|Home | About | Journals | Submit | Contact Us | Français|
Breast cancers contain regions of intratumoral hypoxia in which reduced O2 availability activates the hypoxia-inducible factors HIF-1 and HIF-2, which increase the transcription of genes encoding proteins that are required for many important steps in cancer progression. Recently, HIFs have been shown to play critical roles in the metastasis of breast cancer to the lungs through the transcriptional activation of genes encoding angiopoietin-like 4 and L1 cell adhesion molecule, which promote the extravasation of circulating cancer cells from the lung vasculature, and the lysyl oxidase family members LOX, LOXL2, and LOXL4, which promote invasion and metastatic niche formation. Digoxin, a drug that inhibits HIF-1 activity, blocks primary tumor growth, vascularization, invasion, and metastasis in ex vivo and in vivo assays.
Cancers are characterized by dysregulated growth due to increased cell division and decreased cell death. The resulting increased numbers of cells consume increased amounts of O2 leading to hypoxia (decreased O2 availability), which stimulates angiogenesis (new blood vessel formation). However, the blood vessels that form within cancers are structurally and functionally abnormal, resulting in a marked spatial and temporal heterogeneity in the perfusion of the tumor tissue. Intratumoral hypoxia results when O2 is either diffusion-limited, due to the location of a cancer cell sufficiently distal to a blood vessel that O2 diffusing from the vessel is consumed by cells more proximal to the vessel, or perfusion-limited, due to reduced or absent blood flow through an abnormal tumor vessel . As a result, hypoxic areas that are distributed heterogeneously throughout the tumor mass are observed in approximately half of all locally advanced cancers . Clinical studies in which the partial pressure of O2 (PO2) in human breast cancers was measured directly by an Eppendorf microelectrode revealed a median PO2 of 28 mm Hg, as compared to 65 mm Hg in normal human breast tissue; >50% of all breast cancers studied had a PO2 of less than 2.5 mm Hg . In multiple types of human cancer, the presence of intratumoral hypoxia has been identified as an adverse prognostic factor for patient outcome and this effect is independent of established prognostic parameters such as clinical tumor stage, histological grade, and lymph node status . Remarkably, despite the extensive clinical data linking intratumoral hypoxia to patient mortality, there are currently no approved drugs that target hypoxic cancer cells , indicating a major unmet clinical need in the effort to more effectively treat a major cause of mortality in Western societies.
All human cells adapt to decreased O2 availability through the activity of the hypoxiainducible factors HIF-1 and HIF-2, which are transcriptional activators that regulate the expression of over 800 target genes . HIF-1 is a heterodimer consisting of HIF-1α and HIF-1β subunits [6,7]. HIF-1α is subjected to two O2-dependent post-translational modifications (Figure 1a). First, hydroxylation of proline residue 402 and/or 564 targets the protein for ubiquitination and proteasomal degradation [8–10]; second, hydroxylation of asparagine residue 803 blocks the binding of the coactivator protein p300 . In hypoxic cells, the prolyl and asparaginyl hydroxylation reactions are inhibited, HIF-1α rapidly accumulates, heterodimerizes with HIF-1β, binds to the consensus DNA sequence 5′-RCGTG-3′ within hypoxia response elements located in target genes, and activates their transcription . HIF-1 target genes encode secreted proteins, such as vascular endothelial growth factor (VEGF), that function to increase O2 delivery to cells by stimulating angiogenesis, and intracellular proteins, such as the glycolytic enzymes, which allow cells to survive O2 deprivation by reprogramming their metabolism . HIF-2, a heterodimer consisting of HIF-1β and an O2-regulated HIF-2α subunit, stimulates the expression of some (e.g., angiogenic factors) but not all (e.g., glycolytic enzymes) of the gene products that are regulated by HIF-1 .
Cancer cells co-opt the adaptive responses to hypoxia that are mediated by HIFs (Figure 1b). Given the degree of intratumoral hypoxia that has been demonstrated in human breast cancer, it is not surprising that immunohistochemical studies have demonstrated increased HIF-1α protein levels in the majority of biopsies analyzed from both lymph node-negative  and lymph node-positive  breast cancer patients. In both cohorts, the survival of patients with the highest levels of HIF-1α in their diagnostic biopsies was significantly decreased. The percentage of HIF-1α overexpressing cells increased with disease progression, such that mortality was increased among node-negative patients with >5% HIF-1α overexpressing cells in biopsy sections  and among node-positive patients with >50% overexpressing HIF-1α cells . HIF-1α overexpression also predicted adverse outcome in many subsequent studies of breast cancer patients [16–22]. HIF-1α levels are not increased in benign fibrocystic disease but are already increased in ductal carcinoma in situ, particularly in poorly differentiated lesions . High HIF-2α levels in breast cancer biopsies are also associated with mortality . Finally, increased expression of mRNAs encoded by HIF target genes is associated with breast cancer mortality , thereby linking the overexpression of HIF-1α and HIF-2α with increased HIF transcriptional activity. Among the more than 1000 known HIF target genes are many encoding proteins that play critical roles in angiogenesis, metabolic reprogramming, autocrine growth/survival signaling, epithelial-mesenchymal transition, immortalization, invasion, metastasis, stem cell maintenance, and resistance to radiation and chemotherapy (Table 1; the list is intended to be illustrative rather than comprehensive; for literature citations, see ).
Although it is not possible to see hypoxia in a standard hematoxylin- and eosin-stained biopsy section, one can clearly see its consequences. The presence of necrosis in cancer sections indicates regions of severe intratumoral hypoxia in which oxygenation was insufficient to maintain cell viability. HIF-1α levels are usually highest within the viable cells surrounding areas of necrosis in breast cancers. Although it is commonly believed that hypoxia is only present in the central regions of large tumors, necrosis – and perinecrotic HIF-1α expression – are seen within ductal carcinoma in situ , i.e., severe hypoxia may be present even prior to the development of an invasive carcinoma.
Virtually all patients who die from breast cancer have metastatic disease. Even at the time of initial cancer surgery, metastasis may have already occurred in many patients who eventually die of breast cancer . Metastasis may occur by dissemination of breast cancer cells via either the lymphatic or vascular circulation and it is vascular metastasis that accounts for widely disseminated, malignant disease . Less than 0.1% of cancer cells that enter the vascular system establish a metastatic lesion , indicating that extravasation of cancer cells from the circulation into a tissue microenvironment that is favorable for survival and proliferation is a limiting step in the metastatic process. Increased expression of several genes implicated in breast cancer metastasis to the lungs has been reported in primary tumors of women with lung metastases and in breast cancer subclones selected for lung metastasis in mice [29,30].
The immunohistochemical data from clinical studies described above link high intratumoral HIF-1α and HIF-2α levels to metastasis and patient mortality. Breast cancers arising in conditional knockout mice lacking HIF-1α expression in mammary epithelial cells showed significantly reduced lung metastasis compared to breast cancers arising in wild type mice, demonstrating that HIF-1 promotes breast cancer metastasis . To delineate the role of specific human HIF-1 target genes in defined steps of the metastatic process, we utilized a model system involving orthotopic implantation into the mouse mammary fat pad of MDA-MB-231 cells, which are derived from human breast cancer cells that had metastasized to the lungs . MDA-MB-231 cells exemplify the class of triple-negative breast cancers, which lack expression of the estrogen (ER), progesterone, and HER2 receptors. Whereas ER+ breast cancers respond to anti-estrogen therapies, such as tamoxifen and aromatase inhibitors , and HER2+ cancers respond to trastuzumab, a monoclonal antibody directed against HER2 , there are no similarly targeted and effective therapies for triple-negative breast cancers, which have a poor prognosis with early relapse after chemotherapy .
We stably transfected MDA-MB-231 cells with expression vectors encoding short hairpin RNAs (shRNAs) targeting HIF-1α and HIF-2α and designated the subclone as MDA-MB-231-DKD (for double knockdown of HIF-1α and HIF-2α expression). This subclone and a control subclone, which was transfected with empty vector (designated MDA-MB-231-EV), were injected into the mammary fat pad of severe combined immunodeficiency (SCID) mice. Loss of HIF activity in the DKD subclone was associated with significantly decreased primary tumor growth and an even more dramatic reduction in spontaneous metastasis of cancer cells from breast to lungs .
In order to metastasize, breast cancer cells must intravasate into a blood vessel within the primary tumor and then extravasate out of the blood vessel at a distant metastatic site, such as the lungs. The hypoxia-induced and HIF-dependent expression of VEGF has been shown to increase vascular permeability and thereby promote breast cancer intravasation . To determine whether HIFs specifically promote the extravasation of hypoxic breast cancer cells from pulmonary capillaries into the lung parenchyma, we exposed EV and DKD cells to 20% O2 (PO2 = 140 mm Hg) or 1% O2 (PO2 = 7 mm Hg) for 48 h, injected the cells by tail vein (in order to bypass all steps that are required for cancer cells to access the vasculature), and one week later performed histological analysis to enumerate breast cancer cells that had extravasated out of pulmonary blood vessels into the lung parenchyma. Exposure to 1% O2 (to simulate hypoxia within the primary tumor) increased the extravasation of EV cells, whereas extravasation of DKD cells was significantly impaired .
Two critical requirements for extravasation are that the cancer cell must first adhere to ECs in the lung vasculature (margination) and then disrupt the tight interactions between endothelial cells (ECs) in order to migrate from the luminal to the abluminal side of the blood vessel (extravasation). Exposure of EV cells to 1% O2 for 48 h increased their adherence to ECs, whereas adherence of DKD cells to ECs was significantly impaired. We screened a microarray of metastasis-related genes to identify a cell surface receptor that is induced by hypoxia in MDA-MB-231 cells and might mediate the interaction of breast cancer cells with ECs. Expression of the L1 cell adhesion molecule (L1CAM) was induced by hypoxia in EV but not in DKD cells. Depending on the cell types involved, L1CAM has been shown to mediate cell–cell adherence by homophilic interactions and by heterophilic interactions with integrins, neuropilin 1, and CD24 (Figure 2). Further analysis revealed that HIF-1α, but not HIF-2α, was required for hypoxia-induced L1CAM expression (Figure 1b).
Stable transfection of DKD cells with an expression vector encoding L1CAM (i.e., L1CAM gain-of-function) increased the number of extravasated cancer cells in the lung after tail vein injection. Conversely, stable transfection of MDA-MB-231 cells with expression vector encoding shRNA targeting L1CAM (i.e., L1CAM loss-of-function) dramatically impaired the spontaneous metastasis of cancer cells from breast to lung, whereas primary tumor growth was only modestly reduced . Thus, HIF-1-dependent L1CAM expression in hypoxic breast cancer cells increases their interaction with ECs, thereby facilitating margination, which leads to extravasation, and lung metastasis.
To investigate whether HIF activity promoted the ability of hypoxic breast cancer cells to invade through an EC monolayer, we exposed ECs to conditioned medium generated by culturing the MDA-MB-231-EV and -DKD subclones in the presence of either 20% or 1% O2 for 48 h. When naïve breast cancer cells were tested for their ability to invade through these conditioned EC monolayers, more breast cancer cells invaded through EC monolayers that were exposed to conditioned medium from hypoxic, as compared to non-hypoxic, EV cells. This hypoxia-induced effect was lost when DKD cells were used as the source of conditioned medium. Transendothelial electrical resistance is a biophysical measure of EC–EC interactions and conditioned medium from hypoxic EV cells decreased transendothelial resistance, an effect that was lost in DKD cells. These data suggested that a secreted factor, which was produced by hypoxic breast cancer cells in a HIF-dependent manner, inhibited EC–EC interactions .
Screening of the microarray of metastasis-related genes for a secreted factor that was induced by hypoxia in MDA-MB-231 cells identified angiopoietin-like 4 (ANGPTL4), which is expressed at increased levels in the primary breast cancers of women with lung metastases . ANGPTL4 mRNA and protein expression were hypoxia-inducible in EV but not in DKD cells and a hypoxia-response element containing a HIF binding site was identified 1.5 kb 5′ to the ANGPTL4 gene transcription start site. Both HIF-1 and HIF-2 contributed to ANGPTL4 expression (Figure 1b). Conditioned medium from MDA-MB-231-DKD cells stably transfected with an ANGPTL4 expression vector (i.e., ANGPTL4 gain-of-function) inhibited EC–EC interactions as measured by transendothelial electrical resistance and breast cancer invasion assays. When injected via tail vein, DKD cells with forced expression of ANGPTL4 manifested increased extravasation in the lungs. When MDA-MB-231 cells stably transfected with vector encoding shRNA against ANGPTL4 (i.e., ANGPTL4 loss-of-function) were exposed to hypoxia, the conditioned medium no longer inhibited EC–EC interactions. Following mammary fat pad implantation of ANGPTL4 knockdown cells, primary tumor growth was unaffected but metastasis to the lungs was almost completely eliminated . Thus, HIF-dependent ANGPTL4 expression in hypoxic breast cancer cells inhibits EC–EC interactions (Figure 3b), thereby facilitating extravasation and lung metastasis.
A major advance in our understanding of the metastatic process occurred with the discovery that the metastasis of cancer cells from the primary tumor to the lungs is preceded by, and requires, the recruitment of bone marrow-derived cells (BMDCs) to sites in the lungs that are subsequently colonized by metastatic cancer cells . A second major advance was the discovery that hypoxic breast cancer cells produce lysyl oxidase (LOX), an enzyme that crosslinks extracellular matrix proteins such as collagen (Figure 4a) and promotes breast cancer metastasis . Whereas the role of LOX in metastasis was initially attributed to its remodeling of ECM in the primary tumor (Figure 4b), subsequent studies revealed that LOX can remodel ECM at sites distant from the primary tumor (Figure 4c). Specifically, breast cancer cells in the primary tumor secrete LOX into the circulation, leading to collagen crosslinking in the lungs, which is required for the recruitment of BMDCs (Figure 3a) and, subsequently, cancer cells to the metastatic niche .
The demonstration that LOX played a critical role in lung metastasis suggested that a monoclonal antibody or small molecule inhibitor directed against LOX might be useful for breast cancer therapy . However, in addition to LOX, four LOX-like (LOXL) proteins are encoded in the genome and LOX, LOXL2, and LOXL4 were all expressed, in a HIF-dependent manner, when different metastatic human breast cancer cell lines were subjected to hypoxia . Various combinations of LOX, LOXL2, and LOXL4 were also overexpressed (relative to surrounding normal tissue) in different primary human breast cancers . In MDA-MB-231 cells, in which LOX and LOXL4 expression were induced by hypoxia, knockdown of HIF-1α, HIF-2α, or both blocked hypoxia-induced expression of LOX and LOXL4 and collagen remodeling ex vivo and blocked collagen remodeling and BMDC recruitment in the lungs of tumor bearing mice, which was similar to the effect of knocking down LOX or LOXL4 expression .
MDA-MB-435 is another metastatic triple-negative breast cancer cell line . In MDA-MB-435 cells in which only LOXL2 expression were induced by hypoxia, knockdown of HIF-1α, or both HIF-1α and HIF-2α (but not HIF-2α alone), blocked hypoxia-induced expression of LOXL2 and collagen remodeling ex vivo and blocked collagen remodeling and BMDC recruitment in the lungs of tumor bearing mice, which was similar to the effect of knocking down LOXL2 (but not LOX) expression . These results showed that different LOX family members were expressed in different breast cancer lines but that inhibition of HIF-1α in each of these lines was sufficient to dramatically inhibit metastatic niche formation and subsequent breast cancer metastasis to the lungs.
Multiple drugs have been shown to inhibit HIF-1 activity, block tumor growth and vascularization, and improve responses to chemotherapy and radiation therapy in a variety of mouse models, through different molecular mechanisms of action (Table 2; the list is intended to be illustrative rather than comprehensive). Digoxin was shown to inhibit the growth of human prostate cancer xenografts and its effect was dependent on its inhibition of HIF-1α expression . In the orthotopic model involving implantation of human MDA-MB-231 and MDA-MB-435 cells into SCID mice, treatment of tumor-bearing mice with acriflavine (4 mg/kg) or digoxin (2 mg/kg) by daily intraperitoneal injection significantly inhibited the following: expression of ANGPTL4, L1CAM, LOX, LOXL2, and LOXL4 in the primary tumor; primary tumor growth; collagen remodeling and BMDC recruitment in the lungs (i.e., metastatic niche formation); and metastasis of cancer cells from breast to lungs [36,45]. Most strikingly, treatment of mice bearing MDA-MB-435 primary breast tumors with digoxin reduced the lung metastatic burden by >96% . In contrast to the ability of digoxin to block LOX/LOXL activity in both MDA-MB-231 and MDA-MB-435 cells, the small molecule LOX inhibitor β-aminopropionitrile blocked LOX/LOXL activity in MDA-MB-231 cells, which express LOX and LOXL4, but not in MDA-MB-435 cells, which express LOXL2 .
Digoxin has been used for decades to treat heart disease and a daily oral dosing regimen that is safe and effective (in the context of heart disease) is well established, as are methods for determining serum drug levels. A clinical trial is currently underway in which men with prostate cancer and biochemical evidence of recurrence (rising PSA levels) will be treated with a daily oral dose of 125 mcg or 250 mcg per day of digoxin to achieve target serum levels of 0.8–2.0 ng/mL (ClinicalTrials.gov Identifier: NCT01162135). It should be noted that there are currently no data demonstrating that administration of digoxin to humans at these doses inhibits HIF-1 activity within cancer cells. A pilot clinical trial of topotecan in patients with advanced cancer and HIF-1α overexpression demonstrated on tumor biopsy was recently reported in which HIF-1α protein levels were undetectable in the post-treatment biopsies from 4 of 7 patients studied . HIF-1 inhibitors block the expression of multiple pro-angiogenic factors (including VEGF, stromal-derived factor 1, and stem cell factor) and tumor vascularization in mouse models [45,46], and serial DCE-MRI scans demonstrated reduced tumor perfusion in 7 of 10 patients treated with topotecan .
Intratumoral hypoxia induces HIF-1 activity within many primary breast cancers, which may lead to the expression of multiple gene products that play critical roles in metastasis to the lungs by promoting establishment of the metastatic niche and the subsequent extravasation of circulating tumor cells into the lung parenchyma (Figure 5; see also Box 1). Clinical trials are needed using available drugs (Table 2) to immediately test the hypothesis that the addition of one or more HIF inhibitors to current therapeutic regimens will increase the survival of breast cancer patients, particularly those with triple negative disease, and high HIF-1α levels in their tumor biopsies, by impairing metastasis as well as other critical steps in cancer progression that are mediated by HIF-1 and HIF-2 (Table 1). The National Cancer Institute should create a funding mechanism specifically designed to support clinical trials involving off-patent drugs, which are unlikely to be performed by pharmaceutical companies but have the potential to establish safe, effective, and inexpensive new therapies for cancer patients.
Along with the functions described here, both L1CAM and the LOX family of proteins have been shown to have additional functions in other cell types. In addition to mediating cell-cell interactions, L1CAM-mediated signaling via interactions with surface proteins promotes proliferation and migration, and is associated with an aggressive, metastatic phenotype in many cancers (Figure I). Although it does not have intrinsic kinase activity, L1CAM acts by binding several proteins to favor proliferation and cell migration, including integrins, growth factor receptors, and other protein kinases (represented here by a pair of receptor tyrosine kinases, RTKs). Cell behavior is modified to promote metastasis through, among others, the MAP kinases ERK1 and ERK2, FAK, and NFkappaB. Finally, proteolytic processing of L1CAM at the cell surface produces two soluble domains, the L1 intracellular domain (L1-ICD) and the L1CAM ectodomain, which act independently to modulate cell behavior. After cleavage, the L1-ICD can translocate to the nucleus and modulate gene expression, whereas the soluble ectodomain can interact with surface receptors, similar to the membrane-spanning form of the protein, to stimulate signal transduction. Similar to L1CAM, the LOX family of proteins have been shown to possess functions beyond those observed in the context of breast cancer (Figure II). In addition to remodelling the ECM, the LOX proteins can have both paracrine and autocrine functions that derive from being imported into cells. Imported, processed LOX enzymes can then modify cytoplasmic proteins or localize to the nucleus and participate in transcriptional regulation.
Breast cancer research in the author’s laboratory is supported by the American Cancer Society, Johns Hopkins Institute for Cell Engineering, Komen Foundation, and National Cancer Institute. G.L.S. is the C. Michael Armstrong Professor at the Johns Hopkins University School of Medicine and an American Cancer Society Research Professor.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.