HypoxamiRs and Cancer: From Biology to Targeted Therapy

Abstract

Significance: Hypoxia is a hallmark of the tumor microenvironment and represents a major source of failure in cancer therapy. Recent Advances: Recent work has generated extensive evidence that microRNAs (miRNAs) are significant components of the adaptive response to low oxygen in tumors. Induction of specific miRNAs, collectively termed hypoxamiRs, has become an accepted feature of the hypoxic response in normal and transformed cells. Critical Issues: Overexpression of miR-210, the prototypical hypoxamiR, is detected in most solid tumors, and it has been linked to adverse prognosis in many tumor types. Several miR-210 target genes, including iron-sulfur (Fe-S) cluster scaffold protein (ISCU) and glycerol-3-phosphate dehydrogenase 1-like (GPD1L), have been correlated with prognosis in an inverse fashion to miR-210, suggesting that their down- regulation by miR-210 occurs in vivo and contributes to tumor growth. Additional miRNAs are modulated by decreased oxygen tension in a more tissue-specific fashion, adding another level of complexity over the classic hypoxia-regulated gene network. Future Directions: From a biological standpoint, hypoxamiRs are emerging modifiers of cancer cell response to the adaptive challenges of the microenvironment. From a clinical perspective, assessing the status of these miRNAs may contribute to a detailed understanding of hypoxia-induced mechanisms of resistance and/or to the fine-tuning of future hypoxia-modifying therapies. Antioxid. Redox Signal. 21, 1220–1238.

Introduction

Tumor microenvironment and hypoxia: implications for therapy

Tissue hypoxia arising from a rapidly growing tumor mass with inadequate/dysfunctional blood supply is a feature of virtually all solid cancers (22, 147). The adaptive response to low oxygen encompasses complex biochemical and cellular processes, such as energy metabolism, cell survival and proliferation, angiogenesis, adhesion, motility, and resistance to oxidative stress (140). It is currently widely accepted that hypoxia represents an independent adverse prognostic factor in many tumor types and contributes to the ultimate failure of most anticancer therapies (12, 132). Therefore, a complete understanding of cellular adaptation to oxygen deprivation is key for developing more efficient therapeutic strategies (175).

Cells respond to hypoxia, in part, via a transcriptional program that is orchestrated by an oxygen-monitoring machinery, centered on the hypoxia-inducible factors (HIFs) (168–170). When oxygen tension falls below a critical threshold, HIF-prolyl hydroxylase activity is inhibited, leading to HIFα stabilization and heterodimerization with the β subunit, followed by transcriptional activation of hypoxia-inducible genes (148). Among the few hundred HIF targets identified to date, many are mechanistically involved in cancer formation and progression. Multiple studies have also identified elevated levels of HIF-1α and HIF-2α or both in primary tumors and their metastases (154, 193). In addition to lack of oxygen, HIF up-regulation can be the result of oncogenic pathway activation, loss of tumor suppressor genes such as Von Hippel-Lindau (VHL) (106), or increased abundance in reactive oxygen species (ROS) (26), all of which are constant features of tumor biology. Elevation of HIF-1 and HIF-2 is associated with increased tumor growth in the majority of human tumors analyzed to date, including breast, head and neck squamous cell carcinoma, and ovarian cancers (97, 143).

From a clinical perspective, the importance of hypoxia signaling in tumor progression and prognosis has spurred multidisciplinary efforts to more effectively identify highly hypoxic tumors, as well as to identify patients who are most likely to benefit from hypoxia-modifying therapy. Such methods, including radiological and nuclear medicine markers, immunohistochemistry for HIF-1 or its targets, represent undeniable progress toward this end; however, they are also recognized as having significant limitations (6, 18, 166).

During the past few years, the “classic” protein coding hypoxia-regulated genes have been joined by specific microRNAs (miRNAs), thus adding a new layer of regulation in an already complex response (83). These miRNAs, collectively termed hypoxamiRs, and their role in cancer as biomarkers and potentially biological players in their own right, will be the subject of this perspective.

miRNAs: regulation and roles in cancer

miRNAs are short single-stranded oligoribonucleotides (∼22 nucleotides in length) that regulate gene expression by inhibiting mRNA translation or by triggering cleavage of the target mRNA (162). miRNAs are recognized as important regulators in physiological and pathological settings, including tumorigenesis (17). Genes encoding miRNAs are initially transcribed as longer primary transcripts (pri-miRNAs) (108), which are processed by the nuclear RNase III Drosha, leading to hairpin-shaped pre-miRNAs. Pre-miRNAs are subsequently exported to the cytoplasm and cleaved by the Dicer RNase III into a short miRNA duplex. One strand of this duplex is degraded, while the other is retained as mature miRNA and incorporated into the RNA-induced silencing complex (RISC) in complex with proteins from the Argonaute (AGO) family (144). The mature miRNA guides the RISC to recognize mRNAs based on sequence complementarity, in particular between the “seed region” and the 3′-untranslated regions (3′UTRs) of the target, which generally leads to translation inhibition and/or mRNA degradation (41, 42). Due to the relative shortness of the seed region, the 3′ UTR of a given mRNA may contain multiple miRNA recognition sequences. Conversely, any given miRNA can, at least theoretically, regulate a large number of mRNAs, often hundreds, thus posing significant challenges for the efforts to identify biologically relevant targets.

Deregulated miRNA expression has been demonstrated in virtually all neoplasms. Interestingly, different cancer types tend to exhibit specific miRNA signatures (20, 120), including cancers of colon (125), breast (82), brain (33), liver (127), and lung (181). While the mechanisms behind the specific shifts of profiles in tumors are still being dissected, recent data on miRNA responses to microenvironment stresses and oncogenic alterations have provided critical clues.

During the past 6 years, multiple reports have demonstrated that miRNAs respond to low oxygen challenge and contribute to the regulation of specific genes under hypoxia (21, 47, 54, 56, 77, 102). In the next section, we discuss the current knowledge about the involvement of miRNAs in the hypoxic response in tumors, and debate on potential opportunities for cancer diagnosis, prognosis, and treatment.

Hypoxia-regulated miRNAs: a new paradigm of cellular response to the tumor microenvironment

Taking advantage of novel miRNA profiling techniques, over the past 6 years, groups from diverse fields searched for hypoxia-regulated miRNAs in a variety of cellular contexts. A rather large set of miRNAs, including miR-210, 21, 23, 24, 26, 103/107, and 373, was found to be induced under hypoxic conditions (Fig. 1) (21, 27, 36, 47, 77, 102, 141). Although the miRNAs described earlier were reported in at least two publications, there have been significant differences between the lists reported. Moreover, a large number of miRNAs were found to be down-regulated in hypoxia (101). The complexity was further highlighted by “deep-sequencing” of endothelial cells exposed to hypoxia (164). More than 400 annotated microRNA/microRNA* species were identified with a broad abundance range.

FIG. 1.

Key or experimentally validated miRNAs up- and down-regulated in hypoxia. A large number of hypoxamiRs have been described (>400 to date in various experimental set-ups)—this figure summarizes those discussed in the text. AGO1, argonaute 1; miRNA, microRNA; VEGF, vascular endothelial growth factor; HUVEC, human umbilical vein endothelial cell.

Apart from miR-210, which will be discussed in detail later in this review, there is very little overlap between hypoxia-induced miRNA profile between different cell lines and experimental set-ups, perhaps due to protocol and intrinsic cell line differences (69, 76). With the caveat of differences in the technologies employed by the different groups, the variability of hypoxamiR responses suggests a tissue-specific component of miRNA regulation under low oxygen. In turn, this may contribute to variations in the magnitude of gene expression changes in hypoxia, as well as to differences in cellular viability under these conditions.

Interestingly, while miR-21 is abundantly expressed in response to hypoxia in normal cells such as human umbilical vein endothelial cells (HUVECs), and it is frequently up-regulated in human tumors (82, 121, 156, 159, 165), its status as a hypoxamiR has been debated. Induction of miR-21 under hypoxia has been variably reported in normal and cancer cells (102, 142). When induced, most likely in an HIF-independent fashion, miR-21 should act as a pro-survival gene in the cancer microenvironment, especially in combination with other miRNAs. Polytarchou et al. found that only cancer cells expressing the protein kinase Akt2 had miR-21 induction during hypoxia. This was dependent on the binding of NF-κB, cAMP responsive element-binding protein, and CBP/p300 to the miR-21 promoter, in addition to the regional acetylation of histone H3K9, all of which were under the control of Akt2, and led to hypoxia resistance (134).

The miR-181 family, miR-372/373 and miR-93 have been reported in several studies (36, 100, 178), and they have intriguing clinical implications. For example, miR-181b was identified in a microarray analysis of hypoxia-regulated miRNAs in retinoblastoma cells, and administration of an miR-181b inhibitor suppressed proliferation (178). miR-181b is strongly associated with response to the 5-fluorouracil-based antimetabolite S-1 in colon cancer (79) and gemcitabine in pancreatic ductal adenocarcinoma (19). The induction of miR-373 by hypoxia has been reported in HeLa, MCF-7, and squamous cell carcinoma cell lines (36, 72). Similar to miR-210, the increase in miR-373 levels by hypoxia appears to be HIF-1α-dependent. Forced expression of miR-373 leads to a reduction in the RAD23B nucleotide excision repair protein, as well as in RAD52 (36); therefore, it may synergize with miR-210 in generating DNA damage and genetic instability in the tumor microenvironment.

Hypoxia and the HIF-signaling pathway play an important role in the regulation and sustenance of cancer stem cells and the epithelial-mesenchymal transition (EMT) phenotype. Under hypoxic conditions, the tumor microenvironment generates and sustains major EMT-triggering pathways, such as transforming growth factor-β and Notch signaling pathways (87), and hypoxamiRs are likely to be involved. Mechanistically, this pathway(s) appears to be even more complex than the pathways involving hypoxia-inducible hypoxamiRs, as both positively and negatively regulated miRs have been assigned to it. For example, hypoxia-induced down-regulation of miR-34a has been shown to promote EMT by targeting the Notch signaling pathway in tubular epithelial cells (44).

Conversely, miR-373 has also been found to transactivate E-cadherin gene expression through pairing with complementary promoter sequences (133), although the net effect of this regulatory mechanism on epithelial differentiation specifically under low oxygen remains unclear.

Recent “deep-sequencing” data have confirmed that miR-103/107 are hypoxamiRs (105) which are strongly induced in vascular endothelial cells. These hypoxamiRs are induced by HIF-1α and target AGO1, which anchors the miRNA-induced silencing complex. Interestingly, hypoxamiR targeting of AGO1 resulted in the translational desuppression of vascular endothelial growth factor (VEGF) mRNA and increased angiogenesis (29). Finally, hypoxia-induced miR-103/107 targets tumor suppressors such as death-associated protein kinase and Kruppel-like factor 4 to promote metastasis of colorectal cancer (27). This may be a contributing arm for the well-recognized effect of hypoxia in promoting invasion and metastasis.

Similar to miR-34a, many other miRNA are down-regulated in hypoxia, releasing suppressed genes that may be critical for adaptation to low oxygen. For example, miR-16, a prototypical tumor suppressor miRNA in leukemia and lymphoma, is down-regulated by HIF-1α, and it contributes to overexpression of VEGF in anaplastic lymphoma kinase-positive anaplastic large-cell lymphomas (37). These findings are summarized in Figure 1. Several groups have found that hypoxia and/or HIF-1α down-regulates miR-17/20a and the miR-17-92 cluster; however, the role of c-myc in this process has been directly contradicted by two studies (71, 180), and the functional significance has been complicated by a study showing that miR-20a is up-regulated by HIF-1α (115). It is likely that a balance of miRNAs in vivo is critical for an overall response, which may be cell-type dependent.

While the majority of work to date has focused on HIF as a transcriptional regulator of hypoxamiRs, recent evidence also implicates other transcription factors. For example, the NFκB subunit p50 was shown to contribute to miR-210 up-regulation in hypoxic trophoblasts, suggesting a role for other factors under oxidative and other stresses (190).

Interaction between HIF and miRNA in tumors: feedback loops with potential implications

HIF-1α is itself regulated by multiple miRNAs, creating complex positive and negative feedback loops. Those described so far include the miR-17-92 cluster mentioned earlier (153) and miR-20b, which is induced under hypoxia, implying negative feedback loops to fine-tune the hypoxia response (23, 110); additional players in such feedback include miR-138 (185); and miR-519c (24), but to date, they remain restricted to one publication.

As discussed earlier, Puissegur et al. as well as Kelly et al. showed that high miR-210 levels participate in HIF-1α stabilization during hypoxia (91, 136), which propagates a feed-forward loop of HIF amplification, and may account for a significant proportion of the total HIF protein expressed at any given point in hypoxia. In contrast to this amplification loop, miR-155 induction contributes to an isoform-specific negative-feedback loop for the resolution of HIF-1α activity in cells exposed to prolonged hypoxia (14). MiR-107 induction, in contrast to miR-210, seems to repress the expression of HIF-1α (179); while miR-145 inhibits HIF-2α (188), both with the effect of decreasing tumor growth and angiogenesis.

Other regulatory mechanisms have been described that act on HIF regulators, rather than HIF itself. For example, FIH, the asparagyl hydroxylase inhibitor of HIF activity, is a target of miR-31. Overexpression of this latter miR leads to normoxic stabilization of HIF, for example, in squamous cell carcinoma (121). A novel mechanism of regulation of HIF has been described by Ghosh et al., who found miR-424 induced in hypoxic endothelial cells targeted cullin 2, a scaffolding protein that is critical to the assembly of the ubiquitin ligase system, thereby stabilizing HIF isoforms (55). Given the increasing number of HIF regulators being reported in the literature, it is highly anticipated that the number of miRNAs which can affect the HIF pathway, directly or indirectly, increases significantly.

miR-210: a mirror of HIF in vitro and in tumors

Among the hypoxiamiRs, miR-210 stands out as the only miRNA that all the studies to date agree on (78), being induced in a wide range of cell types in reponse to hypoxia (see Fig. 2 for a range of cancer cells in vitro). Indeed, its induction has been reported in all cells studied to date, except the PEO1 ovarian cancer cell line (and in our hands, this too had induction, albeit at a very low level). This is also drastically different from the case of classic protein-coding genes in which a plethora of mRNAs with diverse functions are induced by hypoxia, with a relatively good overlap between different cell types (39). While miR-210 seems to be a rather HIF-1-specific target (21, 36, 77, 92), HIF-2-dependent regulation of miR-210 has also been reported (191). As is the case for the classic genes, HIF-1 directly binds to a hypoxia-responsive element (HRE) on the proximal miR-210 promoter (77). When the miR-210 core promoter is compared across species, this HRE site is highly conserved, indicating the importance of hypoxia/HIF in regulating miR-210 expression during evolution (34).

FIG. 2.

Comparison of miR-210 levels in normoxia and induction in hypoxia across a range of cancer cell lines. 231—MDA-MB-231; 361—MDA-MB-361; 10A—MCF-10A; DCIS—MCF-10A-DCIS. ***<0.001; **<0.01; *<0.05; paired two tailed t-test. miR-210 levels measured by real-time-polymerase chain reaction after 24 h at 1% oxygen or paired normoxic control, three biological replicates. All cell lines maintained under standard culture conditions. DCIS, MCF-10A-DCIS.

In cancer, in vivo, miR-210 level is correlated with a gene expression signature of hypoxia, termed hypoxia metagene (54, 77). Based on these data, miR-210 expression appears to be an accurate readout of HIF activity in vivo (40, 47, 77), thus opening the door for the use of miRNAs as markers of hypoxia in tumors (Fig. 3 summarizes cancers in which miR-210 is associated with clinicopathological variables; Table 1 includes details and references).

FIG. 3.

Summary of tumor types with which aberrant expression of miR-210 has been described. References and details are given in Table 1.

Table 1.

Association Between miR-210 and Clinicopathological Features in Human Cancer

Cancer	Role	References
Adrenocortical tumours	Higher in adrenocortical carcinoma than benign adenomas	Tombol et al. (156a)
Breast	High level—poor prognosis in early breast cancer	Camps et al. (21)
		Foekens et al. (51)
	High level—poor prognosis in lymph-node negative: estrogen receptor (ER)+ve, ER−ve and triple negative subtypes	Rothe et al. (139), Hong et al. (74)
Diffuse large B-cell lymphoma (DLBCL)	Elevated in serum of patients with DLBCL compared with normal	Lawrie et al. (107)
Glioblastoma	Elevation associated with poor prognosis	Qiu et al. (136a)
Head and neck	High level—poor prognosis	Gee et al. (54)
Hepatocellular carcinoma	Upregulated in hepatocellular carcinoma	Ying et al. (186a)
Lung	Elevated in primary small cell and adenocarcinoma compared with normal	Cho et al. (32a), Miko et al. (125a), Xing et al. (177a)
	Detection in sputum differentiated normal from patients with lung squamous cell carcinoma
Melanoma	Elevated in tumor compared with melanocytic naevi but not associated with prognosis	Satzger et al. (141a)
Ovarian	Deleted in many epithelial ovarian cancers	Giannakakis et al. (56), Vaksman et al. (160a)
	Elevated in effusions compared with primary tumor
Paraganglioma	Identification of HIF-1a/miR-210 axis independent of SDHD mutation—subgroup of head and neck paragangliomas	Merlo et al. (123a)
Pancreatic	Elevated—poor prognosis	Greither et al. (62), Wang et al. (171), Ho et al. (73)
	Circulating miR-210 elevated in serum and plasma
Pediatric osteosarcoma	Elevatation associated with poor prognosis	Cai et al. (19)
Peripheral nerve sheath tumors	Elevated in malignant tumors compared with benign tumors (neurofibromas)	Presneau et al. (134b)
Prostate	Overexpressed in prostate cancer	Porkka et al. (134a)
Renal	Classified tumor subtypes	Fridman et al. (51a), Juan et al. (88), McCormick et al. (123)
	Elevated in clear cell compared with normal
	Associated with improved prognosis
Soft tissue sarcoma	Expression associated with poor survival and age of tumor onset	Greither et al. (63)

ER, estrogen receptor; HIF, hypoxia-inducible factor; SDHD, succinate dehydrogenase complex subunit D.

miR-210 targets: a growing and diverse list

Identification of targets is an essential step toward a mechanistic understanding of hypoxamiRs in cancer. Most approaches include computational predictions using one or more of the online programs, including miRanda (8), TargetScan (113), and Pictar (98). These searches identify complementarity between 3′ UTR sequences of annotated coding genes and the “seed region” sequence of the miRNA (4). In general, the lists of candidates generated by individual programs exhibit a rather limited overlap, and none stands out as a perfectly accurate predictor of real targets. Interestingly, two widely employed algorithms, PicTar and TargetScan, predict relatively few targets for human miR-210, and, conversely, most of the experimentally validated targets are not predicted by any of these programs with a high score. It is also becoming increasingly apparent that “seed” binding is not always sufficient, as other features of the surrounding sequences can affect binding efficacy (112). To further complicate matters, in the case of miR-210, there is recent experimental evidence for a “seedless” target (49), which the prediction programs do not routinely address. Therefore, the addition of an experimental component to the bioinformatic search is crucial for the screening strategy.

In the case of hypoxamiR's targets, the approaches have been based on miRNA manipulation using mimics and antagomirs, in both hypoxia and normoxia, followed by expression profiling and comparisons with the results of computational predictions. This approach is well suited to identify targets that are regulated at the level of mRNA abundance (65, 114). However, since miRNAs frequently regulate the targets primarily by inhibiting protein translation (145), mRNA profiling will certainly miss many authentic miRNA targets. To address this limitation, AGO protein immunoprecipitation methods have been developed that capture the mRNAs recruited to the RISC complex which is enriched for a specific miRNA. Pulldown is followed by microarray or RNA-Seq, leading to targets that are regulated by both translational blockade and message degradation (90). Several groups have successfully pursued this approach in cells overexpressing miR-210 as a part of more integrative strategies in order to identify targets (48, 77). Notably, there were very few targets in common between these two studies, leading to the hypothesis that miR-210 regulates different genes in various cell types.

These results are not necessarily surprising, as comprehensive proteomic studies indicated that miRNAs act as rheostats by performing fine-scale adjustments to the output of hundreds of proteins (3, 145). Thus, only minor changes in protein and/or mRNA levels are expected for most real miRNA targets (e.g., 1.2–1.5-fold). Since microarray or RNA-Seq analyses carry a certain level of noise, these approaches are expected to miss a significant percentage of real targets. On the bright side, as the accuracy of “next-generation” RNA sequencing increases (and the cost decreases), an (almost) complete identification of physiologically relevant hypoxamiR targets becomes a highly realistic goal. Next, we discuss a selection of miR-210 targets that have been independently validated by multiple groups and debate their relevance for cancer biology.

Mitochondrial Metabolism in Cancer: A “Favorite” Target of miR-210?

Under normoxic conditions, mitochondria represent the “cellular energy factories” by generating the majority of ATP through the oxidative phosphorylation pathway using oxygen as a final electron acceptor. When oxygen supply is limited, cells switch to glycolysis for ATP production (the Pasteur effect). HIF-1 plays a critical role in this switch, by up-regulating the expression of most glycolytic enzymes and actively down-regulating mitochondrial respiration and biogenesis (38, 187). Results from several groups have demonstrated that hypoxic induction of miR-210 significantly contributes to this metabolic shift by down-regulating the activity of the mitochondrial electron transport chain (ETC) (Fig. 4). A well-dissected branch of this mechanism is based on the targeting by miR-210 of the iron-sulfur (Fe-S) cluster scaffold protein, ISCU (25, 30, 48, 50). This has been shown in a variety of cell types, including colon cancer, breast cancer, and human pulmonary arterial endothelial cells. ISCU is at the center of an ancient machinery that catalyzes the assembly of iron-sulfur clusters which are critical for the function of aconitase (member of the tricarboxylic acid cycle), and of mitochondrial ETC complexes I, II, and III (157). The relationship between miR-210 and ISCU is robust enough to be detected in cancer samples, a rather rare case for miRNA-mRNA pairs. Indeed, ISCU levels are inversely correlated with miR-210 in multiple tumor data sets (50, 123). Moreover, high ISCU level is generally associated with good prognosis in multiple tumor types, which is the opposite of miR-210, further underlining the biological relevance of this miRNA-target pathway.

FIG. 4.

A model of how miR-210 may coordinate down-regulation of key targets in mitochondrial complexes, leading to a glycolytic phenotype. COX10, cytochrome c oxidase assembly protein; DHAP, dihydroxyacetone phosphate; G3P, glycerol 3-phosphate; GPDH, glycerol-3-phosphate dehydrogenase, GPD1L, glycerol-3-phosphate dehydrogenase-like 1; SDHD, succinate dehydrogenase complex subunit D; ISCU, iron-sulfur cluster scaffold homolog; NDUFA4, NADH dehydrogenase (ubiquinone) 1 alpha subcomplex 4; miR-210—hsa-miR-210, Roman numerals refer to mitochondrial electron transport chain complexes. Several steps in Krebs cycle have been left off out for clarity. *Protein encoded by GPD1L contains a GPDH (NAD+) motif and shares 72% sequence identity with GPD1.

In addition to ISCU, which affects mitochondrial function indirectly, several integral components of the mitochondrial ETC have been found to be miR-210 targets: NADH dehydrogenase (ubiquinone) 1 alpha subcomplex 4 (NDUFA4) and succinate dehydrogenase complex, subunit D (SDHD), in lung cancer cells, (136), and cytochrome c oxidase assembly homolog 10 (COX10) in colon cancer cells (30). Interestingly, SDHD, a subunit of Complex II, is a well-documented tumor suppressor gene (5, 60), suggesting a pro-tumorigenic effect of miR-210.

An intriguing metabolic target experimentally confirmed is glycerol-3-phosphate dehydrogenase 1-like (GPD1L) (48). GPD1L is highly homologous to glycerol-3-phosphate dehydrogenases that transfer electrons from the cytoplasmic NADH to the mitochondrial ETC (16), and it may be itself a key regulator of NAD⁺/NADH ratio (116). Work by Kelly et al. has proposed a feedback mechanism in HEK293 and HeLa cells based on GPD1L repression by miR-210, which inactivates HIF prolyl hydroxylase activity, leading to stabilization of HIF (91). In this model, indirectly supported by data from Puissegur et al. (136), miR-210 is both downstream and upstream of HIF-1. A preliminary analysis of GPD1L expression in vivo shows that miR-210 is inversely correlated with GPD1L (Fig. 5). Lower levels of expression are an adverse prognostic factor in breast, clear cell renal carcinoma (CCRC), and lung adenocarcinoma (Fig. 6). This result remains to be validated at the protein level but suggests that miR-210 regulation of GPD1L is important in vivo. If true, this result stands in contrast to a recent study showing that miR-210 was inversely correlated with ISCU, but was a favorable prognostic marker in CCRC (123). This would need to be further investigated, particularly the impact of VHL mutation on miR-210 regulation. Further details on the methods used here are available in Supplementary Data (Supplementary Data are available online at www.liebertpub.com/ars).

FIG. 5.

Spearman's rank-order correlation test shows a negative correlation between miR-210 and GPD1L in breast, CCRC, and lung adenocarcinoma (data from The Cancer Genome Atlas). Data publicly available from the Cancer Genome Atlas Project (TCGA; http://tcga-data.nci.nih.gov/) were analyzed for mRNA and miRNA expression and correlation coefficient (between miR-210 and GPD1L) was calculated. BRCA, breast cancer (luminal A subtype); CCRC, clear cell renal carcinoma; KIRC, kidney cancer; LUAD, lung adenocarcinoma. Further details in Supplementary Data.

FIG. 6.

GPD1L expression in breast, CCRC, and lung adenocarcinoma (data from The Cancer Genome Atlas). Data publicly available from the Cancer Genome Atlas Project (TCGA; http://tcga-data.nci.nih.gov/) were analyzed for mRNA, miRNA expression. Log-rank test was employed to determine the relationship between expression and overall survival and the Kaplan–Meyer method was used to generate survival curves. The entire population in training/validation cohorts was randomly split (2/3, 1/3) and for each miRNA, we checked for a relation with the survival as follows. In both cohorts, patients were divided into percentiles according to miRNA expression. Using the training set, we considered any cut-off between 25th and 75th to significantly split the samples into two groups and checked for statistical significance in the validation set. We then chose the cut-off value to optimally split the samples in both cohorts. Further details in Supplementary Data.

Since HIF is also induced in tumors as a result of overactive oncogenic signaling even in the absence of hypoxia (94, 194), it is tempting to speculate that elevated miR-210 may also contribute to the Warburg effect in these tumors. This may be achieved by contributing to HIF-1α stabilization to promote aerobic glycolysis, as well as by down-regulating mitochondrial metabolism. Co-ordinated down-regulation of multiple mitochondrial genes has been noted in hypoxic tumors (103), and miR-210 is frequently elevated in highly glycolytic tumors, including glioblastoma (122), and pancreatic cancer. The relative contribution of miR-210 to the Pasteur and/or Warburg effects remains to be elucidated.

Generation of mitochondrial ROS is a well-recognized consequence of electron leakage during electron transport (128). Increased ROS production has been reported in hypoxia, potentially as a result of ETC dysfunction (66). The effect of miR-210 on ROS levels has been the subject of recent debate. Several groups have reported that miR-210 increases oxidative stress, in part, by ISCU suppression (25, 50). In addition, there appears to be a positive feed-forward loop between ROS generation and miR-210, mediated though pathways such as NF-κB (95). Conversely, other groups have reported a protective role of miR-210 against ROS production in a non-cancer context, particularly in human pulmonary artery endothelial cells, raising several possibilities that include cell-specific effects, level of experimental manipulation, and depth and duration of hypoxia.

miR-210: Regulator of Tumor Angiogenesis?

Angiogenesis is a complex multistep process that usually occurs during embryonic development and rarely in the adult under normal conditions (146). Cancer growth is highly dependent on neo-vessel formation to establish nutrient and oxygen supplies for cell viability and proliferation, and tumor hypoxia is a well-recognized trigger of this process (117).

Multiple miRNAs are known or suspected to be involved in the various steps of the angiogenic response, as either positive or negative regulators (172, 177). miR-210 expression was found to correlate closely with VEGF expression, hypoxia, and angiogenesis in breast cancer patients (51). Functionally, transduction of miR-210 in HUVECs using miRNA mimics stimulates the formation of capillary-like structures, as well as VEGF-induced cell migration (47, 119), while inhibition had the opposite effect. Ephrin-A3 (EFNA3) (47, 48, 62, 77) was identified as a candidate mediator for these effects and validated as a target in HUVECs (47). However, regulation of EFNA3 appears to be more complex, as its transcription is usually induced by hypoxia (47), suggesting that a model based on simple miR-210-mediated repression under low oxygen does not hold true. As of early 2013, there is limited information about the roles of EFNA3 in cancer or tumor angiogenesis; therefore, it is premature to state whether or not an axis hypoxia-miR-210-EFNA3 plays a key role in these processes.

More information potentially relevant for cancer angiogenesis was generated in cardiovascular biology models. The tyrosine phosphatase Ptp1b was identified as miR-210 target in a mouse model of myocardial infarction (75). Its human homolog PTP1B is documented to negatively regulate activation of the VEGF receptor VEGFR2, as well as to stabilize cell–cell adhesions through reducing tyrosine phosphorylation of vascular endothelial cadherin (131). Thus, by limiting the expression of EFNA3 and PTP1B, both negative regulators of angiogenesis, miR-210 could positively regulate angiogenesis in hypoxic regions of tumors. Interestingly, a recent study also suggests a positive feedback between VEGF, the best-documented and pharmacologically relevant angiogenic factor, and miR-210 (1). If confirmed in cancer-relevant experimental systems, the studies described earlier may have implications for increasing the efficacy of anti-VEGF therapy, for example by adding miR-210 inhibitors. The major caveat regarding a link between miR-210 and tumor angiogenesis is that a direct correlation between miR-210 expression and tumor angiogenesis (for example, by quantification of microvessel density in miR-210-defective or overexpressing tumors) is still missing. Some of these emerging connections are outlined in Figure 7.

FIG. 7.

A model of ways in which miR-210 and other miRNAs regulated by HIF may regulate angiogenesis. BDNF, brain-derived neurotrophic factor; ELK3, ETS-domain protein (SRF accessory protein 2); HIF, hypoxia inducible factor; HOXA3, homeobox A3; PTPN1 (PTP1B), protein tyrosine phosphatase, non-receptor type 1; VEGF, vascular endothelial growth factor; VEGFR2, vascular endothelial growth factor receptor 2.

miR-210 and the Response to DNA Damage

Genome integrity is challenged by diverse stresses, including mutagens, ROS, ultraviolet light, and chemo- or radio-therapeutic agents. Cellular responses to DNA damage involve a complex network of processes that detect and repair genomic lesions. miRNAs have been demonstrated to participate in these processes (104, 151, 167). Zhang et al. provided direct evidence that more than 20% of examined miRNAs are significantly induced on DNA damage (189). While not standing out as robustly induced by irradiation, at least based on the available data, miR-210, nevertheless, appears to have an impact on this complex process, as it down-regulates RAD52 in breast cancer cells (36, 48), a key component in the homologous recombination-mediated repair of double-strand breaks (7, 150). In addition to suppression of RAD51 in an HIF-dependent fashion (10) in hypoxic regions of tumors, suppression of RAD52 by miR-210 may provide an additional mechanism to help explain compromised homologous recombination repair in hypoxic cells (9). Consistent with this hypothesis, overexpression of miR-210 leads to double-strand DNA breaks in cultured diploid fibroblasts (46), although a possible involvement of RAD52 in this system was not investigated. Conversely, high levels of miR-210 may also be involved in lung cancer cells' radioresistance (64), suggesting that induction of this hypoxamiR both triggers error accumulation in the cancer cell genome and diminishes the effectiveness of radiotherapy. Obviously, the current knowledge is far from sufficient to allow us to speculate the relative contribution of its targets to such effects. Targets with possible relevance on this subject keep emerging. For example, a very recent study reported an intriguing role of a novel type of miR-210 target, which may be involved in response to radiotherapy. The first noncoding RNA candidate target of miR-210, X-inactive specific transcript (XIST) is a potent suppressor of hematological cancer in mice, as its inactivation leads to up-regulation of multiple oncogenes and down-regulation of tumor suppressors (186). This suggests that up-regulation of miR-210 may lead to widespread gene expression changes and even genomic instability, in ways not previously considered, and may be especially relevant in response to radiotherapy (see Fig. 8 for an overview of miR-210 relevance to genomic instability and response to radiotherapy).

FIG. 8.

A model of ways in which miR-210 may affect the response to radiotherapy, based on currently known target genes. ACVR1B, activin receptor 1B; AIFM3, apoptosis-inducing factor, mitochondrion-associated, 3; ATM, Ataxia telangiectasia mutated; BCL2, B-cell lymphoma 2; BHB1b, mitotic checkpoint serine/threonine-protein kinase BUB1 beta; CASP8AP2, caspase-8-associated protein 2; Cdc25B, cell division cycle 25 homolog B; E2F3, E2F transcription factor 3; Fam83d, family with sequence similarity 83, member D; FGFRL1, fibroblast growth factor receptor-like 1; H2AX, histone H2AX; HOXA1, homeobox A1; MDM2, mouse double minute 2 homolog; MNT, MAX-binding protein; PLK1, polo-like kinase 1; PTPN1, tyrosine-protein phosphatase non-receptor type 1.

miR-210 and Apoptosis: Boosting Cancer Cell Viability in Hypoxia?

Tumor microenvironmental stresses, including hypoxia and nutrient deprivation, are well-known triggers of cell death, both apoptosis and necrosis. Evasion from death responses is critical for tumor progression and a well-recognized hallmark of cancer (68). Thus, it is only fitting that miR-210 has been investigated in the context of survival responses, and, in particular, as a possible anti-apoptotic component of the hypoxic response. Generally, the available evidence suggests a predominantly anti-death role of miR-210 in a variety of cell types, with overexpression protecting cells from apoptosis, (75, 92, 102, 129, 176) and down-regulation of miR-210 during hypoxia promoting apoptosis (31, 47, 61, 102, 118, 182). These experiments have been performed in a variety of cells, including non cancerous (such as bone marrow-derived mesenchymal stem cells) and cancerous (such as breast and liver cancer). While significant gaps remain in our understanding of this process, multiple targets have been identified to help explain this effect: PTP1B (75), caspase-8-associated protein-2 (CASP8AP2) (92), and apoptosis-inducing factor, mitochondrion-associated, 3 (AIFM3) (182). However, the major caveat is that with the exception of CASP8AP2 (93), none of the other genes has been confirmed independently. In a recent report, while AIFM3 was found to be regulated by miR-210, its overexpression did not overcome the cytoprotective effects of the miRNA, suggesting that repression of other targets may be necessary (129). Moreover, despite the predominant evidence for an anti-apoptotic role, recent data suggested that miR-210 may also exhibit a pro-apoptotic function, at least in neuroblastoma cells, by targeting the anti-apoptotic gene B-cell lymphoma 2 (BCL2) (32).

In summary, while information supporting an miR-210-mediated blockade of apoptosis in hypoxia is accumulating for various cell types, it is still premature to state that this hypoxamiR represents a major protector against hypoxia-induced cell death in the tumor microenvironment.

miR-210: A Pleiotropic Regulator of the Cell Cycle

A common feature of extended exposure to hypoxia is down-regulation of various cell cycle genes, including cyclins and other positive regulators of cell cycle transition (67). However, this statement needs to be taken with particular caution, as especially in mild hypoxic conditions (5% or 10% oxygen), many cell types tend to proliferate better (99). Over the past few years, down-regulation of cell cycle by overexpressed miR-210 has been one of the more consistent themes about the biological effects of this hypoxamiR. Some of the cell cycle targets of miR-210 are also significant players in cancer biology.

One of the better-characterized cell cycle targets of miR-210 is E2F3, a promoter of G1/S transition (61, 109, 111). E2F3 was first reported and validated as an miR-210 target in ovarian cancer (56); however, the context reported by the authors was unique in the fledgling miR-210 field. Thus, the authors reported genomic loss of miR210 in ovarian cancer, and a resulting de-repression of cell cycle under hypoxic condition as a result of this event. However, this has not been found in other cell lines such as SKOV3 and A2780, which retain miR-210 expression and inducibility in hypoxia. Subsequent studies confirmed E2F3 as an miR-210 target (11, 48, 130), but the relative contribution of E2F3 to tumor cell cycle responses within the hypoxic microenvironment in tumors in vivo remains largely unknown. In addition to E2F3, fibroblast growth factor receptor-like 1 (FGFRL1) was also identified as an miR-210 target involved in cell cycle control in human esophageal cancer and derived cell lines (160), which was consistent with the earlier observation that FGFRL1 is robustly repressed by miR-210 (77). De-repression of FGFRL1 after miR-210 blockade accelerates cell cycle progression, while overexpression of miR-210 leads to cell cycle arrest in G1/G0 and G2/M phases (160). Again, as stated earlier with regard to E2F3, the relative contribution, if any, of this target for the viability of cancer cells in the hypoxic niches remains elusive. The effects of miR-210 on the cells cycle may, in fact, be significantly broader, to include a group of mitosis-related genes, such as Plk1, Cdc25B, Cyclin F, Bub1B, and Fam83D (70). Whether all these represent direct targets or more indirect responders downstream of the genes discussed earlier remains unclear.

Under some circumstances, miR-210 may promote cell cycle progression, for example, by down-regulating MAX-binding protein (MNT) (191), a member of the MYC/MAX/MAD network with a basic-Helix-Loop-Helix-zipper domain, and a well-characterized antagonist of c-MYC (80, 124). Since HIF-1 regulates cell proliferation under hypoxia, in part, by interacting with c-MYC (59), miR-210 may fine-tune cycle progression in hypoxic regions by fine-tuning the balance between major pathways that are involved in the response to the microenvironment. The net outcome (i.e., increased proliferation or cell cycle repression) may vary depending on the cell context, severity of hypoxia, and associated stresses.

miR-210: candidate cancer biomarker

A wealth of studies has firmly established that miRNAs are frequently dysregulated in human cancers (163), and that expression signatures can classify cancer subtypes (53). When used to classify poorly differentiated tumors, miRNA expression profiling outperformed mRNA expression profiling (120), pointing toward considerable biomarker potential. Expression of miR-210 has been consistently associated with poor clinical outcome in several solid tumor types (Table 1), including soft-tissue sarcoma, breast, head and neck, and pancreatic tumors (21, 51, 54, 62, 63, 74, 139).

A rather different scenario with regard to the impact of miR-210 expression seems to unfold in CCRC. This is the prototypical malignancy that is associated with a genetically up-regulated HIF pathway due to common inactivation of the VHL tumor suppressor (13, 135). Mutation and loss of heterozygosity of the VHL gene have been found in 57% and 98% of sporadic RCC cases, respectively (58). The VHL gene product functions as the adaptor subunit of the E3 ubiquitin ligase that targets hydroxylated HIFα for ubiquitination and degradation by the 26S proteasome (84, 85). Therefore, it is not surprising that miR-210 is particularly overexpressed in CCRC (88, 137, 174). However, contrary to the other tumors studied to date, expression of miR-210 in CCRC correlates with a favorable prognosis (123). This finding is certainly surprising but perhaps reflects the special biology of CCRC, given the role of VHL mutation or inactivation. The authors [McCormick et al. (123)] speculated that the association of miR-210 with good clinicopathological factors may be due to a shift to HIF-2 predominance or a further loss of cell differentiation coupled with ongoing mutations, but further work will be required to investigate.

Since hypoxic tumors display innate resistance to radiation and chemotherapy, a multitude of approaches have been tested to increase the treatment efficacy, with rather limited success. First, radiotherapy is often fractionated to take advantage of reoxygenation of surviving cells. Radiotherapy has also been combined with a hyperoxic gas and a vasoactive agent (86). Cytotoxins that specifically target hypoxic cells such as tirapazamine had shown promise in phase II trials (138). However, a phase III trial failed to show a benefit, perhaps because patients were not stratified on hypoxia status before treatment (138). Furthermore, effective biomarkers have yet to be identified that can reliably identify this subset of resistant tumors, enabling targeted treatment for patients at high risk, or “de-escalation” of therapy for patients at lower risk, sparing toxic side-effects on the quality of life.

The overall disappointing translational results using hypoxia-modifying therapies to date highlight the critical importance of novel validated markers of hypoxia for clinical practice, and of an improved understanding of the biology behind hypoxia-induced treatment resistance. Expression of miR-210 (in combination or not with the other hypoxamiRs discussed next) could help select patients at a high clinical risk, as it has been linked to aggressiveness in breast cancer, and, in particular, in the therapeutically challenging “triple negative” (estrogen receptor−ve/progesterone receptor−ve/human epidermal growth factor receptor 2−ve) subgroup. In a study of Japanese patients with triple negative breast cancer, patients whose breast cancers showed low miR-210 expression experienced significantly better disease-free and overall survival than those with high miR-210 expression (158). The impact of miR-210 on the efficacy of various types of therapy may, of course, vary. In particular, radiotherapy could benefit from an in-depth knowledge of miR-210 status. Thus, miR-210 overexpression has been shown to increase radioresistance in human lung cancer cell lines (64), while its down-regulation enhances radiosensitivity in hypoxic human hepatoma cells in vitro (182), and, most recently, in vivo (184). These findings have not yet been translated in humans in vivo but intriguingly, in a series of head and neck patients treated with post-operative radiotherapy, the expression of miR-210 was highly prognostic, suggesting that miR-210 may be a marker of radiotherapy resistance (54). However, whether miR-210 only serves as an “beacon” of tumor hypoxia, or actively promotes a more aggressive phenotype remains unclear (78).

There is an early indication that expression of multiple hypoxamiRs may be a viable alternative as a clinical biomarker of hypoxia as gene expression profiles. For example, a signature of hypoxia-related miRNAs derived by direct data mining of breast cancer data was shown to be a significant independent prognostic factor in breast cancer in multivariate analysis correcting for clinicopathological variables (15). This remains to be prospectively validated. The status of miR-210 expression in tumors may also help drive the choice of targeted therapy. Preliminary in vitro evidence was provided by Chen et al. (30), who reported that overexpressing miR-210 rendered cells significantly more susceptible to killing by 3-bromo-pyruvate, an inhibitor of the glycolytic pathway. Molecules of this class, such as 2-deoxyglucose or dichloroacetate, have been considered promising therapeutic agents; however, they are yet to fulfill their promise in clinical settings. Therefore, miR-210 may help identify subsets of patients who can benefit from such agents in the future.

There seems to be a general agreement that miRNAs are exceptionally stable and can be readily detected in the systematic circulation and other body fluids of healthy subjects and patients with malignant diseases (28, 57, 107, 126, 155, 173). It has been suggested that the high stability of miRNAs may be partially attributed to the exosomal miRNA packaging (161). Pilot studies assessing the use of circulating miRNAs as cancer biomarkers have attracted broad interest in the field and to date, at least 79 miRNAs have been reported as plasma or serum biomarker candidates for solid and hematologic tumors (2). miR-210 has been reported to be increased in the serum from patients with diffuse large B-cell lymphoma (107), CCRC (192), and pancreatic cancer (73, 171). Interestingly, hypoxia has been demonstrated as promoting the release of exosomes from cultured breast cancer cells (96); therefore, one can speculate that the elevated levels of circulating miR-210 may directly reflect the hypoxic state of tumor cells.

Circulating miR-210 levels have also been correlated with sensitivity to trastuzumab (a human epidermal growth factor receptor [EGFR] 2 monoclonal antibody), tumor presence, and lymph node metastases in breast cancer patients (89). This provides proof of concept that plasma miR-210 may also be used to monitor the response to anticancer therapies (35).

miR-210: a viable cancer therapeutic target?

The major and still unsolved dilemma regarding oncogenic miRs is whether or not they can be efficiently targeted in tumors. This question can certainly be formulated for hypoxamiRs, and miR-210 in particular. Recent development of anti-miRNA agents such as locked nucleic acids or peptide nucleic acids represents significant steps for therapeutic targeting of miRNAs in vivo (45, 52, 81, 105, 152). It is conceivable that inactivation of miRNAs involved in hypoxic adaptation, in combination with other anticancer agents, may be a viable strategy to target a tumor compartment that poses significant therapeutic challenges.

Hypoxia and miRNA machinery

Initial work suggested that hypoxia had a minimal effect on miRNA processing machinery (43), but more recent evidence suggests that hypoxia plays a regulatory role. Several genes such as Exportin 5 and AGO2 had an association with hypoxia in mRNA arrays (54). Hypoxia potentiates miRNA-mediated gene silencing through post-translational modification of AGO2 (176). Very recently, EGFR was shown to suppress the maturation of specific tumour-suppressor-like miRNAs in response to hypoxic stress through phosphorylation of AGO2. The association between EGFR and AGO2 was enhanced by hypoxia, leading to a reduction in the binding of Dicer to AGO2 and the inhibition of miRNA processing from precursor miRNAs to mature miRNAs (149).

Concluding Remarks and Cautionary Note

As a concluding remark of the plethora of studies which have investigated miRNAs in hypoxic cancer cells, one can state without any reservation that one miRNA deserves the name hypoxamiR: miR-210. However, putting the “equivalent” sign between hypoxamiRs and miR-210 would likely mean missing out on miRNA responders that in a given tumor type may be even more important than miR-210. On the other hand, taking into account the complexity of HIF regulation, the existence of HIF-independent responses to hypoxia, as well as the identification of hypoxia-downregulated miRs, one runs the risk of expanding the hypoxamiR family up to a point where the notion becomes “over-diluted.” Many more debates in the field are likely to follow in our effort to define some boundaries for a rather fluid notion. Overall, the top hypoxamiRs discussed earlier are quite possibly modifiers of tumor response to the microenvironmental challenges. Some, in particular miR-210, when measured in tumors and in circulating blood, already show promise as biomarkers of treatment response. By understanding, and potentially modifying, the cellular adaptation to oxygen deprivation, hypoxamiRs offer the possibility of both measuring hypoxia and developing more efficient therapeutic strategies to overcome hypoxia-induced mechanisms of resistance.

Abbreviations Used

10A

MCF-10A

231

MDA-MB-231

3′ UTR

3′-untranslated region

361

MDA-MB-361

ACVR1B

activin receptor 1B

AGO

argonaute

AIFM3

apoptosis-inducing factor, mitochondrion-associated, 3

ATM

ataxia telangiectasia mutated

BCL2

B-cell lymphoma 2

BDNF

brain derived neurotrophic factor

BHB1b

mitotic checkpoint serine/threonine-protein kinase BUB1 beta

BRCA

breast cancer (luminal A subtype); breast invasive carcinoma

CASP8AP2

caspase-8-associated protein-2

CCRC

clear cell renal carcinoma

Cdc25B

cell division cycle 25 homolog B

COX10

cytochrome c oxidase assembly homolog 10

DCIS

MCF-10A-DCIS

DHAP

dihydroxyacetone phosphate

EFNA3

ephrin-A3

EGFR

epidermal growth factor receptor

ELK3

ETS-domain protein (SRF accessory protein 2)

EMT

epithelial-mesenchymal transition

ER

estrogen receptor

ETC

electron transport chain

Fam83d

family with sequence similarity 83, member D

FGFRL1

fibroblast growth factor receptor-like 1

G3P

glycerol 3-phosphate

GPD1L

glycerol-3-phosphate dehydrogenase 1-like

GPDH

glycerol-3-phosphate dehydrogenase

H2AX

histone H2AX

HER2

human epidermal growth factor receptor 2

HIF

hypoxia-inducible factor

HOXA1

homeobox A1

HOXA3

homeobox A3

HRE

hypoxia responsive element

HUVEC

human umbilical vein endothelial cell

ISCU

iron-sulfur (Fe-S) cluster scaffold protein

KIRC

kidney cancer

LUAD

lung adenocarcinoma

MDM2

mouse double minute 2 homolog

miR-210

hsa-miR-210

miRNA

microRNA

MNT

MAX binding protein

NDUFA4

NADH dehydrogenase (ubiquinone) 1 alpha subcomplex 4

PLK1

polo-like kinase 1

PR

progesterone receptor

PTPN1 (PTP1B)

protein tyrosine phosphatase, non-receptor type 1

PTPN1

tyrosine-protein phosphatase non-receptor type 1

RISC

RNA-induced silencing complex

ROS

reactive oxygen species

SDHD

succinate dehydrogenase complex, subunit D

VEGF

vascular endothelial growth factor

VEGFR2

vascular endothelial growth factor receptor 2

VHL

Von Hippel-Lindau

XIST

X-inactive specific transcript

Acknowledgments

H.E.G. received support from the Department of Radiation Oncology, Sydney Cancer Center, Royal Prince Alfred Hospital, Sydney. M.I. received support from the American Cancer Society, the NIH/NCI (1R01CA155332-01), and the American Cancer Society (RSG 09-244-01-TBG).