that high SEMA7A is associated with decreased RFS in patients with both LN⁺ and LN⁻ status (Fig. 1F and G). In fact, the hazard ratio for patients whose tumors were ER⁺ (HER2^+/−) SEMA7A⁺ and with LN⁻ status was the highest. We then confirmed SEMA7A protein expression in ER⁺ lines (MCF7, T47D, and BT474), ER⁻ lines (DCIS, MDA-MB-231, and SUM159), and MCF10A breast epithelial cells derived from human fibrocystic disease (23) via immunoblot analysis. SEMA7A expression was observed in two of the three ER⁺ cell lines and one of the three ER⁻ when compared with MCF10A cells (Supplementary Fig. S1C).

To test the hypothesis that SEMA7A expression could be induced by estrogen (E2) stimulation, we treated ER⁺ breast cancer cells in estrogen-deprived conditions with increasing, physiologically relevant (24), concentrations of estrogen. In both MCF7 and T47D cells, E2 resulted in increased SEMA7A protein expression (Fig. 2A and B). Then, to determine whether SEMA7A expression could be promoted by ER-dependent transcription, we performed an in silico analysis of the SEMA7A promoter using the ConTra (conserved transcription factor binding site) web server (25). We identified two highly conserved estrogen response element (ERE) half sites adjacent to two specificity protein 1 (SP1) consensus binding sites (Fig. 2C). SP1 has been shown to facilitate ER binding to ERE half sites in gene promoter regions to promote transcription of estrogen target genes (26); thus, we then utilized a SEMA7A luciferase reporter construct to measure promoter activity with E2 treatment and observed increased luminescence with 10 nmol/L E2 (Fig. 2D). We also observed increased SEMA7A mRNA after treatment with 10 nmol/L E2 (Supplementary Fig. S1D). Next, we performed luciferase reporter assays with constructs containing point mutations in the ERE and SP1 sites to show that both are required for induction of SEMA7A promoter activity by E2 (Fig. 2E and F). To determine whether SEMA7A expression was dependent upon ER, we treated MCF7 and T47D cells with fulvestrant, which decreased expression, but was not sufficient to completely block expression of SEMA7A under normal growth conditions (Fig. 2G). Then, to mimic conditions resulting from AI therapy, we tested whether long-term estrogen deprivation (LTED) of MCF7 and T47D cells would result in downregulation of SEMA7A. Surprisingly, the cells that were deprived of estrogen exhibited marked upregulation of SEMA7A compared with the wild-type (WT) controls (Fig. 2H). Additionally, T47D cells that are tamoxifen resistant (TAMR) exhibit upregulation of SEMA7A (Fig. 2I). These results suggest that E2 induces SEMA7A expression, but that SEMA7A expression can also bypass estrogen dependence. Consistent with this, we observe that MCF7 cells, which normally require estrogen supplementation for growth in vivo, can grow in the absence of estrogen supplementation when SEMA7A is overexpressed (Fig. 2J; Supplementary Fig. S1E and S1F).

To determine if SEMA7A promotes ER⁺ breast cancer progression via increasing tumor cell viability or proliferation, we engineered MCF7 cells to overexpress SEMA7A. We confirmed a 3.6-fold increase in SEMA7A protein (SEMA7A OE; Fig. 3A) and observed increased cell growth in two-dimensional cultures (Fig. 3B), clonogenic assays (Fig. 3C), and increased BrdUrd incorporation in a 3D organoid system that we have previously characterized to model tumor cell growth and invasion into the surrounding tissue (19), which contains matrigel to model the basement membrane and collagen, which is required for tumor cell invasion (Fig. 3D). We also generated two independent shRNA-mediated knockdown (KD) variants of SEMA7A and confirmed decreased SEMA7A protein in MCF7 cells (Fig. 3E; Supplementary Fig. S1G). We observed that SEMA7A KD cells were less confluent over a 72-hour time course (Fig. 3F), exhibited decreased colony formation in a clonogenic assay (Fig. 3G), and SEMA7A KD 3D organoids have significantly reduced proliferation via BrdUrd incorporation (Fig. 3H).

Kaplan–Meier analysis of patients treated with tamoxifen revealed high expression of SEMA7A correlated with decreased RFS (Fig. 4A). To determine if tamoxifen was ineffective for SEMA7A-expressing cells, we cultured MCF7 SEMA7A OE cells in the presence of 4-hydroxytamoxifen. We sought to determine an effective (IC₅₀) dose of 4-hydroxytamoxifen in both empty vector (EV) and OE cells, and we identified a dose, 5 μmol/L, that decreased EV cell viability in full serum and in estrogen-depleted conditions (Supplementary Fig. S2A and S2B). We observed that SEMA7A OE cells grew significantly better than EV controls with 4-hydroxytamoxifen treatment (Fig. 4B; Supplementary Fig. S2C). We also performed a drug titration for fulvestrant and identified 7.5 nmol/L as sufficient to decrease MCF7 EV growth (Supplementary Fig. S2D). We then applied this dose to SEMA7A OE cells. These fulvestrant-treated SEMA7A OE cells grew as well as vehicle-treated EV controls over time and in clonogenic assays (Fig. 4C and D; Supplementary Fig. S3A and S3B). Additionally, exogenous SEMA7A increased clonogenic outgrowth of both MCF7 and T47D cells, despite the presence of fulvestrant (Fig. 4E; Supplementary Fig. S3C). We further observed that MCF7 SEMA7A OE cells can grow in vitro in estrogen-depleted conditions (Supplementary Fig. S3D and S3E). We next tested how SEMA7A OE tumors grow in vivo with endocrine therapy treatments. We chose fulvestrant as we were able to achieve an IC₅₀ dose in the presence of estrogen (Supplementary Fig. S2D). We implanted MCF7 EV and SEMA7A OE cells, allowed tumors to establish, then began administering fulvestrant when the tumors reached 200 mm³. Fulvestrant slowed the growth of control tumors, whereas the SEMA7A OE tumors continued to grow at the same rate (Fig. 4F; Supplementary Fig. S3F). At study completion, we verified SEMA7A overexpression by IHC (Supplementary Fig. S3G). Consistent with a drug-resistant phenotype, SEMA7A OE tumors exhibited increased proliferation as measured by Ki67 staining at study completion (Fig. 4G; Supplementary Fig. S3H) and were overall less necrotic (Fig. 4H; Supplementary Fig. S3I). Interestingly, ER⁺ tumor metastases frequently lose expression of ER under conditions of long-term estrogen deprivation (27), and our SEMA7A OE tumors exhibited decreased ER staining intensity by semiquantitative IHC (Fig. 4I and J), which could be a sign of ER independence and explain the lack of response to fulvestrant. Furthermore, ESR1 and PGR—a known ER target gene—downregulation were confirmed in our MCF7 SEMA7A OE cells (Supplementary Fig. S3J). These results suggest that SEMA7A expression in ER⁺ breast cancer cells renders them less reliant on ER signaling for growth, which also renders them resistant to endocrine therapy and able to disseminate and seed metastases in the presence of tamoxifen, AIs, or fulvestrant. Consistent with this, we observe increased lung metastasis in the SEMA7A OE group (Fig. 4K and L).

Our previous results in ER⁻ breast cancer cells show that SEMA7A induces mesenchymal phenotypes (15), and SEMA7A OE in ER⁺ MCF7 cells resulted in cells adapting a more mesenchymal morphology (Supplementary Fig. S3K). We also observed increased invasion of SEMA7A OE MCF7 cells in a transwell filter assay (Fig. 5A). We next utilized the 3D model described above where we previously demonstrated that SEMA7A promotes invasion of ER⁻ tumor cells (9, 19). Consistent with our hypothesis, SEMA7A OE cells exhibited evidence of invasion as early as 5 days after seeding, when most organoids in the control group remained in rounded, noninvasive spheroid structures (Fig. 5B). Conversely, evidence of organoids with invasive protrusions appeared after 10 days in the control cells and were significantly less abundant in SEMA7A knockdowns (Fig. 5C).

After local invasion and intravasation, metastatic tumor cells exit the primary site and survive in nonadherent conditions, which would induce death of normal epithelial cells via a process termed “anoikis” (28). Thus, we utilized forced suspension cultures to determine whether SEMA7A is upregulated in nonadherent conditions. We observed highest SEMA7A expression after 72 hours in suspension, suggesting that SEMA7A may be required for anoikis resistance (Fig. 5D). To determine if SEMA7A expression promotes cell survival in forced suspension conditions, we measured cell death via cleaved caspase-7 to reveal decreased cell death in SEMA7A OE cells (Fig. 5E). In contrast, KD cells showed a ∼2-fold increase in cell death in both attached and forced suspension conditions (Fig. 5F), suggesting that SEMA7A may mediate protection from cell death both at the primary site and in circulation during metastasis. Collectively, these data suggest that SEMA7A expression promotes local invasion and cell survival in circulation. This may be one mechanism by which SEMA7A promotes relapse in patients.

Based on our observation that SEMA7A drives growth and metastasis of ER⁺ tumors and are resistant to endocrine therapy, we postulated that CDK 4/6 inhibitors, which have recently been approved and proven effective for patients with metastatic ER⁺ breast cancer, could inhibit growth of SEMA7A-expressing cells. As expected, when treated with 2.5 μmol/L of the CDK 4/6 inhibitor palbociclib, MCF7 EV control cells exhibited a significant reduction of growth and clonogenicity in vitro compared with vehicle treatment; however, SEMA7A OE cells grew significantly better with palbociclib treatment (Fig. 6A and B; Supplementary Fig. S4A). We therefore sought to identify therapies that may have efficacy for treating patients with SEMA7A-expressing tumors. Using an Oncomine query to identify potential compounds and targetable pathways, we identified six compounds predicted to elicit sensitivity of SEMA7A⁺ tumors, all of which target proteins in either the MAPK or PI3K/AKT signaling pathways (Supplementary Table S3). Notably, published studies describe these pathways as downstream signaling mediators of SEMA7A activity (2, 5, 8). Thus, we tested whether growth of SEMA7A OE cells is inhibited by commercially available small-molecule inhibitors that target Src and Akt, SU6656 and LY294002, respectively. We observed significantly decreased growth of SEMA7A OE cells with both inhibitors (Supplementary Fig. S4B and S4C). However, as these compounds are not clinically available, we sought to identify a compound that would inhibit the downstream prosurvival function induced by these pathways. Venetoclax is an orally available inhibitor of the antiapoptotic BCL2 protein that has recently demonstrated synergy in combination therapy approaches in patients with ER⁺ breast cancer (29). We first identified that our SEMA7A OE cells expressed BCL2 (Fig. 6C). Then, we tested the sensitivity of SEMA7A OE cells to venetoclax at a clinically relevant dose. Proliferation of SEMA7A OE cells was decreased with venetoclax treatment as compared with vehicle treatment and to EV control cells treated with venetoclax, whereas clonogenic outgrowth was decreased in both EV and SEMA7A OE cells (Fig. 6D and E; Supplementary Fig. S4D). To test whether fulvestrant-resistant SEMA7A OE tumors respond to venetoclax treatment in vivo, we implanted MCF7 EV control or SEMA7A OE cells into NCG mice. Tumors were allowed to grow to an average of 200 mm³, and then we initiated fulvestrant treatment. As expected, the SEMA7A OE tumors continued to grow, whereas tumor growth plateaued in the control group. After separation of the curves was observed, we administered vehicle or venetoclax treatment in combination with fulvestrant. We observed a dramatic decrease in tumor growth, tumor growth rate, final tumor volume, and decreased Ki67⁺ tumor cells in the SEMA7A OE, which was surprisingly not observed in the EV control group (Fig. 6F–H; Supplementary Fig. S4E and S4F). Next, we sought to examine whether BCL2 expression was affected by the fulvestrant treatment, as BCL2 is an ER target gene. To this end, we harvested cells from fulvestrant-treated tumors and cultured them ex vivo in the presence of fulvestrant (Supplementary Fig. S5A and S5B). Immunoblot analysis of fulvestrant-resistant EV (FR-EV) and SEMA7A OE (FR-OE) cells revealed similar levels of BCL2 as the parentals (Fig. 6I). However, the FR-OE cells were enriched for CD44⁺CD24⁻ cell populations (Fig. 6J; Supplementary Fig. S5C–S5E), which exhibit increased capacity for self-renewal, generation of heterogeneous progeny, enhanced invasive phenotypes and drug resistance (30, 31); these results suggest that SEMA7A OE may result in enrichment for drug-resistant stem cells in response to endocrine therapy, which may also promote metastasis. To determine whether there is potential for venetoclax to prevent metastasis of SEMA7A⁺ER⁺ tumors, we measured lung metastasis in our cotreated animals and found a significant decrease with the addition of venetoclax to fulvestrant treatment (Fig. 6K). Thus, our results suggest that venetoclax may be a viable treatment option for inhibiting growth, recurrence, and metastasis of SEMA7A-expressing ER⁺ breast cancers, which are likely to be resistant to endocrine therapies.

Determining whether patients with ER⁺ breast cancer will or will not respond to endocrine therapy and identifying novel treatment options for patients with endocrine-resistant ER⁺ breast cancer is of utmost importance, as patients with ER⁺ disease comprise the largest percentage of breast cancer patients and the largest number of breast cancer–related deaths. The semaphorin family of proteins have multiple, and variable, described roles in cancer, with some demonstrating tumor-suppressive functions and others demonstrating tumor-promotional roles (32). The data presented in this study contribute to an emerging body of evidence describing a role for SEMA7A in tumor progression. Our lab and others have recently begun to characterize SEMA7A as a mediator of breast/mammary, melanoma, and lung cancer (9–14). In this report, we describe a novel role for SEMA7A in ER⁺ breast cancer recurrence and metastasis, where SEMA7A expression confers resistance to endocrine therapy and predicts for poor response in patients with ER⁺ disease. In previously published data, we showed that SEMA7A promotes numerous tumor-promotional phenotypes in ER⁻ preclinical models, including tumor cell growth, motility, invasion, and cell survival (9, 15). We also reported that SEMA7A induces macrophage-mediated remodeling of lymphatic vasculature, as well as tumor cell dissemination via the lymphatics, and others have demonstrated SEMA7A-dependent angiogenesis in ER⁻ breast cancer models (10, 11).

Motility and invasion are known hallmarks of metastatic cells (28). However, as cells venture through the metastatic cascade, they must also adapt to many different microenvironments such as certain hostile conditions, including anchorage-independent conditions during travel through the vasculature or lymphatics. We now show that SEMA7A expression promotes anoikis resistance and consequent anchorage-independent survival and we extend our observations to show that SEMA7A overexpression can promote spontaneous and experimental lung metastasis. Additionally, we suggest that the pleiotropic effects promoted by SEMA7A may give SEMA7A-expressing tumors a selective advantage under the stress of endocrine therapies, which results in long-term hormonal depravation intended to halt tumor cell growth and induce cell death. Our results support that SEMA7A expression allows tumor cells to bypass reliance on ER, rendering them reliant instead upon SEMA7A mediated cell survival. This is further evidenced by our observations that SEMA7A-expressing tumors downregulate expression of ER—a mechanism that warrants further investigation. Furthermore, published studies report that SEMA4C promotes MCF7 tumor growth, viability, metastasis, and renders ER⁺ cells nonresponsive to tamoxifen via downregulation of ER (33), suggesting that this mechanism may extend to other multiple semaphorin family members.

Metastatic disease in patients with ER⁺ breast cancer frequently exhibits endocrine resistance, and many metastases downregulate ER or express mutant forms of the ESR1 gene. Although the prevalence of ER loss has been debated, a recent meta-analysis reveals approximately 23% of ER⁺ tumors have corresponding ER⁻ metastases (27). Further, tumors with low expression of ER (1%–10% positive) are reported to behave more like ER⁻ tumors, and patients with ER-low tumors have significantly lower disease-free survival rates compared with patients with ER high (11%–100% positive) tumors (34). These results are consistent with our observations that SEMA7A expressing cells exhibit prometastatic phenotypes and are also resistant to therapy and downregulate ER. We suggest that initial expression of SEMA7A in hormone receptor positive disease is promoted by cooperation between ER and SP1 at the SEMA7A promoter, as we demonstrate that SEMA7A expression can be promoted by estrogen treatment in a dose-dependent manner and that both the ERE half sites and SP1 sites are required for induction of promoter activity by estrogen. However, SEMA7A expression is not completely inhibited with ER degradation in ER⁺ breast cancer cells, suggesting additional mechanisms are at play to regulate SEMA7A expression. Studies of other semaphorins have described hormone receptor regulation. In ovarian cancer, proproliferative SEMA4D is positively regulated by ER (35). Conversely, expression of SEMA3B and SEMA3F, which are both growth suppressive, in ovarian cancer is positively regulated by E2 and progesterone (36). Finally, SEMA3C is transcriptionally regulated by the androgen receptor (AR) in prostate cancer cells (37) and promotes growth, migration, and invasion in breast cancer (38), is detectable in metastatic lung cancer cells (39), and promotes survival of gastric cancer and glioma stem cells (40, 41). SEMA3C also has a described role in promoting drug resistance in ovarian cancer (42). Consistent with a mechanism for SEMA7A in drug resistance, our observation that ER levels in SEMA7A OE tumors are decreased after fulvestrant treatment suggests that fulvestrant had the intended effect, but that fulvestrant alone was not sufficient to control growth in SEMA7A-expressing tumors, suggesting a bypass mechanism. Taken together, these studies support that future investigations, utilizing chromatin immunoprecipitation, will be necessary to show direct promoter binding of SEMA7A by multiple nuclear hormone receptors, including ER, PR, and AR.

In addition to endocrine therapy resistance, we demonstrate that SEMA7A expression confers resistance to the CDK 4/6 inhibitor palbociclib, which is used in combination with endocrine therapy for first-line treatment of metastatic disease or in subsequent lines following disease progression (43). Our data suggest that patients with SEMA7A⁺ tumors may be at high risk for relapse and poor response to current standard-of-care therapies, and that novel therapeutic strategies could be applied for patients preselected by SEMA7A status. Interestingly, Kinehara and colleagues recently described that SEMA7A promotes resistance to standard-of-care EGFR tyrosine kinase inhibitors in lung cancer through antiapoptotic β1-integrin-mediated ERK signaling; SEMA7A was thus proposed as a predictive marker for response to TKI therapy and potential therapeutic target for resistant tumors (14). This report, along with our findings, suggests that SEMA7A may confer resistance to multiple targeted therapies in numerous tumor types; therefore, direct targeting and therapeutic vulnerabilities of SEMA7A⁺ tumors warrant additional investigation in a clinical setting. Moving toward this goal, results presented herein suggest that SEMA7A⁺ tumors may be targetable via other mechanisms. Specifically, SEMA7A has demonstrated the ability to activate PI3K/AKT and MAPK/ERK signaling (2, 5, 8). We now report that SEMA7A OE cells have increased expression of the antiapoptotic protein BCL2. We therefore tested if SEMA7A⁺ tumor cells exhibit sensitivity to selective inhibitors of these pathways, including the BCL2 inhibitor venetoclax. Our results showing that SEMA7A⁺ tumors cells exhibit reduced growth and viability after exposure to these therapeutics support that targeting downstream effectors of SEMA7A may prove an effective strategy for recurrent SEMA7A⁺ tumors, which are resistant to currently available therapeutic regimes. We also show, for the first time, that when treated with fulvestrant, SEMA7A expressing tumors acquire stem cell markers consistent with a drug-resistant phenotype. Previous reports have described increased BCL2 expression in fulvestrant-resistant MCF7 cells, and fulvestrant sensitivity can be restored with a BCL2 inhibitor or BCL2 knockdown (44). Additionally, the first clinical study of the BCL2 inhibitor venetoclax in solid tumors revealed that patients with metastatic ER⁺ breast cancer received great benefit from this drug in combination with tamoxifen, and a recent report describes a benefit of adding venetoclax to fulvestrant/palbociclib in preclinical ER⁺ models (45). Although we have not explored this triple combination in our current studies, future studies will explore whether venetoclax can reverse the resistance to palbociclib that we observe in our SEMA7A OE cells. Because venetoclax is clinically available, administered orally, and well tolerated, we propose that novel clinical trials could test whether patients with ER⁺SEMA7A⁺ cancer would benefit from combination treatment in both the adjuvant and metastatic settings. Finally, clinical trials assessing SEMA4C as a biomarker for breast cancer diagnosis and relapse are already under way (ClinicalTrials.gov identifiers: NCT03663153 and NCT03662633). Because similar phenotypes are observed in preclinical models with SEMA4C and SEMA7A, we propose that SEMA7A may also warrant investigation in the clinic as a predictive marker of ER⁺ breast cancer recurrence.

No disclosures were reported.

L.S. Crump: Conceptualization, formal analysis, funding acquisition, investigation, writing–original draft, writing–review and editing. G.L. Wyatt: Resources, investigation, methodology, writing–review and editing. T.R. Rutherford: Investigation and methodology. J.K. Richer: Resources, supervision, funding acquisition, writing–review and editing. W.W. Porter: Resources, supervision, funding acquisition, writing–review and editing. T.R. Lyons: Conceptualization, resources, formal analysis, supervision, funding acquisition, investigation, methodology, writing–original draft, writing–review and editing.

This work was supported by Colorado Clinical and Translational Sciences Institute Grant TL1 TR001081 (to L.S. Crump), NIH Grants F31 CA236140 (to L.S. Crump/T.R. Lyons), NIH/NCI R01 CA211696-01A1 (to T.R. Lyons), CU Anschutz RNA Bioscience Initiative RNA-seq Pilot Grant Program (to T.R. Lyons), NIH/NICHD R01-HD083952-01 (to W.W. Porter), NIH/NCI R01 CA187733-01 (to J.K. Richer), NIH/NCI R21-CA197896 (to W.W. Porter), DOD BCRP W81XWH-16-1-0422 (to J.K. Richer), ACS RSG-16-171-010CSM (to T.R. Lyons), and Department of Medicine Outstanding Early Career Scholars Award (to T.R. Lyons). The authors acknowledge the CU Anschutz Cancer Center Protein Production/MoAB/Tissue Culture Shared Resource, the Biorepository Core Facility, and the University of Colorado Cancer Center Support Grant (P30CA046934) for SEMA7A protein production, IncuCyte use, and Aperio Software use. Contents are the authors' sole responsibility and do not necessarily represent official NIH views. The authors also thank the following persons for technical support and feedback in the production of this manuscript: Alan Elder, Alexander Stoller, Veronica Wessells, Michelle Borakove, Virginia Borges, Chloe Young, Sarah Tarullo, Annie Pham, Paris Franz, Craig Jordan, and Brett Stevens. Graphical abstract created with BioRender.com.

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.