|Home | About | Journals | Submit | Contact Us | Français|
Tumor hypoxia is known to activate angiogenesis, anaerobic glycolysis, invasion and metastasis. However, a comparative analysis of the potentially distinct functions of hypoxia in primary tumor growth and organ-specific metastasis has not been reported. Here, we show distinct hypoxia kinetics in tumors generated by the MDA-MB-231 breast cancer sublines with characteristically different primary tumor growth rates and organotropic metastasis potentials. Hypoxia-induced angiogenesis promotes both primary tumor growth and lung metastasis but is non-essential for bone metastasis. Microarray profiling revealed that hypoxia enhances the expression of a significant number of genes in the lung metastasis signature, but only activates a few bone metastasis genes, among which DUSP1 was functionally validated in this study. Despite the different mechanisms by which hypoxia promotes organ-specific metastasis, inhibition of HIF-1α with a dominant negative form of HIF-1α or 2-methoxyestradiol reduced metastasis to both lung and bone. Consistent with the extensive functional overlap of hypoxia in promoting primary tumor growth and lung metastasis, a 45-gene hypoxia response signature efficiently stratifies breast cancer patients with low or high risks of lung metastasis, but not for bone metastasis. Our study demonstrates distinct functions of hypoxia in regulating angiogenesis and metastasis in different organ microenvironments and establishes HIF-1α as a promising target for controlling organotropic metastasis of breast cancer.
Metastasis to vital organs such as bone, lung, liver and brain is responsible for the vast majority of breast cancer deaths (1). It has been well-recognized that metastatic seeding and growth is determined by both the intrinsic genetic properties of tumor cells and the local characteristics of the stromal microenvironment (1, 2). Functional genomic analyses of the in vivo selected organotropic variants of breast cancer cell lines have led to the identification of distinct sets of organ-specific metastasis genes to bone, lung and brain (3-5). Understanding the organ-specific functions of metastasis-related signaling pathways may provide new opportunities for therapeutically targeting metastasis in different organs.
Hypoxia has been increasingly recognized to play a central role in different stages of tumor progression (6). Signaling responses to hypoxia are mediated mainly through the hypoxia-inducible factors (HIFs), including HIF-1, HIF-2 and HIF-3 (7). Among them, HIF-1 is best characterized as being responsible for the regulation of many hypoxia-inducible genes. HIF-1 is composed of the hypoxia responsive subunit, HIF-1α, and the constitutively expressed subunit, HIF-1β. Under normoxia, the oxygen-dependent degradation (ODD) domain of HIF-1α is hydroxylated, facilitating binding of the von Hippel-Lindau E3 ubiquitin ligase and subsequent proteosome-mediated degradation of HIF-1α. When oxygen levels become limited, HIF-1α is free of hydroxylation and escapes degradation. Accumulated HIF-1α enters the nucleus, forms a transcription complex with HIF-1β and drives transcription of downstream genes (8).
HIF-1α overexpression is correlated with distant metastasis and poor prognosis for breast cancer patients (9). Furthermore, HIF-1α overexpression was more frequently observed in metastases than in primary tumor of breast cancer, suggesting an active role of HIF-1α in regulating metastatic progression (9). The functions of hypoxia and HIF-1α in promoting metastasis may overlap with its functions in primary tumor growth, for example, the enhancement of angiogenesis. However, hypoxia is also known to promote metastasis-specific functions, such as epithelial-mesenchymal transition (EMT), invasion and metastatic seeding (10). The distinct functions of hypoxia in promoting metastasis to different organs have not been thoroughly investigated. Here, we integrate in vivo hypoxia imaging, functional inactivation of HIF-1α and transcriptomic analysis to investigate the distinct functions of hypoxia in tumor growth and metastasis in the recently developed organotropic metastasis models of breast cancer.
The MDA-MB-231 sublines were maintained as described (11).
All procedures involving mice, such as housing and care, and all experimental protocols were approved by Institutional Animal Care and Use Committee (IACUC) of Princeton University. Intracardiac, intravenous and mammary gland injections were performed as described (3, 4). Development of metastases in bone and lung was monitored by firefly or Renilla luciferase bioluminescence imaging as described (12-14).
2ME2 (BIOMOL, Inc.) was suspended in 0.5% carboxymethycellulose (Sigma) at 15 mg/ml for in vivo treatment. Mice were inoculated with tumor cells and oral feeding commenced 10 days later with vehicle or 2ME2 daily at 75 mg/kg mouse weight. Primary mammary gland tumors were treated for 20 days. Metastases were treated for 30 days.
SCP2 or LM2 cells were cultured in duplicate under ambient O2 level (21%) for 24 h or low level of O2 (1%) for 6 h and 24 h. RNA was extracted and microarrays were performed with the Agilent Whole Human Genome 4×44k platform as described (15). Microarray data were deposited at the NCBI Gene Expression Omnibus (http://www.ncbi.nlm.nih.gov/geo/) with the accession number GSE17188. Patient microarray samples with metastasis information were obtained from Minn et al. (4). Hierarchical clustering was performed using Genespring GX 7.3 software (Agilent Technologies).
Results were reported as average ± SD (standard deviation) or average ± SE (standard error). Group comparisons were performed using either the nonparametric Mann-Whitney test or unpaired two-sided Student's t-test without equal variance assumption. Kaplan-Meier curve comparison was performed with log-rank test.
Additional materials and methods, including construction of plasmids and stable cell lines, CoCl2 treatment, western and northern blot analyses, histology and immunohistochemistry, detection of hypoxia, quantitative reverse-transcriptase PCR, DUSP1 knockdown, GSEA and derivation of hypoxia response signatures, are listed in the Supplementary Information.
MDA-MB-231 sublines with distinct metastasis organotropisms were previously isolated by in vivo selection in xenograft metastasis models in mice (3, 4, 16). We chose to use the bone-tropic subline SCP2 and the lung-tropic subline LM2 to perform our study, because of their distinctive proclivities to metastasize to different target organs: following intracardiac injection, SCP2 forms osteolytic bone lesions much more efficiently than LM2, whereas, following intravenous injection, LM2 generates numerous lung metastasis nodules but SCP2 rarely forms any (Fig. 1A) (11). SCP2 and LM2 have similar growth rates in vitro (11). Interestingly, SCP2 and LM2 display significant differences in primary tumor growth when injected into mammary fat pads. While LM2 forms rapidly growing and well-vascularized primary tumors, SCP2 gives rise to tumors of much smaller size (Fig. 1B). The distinct growth and metastatic properties of the two cell lines that share a closely-related genetic background allow us to comparatively investigate the potential organ-specific contribution of hypoxia to the growth of primary tumor and metastases in different organs.
First, we evaluated hypoxic response of SCP2 and LM2 in vitro. HIF-1α protein was not detectable by immunoblotting in either subline under ambient oxygen conditions (~21%), but became highly abundant under low oxygen conditions (1%) (Fig. 1C). LM2 showed a noticeably stronger response to hypoxia in vitro. To investigate the status of hypoxia and HIF-1α in tumor tissues, we detected hypoxia with Hypoxyprobe (Supplementary Fig. S1) and performed immunohistochemistry (IHC) of HIF-1α on primary tumors formed by LM2 and SCP2, lung metastasis formed by LM2 and bone metastasis formed by SCP2 (Fig. 1D). All tumor samples were harvested at the end of the in vivo experiments shown in Fig. 1A-B. Hypoxia was detected in primary tumors and metastases formed by both cell lines (Supplementary Fig. S1), suggesting hypoxia as the cause for in vivo HIF-1α activation, although we cannot rule out the influence of other signaling pathways on the HIF-1α level. In primary tumors, HIF-1α was positively stained in cell nuclei for both LM2 and SCP2 tumors with an appreciably stronger staining in SCP2 tumors (Fig. 1D, upper images). Considering the intrinsically weaker HIF-1α response in SCP2 in vitro (Fig. 1C), the more intense HIF-1α staining of SCP2 tumors suggested that these tumors experienced more hypoxia stress. Consistently, bone metastases formed by SCP2 have a stronger HIF-1α staining than lung metastases formed by LM2 (Fig. 1D, lower images). It was also noted that LM2 lung metastases contained prominent intratumoral blood vessels whereas SCP2 bone metastases contained very few.
HIF-1α immunostaining only shows tumor cell responses to the hypoxia condition at a static time point. In order to analyze dynamic changes in the hypoxia response by tumor cells longitudinally in vivo, we adopted a recently developed approach of real-time bioluminescence imaging (BLI) of HIF-1α prolyl hydroxylase activity using the reporter construct ODD-Fluc (17). When the ODD peptide is fused with firefly luciferase (Fluc), the stability of ODD-Fluc fusion protein mimics that of HIF-1α and serves as an accurate reporter of the intracellular HIF-1α level and an indirect reflection of hypoxia activity (17). Unlabelled SCP2 and LM2 were stably transfected with either a constitutively expressed Fluc reporter or the ODD-Fluc reporter. To test the sensitivity of the reporter in detecting hypoxia response in vivo, we used the hypoxia mimetic CoCl2 to create acute hypoxia-like conditions in tumors. We first detected dose-dependent and time-dependent stabilization of HIF-1α, as well as a dose-dependent increase of ODD-Fluc activity, by CoCl2 treatment in vitro (Supplementary Fig. S2). Next, we determined the efficacy of the reporter system in vivo by treating mammary tumor-bearing mice with a single dose of CoCl2. Tracking the Fluc activity with BLI after the injection showed peak luciferase signals at 20 h post-treatment for ODD-Fluc sublines, indicating response kinetics similar to the in vitro treatment (Supplementary Fig. S3). As a negative control, the Fluc-labeled sublines showed little fluctuation of BLI signals after CoCl2 treatment.
After validating the ODD-Fluc hypoxia reporter system in vivo, we characterized the dynamic changes of hypoxia activity during the primary tumor development of SCP2 and LM2. To quantify the unit cellular response to hypoxia, the ODD-Fluc sublines were further labeled with a constitutively expressed Renilla luciferase (Rluc) (11). Previous work has demonstrated no cross-reactivity of Rluc and Fluc to their non-corresponding substrates, D-luciferin and coelenterazine (14). The dual-luciferase imaging system allowed direct comparison of relative hypoxia activity of tumors with different sizes by normalizing Fluc (absolute hypoxia signaling strength) to Rluc (tumor size). When Fluc/Rluc ratio was plotted at different time points after tumor inoculation, distinct hypoxia response kinetics of SCP2 and LM2 were observed (Fig. 2A). At the early phase of tumor growth (day 18), LM2 cells experienced higher hypoxia stress as LM2 showed greater normalized Fluc signal than SCP2. However, at later time points (day 34 and day 41), a more hypoxic environment was detected in SCP2 tumors than in LM2 tumors. This result is consistent with the end-point HIF-1α IHC analysis (Fig. 1D).
Both groups of primary tumors harvested at the end of the experiment contained necrotic centers as revealed by H&E staining (Fig. 2B). However, LM2 tumors showed a much thicker growth zone than SCP2 tumors, suggesting the ability of LM2 tumors to sustain continuous growth. This was confirmed by Ki67 and TUNEL stainings (Fig. 2C). LM2 tumors contained significantly more Ki67-positive nuclei than SCP2 tumors (Supplementary Fig. S4). Neither of the tumors showed significant levels of apoptosis in the non-necrotic regions. Taken together, these results suggested that LM2 cells possess an overall proliferation advantage over SCP2 cells in primary tumors. In general, continuous tumor growth is dependent on adequate oxygen and nutrient supply provided through blood circulation. Therefore, we determined whether intratumoral blood vessel density was different in the two tumors. Both CD31 IHC and in vivo dextran labeling showed vessels in LM2 tumors were more abundant and with larger inclusion areas, whereas those in SCP2 tumors were fewer and narrower (Fig. 2D). Because functional tumor angiogenesis is largely triggered by hypoxia stress, and hypoxia level, in turn, is alleviated by the successful formation of neovasculature, the hypoxia kinetics observed in SCP2 and LM2 primary tumors may reflect the stronger angiogenic response of LM2 cells to hypoxia.
To investigate the potential differences in hypoxic activities in organotropic metastasis, SCP2 and LM2 cells harboring the dual-reporter system were subjected to in vivo bone and lung metastasis assays. During the progression of bone metastasis, we observed that the trends of normalized Fluc activity in SCP2 and LM2 were quite similar to those in the primary tumors: the Fluc/Rluc signal ratio decreased continuously in LM2 bone metastasis, compared to a sustained high level of hypoxia in SCP2 bone metastasis (Fig. 3A). Again, the difference of the hypoxia reporter activity at the end of experiment was consistent with the end-point HIF-1α IHC analysis (Fig. 1D). Since SCP2 failed to generate any detectable lung metastasis signal, we could only analyze hypoxia activity kinetics for LM2 lung metastasis. The normalized Fluc curve also showed a continuous decrease (Fig. 3B).
Since the trends of hypoxia kinetics were consistent between metastases and primary tumors formed by LM2 and SCP2, we hypothesized that the different angiogenic properties of the two cell lines might also explain the different hypoxic dynamics seen in metastases. Indeed, CD31 staining showed very few blood vessels in SCP2 bone metastases, whereas LM2 bone metastases, although much less frequent, possessed rich pools of vasculature (Fig. 3C). Similarly, LM2 cells recruited conspicuous blood vessels when forming lung metastases (Fig. 3D). SCP2 did not generate any appreciable tumor mass in the lung to allow for a similar analysis. It is unclear how SCP2 was able to generate fast growing osteolytic bone metastases without necrosis despite the lack of hypoxia-induced angiogenesis. The relatively small size of bone metastases and the unique characteristics of the bone microenvironment may allow SCP2 tumors to sustain growth and avoid necrosis despite the lack of productive angiogenesis. Nonetheless, HIF-1α accumulation in SCP2 may have angiogenesis-independent functions in promoting bone metastasis. Likewise, hypoxia response in LM2 may have lung metastasis-promoting function beyond the activation of angiogenesis. Inhibition of HIF-1α function in vivo was necessary to test the pleiotropic effect of hypoxia in tumor growth at different organs.
2-methoxyestradiol (2ME2) downregulates HIF-1α, inhibits the HIF-induced VEGF activation and angiogenesis, and has been used in vivo to block HIF-1α activity (18-20). Treating cells under hypoxia with 2ME2 abolished the induction of HIF-1α in vitro (Fig. 4A). Therefore, 2ME2 was tested for its effect on primary tumors and metastases. 2ME2 significantly reduced the growth of LM2 primary tumors, but had little effect on SCP2 tumors (Fig. 4B). Because SCP2 tumors generally stop expanding after reaching a certain size (<100mm3), we cannot directly determine whether 2ME2 would inhibit SCP2 tumor growth if SCP2 grew to a similar size as LM2. Nevertheless, this result indicated that 2ME2 could inhibit the growth of well-vascularized tumors. IHC confirmed lower HIF-1α levels in both LM2 and SCP2 tumors after treatment with 2ME2 (Supplementary Fig. S5A-B). Microvessel density and architecture analysis by CD31 IHC showed narrower vessels in LM2 tumors by 2ME2 treatment, even though 2ME2 did not seem to significantly affect vessel density (Fig. 4B and Supplementary Fig. S5C). Consistently, 2ME2 did not affect the already compressed vessels in SCP2 tumors (Fig. 4B). Overall, these results suggest that the primary effect of 2ME2 on primary tumor growth is through targeting the function of HIF-1α in angiogenesis
If angiogenesis were also the primary contribution of HIF-1α to promote metastasis development, we would expect 2ME2 could reduce LM2 lung metastasis burden but have relatively little effect on SCP2 bone metastasis. To our surprise, both LM2 lung metastasis and SCP2 bone metastasis were significantly inhibited by 2ME2 treatment (Fig. 4C-D), suggesting angiogenesis-independent functions of HIF-1α in promoting bone metastasis. Likewise, we also cannot rule out the possibility of non-angiogenic functions of HIF-1α and hypoxia in promoting lung metastasis.
One potential caveat of using 2ME2 to inhibit HIF-1α is the potential non-specific effects. To block the function of HIF-1α more specifically, a FLAG tagged dominant negative (DN) form of HIF-1α (21) was stably expressed in SCP2 and LM2 (Fig. 5A) and the cell lines were subjected to in vivo analysis of tumor growth and metastasis. We observed similar effects of DN-HIF-1α on primary tumor and metastasis growth as 2ME2 treatments (Fig. 5B-D). LM2 primary tumor growth was inhibited by DN-HIF-1α, while SCP2 growth was unaffected within the tested period (Fig. 5B). LM2 lung metastasis and SCP2 bone metastasis were both significantly inhibited (Fig. 5C-D). These results further confirm that inhibition of HIF-1α reduces tumor growth in a site-independent manner.
To identify hypoxia target genes that may play a role in promoting organ-specific metastasis, LM2 and SCP2 were cultured under ambient (21%) or low (1%) oxygen condition for 6 h or 24 h before microarray profiling. Gene set enrichment analysis (GSEA) was used to examine the enrichment for gene sets representing the general HIF-1 target genes (22), lung metastasis gene signature (LMS) (4) and bone metastasis gene signature (BMS) (3) by the hypoxic condition. The BMS and LMS include genes that are up-regulated in the highly metastatic cells (candidate metastasis enhancers) and those down-regulated in highly metastatic cells (candidate metastasis suppressors). The up-regulated and down-regulated subsets were tested separately. With an FDR q-value of 0.2 as the significance cutoff (more stringent than the recommended 0.25), hypoxic LM2 and SCP2 both enriched for the general HIF-1α targets (Supplementary Fig. S6A-C), such as VEGF (Supplementary Fig. S6D), indicating that both cells were able to turn on the canonical HIF-1α transcription programs. Although VEGF is activated by hypoxia at a slightly stronger level in LM2 than in SCP2, this could not fully account for the dramatic differences in angiogenesis in the two tumors. SCP2 may possibly express a strong anti-angiogenic factor(s) that buffers the pro-angiogenic activity of HIF-1α-induced factors. When metastasis gene signatures were tested in GSEA, the up-regulated subset of the LMS, but not the down-regulated subset, was found to be significantly enriched in hypoxic LM2 transcriptome (Supplementary Fig. S6A-C, Supplementary Table S1). This finding suggests that hypoxia may have lung metastasis-promoting functions beyond the activation of angiogenesis. Among genes in the LMS, ANGPTL4, an essential gene in lung metastasis formation (23), has the strongest response to hypoxia (Supplementary Fig. S6D).
The hypoxic SCP2 gene profile did not enrich for either the up-regulated or the down-regulated subset of the BMS (Supplementary Fig. S6A-C, Supplementary Table S2). Nevertheless, several putative bone metastatic genes were activated by hypoxia, including CXCR4 and DUSP1 (Supplementary Fig. S6D). Hypoxic responses of these two genes were much weaker or absent in LM2 cells. The functional importance of CXCR4 in bone metastasis has been established before (24, 25). DUSP1 is overexpressed in breast carcinomas and implicated in breast cancer chemoresistance (26). We previously identified DUSP1 as one of the 11 mostly significantly up-regulated genes in the bone-metastatic sublines of MDA-MB-231 (3). Northern blot analysis confirmed the dramatic overexpression of DUSP1 in bone-tropic cells (Supplementary Fig. S7A). However, the functional importance of DUSP1 in bone metastasis had not been investigated. Using shRNA-mediated gene silencing, we knocked down DUSP1 expression in SCP2 and observed a significant decrease in bone colonization (Supplementary Fig. S7B-C). Furthermore, when we examined the potential clinical importance of DUSP1 in bone metastasis by analyzing its expression level in a published microarray dataset (4), higher DUSP1 level in primary tumors was detected in patients with bone metastases but not lung metastasis (Supplementary Fig. S7D). Taken together, these results suggest that hypoxia has an organ-specific metastasis-promoting function in addition to its well-appreciated role in eliciting angiogenesis.
HIF-1α overexpression and hypoxia-induced gene expression changes have been associated with poor prognosis and metastasis of breast cancer patients (27, 28). However, the prognostic value of the hypoxia gene signature in organ-specific metastasis has not been investigated. To test this, we compared the expression profiles of LM2 or SCP2 in normoxic or hypoxic conditions and identified differentially expressed genes as hypoxia response signatures for each cell line (see gene lists in Supplementary Table S3 and S4). LM2- and SCP2-specific hypoxia response signatures were used to perform unsupervised hierarchical clustering of tumor samples in the Minn et al. dataset (4), which contains organ-specific metastasis status of all patients in a five year follow-up period. Both the LM2 and SCP2 hypoxia signatures were able to segregate patients into two major clusters with distinctively different outcomes in lung metastasis while the same signatures were not able to identify patients with high risk for bone metastasis (Supplementary Fig. S8 and S9). This result suggests that different MDA-MB-231 variants may share a common hypoxia transcriptomic program that can be used to predict organ-specific metastasis to lung. Indeed, using the overlapped hypoxia gene signature containing 45 genes with 70 probe sets (Supplementary Table S5), we successfully classified patients into two subgroups with distinct incidence rates for lung metastasis but not bone metastasis (Fig. 6A). Kaplan-Meier curve analyses further confirmed that patients with tumor samples harboring positive hypoxia signatures displayed significantly higher risks of developing lung metastasis than patients without this signature (P = 0.0003 by log-rank test) (Fig. 6B). Importantly, the two subgroups of patients as classified by the common hypoxia gene signature did not show any statistical significance in the risk of developing bone metastasis (P = 0.3037). Taken together, our 45-gene hypoxia response signature demonstrated prognostic power for lung metastasis but not for bone metastasis.
A major bottleneck in metastasis research has been the lack of tools to investigate tumor-stromal interactions in real time in living animals. Recent development of noninvasive real-time imaging methods for various signaling pathways has greatly improved our understanding of their in vivo activities during the development of cancer (14, 17, 29, 30). In this study, we adopted and engineered a dual-luciferase imaging system in order to understand the dynamics of intratumoral hypoxia stress in primary tumors and metastases, as well as the impact of hypoxia on tumor growth in different organ microenvironments. Using such an imaging approach, we found that the intracellular hypoxia kinetics correlated with angiogenic abilities at both the primary site and the distant organs. This finding is consistent with the canonical paradigm of the relationship between hypoxia and angiogenesis.
The distinct behaviors of LM2 and SCP2 in primary tumor growth raise an interesting question: would growth of one cell line be influenced by the presence of the other cell line? We addressed this question by labeling LM2 and SCP2 with different luciferases and co-injecting into mammary glands at variable ratios. When total cell number injected was kept constant (106), the tumor volume declined as the fraction of LM2 in the mixture decreased (Supplementary Fig. S10A), consistent with the fact that LM2 is more tumorigenic in mammary glands than SCP2. Surprisingly, when each cell line was detected separately by BLI and normalized to the initial injection amount, decreased LM2 growth potential corresponded to its lower percentage in the implanted mixture, whereas SCP2 showed the opposite trend (Supplementary Fig. S10B-D). Perhaps the most striking observation was that at week 3 after tumor implantation, the absolute number of SCP2 cells (measured by Fluc BLI) in tumors initiated by the 10% SCP2/90% LM2 mixture became higher than that in tumors containing 50%, 90% and 100% of SCP2 during the initial implantation (Supplementary Fig. S10C, right panel). This result suggested that the net output of balancing the angiogenic activity (factors from both LM2 and SCP2, e.g. VEGF) and the anti-angiogenic activity (unidentified factor produced by SCP2) favors SCP2 growth but suppresses LM2 growth, as compared with their respective homogenous tumors. An important implication of this finding is that the primary tumor growth properties of SCP2 and LM2 are substantially influenced by the tumor microenvironment, instead of being completely dominated by the cell autonomous mechanisms.
It is interesting to note the apparent lack of extensive angiogenesis in the aggressive bone metastases generated by SCP2 confirmed by CD31 IHC, von Willebrand factor IHC, or dextran labeling (data not shown). This finding is in contrast to previous studies showing the positive CD31 staining in bone metastases formed by MDA-MB-231 and linking the HIF-1α activity in bone metastasis directly to angiogenesis (20, 21). We believe the difference can be explained by the heterogeneous nature of the parental MDA-MB-231 cell line. By using the relatively homogenous organotropic sublines, we were able to discover pleiotropic functions of hypoxia in promoting bone and lung-specific metastasis beyond regulating angiogenesis.
Intriguingly, although inhibition of HIF-1α reduces both bone and lung metastases, the hypoxia response signature is prognostic of lung but not bone metastasis in breast cancer patients. This paradox could be explained by the different mechanisms through which hypoxia contributes to tumor growth at the different organs. Development of lung metastasis and the growth of the primary tumors both rely on productive angiogenesis in response to hypoxia. Such a function could be selected for in the primary tumor during tumor progression and consequently promote lung metastasis after tumor dissemination. Furthermore, the LMS genes are highly enriched in hypoxic conditions. As HIF-1α expression has been found to progressively increases during human breast carcinogenesis (31), elevated HIF-1α may induce activation of LMS genes, such as ANGPTL4, in highly aggressive breast carcinoma. Such activation may subsequently prime tumor cells to effectively generate pulmonary metastases. Thus, it is not surprising that the hypoxia response signature correlates with higher risk of lung metastasis. On the other hand, only a limited number of genes (such as CXCR4 and DUSP1) among the hypoxia response signature contribute directly to bone metastasis, while many other common hypoxia regulated genes, including various angiogenic factors, may have relatively little direct influence on bone metastasis. Therefore, the overall hypoxia response gene signature was not able to stratify patients for risk of bone metastasis based on the gene expression profiles of their primary tumors. Nevertheless, this finding does not exclude the targeting of HIF-1α as a viable strategy for the treatment of bone metastasis, as our current study clearly indicated the inhibitory effect of DN-HIF1α or 2ME2 on bone metastasis.
Overall, we demonstrated the pleiotropic functions of hypoxia/HIF-1α at different stages of breast cancer progression and metastasis (Fig. 6D). In the primary tumor, hypoxia-induced angiogenesis promotes neovascularization and continuous expansion of the tumor mass. Furthermore, hypoxia transcriptional program may enhance the expression of LMS genes, which empowers departing tumor cells with the ability to effectively colonize the lung microenvironment. On the other hand, although hypoxia-induced angiogenesis appears to be non-critical for the aggressive growth of bone metastasis in the cell system studied, hypoxia does contribute to bone metastasis by activating the expression of a few bone metastasis genes, such as DUSP1 and CXCR4. Therefore, HIF-1α in breast cancer may be targeted to control the metastasis in multiple distant organs.
We thank the members of our laboratory for helpful discussions and M. Blanco for helpful comments on the manuscript. We thank C. DeCoste, M. Bisher, D. Storton, and J. Buckles at Princeton research facilities for technical assistance, S. Gambhir for triple-reporter plasmid, and T. Yoneda for the FLAG-DN-HIF-1α plasmid. Y. K. is an investigator of Champalimaud Metastasis Center at Princeton University and a Department of Defense Era of Hope Scholar Award (BC051647) recipient. This research was additionally supported by the Brewster Foundation, a research scholar grant from the American Cancer Society (RSG MGO-110765) and a R01 grant from the National Institutes of Health (R01CA134519). X.L. is a recipient of a Harold W. Dodds Fellowship from Princeton University.
Grant support: Department of Defense Era of Hope Scholar Award (BC051647), Brewster Foundation, American Cancer Society (RSG MGO-110765) and a R01 grant from the National Institutes of Health (R01CA134519). Y. Kang is an investigator of Champalimaud Metastasis Center at Princeton University and X.L. is a recipient of a Harold W. Dodds Fellowship from Princeton University.
Potential Conflicts of Interest: No potential conflicts of interest were disclosed.