Signaling in and out: long-noncoding RNAs in tumor hypoxia

Over the past few years, long non-coding RNAs (lncRNAs) are recognized as key regulators of gene expression at chromatin, transcriptional and posttranscriptional level with pivotal roles in various biological and pathological processes, including cancer. Hypoxia, a common feature of the tumor microenvironment, profoundly affects gene expression and is tightly associated with cancer progression. Upon tumor hypoxia, the central regulator HIF (hypoxia-inducible factor) is upregulated and orchestrates transcription reprogramming, contributing to aggressive phenotypes in numerous cancers. Not surprisingly, lncRNAs are also transcriptional targets of HIF and serve as effectors of hypoxia response. Indeed, the number of hypoxia-associated lncRNAs (HALs) identified has risen sharply, illustrating the expanding roles of lncRNAs in hypoxia signaling cascade and responses. Moreover, through extra-cellular vesicles, lncRNAs could transmit hypoxia responses between cancer cells and the associated microenvironment. Notably, the aberrantly expressed cellular or exosomal HALs can serve as potential prognostic markers and therapeutic targets. In this review, we provide an update of the current knowledge about the expression, involvement and potential clinical impact of lncRNAs in tumor hypoxia, with special focus on their unique molecular regulation of HIF cascade and hypoxia-induced malignant progression.

HIF-1α and HIF-2α regulate both common as well as distinct sets of target genes. Meanwhile, HIF-3α, an isoform lacking the transactivation domain, has a dominant negative effect on HIF-dependent gene transcription [7,8].
In the presence of sufficient oxygen, HIF-1α subunits are post-translationally modified by a family of dioxygenases (prolyl hydroxylase domain-containing dioxygenases PHD1, 2 and 3, also known as EGLN1-3, Egl-9 family hypoxia inducible factor 1-3,). Upon hydroxylation, HIF-1α subunits are recognized by the E3 ubiquitin ligase, VHL (von Hippel-Lindau tumor suppressor protein), leading to the poly-ubiquitination and subsequent rapid degradation through the ubiquitin-proteasome pathway (Fig. 1a). Under hypoxic conditions, the PHD dioxygenase activity is inhibited, and the accumulated HIF-1α subunit translocates into the nucleus, dimerizing with HIF-1β and binding to the HREs (hypoxia response elements; the consensus 5′-(A/ G)CGTG-3′ nucleotide sequence) within the promoter regions of HIF target genes to stimulate downstream transcriptional activation of multiple hypoxia responsive genes (Fig. 1a), eliciting a wide spectrum of cellular adaptations, such as decreased apoptosis, enhanced angiogenesis, proliferation, migration and invasion [1,[9][10][11]. In addition to protein coding genes, it has been widely acknowledged that the non-coding transcriptome is also responsive to hypoxia and play critical roles in the hypoxic response and HIF-1 associated cancer progression [12][13][14][15][16].
Over the last few years, a large number of dysregulated lncRNAs have been associated with numerous diseases, including cancer [34][35][36][37]. While a few cancer-associated lncRNAs have been well characterized [27,38], the functions of most remain largely unknown. Dysregulation of many cancer-associated lncRNAs is linked to both clinicopathological features and survival outcomes of patients, suggesting that functional annotation of these lncRNAs will eventually identify new venues for early diagnosis and therapy of cancer [39]. Several studies have shown that the modulation of lncRNAs in response to hypoxia could play a regulatory role in HIF signaling cascade [14-16, 40, 41]. Here, we refer to these unique transcripts as "hypoxia-associated lncRNAs" (HALs). These RNA molecules are involved in multiple hypoxiadriven cancer progression pathways. In this review, we provide an updated summary of the tumor HALs, with a specific emphasis on the crosstalk between these lncRNA species and cellular hypoxia response (Table 1 and Additional file 1: Table S1). We address current models describing the functional involvement of these new players in cancer progression, highlighting their relevant clinical potential as cancer biomarkers or therapeutic targets. Our discussion is centered on tumor hypoxia. For the functional roles of lncRNAs in hypoxia-induced kidney/hepatic/myocardial injury and neuromuscular or cardiovascular diseases, interested readers are referred to a number of comprehensive reviews published in recent years [127][128][129][130][131][132].

Review
LncRNAs as emerging driving forces in cancer progression upon tumor hypoxia Given the pivotal roles of lncRNA in hypoxia-associated tumorigenesis pathways, multiple approaches have been applied in the identification of hypoxia-regulated lncRNAs [87,90]. A comprehensive analysis coupling RNA-seq with ChIP-seq [12] revealed the extensive involvement of HIF-1α and HIF-2α in the transcriptional regulation of lncRNAs upon hypoxia. In recent years, the rapid expansion of research on lncRNAs has provided additional insights into those associated with cellular hypoxia response. Table 1 presents an updated list of these hypoxia-associated lncRNAs (HALs). Upon hypoxia, most HALs are up-regulated. HIF could directly promote the expression of these hypoxiainducible lncRNAs through binding to the HREs (hypoxia response elements) located in their promoter (Table 1) [41]. lncRNA-LET [93], CF129 [54] and CRPAT4 [56] are among the few which are down-regulated in hypoxic conditions. Notably, lncRNA-SARCC is able to respond to hypoxic stress differentially in a VHL-dependent manner [94].
Most of the HALs identified have impacts on cancer progression, although the mechanistic details are not all clear. Table 1 shows an overview of the tumor HALs. We summarize in the table, their potential molecular target related to hypoxic responses as well as their reported functions and signaling pathways. These HALs Fig. 1 Regulations of HIF-1 activity by HALs. a Regulation of HIF-1. Under normoxia (green arrows), HIF-1α subunit is hydroxylated by PHDs (prolyl hydroxylase domain proteins). Hydroxylation residues within HIF-1α facilitates interaction of HIF-1α with the E3 ubiquitin ligase VHL protein, targeting HIF-1α for polyubiquitination and subsequent proteasome-dependent degradation. Upon hypoxia (red arrows), the PHDs and other prolyl hydroxylases are inhibited, leading to HIF-1α stabilization and translocation into nucleus. After dimerization with its transcriptional partner HIF-1β and recruitment of co-activators (e.g. CBP/ p300), the HIF-1 heterodimer binds the HRE (hypoxia response element) of target genes to regulate transcription. b Transcriptional co-activator. Hypoxiainduced LncHIFCAR could directly interact with HIF-1α and facilitate the recruitment of HIF-1α and p300 cofactor to the target loci, thereby upregulating HIF-1 target genes. c Recruitment of transcription factor. HIF-1α-induced LncRNA-MTA2TR could recruit ATF3 to the promoter area of MTA2, thereby transcriptionally upregulating the expression of oncogenic MTA2. MTA2 can subsequently enhance HIF-1α protein accumulation via deacetylation, forming a feedback loop to amplify HIF-1 signaling. d mRNA stability control. The expression of lncRNA-LET is repressed through hypoxia-induced HDAC3, which reduces the histone H3 and H4 acetylation at the LncRNA-LET promoter. Decreased lncRNA-LET expression reduces the lncRNA-LET-mediated degradation of HIF-1α negative regulator, NF90, leading to HIF-1α accumulation. e ceRNA/miRNA sponge. Hypoxia-induced H19 could upregulate HIF-1α expression by absorbing miRNA let-7 and nullifying let-7-mediated HIF1A mRNA suppression. f Molecular decoy. lincRNA-p21 is able to disrupt the interaction between HIF-1α and its negative regulator VHL via separate binding to both HIF-1α and VHL, thereby blocking VHL-dependent HIF-1α degradation. g Complex scaffold. LINK-A-mediated recruitment and enzymatic activation of BRK and LRRK2 kinases could facilitate phosphorylation of HIF-1α at specific residues. These phosphorylation modifications prevent subsequent HIF-1α degradation and enhance the association between HIF-1α and cofactor p300, thereby upregulating HIF-1 target genes. See text for a more detailed discussion

Unclear mechanism
(HAS2-AS1-mediated hypoxia-induced EMT is dependent on cell-adhesion molecule CD44 and RHAMM) [75] HIF2PUT N.D.     (Table 2). Notably, several HALs, such as UCA1, PVT1, H19 and MALAT1, might adapt more than one action mode in different cancer types. In the discussion below, we highlight the selected few HALs to illustrate their mechanisms of actions.

HAL-mediated epigenetic and transcriptional regulation
A large number of lncRNAs are localized in the nucleus, participating in various biological processes, including chromatin organization, nuclear structure, transcriptional and post-transcriptional regulation of gene expression. With regard to chromatin organization, the pangenomic investigations of RNA-protein interactions have shown that two hypoxia-inducible, oncogenic antisense RNAs ANRIL (also known as CDKN2B antisense RNA 1) and HOTAIR (HOX transcript antisense RNA) [50,80] could interact with different histone-modifying complexes, and have thus been proposed to impact the chromatin modification and transcriptional state [138]. However, whether these two antisense RNAs are involved in modulating gene expression in response to hypoxia via epigenetic modification or chromatin reorganization remains to be characterized. In addition, WT1-AS could mediate hypoxia-induced upregulation of oncogenic transcription factor WT-1 in cis through modulating histone H3K4 and H3K9 methylation around the transcription start site of WT1 mRNA, contributing to acute myeloid leukemia (AML) progression [124]. Similarly, in gastric cancer, lncRNA-AK058003, which could be profoundly induced by hypoxia, resides upstream of SNCG (synuclein gamma, a synuclein family member, promotes migration, invasion and metastasis) and enhances SNCG expression in cis through demethylation of SNCG promoter CpG islands, thereby driving hypoxia-induced metastasis [90]. In the context of nasopharyngeal carcinoma (NPC), up-regulated PVT1 could serve as a scaffold for a transcriptional activator, the histone acetyltransferase KAT2A, to activate transcription of NF90. NF90, a RNA-binding protein, has been reported to stabilize many target mRNAs, including HIF1A mRNA. Indeed, the upregulated NF90 increased HIF1A mRNA stability and promoted malignant transformation of NPC cells [111]. In addition, in hypoxia-injured pheochromocytoma cells, up-regulated MEG3 (maternally expressed gene 3) could recruit methylation proteins DNMT3a, DNMT3b and MBD1 to facilitate TIMP2 promoter methylation, which in turn inhibited the expression of this cell cycle arrest inducer TIMP2. Moreover, a HIF-1α negative regulator, OS-9, is reported to facilitate HIF-1α hydroxylation and subsequent proteasomal degradation through tethering the interaction between HIF-1α and prolyl hydroxylases (PHDs) [139]. Interestingly, in pancreatic ductal adenocarcinoma (PDAC), another  [137] lncRNA ENST00000480739 could inhibit HIF-1α by upregulating OS9 (osteosarcoma amplified-9) expression through enhancing the acetylation of H3K27 within OS9 gene promoter [60]. Of note, in PDAC, the level of ENST00000480739 is markedly downregulated, and negatively correlated with lymph node metastasis, in agreement with its negative regulatory role in HIF-1 signaling [60]. As ENST00000480739 resides upstream of the OS9 promoter region, this lncRNA also act in cis to induce OS9 transcription. Apart from chromatin structure remodeling, a series of HALs could modulate transcription and thereby fine-tune the HIF network. For instance, lncRNA HIF2PUT (HIF-2α promoter upstream transcript), RERT-lncRNA and hypoxia-inducible BC005927 are all found to act in cis to up-regulate neighboring protein-coding genes HIF2A (encodes HIF-2α), EGLN2 (encodes prolyl hydroxylase PHD1) and EPHB4 (encodes Ephrin type-B receptor 4, a metastasis-related gene), at the transcriptional level, respectively [51,76,116].
Moreover, HALs could directly act on specific transcription factors through physical interactions to modulate their transactivation activities. We recently identified a hypoxia-inducible lncRNA LncHIFCAR (long noncoding HIF-1α co-activating RNA, also known as MIR31HG) acting as a HIF-1α co-activator via direct interaction with HIF-1α, thereby enhancing the binding of HIF-1α and cofactor p300 to the target loci (Fig. 1b). As the abundance of the HIF complex increases, the hypoxia-induced HIF-1 signaling cascade is augmented to further promote subsequent cancer progression [89]. Meanwhile, in pancreatic cancer, HIF-1α-induced lncRNA-MTA2TR (MTA2 transcriptional regulator RNA) transcriptionally upregulates the expression of oncogenic MTA2 (metastasis associated protein 2) by recruiting ATF3 (activating transcription factor 3) to the promoter area of MTA2 [102]. Subsequently, MTA2 can enhance the accumulation of HIF-1α protein via MTA2-mediated HIF-1α deacetylation and stabilization, which further activates HIF-1α transcriptional activity, forming feedback loops to augment HIF-1 signaling [102] (Fig. 1c). In addition, through binding to PSF (PTB-associated splicing factor), hypoxia-induced lncRNA MALAT1 released PSF from its downstream proto-oncogene GAGE6 (proto-oncogene G antigen 6) and activated its transcription, thereby promoting proliferation, migration and invasion of lung adenocarcinoma cells [98,99]. Given the extraordinary variety of transcriptional regulatory machinery discovered in the cell, it is anticipated that more lncRNAs-mediated regulation on hypoxia-induced transcriptional program will be unraveled in the imminent future.

HAL-mediated post-transcriptional control
HALs also participate in post-transcriptional regulation including mRNA stability and miRNA-mediated gene silencing.
mRNA stability control Three HALs, lncRNA-LET (Long noncoding RNA Low Expression in Tumor), DANCR (Differentiation Antagonizing Non-Protein Coding RNA) and HIF1A-AS2 (HIF1A Antisense RNA 2; also known as aHIF), have all been reported to affect HIF1A mRNA stability. lncRNA-LET expression is generally suppressed in various types of tumors, whereas hypoxia-induced HDAC3 (histone deacetylase 3) could repress its expression by reducing the histone acetylation of the lncRNA-LET promoter region [93,140]. Mechanistically, lncRNA-LET is bound to NF90 (nuclear factor 90), which increases NF90 degradation by the proteasome. As RNA binding protein NF90 could stabilize HIF1A mRNA [93,141], the downregulation of lncRNA-LET upon hypoxia plays a key role in the stabilization of NF90 protein, thereby increasing HIF-1A mRNA stability upon hypoxia and accordingly hypoxia-induced cancer cell invasion [93] (Fig. 1d). Likewise, in nasopharyngeal carcinoma, another oncogenic lncRNA DANCR was upregulated and associated with lymph lode metastasis and poor survival [57]. Through interaction with the NF90/ NF45 complex, DANCR could increase HIF1A mRNA stability, leading to metastasis and disease progression.
In addition, another hypoxia-inducible antisense lncRNA HIF1A-AS2, was shown to be up-regulated in various tumors [42,43,46,142,143] and could differentially regulate HIF-1α and HIF-2α expression during long-term hypoxic conditions [43,47]. Upon acute hypoxia, HIF-1α and HIF-2α were similarly induced. Interestingly, during prolonged hypoxia, these two proteins were differentially regulated as HIF-1α protein level gradually decreased due to a reduction in its mRNA stability, whereas HIF-2α protein remained upregulated. Meanwhile, long-term hypoxia also induced an increase in HIF1A-AS2, whose gene promoter harbors functional HREs. During prolonged hypoxia, HIF1A-AS2 could bind to its sense counterpart, the HIF-1A mRNA 3′-UTR, and possibly expose the AU-rich elements in this region, thereby destabilizing HIF-1A mRNA to convey target gene specificity [43,47]. Paradoxically, HIF1A-AS2 was also shown to sequester miR153-3p (see next section) to enhance HIF-1A expression [44]. Thus, the mode of action of HIF1A-AS2 is complex and likely context-dependent.
miRNA sponges A wealth of lncRNAs adapt a wellcharacterized, common mechanism, "ceRNA (competing endogenous RNA)" or "RNA sponges", to repress miRNA-mediated gene silencing. The ceRNAs compete for shared miRNAs, sequester these miRNAs and diminish their silencing effect on target mRNAs.
HAL-mediated control of protein activity, stability and/or higher-order complex formation In addition to acting as ceRNAs to modulate gene expression through interaction with miRNAs, HALs have multiple molecular modes to act at the protein level to further modulate gene expression. One of the hypoxiainduced lncRNAs, PVT1 (plasmacytoma variant translocation 1), was implicated in cervical cancer progression, likely through its interaction with a multifunctional shuttling protein, nucleolin [115]. In multiple cancer cell lines, HIF-1-induced lincRNA-p21 provides another example as to how HALs modulate hypoxia response by protein sequestration. Through separate binding to HIF-1α and VHL, lincRNA-p21 could increase HIF-1α accumulation by disruption of the VHL/HIF-1α interaction and subsequent attenuation of VHL-mediated HIF-1α ubiquitination and degradation [85] (Fig. 1f). Another HIF-1α binding lncRNA CASC9 (cancer susceptibility candidate 9) is highly expressed in nasopharyngeal carcinoma (NPC) tissues. CASC9 could interact with and stabilize HIF-1α, promoting the glycolysis and tumorigenesis of NPC cells [53].
Nevertheless, in addition to fine-tuning the activity of one single protein, HALs can also dynamically modulate higher-order protein organizations by serving as scaffolds or molecular decoys. In mesenchymal glioblastoma stemlike cells, through direct binding to two RNA binding proteins, DHX9 (ATP-dependent RNA helicase A) and IGF2BP2 (insulin-like growth factor 2 mRNA-binding protein 2), lncRNA HIF1A-AS2 could facilitate the interaction between this protein complex and their mRNA target HMGA1 (high mobility group AT-hook 1), thereby enhancing HMGA1 expression as well as the downstream molecular response to hypoxic stress [46,47].
In triple-negative breast cancer (TNBC), LINK-A (long intergenic non-coding RNA for kinase activation) has a critical role in the growth factor-induced HIF-1α signaling under normoxic conditions [88]. LINK-A is required for the recruitment of BRK (breast tumor kinase) and subsequent enzymatic activation, which is stimulated by HB-EGF (Heparin-binding EGF-like growth factor) signal. HB-EGF mediates the heterodimerization of EGFR (epidermal growth factor receptor) and GPNMB (transmembrane glycoprotein NMB) to form 'EGFR:GPNMB' complex. Due to its direct interaction with BRK and LRRK2 (leucine-rich repeat kinase 2), LINK-A could recruit these two kinases to EGFR:GPNMB heterodimer, thereby inducing their kinase activities, resulting in HIF-1α phosphorylation: BRK-mediated HIF-1α phosphorylation at Tyr 565 , a phosphorylation preventing the adjacent Pro 564 hydroxylation of HIF-1α and subsequent HIF-1α degradation under normoxic conditions; and LRRK2-mediated HIF-1α phosphorylation at Ser 797 , which facilitates the interaction of HIF-1α with the transcriptional cofactor p300 [88] (Fig. 1g). In TNBC samples, both LINK-A abundance and HIF-1 signaling activation are correlated with cancer progression and shorter survival, revealing potential therapeutic targets for TNBC [88].
An additional novel function of lncRNAs is their structural role in the assembly of nuclear domains. For instance, MALAT1 (metastasis-associated lung adenocarcinoma transcript 1, also known as NEAT2) and NEAT1 (nuclear enriched abundant transcript 1) are located in two well-characterized nuclear bodies, nuclear speckles and paraspeckles, respectively. Also known as SC35 splicing domains, nuclear speckles are membrane-less compartments and their formation involves "phase-separation" mediated by aggregated lncRNAs and proteins. Being an abundant component of the nuclear speckles, MALAT1 associates with numerous splicing factors and other SR (serine/arginine-rich) proteins, and is required for their correct localization to the nuclear speckles, although the overall nuclear speckle assembly is not dependent on the abundance of MALAT1 [144,145]. So far, the functional involvement of MALAT1 in RNA splicing in response to hypoxia remains to be determined. In contrast, lncRNA NEAT1 is shown to be an essential architectural component of nuclear paraspeckles [144,145]. The precise function of paraspeckles remains largely elusive, but proposed to regulate gene expression via the retention of hyper-edited RNA and other multifunctional factors in the nucleus [104]. Given the functional involvement of both MALAT1 and NEAT1 in nuclear structure, further investigation of the extent to which these nuclear structures and their associated transcription reprogramming respond to hypoxia will deepen our understanding of the cellular dynamic response to hypoxia.

HAL-mediated control of hypoxia response via unclear mechanism
As listed in Table 1, most of the HALs identified with profound impact on tumorigenesis have not yet been examined in mechanistic detail. However, other reports regarding the same lncRNA with functional characterization might reveal clues about their biological roles in response to hypoxia. For instance, lncRNA PCGEM1 was found to be overexpressed in gastric cancer, and could be induced by hypoxia [110]. In gastric cancer cells, PCGEM1 could promote the invasion and metastasis through activating the expression of SNAI1, a key transcription factor of EMT, though the underlying mechanism remains elusive [110]. Notably, in prostate cancer, our group previously reported that the oncogenic PCGEM1 could promote chromatin recruitment of c-Myc and enhances its transactivation activity through direct physical interaction [146]. As SNAI1 is a wellcharacterized downstream gene of c-Myc, the possible functional role of the PCGEM1/c-Myc/SNAI1 signaling axis in hypoxia-associated cancer progression warrants further investigation.
In summary, as noted in the above sections, given the relatively large size and the structural flexibility of lncRNAs, it is to be expected that they interact with multiple RNA or protein components and have multifunctions, perhaps in a context-dependent manner. As such, their roles in hypoxia responses and in tumor progression may differ appreciably in different cancer types.

LncRNAs as predictive biomarkers and therapeutic targets for hypoxic tumor Extracellular vesicles-containing HALs and their biologic effects on tumorigenesis
Extracellular vesicles are effective devices for transporting biomolecules among various cells types [147,148]. Based on the difference in size and biogenesis, cellderived extracellular vesicles can be broadly divided into two main categories: exosomes (30-100 nm in diameter) and microvesicles. Together with proteins and other non-coding RNAs, emerging evidence has shown that lncRNAs are packaged into exosomes [149,150], and the abundance of lncRNAs in exosomes correlates with their expression level in the cell of origin [151]. Through exosomal transfer, several lncRNAs are shown to potentiate cell responses to hypoxia between cancer cells [87], as well as between cancer cell and the associated microenvironment [150]. Table 2 summarizes hypoxiaassociated lncRNAs identified extracellularly. For example, linc-ROR was found abundant in tumor cells as well as in exosomes derived from tumor cells [87]. It is increased both in cells or exosomes during hypoxia, and it up-regulates HIF-1α expression by absorbing miR-145. By co-culture systems, linc-ROR-containing exosomes increase HIF-1a transcription in recipient cells [87]. Hypoxia can shape and fine tune specific macrophage phenotypes in the tumor milieu that are known to promote tumor progression [152]. Chen et al found lncRNA HISLA (also known as LINC01146), secreted by tumorassociated macrophages, stabilized HIF-1α and enhanced aerobic glycolysis in cancer cells, leading to contagious metabolic reprogramming within tumor regions [150]. PVT1, a lncRNA that often co-amplifies with c-myc and functions as miRNA sponge to upregulate HIF-1α expression [153,154], is another example of exosomal transfer between TAMs (tumor associated macrophages) and cancer cells. PVT1 is detected in exosomes derived from colon cancer cells, particularly in more aggressive cells [136]. In granulocytic myeloid-derived suppressor cells (G-MDSCs), PVT1 was up-regulated by HIF-1α under hypoxia and contributed to immunosuppression, given its depletion reduced the suppression of these cells on T-cells and delayed tumor progression [155]. Other exosomal-transferred lncRNAs that are implicated in cancer cells during hypoxia include UCA1 in bladder cancer for promoting tumor growth and EMT [137], and CCAT2 for glioma's resistance to apoptosis and angiogenesis [134].
The functions of lncRNAs in exosomes for tumor progression await to be explored given a significant level of non-coding RNAs are revealed in exosomes (and elevated upon hypoxia) whereas only a small fraction has been studied [149,150,156]. Accordingly, it is conceivable that multiple tumor phenotypes and signaling pathways are affected upon exosomal loading. Indeed, by microarray analyses, Mao et al showed hundreds of lncRNAs, together with other transcripts, are changed in endothelial cell recipients of exosomes derived from squamous cancer cells [157]. Importantly, they found exosomes obtained from hypoxic condition facilitate angiogenesis and metastasis better than those obtained from normoxic condition in a xenograft model. Similar effects between normoxic exosomes and hypoxic exosomes on angiogenesis were found in a mouse xenograft model of glioblastoma, with additional effect on accelerating tumor expansion at later stage [158]. The elevation in transcripts by exosomes could result from direct gene transfer, or sequential effects mediated by the transferred genes. By which mechanism lncRNAs are selected to be packaged in the exosomes upon stimuli is not known; nevertheless, these studies revealed exosomes as a means by which hypoxia in the tumor microenvironment facilitates tumor cells to spread and progress.

Diagnostic potential of HALs
Several HALs with known oncogenic functions have been detected in patient-derived exosomes, including H19 in serum from patients with bladder cancer [159], HOTAIR in urinary exosomes from patients with urothelial bladder cancer [156], UCA1 in serum from bladder cancer patients [137], and HIF1A-AS2 in patients with endometriosis [133]. Future studies aimed at identifying hypoxia-responsive transcripts in extracellular vesicles would surely reveal more players in this aspect. Bearing differential expression patterns between normal and malignant stages and/or tumor size, oncogenic lncRNAs that can be detected extracellularly would potentially serve as non-invasive biomarkers for early detection, prognosis prediction, and disease surveillance. PCA3, up-regulated in > 90% of men with prostate cancer, is an example of this [160]. A urine-based assay has been approved by the United States Food and Drug Administration (FDA) since 2012 as an alternative diagnostic test for patients undergoing repeat prostate biopsy or with previous negative prostate biopsy.
As described above, there is considerable evidence indicating hypoxia as a progression factor for tumor development [161]. Hypoxia promotes angiogenesis, tumor metastasis, immune evasion and therapy-resistance. The oxygenation status of tumor was reported to influence local tumor response to radiation treatment, as well as overall survival in a variety of tumors [162][163][164]. Chemotherapeutic drugs, such as Docetaxel and Sorafenib, also tend to be more effective in normoxic conditions [165,166]. The hypoxic regions in tumors are infiltrated with cells which promote tumor tolerance (regulatory Tcells, myeloid-derived suppressor cells, and macrophages), while antitumor T-cells are devoid and inhibited by HIF-1α-mediated accumulation of extracellular adenosine [167][168][169]. PD-L1 (Programmed death-ligand 1), a ligand expressed by tumor cells or myeloid-derived suppressor cells to suppress T-cell's anti-tumor immunity, is upregulated by and a direct target of HIF-1α during hypoxia [170]. It has become increasingly apparent that hypoxia in tumors fosters immune suppression and prevents effective immunotherapy. Considering the ill-effects of hypoxia, it is important to detect and to overcome tumor hypoxia even before therapy starts, for the best of patient care.
By far, while there has been a great deal of interest in methodologies to measure hypoxia in patients, an efficient, non-invasive, while sensitive method to detect small regions of hypoxia that frequently occur in the tumors is still lacking [163]. A few metabolic markers (HIF-1α, HIF-2, CA9 and GLUT1) have been used to assess low oxygen tensions by immunohistochemistry [171,172]; however, the application of them in clinic is limited given that their expressions can be triggered by factors other than hypoxia and that biopsies only represent a small sampling of the tumor. As exosome composition mirrors the hypoxia status of tumors [158], a hypoxia signature may be formulated based on the exosomal hypoxia-responsive transcripts including HALs to evaluate oxygenation in the body for clinical exploitation, once our knowledge is advanced.
Therapeutic potential of HALs (targeting hypoxia in cancer therapy, a lncRNA perspective) Several approaches have been proposed to target hypoxia in tumor [161,163]. These include drugs that induce cell death selectively in hypoxic cells, e.g. hypoxia-activated prodrugs, or drugs sensitizing hypoxic cells to radiation. Since the adaptive response to hypoxia mainly orients from the transactivation of HIF signaling, some approaches seek to block hypoxia-induced responses by targeting HIFs and the related signaling, or to target pathways that also play pivotal roles in hypoxia adaptation, such as signaling involving mTOR, DNA damage response, and the unfolded protein response. In that regard, HALs that are elevated upon hypoxia and contribute to tumor progression in pre-clinical studies could potentially serve as molecular targets, e.g. PVT, LncHIFCAR, etc. (see Table 1) [41]. By contrast, HALs that are repressed in order to magnify hypoxia response, such as lncRNA-LET, could be induced for therapeutic intervention.
Various strategies have been developed to modulate RNAs. Silencing lncRNAs by small interfering RNAs, antisense oligonucleotides (ASOs), or ribozymes and deoxynucleotides are well demonstrated in pre-clinical studies. Until now, three ASOs and one aptamer therapies have been approved by the FDA for diseases and a handful of others are in clinical trials. The development of short oligonucleotides that fold into three-dimensional structures, aptamers, offers a greater specificity as they target specific structure regions to either mediate RNA degradation or disrupt functional interactions between binding partners [173]. Small molecules that bind to lncRNA and hinder its interaction surface have similar advantages. Additionally, peptide nucleic acids (PNA)-based approach against lncRNAs have been described. HOTAIR-targeting PNAs conjugated with pH-low insertion peptide (pHLIP) successfully delivered the anti-lncRNA to the acidic tumor. It blocked the interaction between HOTAIR and EZH2, subsequently inhibited HOTAIR-EZH2 activity and resensitized resistant ovarian tumors to platinum [174].
In any case, an issue all hypoxia-based therapeutics need to consider is the poorly perfused tissue in tumors. In response to their rapid growth, tumor cells secret pro-angiogenic factors such as VEGF to induce vascular formation, yet the constant stimulation leaves tumor vasculature ill-formed and leaky [175]. Simultaneous blockade of HIFs and pro-angiogenic factors has been proposed for targeting tumor hypoxia, in that targeting the angiogenic factors may allow vasculature to mature, resulting more effective blood supply and drug delivery. Another strategy is to relieve oxygen demand by drugs that alleviates oxygen tension in tumors. Papaverine, an FDA-approved drug as a smooth muscle relaxant, was found to inhibit mitochondria complex I and enhance the response to radiotherapy, while well-oxygenated normal tissues were not sensitized [176]. Accordingly, lncRNAs that regulate mitochondria respiration may be considered for targeting tumor hypoxia as an adjuvant treatment.

Conclusion and future perspectives
Decades of intensive scientific research on hypoxia and HIF biology has greatly contributed to our understanding of oxygen homeostasis. Over the past few years, a substantial increase in our knowledge of the noncoding transcriptomes, while putting on an additional layer of complexity in hypoxia regulation and responses, has advanced our comprehension of hypoxic biology. This review has presented an update of our current insights regarding lncRNAs involved in hypoxia-associated processes, highlighting the diverse mechanisms and functions of hypoxia-associated lncRNAs (HALs). These novel action modes unveil the unanticipated predominance of HALs in the regulation of gene expression under hypoxic conditions and outline the elaborate network among the different types of RNA transcripts, chromatin DNA and protein factors. However, advancement in analytical methodologies and in structural and genomic technologies of RNA are required to open up new important directions for in-depth investigation. For the state-of-the-art methodologies to unveil the functions of lncRNAs, readers are directed to two excellent recent reviews [177,178], as well as those in this special issue.
The role of HIF in hypoxia responses has been the central topic of most investigations. Indeed, HIF has been shown to be a central regulator of the coding and non-coding transcriptome and tightly associated with cancer risk [12,[179][180][181]. Most HALs, in particular, are highly responsive to hypoxia and HIF and, in turn, participate in the regulation of the protein-coding genome either in cis or in trans to offer multiple routes to HIFmediated gene regulation, implementing both positive and negative feedback loops that either strengthen or repress the hypoxia response. Most notably, the extracellular vesicles-containing HALs could evoke peculiar response to specific cell population, affecting nearby cells and those at a great distance, diversifying the hypoxia response far beyond the previously recognized. The cellular adaptation to hypoxia requires the precisely coordinated regulatory network to cope with the acute, transient and dynamic oxygen deprivation stress in local regions, whereas lncRNAs, with their flexible structure for interaction and quick biogenesis nature, could be uniquely suited to provide rapid, precise and reversible responses to this insult. It is clear that HALs and their downstream targets are shown to confer a series of biological effective responses to hypoxia. Feasibly, this extensive molecular crosstalk between lncRNA and hypoxic signaling cascades may undergo co-evolution to maintain such an exquisite, orchestrated program. Thus, for a comprehensive understanding of hypoxiaassociated tumor biology, it is of relevance to characterize the long non-coding transcriptome involved in hypoxia adaptation.
Given the prominent pathological roles of HALs in hypoxia-associated cancer progression, these RNAs could be exploited as useful indicators to define the cancer intrinsic subtypes to aid in precision medicine. Importantly, HALs are often tissue specific and respond to hypoxia in a cell context dependent manner. As such, they are excellent markers for tissue and tumor hypoxia responses. Compared with other bio-molecules, lncRNAs are ideal biomarkers that provide specificity, stability, sensitivity and easy accessibility [38]. Most notably, cell-free lncRNAs or those packed in extracellular vesicles can be detectable in various body fluids [182]. Hence, the genome-wide annotation of tissue-specific HAL signatures could guide development of promising, non-invasive biomarkers for early diagnosis, prognosis and prediction. Although most lncRNA-targeted treatments are still in their infancy stages, the recent success in RNA-based therapeutics holds promises for future technical innovations. With in-depth characterization of the interplay among hypoxia microenvironment and lncRNA function, more HALs could surely accelerate the design of therapeutics for tumor patients, enabling the targeting of the previously undruggable transcriptome in the near future.