Hypoxia Selectively Enhances Integrin α₅β₁ Receptor Expression in Breast Cancer to Promote Metastasis

This article is featured in Highlights of This Issue, p. 633

Increased cell proliferation and oxygen consumption result in lower oxygen availability in solid tumors as compared with normal tissue (1, 2). Intratumoral hypoxia has been associated with invasion, metastasis, treatment failure, and patient mortality (3, 4). Cancer cells survive and adapt to hypoxic conditions, in part, through the activation of hypoxia-inducible factor 1 (HIF-1) and HIF-2, which induce the expression of gene products involved in angiogenesis, glucose utilization, invasion, and metastasis (5). HIF-1 is a heterodimeric protein composed of a constitutively expressed HIF-1β subunit and an O₂-regulated HIF-1α subunit (6, 7). In mouse models, inhibition of HIF expression impedes breast cancer growth, angiogenesis, and metastasis (8–12). In human breast cancer biopsies, increased HIF-1α protein levels are associated with an increased risk of metastasis and mortality (13–17). Our previous work implicated hypoxia in alterations of the composition and stiffness of the intratumoral extracellular matrix (ECM), which promoted metastasis (18–20). The ECM provides intracellular signaling cues and also provides contact points for cell migration. Specific cell–ECM interactions are critical for cell survival, and altered cell–matrix adhesion is a classic hallmark of neoplasia (21, 22). We hypothesized that hypoxia could coordinately induce both ECM alterations and the expression of integrins that interact with the altered ECM (23). The major constituents of the ECM are collagens, elastin, fibronectin, and laminins. Collagen is the most abundant fibrous protein within the interstitial ECM. Fibronectin is required for ECM organization and fibrillogenesis and mediates cell attachment. Integrin dimers, as well as discoidin domain receptors, and syndecans mediate adhesion to the ECM (24, 25).

Twenty-four distinct integrin heterodimers are expressed in mammals as a result of combinatorial association of 18 α and 8 β subunits. ECM ligands bind to the α subunit and activate intracellular signaling events via the β subunit to integrate extracellular and intracellular events necessary for cell motility and invasion (26, 27). Integrin expression patterns are cell-type specific and vary with microenvironmental context (26, 27). Many integrins are expressed at low or undetectable levels in adult epithelia, but are upregulated in tumors (28). Integrins expressed in epithelia include several β₁ integrins (α₁β₁, α₂β₁, α₃β₁, α₅β₁, α₆β₁, and α₉β₁), α₆β₄, α_vβ₅, and α_vβ₆ (29).

Hypoxia has been shown to enhance ITGB1 (integrin β₁) expression in fibroblasts (30), α₅β₁ and α_vβ₅ in endothelial cells (31), α_vβ₃ in melanoma cells (32), β₂ expression in leukocytes (33, 34), and α₅ expression in colon cancer cells (35). In human mesenchymal stem cells, hypoxia induces the expression of various integrins (α₁, α₃, α₆, α₁₁, α_v, β₁, and β₃; ref. 36). Our goal was to determine how hypoxia regulates integrin expression in human breast cancer and to investigate whether targeting integrin receptors might be an effective method for blocking the metastatic response potentiated by changes to the ECM during tumor progression. Several integrin inhibitors are advancing through clinical trials with positive clinical findings (28). To date, no study has investigated the induction of each individual integrin subunit in a large cohort of patients and/or in a large panel of breast cancer cell lines.

To determine whether patients with highly hypoxic tumors might benefit from blocking integrin receptors and to determine the integrin subunit(s) that would be suitable as a means to target hypoxic cells, we examined a large cohort of breast cancer patients and a panel of 20 well-characterized cell lines for integrin expression. Initially, we utilized a bioinformatics approach and interrogated two datasets: The Cancer Genome Atlas (TCGA; ref. 37) and the Gene Expression-Based Outcome for Breast Cancer Dataset (GOBO; ref. 38) and identified candidate integrins that might be induced under hypoxic conditions based on a significant correlation of their expression pattern in human breast cancers with that of known HIF target genes. The results were confirmed using a panel of 20 breast cancer cell lines exposed to hypoxic conditions. Expression of integrin α₅ subunit (ITGA5) was uniquely hypoxia-induced in 19 of the 20 cell lines assayed and had the strongest correlation with the HIF gene signature. HIFs were required for transcriptional activation of the ITGA5 gene. Surface expression of the α₅β₁ receptor was required for three-dimensional (3D) cell migration and migration of cells within a multicellular spheroid, but surprisingly did not alter 2D cell migration. Inhibition of α₅β₁ expression abrogated invasion and motility of cells within a spheroid embedded in a collagen and fibronectin matrix. Importantly, inhibition of α₅β₁ expression decreased metastasis in mouse models of breast cancer, suggesting that α₅ inhibition may be an effective treatment strategy for breast cancer patients.

All cell lines except SUM159 and SUM149 were obtained from the ATCC and cultured as described by the ATCC. The SUM149 and SUM159 cells were gifts from the Sukumar laboratory and were authenticated by STR sequencing and confirmed to be mycoplasma free. Hypoxic cells were maintained at 37°C in a modular incubator chamber (Billups–Rothenberg) flushed with a gas mixture containing 1% O₂, 5% CO₂, and 94% N₂.

Female 5- to 7-week-old NOD/SCID or BALB/c (Charles River Laboratories) mice were used according to protocols approved by the Johns Hopkins University Animal Care and Use Committee. Mice were anesthetized, and 2 × 10⁶ MDA-MB-231 cells or 5 × 10⁵ 4T1 cells were injected into the mammary fat pad (MFP). Tumors were measured in three dimensions (a, b, and c), and volume (V) was calculated as V = abc × 0.52. Tumors, ipsilateral axillary lymph nodes, and lungs were harvested, formalin fixed, paraffin embedded, and used for IHC staining. Lung tissue was used to isolate genomic DNA for qPCR to quantify human HK2 and mouse 18S rRNA gene sequences.

Aliquots of whole-cell lysates were prepared in NP-40 buffer (150 mmol/L NaCl, 1% NP-40, 50 mmol/L Tris-HCl, pH 8.0) and fractionated by 8% SDS-PAGE. Antibodies against HIF-1α and ITGA5 (BD Biosciences), HIF-2α (Novus Biologicals), β-actin, and ITGB1 (Santa Cruz Biotechnology) were used.

Paraffin-embedded tissue sections were dewaxed and hydrated. LSAB+ System (Dako) was used for ITGA5, HIF-1α, and vimentin IHC staining according to the manufacturer's instructions. Inflated lung sections were stained with hematoxylin and eosin (H&E) to detect metastatic foci as described previously (11, 12). Image analysis of vimentin-stained lymph node tissue sections was conducted as described previously (20).

The pLKO.1-puro lentiviral vectors encoding shRNA targeting human and mouse ITGA5 were purchased from Sigma-Aldrich. The pLKO.1-puro lentiviral vectors encoding shRNA targeting human HIF-1α and HIF-2α were described previously (39). The recombinant vectors were cotransfected with plasmid pCMV-dR8.91 and a plasmid-encoding vesicular stomatitis virus G protein into 293T cells using Polyjet. Filtered viral supernatant collected 48 hours posttransfection was added to MDA-MB-231 cells with 8 μg/mL polybrene (Sigma-Aldrich). Puromycin (0.5 μg/mL) was added to the medium of cells transduced for selection. Following selection, cells were pooled together for use.

Mice were euthanized, and India ink (15%) was injected into the lungs through the trachea. The lungs were fixed in Fekete's solution (100 mL of 70% alcohol, 10 mL of formalin, and 5 mL of glacial acetic acid) at room temperature.

Total RNA was extracted from cells using TRIzol (Invitrogen) and treated with DNase I (Ambion). One microgram of total RNA was used for first-strand DNA synthesis with the iScript cDNA Synthesis System (Bio-Rad). qPCR was performed using human-specific primers and iTaq SYBR Green Universal Master Mix (Bio-Rad). The expression of each target mRNA relative to 18S rRNA was calculated on the basis of the threshold cycle (C_t) as 2^−Δ(ΔC_t), where |$\Delta {C_{\rm{t}}} = {\rm{ }}{C_{\rm{t}}}_{_{{\rm{target}}}} - {\rm{ }}{C_{\rm{t}}}_{_{{\rm{18S}}}}$| and |$\Delta ( {\Delta {C_{\rm{t}}}} ){\rm{ }} = {\rm{ }}\Delta {C_{\rm{t}}}_{_{{\rm{test}}}} - {\rm{ }}\Delta {C_{\rm{t}}}_{_{{\rm{control}}}}$|⁠. Primer sequences are shown in Supplementary Table S1.

MDA-MB-231 cells were cross-linked with formaldehyde quenched with glycine and lysed with SDS lysis buffer (1% SDS, 10 mmol/L EDTA, and 50 mmol/L Tris, pH 8.1). Chromatin was sheared by sonication, and lysates were precleared with salmon sperm DNA/protein A-agarose slurry (Millipore) and incubated with antibody against HIF-1α (Santa Cruz Biotechnology), HIF-1β (Novus Biologicals), or HIF-2α (Novus Biologicals), or with IgG (Santa Cruz Biotechnology or Novus Biologicals) as described previously (40).

Cells were fixed with 4% formaldehyde and permeabilized with 0.1% (v/v) Triton X-100 and incubated with an antibody against vinculin (Sigma-Aldrich) and integrin α₅ (BD Biosciences). Following washes in PBS, cells were incubated with phalloidin (Invitrogen) at 1:150 dilution, anti-mouse Alexa Fluor 488 and anti-rabbit Alexa Fluor 568 antibodies (Life Technologies) at 1:200. Fluorescent images were obtained using a Nikon A1 confocal microscope with 60× oil immersion lens.

Cells were trypsinized, washed with PBS, collected in FACS staining buffer, incubated with Fc Block (BD Biosciences), followed by fluorescent labeling using an allophycocyanin-conjugated antibody against integrin α₅β₁ (BD Biosciences). Samples were analyzed using a FACSCalibur flow cytometer.

Cell culture plates were left uncoated or coated with 10 μg of fibronectin for 2D motility measurements. For 3D migration measurements, collagen matrices were prepared with soluble rat tail type I collagen in acetic acid (Corning) to achieve a final concentration of 1 mg/mL collagen supplemented with 10 or 50 μg/mL of fibronectin. Cell movement over time was imaged using a Cascade 1K CCD camera (Roper Scientific) mounted on a Nikon TE2000E phase-contrast microscope equipped with a 10× objective and controlled by NIS-Elements AR imaging software. Images were taken every 5 minutes for 13 hours. Cells in the time-lapse movies were tracked using MetaMorph to calculate x and y coordinates at each time interval.

Anisotropic persistent random walk (APRW) model analysis was performed as described in detail using MATLAB (code available; ref. 41). 2D and 3D cell trajectory data were used to statistically profile cell migration using the mean squared displacement (MSD), which can be obtained from [x(t), y(t)] coordinates of cells with time (t). MSD(τ) = [x (t + τ) − x(t)]² + [y(t + τ) − y(t)]², where τ = 5 minute*frame number. Values of persistence and speed are obtained from APRW model fitting and expressed as speed (S) and persistence (P) of cells.

TCGA breast cancer data (37) were obtained from the online data portal (https://gdc.cancer.gov/). GOBO data were collated from 7 independent studies available in the GEO database and obtained as *.rdata files from the authors (38). Pearson correlation and Kaplan–Meier analyses were conducted using GraphPad Prism Software.

Cell spheroids were formed in round-bottom 96-well tissue culture plates as described previously (42). Cells (1 × 10⁴ cells/mL) were resuspended in spheroid formation media, [DMEM and Methocult H4100 (3:1)] and centrifuged at 1,800 rpm for 30 minutes. Following 72-hour incubation, spheroids were embedded into 2 mg/mL collagen containing DMEM and soluble rat tail type I collagen (Corning) with the addition of fibronectin (50 μg/mL). Spheroids were imaged using Cytation 5 (BioTek) at 10×.

We previously reported that hypoxia promotes collagen biogenesis in the hypoxic region of breast tumors by enhancing the transcription of specific enzymes that mediate fibrillar collagen formation (18–20, 23). To investigate whether hypoxia alters integrin expression in breast cancer cells, we analyzed gene expression data from two independent databases (TCGA and GOBO) correlating the level of all 18 integrin α and β subunits with an HIF meta-gene signature. The HIF signature was comprised of 9 genes (P4HA1, P4HA2, PLOD1, PLOD2, LOX, LOXL2, ANGPTL4, VEGF, and SLC2A1), which were selected on the basis of their level of induction under hypoxia in a panel of breast cancer cells and were prognostic of survival in a retrospective study of breast cancer patients (43) (Supplementary Fig. S1A). A Pearson correlation coefficient was calculated for each integrin by comparing the averaged mean-centered expression of each gene in the hypoxia signature with the expression of each integrin subunit (Fig. 1A and B). The correlation coefficient was well matched between the two datasets, with ITGA5 showing the highest correlation with the HIF signature in both TCGA and GOBO datasets (Fig. 1B). We chose to further investigate four genes encoding integrin subunits (ITGA1, ITGA5, ITGA11, and ITGB1) that showed the highest level of correlation with the HIF signature.

We analyzed ITGA1, ITGA5, ITGA11, and ITGB1 mRNA levels in a panel of 20 breast cancer cell lines (in triplicate) following exposure to 20% or 1% O₂ for 24 hours (Fig. 1C–F). ITGA5 mRNA levels were significantly induced (Student t test; α > 0.05) by 2- to 10-fold in 19 of the 20 cell lines examined (Fig. 1C). ITGA5 mRNA levels were a minimum of 10-fold lower in luminal as compared with basal breast cancer cell lines (Fig. 1C). ITGB1 mRNA levels were induced by hypoxia in 12 of 20 cell lines with modest but reproducible induction of up to 2.5-fold (Fig. 1D). ITGA1 (α₁ subunit of the α₁β₁ collagen receptor) and ITGA11 (α₁₁ subunit of the α₁₁β₁ collagen receptor) expression was induced in 4 (ITGA1) and 10 (ITGA11) of 20 cell lines (Fig. 1E and F). In most cell lines, ITGA11 mRNA expression was not detectable until cells were exposed to hypoxic conditions.

To exclude the possibility that additional integrins might be induced by hypoxia but were not identified using our bioinformatics approach, we examined the expression of four additional integrin subunits that are expressed in epithelial cells but that did not correlate with the HIF signature (ITGA2, ITGA3, ITGA4, and ITGA6). Exposure to hypoxia did not induce their mRNA expression (Supplementary Fig. S1B). P4HA1 mRNA levels were used as a positive control for the experiment. Taken together, the analysis suggests that our HIF meta-gene signature is a reasonable method for identifying potential hypoxia-inducible genes encoding integrin subunits in breast cancer cells and that ITGA5 is the only integrin subunit robustly expressed across a wide range of breast cancer cells.

ITGA5 protein expression was increased 1.9- to 2.8-fold in MCF-7, ZR.75.1, MDA-MB-231, and MCF10A cells by exposing cells to hypoxic conditions or by treating cells with dimethyloxallyl glycine (DMOG), a 2-oxoglutarate (2-OG) analogue that stabilizes HIF protein levels (Fig. 2A). IHC staining of tumors derived from the orthotopic transplantation of MDA-MB-231 cells revealed intense ITGA5 staining in perinecrotic (hypoxic) regions of orthotopic tumors, which colocalized with HIF-1α staining (Fig. 2B). To determine whether HIF-1α or HIF-2α were required for ITGA5 or ITGB1 expression under hypoxic conditions, we generated MDA-MB-231 subclones that were stably transfected with an expression vector encoding a nontargeting control (NTC) short hairpin RNA (shRNA) or expression vector encoding an shRNA targeted against HIF-1α (sh1α) or HIF-2α (sh2α). The hypoxic induction of ITGA5 and ITGB1 mRNA (Fig. 2C and D) and protein (Fig. 2E) expression was blocked when expression of HIF-1α or HIF-2α (for ITGA5 only) was inhibited by shRNA. RNA from tumors generated with MDA-MB-231 subclones expressing sh1α or sh2α also showed decreased ITGA5 and ITGB1 expression compared with tumors generated with the NTC subclone (Fig. 2F and G). MCF-7 subclones expressing sh1α or sh2α (Supplementary Fig. S2A) were also generated and displayed decreased cell surface expression of α₅β₁ at 20% and 1% O₂ compared with NTC cells (Fig. 2H; Supplementary Fig. S2B). Taken together, the results show that HIF-1α and HIF-2α are required for hypoxic induction of ITGA5, whereas only HIF-1α is required for ITGB1 induction.

The human ITGA5 gene was examined for matches to the consensus HIF-binding site sequence 5′-(A/G)CGTG-3′. Three candidate sites were identified and interrogated by chromatin immunoprecipitation (ChIP) assays of MDA-MB-231 cells exposed to 20% or 1% O₂ for 16 hours (Fig. 3). Antibodies against HIF-1α, HIF-2α, HIF-1β, and rabbit Ig (IgG) were used for ChIP. Site 1 was located 1.0 kb 5′ upstream of the transcriptional start site (Fig. 3A) and showed significant hypoxia-induced binding of HIF-1α, HIF-2α, and HIF-1β (Fig. 3B). Site 2 and site 3, located within the first intron of ITGA5 (Fig. 3A), were not bound by HIF-1α, HIF-2α, or HIF-1β (Fig. 3B).

To test whether site 1 functions as a hypoxia response element (HRE), a 60-bp fragment encompassing site 1 (Fig. 3A) was inserted downstream of SV40 promoter and a firefly luciferase coding sequences in the reporter plasmid pGL2-promoter. MDA-MB-231 cells were cotransfected with pGL2–ITGA5-HRE and pSV-Renilla and exposed to 20% or 1% O₂ for 24 hours. Exposure to hypoxic conditions resulted in a significant increase in firefly luciferase activity compared with a reporter plasmid containing a mutated ITGA5 HRE (Fig. 3C). Taken together, the ChIP and reporter data indicated that HIF-1 and HIF-2 bind to site 1 and activate ITGA5 transcription under hypoxic conditions.

To identify the functional aspects of α₅β₁ cell surface expression, we generated MDA-MB-231 subclones that were stably transfected with a vector encoding either of two different shRNAs targeted against ITGA5 (shA5-1 and shA5-2). Decreased mRNA, protein, and cell surface expression of the α₅β₁ receptor was confirmed for both subclones under normal as well as hypoxic conditions (Supplementary Fig. S3A–S3C). To determine whether ITGA5 is required for focal adhesion formation, subclones were stained with Alexa Fluor 647–conjugated phalloidin to detect polymerized actin (F-actin) and anti-vinculin antibody to detect focal adhesions. ITGA5 staining confirmed localization of ITGA5 to focal adhesions in control cells and decreased focal adhesion formation in shA5-1 cells (Fig. 4A and B).

Focal adhesion size has been shown to correlate with 2D cancer cell motility (44, 45). To determine whether ITGA5 expression promotes 2D cancer cell motility, we monitored the motility of NTC and shA5-expressing cells on uncoated or fibronectin-coated surfaces using time-lapse microscopy. Individual (x, y) coordinates were determined for each cell every 5 minutes for 13 hours to construct cell trajectory maps (Fig. 4C). MSDs of individual cells were calculated and fit to an APRW model of cell motility (41, 46). The cell trajectories were fit to an APRW model of cell motility, to calculate velocity (S) and persistence (P). The parameter S can be interpreted as the rate at which a cell moves, the parameter P, or persistence time, represents a measure of the average time period between significant changes in the direction of movement.

Both velocity and directional persistence are used to calculate diffusivity (D = S²P/4), which describes overall cell movement within a specific time frame. Cell diffusivity and persistence time were increased on fibronectin-coated as compared with uncoated surfaces as has been previously reported, but decreasing the level of ITGA5 expression only moderately decreased cell migration on 2D surfaces (Fig. 4D and E).

Decreased α5β1 surface expression reduces 3D cell migration in collagen as well as single-cell migration in a complex multicellular spheroid.

The most abundant ECM proteins in mouse metastatic breast tumors are fibronectin and collagens (47). Collagen is abundant in human breast cancer and collagen organization influences tumor progression (48–50); increased fibronectin also correlates with disease progression and mortality (51). To examine the effect of the ITGA5 knockdown on 3D cell motility, MDA-MB-231 control or knockdown cells were embedded in 3D collagen matrices containing 10 or 50 μg/mL of fibronectin to mimic ECM found in breast cancer. The diffusivity and persistence time of ITGA5 knockdown cells migrating in 3D collagen matrices was decreased compared with control cells (Fig. 5A–D) or cells plated on 2D substrates (Fig. 4C and D). A modest decrease in cell motility was noted for NTC cells migrating in matrices containing 50 μg/mL versus 10 μg/mL of fibronectin, which may be the result of a decrease in pore size caused by increasing fibronectin concentrations (Supplementary Fig. S4A and S4B).

To determine whether ITGA5 expression can also promote migration within a tumor mass, we generated multicellular spheroids of target cells (NTC, shA5-1, or shA5-2) that were labeled with a red fluorescent carbocyanine tracer and mixed them with nonlabeled naïve cells at a ratio of 1:10. The multicellular spheroids were embedded into collagen gels containing 50 μg/mL of fibronectin. This allowed individually labeled NTC, shA5-1, or shA5-2 cell trajectories to be monitored over a 15-hour time course at single-cell resolution (Fig. 5E). The results demonstrate that decreasing ITGA5 expression reduces the diffusivity and persistence time of cells migrating within a multicellular spheroid mass (Fig. 5F and G).

To examine the role of ITGA5 in invasion at the edge of a multicellular spheroid, we generated spheroids of NTC or shA5 cells and embedded them in collagen gels containing 50 μg/mL of fibronectin. The spheroids were imaged on days 0, 3, 5, 7, and 10 to determine the invasion distance into the ECM (Fig. 5H; Supplementary S4C). The results show that α₅β₁ expression dramatically enhances cell invasion (Fig. 5I). Results from Fig. 5 suggest that ITGA5 expression enhances migration within a multicellular spheroid, allowing cells to more quickly access the cell–ECM boundary, where ITGA5 expression promotes invasion into the surrounding matrix.

As migration and invasion are two critical steps in the metastatic cascade, ITGA5 knockdown subclones were injected into the MFP of NOD/SCID mice to assess tumor growth and metastasis. The in vitro and in vivo proliferation rates of the knockdown and control subclones were similar, and tumors formed at a similar rate (Fig. 6A; Supplementary Fig. S5A). We also compared the overall metastatic burden resulting from control tumors or tumors lacking ITGA5 by isolating genomic DNA from the mouse lung for analysis of human DNA by qPCR (Fig. 6B). Inhibition of ITGA5 expression in tumors significantly decreased lung metastasis. Lung metastasis was also assessed histologically by H&E staining (Fig. 6C). The number of metastatic foci per field of view in each section (Fig. 6D) confirmed the qPCR results (Fig. 6B).

We assessed breast cancer cell infiltration of the ipsilateral axillary lymph node by IHC using an antibody that specifically recognizes human vimentin. The lymph nodes of mice bearing control tumors were enlarged and completely infiltrated with breast cancer cells, whereas normal follicular structure and size were maintained in mice bearing ITGA5-knockdown tumors (Fig. 6E and F; Supplementary Fig. S5B).

To investigate the role of ITGA5 in an immunocompetent model, we used 4T1 mouse mammary carcinoma cells, which form primary tumors and metastases similar to human triple-negative breast cancer (TNBC) after implantation into the MFP of syngeneic BALB/c mice. We generated 4T1 subclones that were stably transfected with a vector encoding NTC shRNA or shRNA that inhibited the expression of mouse ITGA5 (shA5-1, shA5-2; Fig. 7A). After orthotopic implantation in the MFP of female BALB/c mice, the growth of primary tumors derived from shA5-1 or shA5-2 subclones did not show a significant decrease in growth compared with tumors derived from NTC subclone cells (Fig. 7B). The number of metastatic nodules in lungs (Fig. 7C–E) harvested from mice implanted with shA5 cells was significantly decreased compared with mice implanted with NTC cells, whereas metastatic foci in axillary lymph nodes were similar between groups (Fig. 7F). Interestingly, the lymph node volume of nontumor-bearing BALB/c mice is 2 to 3 times larger than lymph nodes from nontumor-bearing NOD/SCID mice. Taken together, this suggests that decreased lymph node seeding of ITGA5-deficient cells only occurs in the absence of competent immune cells.

To determine whether our results were relevant for human cancer, we analyzed microarray expression data obtained from more than 1,800 primary human breast cancers, stratifying patients by high (above the median level) or low (below the median level) expression of ITGA5 or ITGB1 for survival analysis. High ITGA5 but not ITGB1 mRNA levels were associated with decreased patient survival (Fig. 7G; Supplementary Fig. S6A) and were independent of estrogen receptor or lymph node status (Supplementary Fig. S6B). By further stratifying patients into basal, luminal, and HER2⁺ subtypes based on the PAM50 score for each patient (38), we found that only patients with basal or HER2⁺ subtypes of breast cancer showed a significant correlation between increased ITGA5 expression and patient mortality (Fig. 7H and I). On the basis of results from χ² testing, ITGA5 expression is not associated with lymph node status in this dataset. Thus, expression levels of ITGA5 are prognostic for distant metastasis-free survival in patients but not lymph node infiltration, which is consistent with the observed role of ITGA5 in promoting lung metastasis in orthotopic transplantation models of TNBC.

Intratumoral hypoxia induces the expression of HIF-1α, which is associated with an increased risk of metastasis, relapse, and mortality in clinical studies (23). Many studies have highlighted the role of hypoxia and HIF-1 in inducing the expression of multiple genes that are involved in ECM processing and promote metastasis in tumor models (18–20, 23, 52–56), suggesting that ECM-binding proteins, such as integrins, could be induced under hypoxic conditions. Given that several integrin inhibitors are advancing through clinical trials, identifying which integrin is induced by hypoxia could provide a useful approach to specifically target hypoxic cells. Therefore, we investigated the role of hypoxia and HIFs in the regulation of each individual integrin subunit. Our results showed that ITGA1, ITGA5, ITGA11, and ITGB1 were well correlated with an HIF meta-gene signature. Surprisingly, ITGA1, ITGB1, and ITGA11 were induced in less than half of the cell lines tested. ITGA5 mRNA expression was (i) induced in 19 of 20 breast cancer cell lines tested; (ii) prognostic for patient survival; and (iii) increased 50-fold in basal compared with luminal breast cancers. The findings suggest a unique and potentially complex pattern of hypoxia-regulated induction of integrin subunits, which can only be recognized by assessing a large panel of cell lines to determine the robustness of the hypoxic response across cell lines that harbor extensive genetic and phenotype differences.

ITGA5 modulated single-cell migration within a complex multicellular spheroid as well as single-cell migration within 3D collagen–fibronectin gels but did not alter cell migration in 2D. The difference between 2D and 3D motility has been observed previously in multiple contexts. For example, previous studies suggest that migrating breast, pancreatic, and prostate cancer cells do not display lamellipodial structures in 3D but instead have unique dendritic protrusions (57). Recent work shows that Rab coupling protein (RCP, also known as Rab11-FIP1) interacts with, and controls, the recycling of α₅β₁. In 3D matrices, RCP–α₅β₁ trafficking increased local invasion by enhancing dynamic actin spike protrusions unique to cells in 3D to promote migration (58). Our previous results have shown that hypoxia induces RhoA and ROCK1 expression to enhance motility under hypoxic conditions (59). Determining how and whether ITGA5 signals to RhoA and ROCK1 in a complex 3D environment may increase our understanding of the functional effects of the coordinate regulation of ECM, integrins, and the Rho signaling pathway in the hypoxic tumor microenvironment.

We determined that HIF-1 and HIF-2 bind to an HRE in the ITGA5 gene to activate its transcription, leading to increased ITGA5 mRNA and protein expression. Interestingly, HIF-1α, but not HIF-2α, was also required for the hypoxia-induced expression of ITGB1 (the obligate partner of ITGA1). The findings suggest that integrin inhibitors, which are being tested in clinical trials (28, 60), may hold promise for patients with hypoxic tumors and patients with increased levels of ITGA5 due to oncogene activation by ERBB2 (61), c-Myc (62), or by miRNA regulation (63). ATN-161 (Ac-PHSCN-NH2), a peptide that binds to α₅β₁ and α_vβ₃, was shown to block breast cancer growth and bone colonization after intracardiac injection of MDA-MB-231 cells (64). Preclinical studies assessing liver and prostate metastasis also demonstrated a benefit using ATN-161 (65, 66). Both ATN-161 and volociximab, a mAb that blocks α₅β₁, demonstrated good tolerability in phase I and II trials (67, 68).

In addition to α₅β₁, blocking integrin αv using intetumumab, an mAb with encouraging results in phase I and II trials, has been shown to prevent brain metastasis in a 231BR-HER2 hematogenous breast cancer to brain metastasis model. Cilengitide, an inhibitor of α_vβ₃ and α_vβ₅, was the first integrin inhibitor to reach phase III clinical trials (60). In mouse models of breast to bone metastases, abrogating integrin β₃ expression prevented bone metastasis. In our studies, we did not observe induction of α_v or β₃ expression under hypoxic conditions. However, hypoxia has been shown to regulate α_v and β₃ transcription in endothelial cells (31) and α_vβ₃ cell surface expression in melanoma cells (32). It is also possible that additional integrin subunits may play a role in modulating metastasis, for example, the laminin-binding integrin α₆β₁ has been shown to sustain the stem-like state of breast cancer cells (69). Integrin α₆ was not induced by hypoxia in our studies.

Given (1) our finding that inhibition of ITGA5 expression prevents spontaneous lung metastasis in mouse models, (2) several pathways converge to promote ITGA5 expression during tumor progression, and (3) recent evidence suggesting that reducing ITGA5 levels in the presence of a PI3K inhibitor (navitoclax) can enhance apoptosis in PTEN-mutant cancer cells (70), further studies are warranted. Integrin blockers should be considered as potential therapeutics, especially in patients with hypoxic tumors.

No potential conflicts of interest were disclosed.

Conception and design: D. Wirtz, D.M. Gilkes

Development of methodology: J.A. Ju, I. Godet, S.J. Lee, L. Xiang, M.H. Lee, D. Wirtz, D.M. Gilkes

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): I.C. Ye, J. Byun, S.J. Lee, M.H. Lee, P.-H. Wu, D. Wirtz, D.M. Gilkes

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J.A. Ju, I. Godet, I.C. Ye, J. Byun, S.J. Lee, P.-H. Wu, D. Wirtz, D.M. Gilkes

Writing, review, and/or revision of the manuscript: J.A. Ju, I. Godet, H. Jayatilaka, D. Samanta, D. Wirtz, G.L. Semenza, D.M. Gilkes

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): I.C. Ye, J. Byun, D. Wirtz

Study supervision: H. Jayatilaka, D. Wirtz, D.M. Gilkes

The authors thank Karen Padgett of Novus Biologicals for providing HIF-2α antibodies.

This work was supported by NCI grant U54-CA143868 (to G.L. Semenza and D. Wirtz), U54-CA210173 (D. Wirtz) and R00-CA181352 (to D.M. Gilkes). D.M. Gilkes is a 2016 V Scholar awardee.

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.