Stressing the Regulatory Role of Long Non-Coding RNA in the Cellular Stress Response during Cancer Progression and Therapy

Cellular stress response is an important adaptive mechanism for regulating cell fate decision when cells confront with stress. During tumorigenesis, tumor progression and the course of treatment, cellular stress signaling can activate subsequent response to deal with stress. Therefore, cellular stress response has impacts on the fate of tumor cells and tumor responsiveness relative to therapeutic agents. In recent years, attention has been drawn to long non-coding RNAs (lncRNAs), a novel class of RNA molecules with more than 200 nucleotides in length, which has little protein-coding potential and possesses various functions in multiple biological processes. Accumulating evidence has shown that lncRNAs are also engaged in the regulation of cellular stress response, particularly in cancers. Here, we summarize lncRNAs that have been reported in the adaptive response to major types of cellular stress including genotoxic, hypoxic, oxidative, metabolic and endoplasmic reticulum stress, all of which are often encountered by cancer cells. Specifically, the molecular mechanisms of how lncRNAs regulate cellular stress response during tumor progression or the development of therapy resistance are emphasized. The potential clinical applications of stress-responsive lncRNAs as biomarkers will also be discussed.


Introduction
During tumorigenesis and tumor progression, cancer cells require a sufficient supply of nutrients, oxygen and growth factor stimulation to support their rapid growth. Nevertheless, the uncontrolled proliferation, confined living space and poor vascularization of solid tumor often restrict the continuous supply of nutrients and oxygen in the tumor microenvironment, generating different types of cellular stress including hypoxic stress, oxidative stress, endoplasmic reticulum (ER) stress and metabolic stress [1,2]. Moreover, dysfunctional DNA repair pathways in cancer cells lead to the accumulation of DNA damages and the occurrence of genotoxic stress [3]. Cancer cells have evolved a variety of cellular mechanisms to confront different types of stress. Currently, plenty of therapeutic agents have been developed for treating cancer, and some of them are stress-inducing agents that trigger cancer cell death by provoking stress [4,5]. When encountering these stimulations, type-specific stress-responsive pathways are activated to overcome stresses and may further confer therapeutic resistance in cancer [6,7]. Various molecules including long non-coding RNAs (lncRNAs) participate in these stress-response pathways.
LncRNAs is a group of RNA transcripts longer than 200 nucleotides with little proteincoding capacity. They compose of a large proportion of human transcriptome and have been demonstrated to regulate many biological processes and the development of human

Genotoxic Stress
Genomic instability in cancer cells is associated with the accumulation of endogenous DNA damages or the exposure to exogenous genotoxic agents and may further promote cancer progression. A comprehensive process denoted DNA damage response (DDR) is activated to resolve the genotoxic stress. DDR is composed of DNA lesion sensing, transducing damaging signals and transcriptionally activating cellular processes including cell cycle arrest, DNA repair, senescence and apoptosis. Several types of DNA damage have been reported, and DNA double-strand breaks (DSBs) are the most lethal for cancer cells. To overcome these damages, non-homologous end joining (NHEJ) and homologous recombination (HR) are the most dominant mechanisms that repair the lesions on the helix backbone. The repair process is exerted by different components involved in DDR machinery, including the sensors (MRN complex and 9-1-1 complex), the kinases (ATM, ATR and CHK1/2) and the DNA repair mediators (γH2AX, 53BP1, RAD51 and BRCA1), which are all indispensable in maintaining genomic integrity [7,22,23]. In addition, the leading transcriptional regulator in DDR response, p53, participates in multiple cellular processes upon DNA damage. p53 not only suppresses cell growth but also enhances apoptosis and transcriptionally activates the components involved in DNA repair processes. [24,25].
LncRNAs have been reported to play a critical role in gating genomic stability upon DNA damage. After being transcriptionally regulated by specific molecules, lncRNAs link different biological processes or collaborate with other regulators in DDR [26][27][28] ( Figure 1). Several lncRNAs are directly induced by the binding of p53 at their promoter regions and promote apoptosis or gene suppression in a p53-dependent manner [29][30][31]. The p53-induced lncRNAs activate apoptosis by binding to PARP1 or downregulating cell proliferation through the inhibition of MYC [32,33]. In addition to p53, lncRNAs are regulated by other DDR-mediated transcription factors. ERIC is transcriptionally activated by E2F1 following DNA damage and further inhibits the DNA damage-induced apoptosis in a p53-independent manner [34]. Moreover, NF-κB directly induces HOTAIR under the genotoxic stress, and elevated HOTAIR promotes DDR, cellular senescence, IL-6 secretion and sustains NF-κB activation [35].
LncRNAs impact the distinct DSB repair process upon DNA damage [26,27]. DlinRNAs are transcribed in the bidirectional forms after the MRN complex-mediated recruitment of RNA Pol II, and they further act as the precursors of small non-coding RNA DDRNAs to facilitate the recruitment of DNA repair proteins such as 53BP1 to the DSB sites [36]. LINP1 has been reported to strengthen the binding efficiency of Ku70/80 complex and DNA-PKcs to the break sites to promote NHEJ after being activated by AP1 upon the formation of DSBs [37]. Furthermore, the HR repair pathway can also be regulated by lncRNAs. HITT, which is transcriptionally activated by early growth response 1 (EGR1), inhibits the recruitment of ATM to the damage sites and further deregulates the HR repair signaling [38].
The development of anti-cancer treatments with genotoxic agents such as ionizing radiation and chemotherapeutics is based on the inability of cancer cells to efficiently repair DNA damages. However, cancer cells with intrinsic or acquired resistance to DNA damage-induced cytotoxicity are less sensitive to genotoxic stress, rendering poor clinical outcomes in patients undergone genotoxic therapies [27,39]. Evidence has piled up to show that lncRNAs play a role in the resistance to genotoxic therapies. BORG is highly responsive to multiple chemotherapeutic agents such as doxorubicin or 5-fluorouracil (5-FU) and leads to chemoresistance by sustaining the activation of NF-κB [40]. MALAT1 is co-activated by NF-κB and p53 following the treatment of temozolomide (TMZ) in glioblastoma (GBM) and promotes resistance to TMZ [41]. SBF2-AS1 is activated by zinc finger E-box binding homeobox 1 (ZEB1) and regulates the XRCC4-mediated DSB repair machinery by sponging miR-151a-3p. This protective effect is amplified by exosomal SBF-AS1 secretion to influence the neighboring GBM cells [42]. Moreover, lncRNAs can control cell fate depending on the degree of stress. Nrf2-lncRNA is induced by p53 upon weak DNA damage and promotes the expression of itself and Plk2/p21 cip1 by competing with miR-128 and miR-224, further leading to the upregulation of nuclear factor erythroid 2-related factor 2 (NRF2) and cell growth. In contrast, stronger DNA damage dampens the p53-mediated transcriptional enhancement of Nrf2-lncRNA [43].

Hypoxic Stress
Hypoxia is defined as a reduction in oxygen supply from vessels and accompanies several pathological states including cancer. As a hallmark of tumor microenvironment, hypoxia modulates multiple aspects of tumor progression, such as promoting survival signaling response, metastasis, anti-apoptosis, central metabolic reprogramming and further impacts therapeutic outcomes. Under hypoxic stress, various transcription factors are upregulated to activate downstream cellular signals. Hypoxia-inducible-factor-1 (HIF-1) complex is one of the major mediators for regulating oxygen-dependent transcriptional responses. The activation of HIF-1αregulates multiple biological pathways, including glycolysis, metastasis and cell immortalization [44][45][46][47].
In addition to the largely revealed transcriptomes of protein-coding genes in response to hypoxic tumor microenvironments, accumulating evidence also suggests the involvement of lncRNAs under hypoxia [48,49] (Figure 2). For example, HIF-1α induces specific lncRNAs such as PCGEM1 or BX111 by directly binding or recruiting Pol II to the hypoxiaresponsive element (HRE) at the promoter of target lncRNAs. The upregulated lncRNAs further enhance molecules involved in the EMT to promote migration and invasion of cancer cells [50,51]. HIF-1α has been demonstrated as a key mediator of aerobic glycolysis under hypoxia [52] and studies have shown that lncRNAs also participate in this process.
Upon the induction by HIF-1α, lncRNAs can translocate to the cytoplasm for diverse functions. AC020978 directly interacts with and stabilizes pyruvate kinase isozymes M2 (PKM2) to promote PKM2 nuclear translocation and further facilitates PKM2-enhanced HIF-1α transcriptional activity [53]. LincRNA-p21 or HIF1A-AS1 can provide positive feedback on the expression of HIF-1α by preventing VHL-mediated ubiquitylation or by phosphorylating the RNA-binding protein to promote the translation of HIF-1α, respectively [54,55]. These reciprocal feedback loops between HIF-1α and the target lncRNAs further upregulate both lncRNAs and the glycolytic genes such as GLUT1, LDHA and PKM2 to promote glycolysis. Moreover, lncRNAs have been reported to function as ceRNAs to regulate target genes by modulating the expression or the activity of miRNAs [15,56,57]. In hypoxic hepatocellular carcinoma (HCC), NEAT1 sponges miR-199a-3p to promote cell proliferation by increasing the expression of UCK2 [58]. In addition, after the transcriptional activation by HIF-1α in breast cancer (BC), BCRT acts as a miRNA sponge to inhibit miR-1303. The reduced level of miR-1303 prevents PTBP3 from degradation and further promotes cell proliferation and metastasis [59]. positive feedback on the expression of HIF-1α by preventing VHL-mediated ubiquitylation or by phosphorylating the RNA-binding protein to promote the translation of HIF-1α, respectively [54,55]. These reciprocal feedback loops between HIF-1α and the target lncRNAs further upregulate both lncRNAs and the glycolytic genes such as GLUT1, LDHA and PKM2 to promote glycolysis. Moreover, lncRNAs have been reported to function as ceRNAs to regulate target genes by modulating the expression or the activity of miRNAs [15,56,57]. In hypoxic hepatocellular carcinoma (HCC), NEAT1 sponges miR-199a-3p to promote cell proliferation by increasing the expression of UCK2 [58]. In addition, after the transcriptional activation by HIF-1α in breast cancer (BC), BCRT acts as a miRNA sponge to inhibit miR-1303. The reduced level of miR-1303 prevents PTBP3 from degradation and further promotes cell proliferation and metastasis [59].  It has been reported that the expression of specific lncRNAs can be modulated by DNA methylation at their promoter regions [60,61]; furthermore, the hypoxic status can induce abnormal DNA methylation and acetylation, which indicates that HIF-1α not only directly binds to the promoter regions of target lncRNAs but also modulates lncRNAs with epigenetic modifications [48]. In renal cell carcinoma, lncHILAR expression is upregulated via H3K4me1 and H3K27ac enrichments at the promoter region of lncHILAR. In addition, the elevated lncHILAR acts as the ceRNA for miR-613/206/1-13p and further activates the Jagged1/Notch/CXCR4 signaling pathway to promote metastasis under hypoxia [62]. On the other hand, tumor-suppressive lncRNAs such as CF129 or LET are inhibited by HIF-1αmediated histone deacetylase (HDAC) recruitment. HIF-1α suppresses lncRNA expression by promoting HDAC1 or HDAC3 recruitment to promoter regions. The downregulated lncRNAs stabilize downstream targets by inhibiting proteins from proteasomal degradation and result in a positive feedback loop to enhance the HIF-1α response [63,64].
In addition to the major role of HIF-1α in regulating hypoxic stress signaling, other hypoxia-responsive transcription factors also play important roles in the hypoxic response [47]. In BC, NEAT1 is transcriptionally upregulated by HIF-2α under hypoxia and regulates the nuclear structure, further contributing to BC cell proliferation [65]. LHFPL-AS2, a tumor-suppressive lncRNA in lung cancer, is downregulated under hypoxia due to the reduction in EGR1 expression, resulting in the transcriptional repression of thioredoxininteracting protein (TXNIP) and promoting tumor metastasis [66].
In addition to the NRF2 pathway, lncRNAs can also regulate oxidative stress response through multiple signaling pathways and affect different biological processes, such as apoptosis, autophagy or inflammatory pathways. LncRNA H19 and HULC are found to activate inflammatory responses through miRNA sponging under oxidative stress. H19 and HULC are upregulated under H 2 O 2 and glucose oxidase treatment. H19 sponges let-7 to increase IL-6 expression and HULC sponges miR-372 to upregulate CXCR4, which promotes the migratory ability of cholangiocarcinoma cells [72]. LncRNA XIST is shown to be upregulated for promoting the migratory ability of osteosarcoma cells under H 2 O 2 treatment. Mechanistically, XIST sponges miR-153 to increase the expression of Snail and further promotes migration [73]. As 2 O 3 -induced oxidative stress upregulates the expression of lncRNA ROR, which reduces p53 expression to prevent apoptosis [74]. H 2 O 2 -induced oxidative stress downregulates lncRNA SCAMP1 to enhance the expression of miR-429. miR-429 then targets ZEB1 and c-Jun for degradation and promotes autophagy, which eventually contributes to the increase in apoptosis and the decrease in tumorigenesis [75]. Oxaliplatin is a chemotherapeutic drug that can arouse oxidative stress in cells [76]. In oxaliplatin-resistant HCC cells, the expression of lncRNA LINC01134 is increased by lysine-specific demethylase 1 (LSD1)-mediated demethylation. LINC01134 recruits SP-1 to the SQSTM1/p62 promoter for enhancing its expression, which improves mitochondrial function to reduce ROS production and finally results in the resistance to oxaliplatin [77]. LncRNA NORAD is also reported to contribute to oxaliplatin resistance. Oxaliplatin promotes CREBBP-mediated H3K27ac on NORAD promoters to upregulate NORAD expression. NORAD then sponges miR-433-3p to increase the expression of autophagy-related genes ATG5 and ATG12, which helps in alleviating DNA damage and oxidative stress and eventually confers oxaliplatin resistance in gastric cancer (GC) cells [78]. Some lncRNAs directly regulate the accumulation of endogenous oxidative stress. LncRNA THAP9-AS1 is highly expressed in osteosarcoma tissues and increases methylation at the promoter region of the suppressor of cytokine signaling 3 (SOCS3), which activates JAK2/STAT3 signaling to reduce oxidative stress [79]. Conversely, lncRNA GAS5 promotes the accumulation of oxidative stress and apoptosis to inhibit the growth of melanoma cells. Mechanistically, GAS5 binds E2F4 to prevent the binding of E2F4 on the EZH2 promoter region, causing a decrease in EZH2 expression and EZH2-mediated CDKN1C downregulation. This finally leads to the enhancement of oxidative stress and apoptotic cell death [80].
Ferroptosis, a new type of regulated cell death induced by oxidative stress, has been recently implicated to regulate tumor initiation or progression [81]. It is mainly initiated by the accumulation of iron-mediated ROS or lipid peroxides [82] and is shown to be regulated by lncRNAs in the following studies. Erastin, a ferroptosis inducer, upregulates lncRNA NEAT1 through p53 in HCC cells. NEAT1 sponges miR-362-3p to increase myo-inositol oxygenase (MIOX) expression and promotes ferroptosis, which is shown to reverse erastin resistance in HCC [83]. On the other hand, NEAT1 inhibits erastin-induced ferroptosis by not only increasing GPX4 and SLC7A11-mediated antioxidative pathways but also decreasing ACSL4-mediated lipid peroxidation, which may confer the resistance to erastin in non-small cell lung cancer (NSCLC) cells [84]. LncRNA GABPB1-AS1 is upregulated to bind its sense mRNA, GABPB1, under erastin treatment, and this results in the decrease in GABPB1 translation. GABPB1 downregulation inhibits peroxiredoxin-5 (PRDX5) expression to abrogate antioxidative capacities, which finally increases the accumulation of lipid ROS and promotes ferroptosis in HCC cells [85]. LncRNA LINC00336 is upregulated through ELAVL1-mediated RNA stabilization. LINC00336 sponges miR-6852 to enhance cystathionine-β-synthase (CBS) expression, which produces more antioxidant glutathione to alleviate oxidative stress and inhibit ferroptosis [86]. Cadmium (Cd) is found to suppress ferroptosis for promoting prostate tumor growth by upregulating lncRNA OIP5-AS1. OIP5-AS1 sponges miR-128 to increase the expression of miR-128-targeted SLC7A11, which is involved in the production of glutathione. Therefore, OIP5-AS1 is able to reduce Cdinduced ferroptosis through the miR-128/SLC7A11 axis [87].

Metabolic Stress
Cancer cells are featured of their high demand of nutrients since they have to proliferate and grow rapidly. Nonetheless, the limited vascularization in the core of a solid tumor often results in the deficiency of nutrients in the tumor microenvironment [88,89]. The inability of consuming enough nutrients due to the insufficient nutrient supply or the high demand of nutrients causes stress to the cancer cells, which is denoted as metabolic stress [90]. During tumor development, cancer cells often undergo metabolic reprogramming in order to support their rapid growth or survive from the metabolic stress [91]. Although a number of studies have already indicated the role of lncRNAs in regulating metabolic reprogramming [92] and cancer metabolism [93,94], here, we are focusing on how cancer cells respond or confront to metabolic stress through these nutrient deficiencyresponsive lncRNAs (Figure 4). found to suppress ferroptosis for promoting prostate tumor growth by upregulating lncRNA OIP5-AS1. OIP5-AS1 sponges miR-128 to increase the expression of miR-128-targeted SLC7A11, which is involved in the production of glutathione. Therefore, OIP5-AS1 is able to reduce Cd-induced ferroptosis through the miR-128/SLC7A11 axis [87].

Metabolic Stress
Cancer cells are featured of their high demand of nutrients since they have to proliferate and grow rapidly. Nonetheless, the limited vascularization in the core of a solid tumor often results in the deficiency of nutrients in the tumor microenvironment [88,89]. The inability of consuming enough nutrients due to the insufficient nutrient supply or the high demand of nutrients causes stress to the cancer cells, which is denoted as metabolic stress [90]. During tumor development, cancer cells often undergo metabolic reprogramming in order to support their rapid growth or survive from the metabolic stress [91]. Although a number of studies have already indicated the role of lncRNAs in regulating metabolic reprogramming [92] and cancer metabolism [93,94], here, we are focusing on how cancer cells respond or confront to metabolic stress through these nutrient deficiencyresponsive lncRNAs (Figure 4).  and LOC730101 to participate in the regulatory of AMPK signaling. Another group of lncRNAs, such as HAND2-AS1 and HOXC-AS3, was upregulated under glucose starvation to modulate glycolysis. On the other hand, glucose deprivation induces GAS5 and Linc01564 to regulate metabolic plasticity in response to stress stimulation. GIRGL, GLS-AS and UBA6-AS1 are shown to be glutamine deprivation-responsive lncRNAs, which regulates glutamine metabolism or maintains PARP1 function to determine cancer cell fate. ATF4 (activating transcription factor 4), PHGDH (phosphoglycerate dehydrogenase), MDH2 (malate dehydrogenase 2), FH (fumarate hydratase), CS (citrate synthase), LKB1 (liver kinase B1), AMPK (AMP-activated protein kinase), MACC1 (metastasis-associated in colon cancer protein 1), SP-1 (specificity protein 1), SIRT6 (sirtuin 6), HIF1α (hypoxia inducible factor 1 subunit alpha), PKM2 (pyruvate kinase M2), GRP78 (glucose-regulated protein), IDH1 (isocitrate dehydrogenase 1), FOXO (forkhead box O3), GLS (glutaminase), GCN2 (general control nonderepressible 2) and PARP1 (poly(ADP-ribose) polymerase 1). This figure was created with BioRender.com.
As glucose is the major source of nutrient for most cancer cells, a significant portion of studies pays attention to how lncRNAs participate in metabolic stress response under glucose deficiency. AMP-activated protein kinase (AMPK) signaling is an important pathway activated by low glucose availability or nutrient deficiency [95] to regulate the growth, differentiation or metabolism of cancer cells in response to metabolic stress [96], and there are several lncRNAs reported to be induced upon glucose deprivation to participate in the AMPK signaling pathway. LncRNA LOC730101 is upregulated by AMPK activation to promote cell survival and to inhibit apoptosis under glucose deprivation in osteosarcoma [97]. Liver kinase B1 (LKB1) is a kinase that activates AMPK through phosphorylation [98]. This LKB1/AMPK axis was found to induce lncRNA MITA1 by promoting the DNA methylation of MITA1 under glucose deficiency, and the upregulation of MITA1 can further increase SNAI2/Slug expression to promote the metastasis of HCC cells [99]. In addition to be induced by AMPK signaling, another two lncRNAs are reported to activate AMPK under metabolic stress. NBR2 is a lncRNA induced by LKB1-AMPK axis under glucose deprivation, and it can, in turn, bind to AMPK and promotes AMPK activity to form a feed-forward loop. The activation of AMPK signaling under this circumstance inhibits cell cycle progression and autophagy, resulting in reduced tumor growth [100]. Glucose starvation upregulates lncRNA MACC1-AS1 to activate the AMPK/Lin28 axis, which improves mRNA stability and the protein expression of MACC1. The upregulation of MACC1 mediated by MACC1-AS1 finally triggers MACC1-mediated metabolic plasticity, which enhances glycolysis and antioxidant capacity to promote tumor progression of GC [101].
Moreover, lncRNAs are upregulated under glucose starvation to regulate glycolysis through multiple mechanisms. The upregulation of lncRNA TP53TG1 by glucose deprivation increases the mRNA expression of GRP78 and IDH1 and inhibits PKM2 expression, which may contribute to a decrease in glycolysis. The alteration of these genes caused by TP53TG1 further promotes cell growth and migration in glioma cells [102]. LncRNA FILNC1 is upregulated by transcription factor FOXO under glucose deprivation for interacting with AU-binding factor 1 (AUF1), which prevents the binding of AUF1 on the MYC/c-Myc promoter and finally leads to a reduction in c-Myc expression. c-Myc is responsible for upregulating genes involved in glycolysis [103], so the downregulation of c-Myc caused by FILNC1 inhibits glycolysis and ultimately represses renal tumor growth [104]. In osteosarcoma, glucose deficiency enhances the expression of lncRNA HAND2-AS1, which binds to Fructose-1, 6-bisphosphatase 1 (FBP1) to sequester the interaction between FBP1 and HIF1α mRNA, thereby influencing glycolysis-related genes [105]. The loss of FBP1 binding mediated by HAND2-AS1 leads to a decrease in HIF1α translation, which attenuates glycolysis and further reduces tumor growth [106]. Finally, lncRNA HOXC-AS3 is induced by transcription factor SP1 under glucose starvation. HOXC-AS3 complexes with Sirtuin 6 (SIRT6), a histone deacetylase, to antagonize SIRT6 function, which results in the increase in HIF1α activity and the expression of glycolysis-related genes. This HOXC-AS3-mediated metabolic reprogramming eventually assists BC cells in adapting to metabolic stress [107]. In addition to glycolysis, lncRNAs also regulate other metabolic pathways when glucose is deficient. LncRNA GAS5 is upregulated to promote the binding of malate dehydrogenase (MDH2) and histone deacetylase SIRT3 under glucose deprivation, which further results in a decrease in MDH2 acetylation. The reduction in acetylation loosens the interaction between fumarase (FH)-MDH2-citrate synthase (CS) complex, which reduces TCA flux and finally limits breast tumor growth [108]. Activating transcription factor 4 (ATF4) mediates the upregulation of lncRNA Linc01564 under glucose starvation. Linc01564 works as a miRNA sponge to sponge PHGDH-targeted miR-107 and miR-103a-3p, which increases the expression of PHGDH and further enhances serine synthesis pathway. More glutathione is produced by the enhanced serine synthesis pathway to relieve oxidative stress, helping cells survive from glucose starvation-induced metabolic stress [109].
Apart from glucose, cancer cells require amino acids, especially glutamine, to support their rapid growth. As a result, amino acid deficiency is also a common phenomenon observed during tumor development [89]. LncRNA GLS-AS is downregulated by c-Myc under glutamine deprivation. Under normal conditions, GLS-AS inhibits glutaminase (GLS) expression through Dicer-dependent RNA silencing, so the decrease in GLS-AS increases GLS expression under glutamine deficiency to confront with metabolic stress. In addition to helping cells survive from stress, the upregulation of GLS reciprocally maintains c-Myc stability to inhibit GLS-AS expression, which reinforces GLS-mediated cell survival under stress in pancreatic cancer [110]. Glutamine deprivation activates transcription factor c-Jun to upregulate lncRNA GIRGL, which inhibits cell proliferation under metabolic stress in colon cancer cells. Mechanically, GIRGL promotes the interaction between CAPRIN1 and GLS1 mRNA, resulting in the inhibition of GLS1 translation and GLS1-mediated glutamine metabolism to inhibit cell growth [111]. Finally, an amino acid restriction-induced lncRNA UBA6-AS1 is reported to help cells adapt to metabolic stress in BC cells. UBA6-AS1 is upregulated through the GCN2/ATF4 axis to maintain PARP1 expression and function, which ultimately leads to a trade-off between migratory ability and cell survival under glutamine or arginine deprivation [112].

ER Stress
ER is an organelle responsible for calcium storage, lipid metabolism, protein synthesis and protein folding [113]. When cells encounter stress stimuli or ER dysfunction, a portion of proteins become misfolded or unfolded, and the aberrant accumulation of these misfolded or unfolded proteins in the ER leads to the occurrence of ER stress and further activates ER stress response, which is the so-called unfolded protein response (UPR) [114]. The activation of UPR attenuates protein synthesis or induces autophagy to recover cell homeostasis or triggers cell death in response to stress stimulation [115].
The UPR is initiated by three main ER transmembrane protein sensors: protein kinaselike ER kinase (PERK), inositol-requiring transmembrane kinase endoribonuclease-1α (IRE1α) and activating transcription factor 6 (ATF6). These sensors are bound by binding Ig protein (BiP), also known as GRP78, under homeostatic conditions to suppress their activation. As the misfolded or unfolded peptides accumulate in the ER lumen, BiP will in turn bind to the unfolded peptides and further release UPR sensors for activation to initiate the stress response cascades [116]. PERK is a kinase that phosphorylates eukaryotic translation initiation factor 2 subunit-α (eIF2α) at serine 51, which results in the attenuation of global protein synthesis but selectively increases the translation of an important stressresponsive transcription factor, ATF4. ATF4 then plays a dual role in promoting either cell survival or cell death depending on the types or duration of stress [117]. IRE1α is a ribonuclease activated by the unfolded protein aggregates to splice X-box binding protein 1 (XBP1) mRNA to generate the functional transcription factor, spliced XBP1 (XBP1s). XBP1s can promote the transcription of genes that participate in protein folding and ER-associated degradation (ERAD) to eliminate the unfolded proteins and further relieve ER stress [118]. ATF6 is a transmembrane transcription factor that is retained on ER membranes by BiP under normal condition. Once released by BiP, ATF6 is transported to Golgi apparatus and cleaved by peptidases to generate an active transcription factor, cleaved ATF6α [119]. Cleaved ATF6α enters into the nucleus to promote transcription of XBP1 as well as genes involved in protein folding and ERAD, which helps resolve ER stress [120].
In multiple cancers, lncRNAs have been demonstrated to participate the UPR cascades in different aspects ( Figure 5). Some lncRNAs are involved in the initiation of UPR by regulating the expression or the activation of UPR sensors. The overexpression of lncRNA MEG3 [121,122] and GAS5 [123] increases the expression of GRP78, IRE1, PERK, and ATF6, which may further trigger apoptosis to suppress tumor growth. LUCRC, highly expressed in colorectal tumor tissue to promote proliferation, migration and invasion of tumor cells, is required for the expression of BiP and the cleavage of XBP1 and ATF6 to evoke UPR [124]. CASC2 increases PERK mRNA stability to trigger the PERK/pS51-eIF2α/CHOP (C/EBP homologous protein) axis, which induces apoptosis in irradiated NSCLC cells [125]. MIR503HG, a miRNA sponge, increases TUSC3 expression by sponging miR-224-5p. The increase in TUSC3 further inhibits ATF6-mediated UPR and decreases cancer cell viability, proliferation and epithelial-mesenchymal transition (EMT) [126]. LINC00963 can sponge miR-320a, which directly targets XBP1 3 -UTR to abrogate XBP1 expression. As a result, when LINC00963 sponges miR-320a, XBP1-mediated UPR enhances apoptosis in GC [127]. On the other hand, ER stress can act as a stimulator of some lncRNAs. GAS5 is a lncRNA that is suppressed by ZBTB7A under normal conditions in osteosarcoma cells. When the UPR is activated by ER stress inducers, tunicamycin (Tm) and thapsigargin (Tg), ZBTB7A is downregulated and results in the increase in GAS5 to promote apoptosis [128]. MALAT1, one of the most well-characterized lncRNAs, is found to be induced by Tg and may be regulated by XBP1 and ATF4 to increase cell migratory abilities under ER stress in colorectal cancer cells [129]. GOLGA2P10 is found to be increased in HCC and is regarded as an oncogenic lncRNA. Under Tm or Tg treatment, GOLGA2P10 is upregulated by the PERK/ATF4/CHOP axis to increase BCL-xL expression and BAD phosphorylation, which decreases ER stress-induced apoptosis to assist cells in confronting ER stress [130]. LncRNA HITTERS, which is also upregulated by Tm or Tg, functions as an RNA scaffold to promote the interaction of the MRN complex. As ER stress can trigger DNA damage and may further induce apoptosis, the increase in MRN complex interactions can maintain DNA repair efficacy to prevent ER stress-induced apoptosis, which suggests that HITTERS is a prosurvival molecule in UPR [131]. In addition to participating in the regulation of UPR, some lncRNAs have been shown to either sensitize or counteract the efficacy of ER stress-inducing chemotherapeutic or targeted therapeutic agents. A potent anti-cancer agent resveratrol initiates UPR and enhances the expression of lncRNA H19. The combination of resveratrol and knockdown of H19 increases the expression of ER stress-related genes, promotes cell death and decreases EMT-related genes, suggesting the role of H19 in resveratrol-resistant GC cells [132]. Sorafenib, as a first line protein kinase inhibitor for treating advanced HCC [133], can induce UPR signaling to trigger apoptotic cell death with the assistance of lincRNA-p21 [134]. This study indicates that lincRNA-p21 is a tumor suppressive lncRNA in HCC and contributes to the activation of UPR signaling and apoptosis under sorafenib treatment. Another lncRNA ZFAS1 is shown to be upregulated by PERK/ATF4 axis after sorafenib treatment in sorafenib-resistant HCC cells, which suggests that ZFAS1 may be a therapeutic or prognostic biomarker of sorafenib treatment in HCC patients [135]. 5-FU is one of the most commonly used anti-cancer drugs which is mis-incorporated into RNA and DNA and further leads to cell death [136]. In 5-FU resistant cells, lncRNA MIAT is found to be upregulated. Mechanically, 5-FU evokes UPR and upregulates MIAT through the GRP78/octamer-binding transcription factor 4 (OCT4) pathway to confer 5-FU resistance in BC cells [137].

Stress-Responsive lncRNAs as Emerging Biomarkers for Assessment of Prognosis and Prediction of Therapy Response in Cancer
During cancer progression and therapy, cellular stress responses have been designated to regulate cell fate decision and susceptibility to therapeutic agents. It is an emerging fact that lncRNAs play indispensable roles in these processes. LncRNAs participate in not only the initiation or activation of stress sensors but also the regulation of biological pathways in response to various stress stimulations. In this review, we summarized the regulatory role of lncRNAs and emphasized the molecular mechanisms of how lncRNAs participate in the five major stress conditions including genotoxic, hypoxic, oxidative, metabolic and ER stress. The involvements of lncRNAs in the stress response in cancer and their associations with tumor progression and the effectiveness of cancer therapeutics are highlighted (Table 1). We then propose that these stress-responsive lncRNAs are ideal for being developed into cancer biomarkers for assessing disease prognosis and predicting therapy response. While a significant amount of studies has reported lncRNAs that are either upregulated or downregulated in various types of cancer compared to normal tissues [21,138], the identification of stress-responsive lncRNAs in cancer reveals a new layer of regulation and paves the way for biomarker development. The expression of lincRNA-p21 is deregulated in many cancers including skin, prostate, colorectal cancer and chronic lymphocytic leukemia [139], and it has been proposed to serve as a marker for diagnosing cancer. Moreover, lincRNA-p21 has also been found to play diverse roles in genotoxic, hypoxic and ER stress responses during cancer progression [29,54,134], indicating that when combining with specific stress-responsive factors, lincRNA-p21 has the potential to serve as a biomarker for precisely predicting disease prognosis or therapy response. NEAT1 participates in cancer progression and therapeutic response under various conditions, and it seems to be an ideal prognostic marker for assessing hypoxic tumors due to its tumor-promoting role under hypoxia [58,65]. Moreover, NEAT1 may also become a promising biomarker when assessed in combination with other markers. Based on the studies demonstrating the roles of NEAT1 in regulating cancer response to the ferroptosis inducer, erastin [83,84], it is plausible that by the co-evaluation of NEAT1 and miR-362-3p expression, one can predict the tumor response to erastin.
Many cancer therapeutic drugs, such as oxaliplatin and As 2 O 3 and erastin, 5-FU and TMZ abrogate tumor growth by inducing stress, suggesting that potential therapeutic targets can be developed for reversing the lncRNA-mediated resistance of these drugs. In addition, lncRNAs associated with the resistance to these cancer therapeutic drugs, for example, BORG, MALAT1, SBF2-AS1, HIF1A-AS1, LINC01134, NEAT1, LAMTOR5-AS1, NORAD, lincRNA-p21, ZFAS1 and MIAT, may also serve as promising biomarkers for the diagnosis of therapy-resistant cancers or the prediction of therapy outcome. SBF2-AS1 and MALAT1 are associated with TMZ-resistant GBM [41,42]. Specifically, SBF2-AS1 is highly present in the exosomes secreted by TMZ-resistant GBM cells, suggesting that the detection of serum SBF2-AS1 holds hope for diagnosing therapy-resistant GBM [42]. Hypoxic renal cancer cells secrete exosomes containing lncHILAR to promote cell invasion and migration [62]. Therefore, serum lncHILAR can be a biomarker to evaluate disease pro-gression and patient outcome. In chemo-resistant gastrointestinal cancers, LINC01134 [77], NORAD [78] and HIF1A-AS1 [55] are all elevated and could be used for predicting responsiveness to chemotherapies. BORG [40] and MIAT [137] are both associated with the chemoresistance of BC and may reflect the clinical outcomes of patients treated with a commonly used chemotherapy drug, 5-FU. Finally, cancer metabolism has gradually been regarded as potential therapeutic targets [140] or the causes of therapeutic resistance [141] in cancer treatments. Since several lncRNAs are found to regulate cell survival under metabolic stress, potential therapeutic strategies or prognostic markers can be developed based on these studies. AMPK is a promising target for cancer treatment, and therapeutic strategies have been developed for activating or inhibiting AMPK [142]. There are several lncRNAs including NBR2, MITA1, LOC730101 and MACC1-AS1 associated with AMPK activity and may serve to assess the response of AMPK modulation in cancer. The inhibition of GLS has also been proposed for suppressing tumor growth and metastasis [140]. GIRGL and GLS-AS are reported to inhibit GLS expression to impede tumor progression [110,111], suggesting that these lncRNAs have the potential to become therapeutic targets. UBA6-AS1 participates in the amino acid deficiency-induced integrated stress response and regulates the activity of PARP1 for BC cell survival. The level of UBA6-AS1 correlates with the cellular response to PARP inhibitors [112], implying that UBA6-AS1 may be considered as a biomarker for predicting breast tumor response to PARP inhibitors.
Taken together, increasing number of studies has uncovered and recognized the importance of lncRNAs in the regulation of cellular stress response during cancer progression and therapy. It would be worthwhile to further investigate and explore the possibility of developing these stress-responsive lncRNAs into biomarkers that can be applied clinically for assisting disease stratification in the future.