Role of the lncRNA–p53 regulatory network in cancer

Abstract

Advances in functional genomics have led to discovery of a large group of previous uncharacterized long non-coding RNAs (lncRNAs). Emerging evidence indicates that lncRNAs may serve as master gene regulators through various mechanisms. Dysregulation of lncRNAs is often associated with a variety of human diseases including cancer. Of significant interest, recent studies suggest that lncRNAs participate in the p53 tumor suppressor regulatory network. In this review, we discuss how lncRNAs serve as p53 regulators or p53 effectors. Further characterization of these p53-associated lncRNAs in cancer will provide a better understanding of lncRNA-mediated gene regulation in the p53 pathway. As a result, lncRNAs may prove to be valuable biomarkers for cancer diagnosis or potential targets for cancer therapy.

Introduction

The human genome is pervasively transcribed. Of interest, protein-coding genes account for only <2% of the human genome, whereas the vast majority of transcripts are non-coding RNAs. Among them are long non-coding RNAs (lncRNAs) (Birney et al., 2007; Kung et al., 2013). In contrast to well-studied small non-coding RNAs microRNAs, lncRNAs are poorly characterized. However, increasing evidence has linked mutations and dysregulations of lncRNAs to diverse human diseases, suggesting that they are functional and of clinical relevance. In support of this notion, analysis of 1600 lncRNA intervals that are predicted from chromatin-state maps from mouse genome reveals relatively high conservation (Marques and Ponting, 2009).

Although lncRNAs are arbitrally defined as a group of RNAs with a molecular weight of >200 bases in length, they may vary from several hundred bases to tens of kilobases (kb); they may be located in isolation from protein-coding genes, or interspersed nearby or within protein-coding genes. Increasing evidence indicates that lncRNAs can function as master gene regulators, regulate a large number of genes, and impact a variety of cellular pathways. In this review, we discuss the role of lncRNAs in the p53 regulatory network in cancer, focusing on how lncRNAs regulate p53 and how lncRNAs are regulated by p53.

LncRNAs as master gene regulators

As a matter of fact, lncRNAs were discovered and reported a long time ago. For example, the growth arrest specific 5 (GAS5) was originally isolated from mouse NIH 3T3 cells using subtraction hybridization over two decades ago (Schneider et al., 1988). However, the functional studies of GAS5 did not start until 20 years later (Mourtada-Maarabouni et al., 2009; Kino et al., 2010). This is mostly likely due to the lack of sufficient knowledge or technology. As the functional genomics advances, the pace of lncRNA studies has been picked up. It becomes evident now that lncRNAs are an important part of our genome and at least some of them could play a critical role in regulation of gene expression and cellular functions. Although lncRNAs can be transcribed through RNA polymerase III, most of them, like protein-coding genes, are transcribed by RNA polymerase II; they are spliced products via canonical genomic splice site motifs, frequently ended with a poly A tail. Moreover, they are often regulated by well-established transcription factors and expressed in a tissue-specific manner (Prensner and Chinnaiyan, 2011). A variety of previously identified RNA species belong to this group, including antisense RNAs, transcribed ultraconserved regions (T-UCR), and pseudogenes. For example, antisense RNAs are transcribed on the opposite strand from a protein-coding gene and frequently overlap with the corresponding gene (He et al., 2008); pseudogenes such as PTENP1 can function as a microRNA decoy (Swami, 2010). Overexpression of the PTENP1 leads to an increased level of PTEN and causes cellular growth inhibition, an effect dependent on the presence of mature microRNAs (Poliseno et al., 2010). The most important group of lncRNAs are probably long intergenic non-coding RNAs (lincRNAs). They are distributed as numerous ‘foci’ of transcription that are separated by long stretches of intergenic space (Carninci et al., 2005). Another interesting group of recently identified lncRNAs are enhancer RNAs (eRNAs) that are capable of positively regulating transcription in cis or in trans (Natoli and Andrau, 2012); most of them, if not all, are nonpolyadenylated RNAs with sizes up to ∼2 kb. The recent updated NONCODE database lists over 70000 lncRNAs and over 30000 of them are human lncRNAs (Bu et al., 2012). Evidently, this number is even larger than that for protein-coding genes and is expected to keep growing, thus providing further evidence of the complexity of our genome.

Despite relatively newly identified or characterized, lncRNAs have been shown to play a critical role in regulation of a variety of cellular functions and disease processes including stem cells and cancer metastasis (Khalil et al., 2009; Gupta et al., 2010; Tsai et al., 2010; Guttman et al., 2011; Hung et al., 2011; Prensner et al., 2011). This may have to do with their ability to interact with DNA, RNA, or protein to regulate gene expression. In this regard, it has been proposed that lncRNAs may serve as (i) signals for transcription; (ii) decoys to titrate transcription factors; (iii) guides so that chromatin-modifying enzymes can be recruited to target genes; and (iv) scaffolds to bring together multiple proteins to form ribonucleoprotein complexes (Mercer et al., 2009; Wang and Chang, 2011; Ulitsky and Bartel, 2013). Evidently, the mechanism of lncRNA-mediated gene regulation may be much more sophisticated than that of the microRNA-mediated gene regulation.

One of the most important functions of lncRNAs is probably the epigenetic regulation of gene expression involving chromatin remodeling and histone modification (Lee, 2012). In this regard, several lncRNAs have been shown to interact with the chromatin remodeling complex so that they can regulate a large number of genes. In particular, lncRNAs can regulate chromatin by changing three-dimensional genome architecture. For example, X-inactive specific transcript (Xist) is one of the first identified and best studied lncRNAs (Brown et al., 1991, 1992). Xist binds broadly across the X-chromosome, causing X-chromosome silencing (Engreitz et al., 2013) through interaction with the polycomb repressive complex 2 (PRC2) (Zhao et al., 2008). Similarly, HOTAIR plays a critical role in cancer through epigenetic regulation mechanisms. HOTAIR is a 2.2 kb gene in the HOXC locus, which, however, can repress transcription in trans of HOXD genes. This repressive action is mediated by the interaction of HOTAIR with PRC2 (Takagi et al., 2005). In another case, PCAT-1, as a prostate-specific regulator of cell proliferation, is a target of PRC2, functioning as a transcriptional repressor in a subset of prostate cancer patients (Prensner et al., 2011). Of interest, HOTAIR can serve as scaffolds by providing binding surfaces to assemble selected histone modification enzymes, thereby specifying the pattern of histone modifications on target genes (Tsai et al., 2010). These findings suggest that lncRNAs manipulate the epigenetic machinery to remold the epigenetic landscape, leading to cancer.

Gene regulation by lncRNAs can also occur at the posttranscriptional levels. LncRNAs may function as an endogenous ‘sponge’ and downregulate a series of microRNAs. For example, linc-MD1 (Cesana et al., 2011) is able to interact with miR-133 and miR-135 to regulate the expression of MAML1 and MEF2C, transcription factors that activate muscle-specific gene expression. Similarly, HULC can downregulate miR-372 through its interaction with miR-372 (Wang et al., 2010). Since HULC is highly upregulated in liver cancer (Panzitt et al., 2007) and plays an important role in tumorigenesis, this may explain why miR-372 is often downregulated in tumor specimens. In consistent with these findings, we have recently demonstrated that there is a reciprocal repression between miR-211 and loc285194 (Liu et al., 2013). We have also shown that GAS5 is a direct target of miR-21; in particular, GAS5 may function as an endogenous sponge for miR-21 because manipulation of GAS5 levels leads to a negative alteration of miR-21 (Zhang et al., 2013b). Therefore, lncRNAs also join the ‘competitive endogenous RNA (CeRNA)’ regulatory system (Salmena et al., 2011) where microRNA response elements (MREs) may serve as letters of a new language through which microRNAs may regulate not only protein-coding genes, but also non-coding genes such as lncRNAs. LncRNAs can also regulate pre-mRNA process and alternative splicing. For example, metastasis-associated lung adenocarcinoma transcript 1 (MALAT1) is able to interact with splicing factors SR proteins and influences their distribution in nuclear speckle domains (Tripathi et al., 2010). Depletion of MALAT1 or ectopic expression of splicing factor SRSF1 causes a similar change pattern of endogenous pre-mRNAs, which may be related to its ability to regulate cellular levels of phosphorylated forms of SR proteins. In addition, lncRNAs are also involved in posttranscriptional regulation of gene expression. For example, lncRNAs may form complementary base pairing with the target mRNA. The formation of RNA duplexes between complementary non-coding RNA and mRNA may mask key elements within the mRNA required to bind trans-acting factors, potentially affecting any step in posttranscriptional gene expression including pre-mRNA processing, splicing (Beltran et al., 2008), and translation (Carrieri et al., 2012).

Therefore, there are various mechanisms for lncRNAs to regulate gene expression. LncRNAs can regulate any step of gene expression, including transcription, mRNA stability, translation, alternative splicing, and protein–protein interaction. Accordingly, lncRNAs may play a bigger role in regulation of gene expression and cellular functions.

Role of lncRNAs in cancer

Like protein-coding genes or microRNAs, lncRNAs can function as oncogenic and tumor-suppressor genes, thus impacting one or more of the cancer hallmarks (Hanahan and Weinberg, 2011). Alterations in the primary structure, secondary structure, and expression levels of lncRNAs as well as their cognate RNA-binding proteins underlie diseases ranging from neurodegeneration to cancer. The involvement of lncRNAs in human diseases could be more prevalent than previously appreciated (Wapinski and Chang, 2011). In this regard, a number of lncRNAs have been shown to play a role in cancer. For example, HOTAIR is remarkably overexpressed in breast tumors and the expression of HOTAIR in primary breast tumors is a strong prognosis marker of patient outcomes such as metastasis and patient survival (Gupta et al., 2010). Enforced expression of HOTAIR causes altered histone H3 Lys27 (H3K27) methylation pattern and increases invasiveness. In contrast, the depletion of HOTAIR results in the opposite cellular phenotype. ANRIL is upregulated in prostate cancer and is required for the repression of the tumor suppressors INK4a/p16 and INK4b/p15 (Yap et al., 2010; Kotake et al., 2011). MALAT1 plays a role in cell migration and tumor metastasis, as demonstrated by knockout of MALAT1 in lung cancer cell lines (Gutschner et al., 2013). This may have to do with its ability to regulate the expression of cell cycle genes, which is required for G1/S and mitotic progression (Tripathi et al., 2013). Furthermore, depletion of MALAT1 leads to activation of p53 and its target genes. In contrast, lincRNA-p21 serves as a repressor in p53-dependent transcriptional responses by directing the recruitment of hnRNP K to its genomic targets (Huarte et al., 2010). Our own work suggests that while RoR (Zhang et al., 2013a) may function as an oncogene through suppression of p53 in response to DNA damage, loc285194 (Liu et al., 2013) and GAS5 (Zhang et al., 2013b) may play a tumor suppressive role through the ‘competitive endogenous RNA’ (CeRNA) mechanism (Salmena et al., 2011). Together, these findings highlight the significance of RNAs in cancer. Thus, a better understanding of how lncRNAs regulate gene expression in cancer will aid in the development of novel strategies for cancer therapy.

Regulation of lncRNAs

As discussed earlier, lncRNA may function as oncogenes or tumor suppressors. Thus, lncRNAs are often dysregulated in cancer specimens. So a question is how lncRNAs are regulated in cancer cells. Like protein-coding genes, lncRNAs are mainly transcribed by RNA polymerase II and they can be regulated by various other factors in addition to the core transcription complex. For instance, a number of lincRNAs are transcriptionally regulated by key transcription factors such as p53, NFκB, Sox2, Oct4, and Nanog (Guttman et al., 2009). Apparently, p53 is a major transcription factor that can regulate a large number of genes including lncRNAs which will be discussed in more details below.

In addition, other transcription factors have been implicated in regulation of lncRNA expression. For example, several lincRNAs have been shown to be elevated in induced pluripotent stem cells (iPSCs) compared with embryonic stem cells (ESCs), suggesting that their activation may promote the emergence of iPSCs (Loewer et al., 2010). Importantly, these lincRNAs are direct targets of key pluripotency transcription factors. lincRNA–RoR can modulate reprogramming, suggesting critical functions of lincRNAs in the derivation of pluripotent stem cells (Loewer et al., 2010). Pluripotency factors OCT4, SOX2, and KLF4 directly target the promoter region of RoR. Since stemness is directly related to cancer initiation, progression, and metastasis, a better understanding of how lncRNAs are regulated is important to cancer biology. On the other hand, lncRNAs are also subject to transcriptional repression. For example, rhabdomyosarcoma 2-associated transcript (RMST) is necessary for neuronal differentiation and its expression is repressed by REST, a master transcription factor controlling key neuronal genes through which RMST binds to promoters of SOX2 target genes and activates gene transcription (Ng et al., 2013).

LncRNA-associated ribonucleoprotein complexes

Interaction of lncRNAs with proteins to form ribonucleoprotein complexes is critical for lncRNAs to exert their function as gene regulators. There are many reports on interactions of lncRNAs with proteins. The well-known example is the interaction of lncRNAs with chromatin remodeling enzymes such as EZH2 and PRC2 (Gupta et al., 2010; Prensner et al., 2011). Furthermore, as a H3K27 methyltransferase, EZH2 is a phosphorylated protein when it is active; the phosphorylation at threonine residue 345 has been shown to be critical to its interaction with HOTAIR (Kaneko et al., 2010). Similarly, linc-HOXA1 RNA represses Hoxa1 by interaction with the protein PURB as a transcriptional cofactor (Maamar et al., 2013). LincRNA-p21 has been shown to be a p53 transcriptional target (Huarte et al., 2010). When lincRNA-p21 binds to hnRNP K to form a ribonucleoprotein, this complex mediates global gene repression and apoptosis in the p53 pathway. Thus, lincRNA-p21 serves as a p53 effector. In addition, both CDKN1A and PANDA (P21 associated ncRNA DNA damage activated) are p53 transcriptional targets (Hung et al., 2011). While CDKN1A mediates cell cycle arrest, PANDA is able to block apoptosis through interaction with the transcription factor NF-YA to limit expression of pro-apoptotic genes. In prostate cancer, CTBP1-AS can interact with PSF to cause the global androgen-mediated gene repression (Takayama et al., 2013); similarly, PCGEM1 and PRNCR1 can interact with androgen receptor (AR) to activate AR signaling and cause castration resistance (Yang et al., 2013). We have previously shown that hnRNP I is a binding partner for RoR (Zhang et al., 2013a). Of interest, only phosphorylated form of hnRNP I in the cytoplasm can interact with RoR. Together, these examples signify the importance of lncRNA–ribonucleoprotein complexes in gene regulation and also support the notion that identification of lncRNA-associated ribonucleoprotein complexes is a critical step to the understanding of lncRNA-mediated gene expression.

LncRNAs and p53 regulatory network

It is well known that as a master transcription factor, p53 regulates a large set of genes and at the same time, p53 itself is under sophisticated regulations involving mRNA stability, translation, and posttranslational modifications (Halaby and Yang, 2007; Kruse and Gu, 2009). A well-known negative p53 regulator is MDM2, which is a major E3 ubiquitin ligase controlling the p53 stability through the ubiquitin–proteasome pathway while allowing its rapid increases in response to stress (Kruse and Gu, 2009; Meek, 2009). Upon induction, p53 binds to p53 response elements in the promoter of p53 target genes, leading to their transcription activation. Among them, p21 (Cip1) is a well-known regulator of cell cycle progression (Harper et al., 1993). Now lncRNAs have also joined this p53 regulatory network (Figure 1). In this regard, several lncRNAs have been shown to regulate p53 and thus, they serve as p53 regulators. On the other hand, p53 is able to regulate lncRNA expression. For the sake of simplicity of our discussion, we categorize lncRNAs into two groups. Group 1 are those lncRNAs serving as p53 regulators such as MALAT1, maternally expressed gene 3 (MEG3), p53-eRNAs, and Wrap53. However, unlike MALAT1 or MEG3, Wrap53 and p53-eRNAs are capable of regulating p53 mRNA stability and impacting transcription of p53 target genes. Group 2 lncRNAs serve as p53 effectors, including lncRNA-p21, PANDA, H19, and loc285184. RoR is unique because in addition to be serving as a p53 effector, RoR can also repress p53 in response to DNA damage, forming an autoregulatory feedback loop. Together, lncRNAs add another layer of regulation to this complex p53 tumor suppressor network. Although a number of microRNAs are direct p53 targets as well (He et al., 2007; Hermeking, 2012), in this discussion we will focus on lncRNAs.

Figure 1.

LncRNAs associated with the p53 pathway. Like protein-coding genes such as MDM2, lncRNAs can serve as p53 regulators, which include MALAT1, MEG3, p53-eRNAs, and Wrap53. On the other hand, lncRNAs can also serve as p53 effectors, including lincRNA-p21, PANDA, H19, and loc285184, similar to the protein-coding gene p21.

LncRNAs as regulators of p53

These lncRNAs are capable of directly or indirectly regulating p53. Although gene regulation can occur at the transcriptional or posttranscriptional levels, to date, the evidence of transcriptional regulation of p53 by lncRNAs is still lacking, probably because the p53 mRNA level often remains consistent in the cell. Available evidence points at posttranscriptional regulation by lncRNAs, which may be able to either activate or suppress p53. In addition, we will briefly touch on other potential p53-associated lncRNAs at the end of this section as evidence is still sketching.

MALAT1

MALAT1 (also referred to as Neat2) was originally identified to be overexpressed in lung cancer (Ji et al., 2003) and subsequently reported to be upregulated in many other cancer types including breast cancer, colon cancer, and hepatocarcinoma (Yamada et al., 2006; Lin et al., 2007; Guffanti et al., 2009). Ectopic expression of MALAT1 in various cell lines enhances cell proliferation and promotes tumor formation in nude mice. In support of this notion, a recent study with MALAT1 knockout model in human lung tumor cells suggests that MALAT1 promotes metastasis through regulation of genes including a set of metastasis-associated genes (Gutschner et al., 2013), which appears to be unrelated to the ability of MALAT1 to regulate alternative splicing of pre-mRNAs (Bernard et al., 2010).

The expression of MALAT1 in normal tissue has been extensively studied. Although MALAT1 RNA is expressed in various tissues, it is highly abundant in neurons. This high expression appears to come from upregulation of MALAT1 during differentiation, as in vitro differentiation of neural stem cells show significant upregulation of MALAT1 expression in neuronal and glial differentiated progeny (Mercer et al., 2010). MALAT1 has recently been shown to regulate alternative splicing of endogenous target genes by modulating SR splicing factor phosphorylation, affecting their levels, the distribution, and ratio of phosphorylated to dephosphorylated pools (Bernard et al., 2010). MALAT1 is colocalized with pre-mRNA-splicing factor SF2/ASF and CC3 antigen in the nuclear speckles and thus, has been implicated in regulation of mRNA splicing. It is enriched in nuclear speckles only when RNA polymerase II-dependent transcription is active. Together, these functional studies suggest the importance of MALAT1 in regulation of cellular functions as well as disease processes.

Of great interest, MALAT1 serves as a p53 repressor and is expressed in a cell cycle-dependent manner, with low levels during G1 and G2 and high levels during G1/S and M phases (Tripathi et al., 2013). Further studies suggest that MALAT1 controls cell cycle progression in human cells and MALAT1-depleted human diploid fibroblasts fail to progress through the G1 phase (Tripathi et al., 2013). Importantly, MALAT1 depletion results in the activation of p53 and its downstream target genes, including p21. However, transient overexpression of MALAT1 in human diploid fibroblasts does not alter p53 levels, suggesting that p53 activation in MALAT1-depleted cells could be a part of specific stress response. p53 activation is essential for the S phase defects observed in MALAT1-depleted human cells, possibly by double-stranded DNA damage (Tripathi et al., 2013).

MALAT1 has also been implicated in regulation of the activity of E2F1 transcription factor by modulating the PC2 polycomb protein-mediated sumoylation of E2F1 (Yang et al., 2011). However, it is not clear whether reduction in E2F1 activity or activation of p53 can elicit cell cycle arrest. Time course experiments suggest that induction of p53 upon MALAT1 depletion could be a consequence of double-stranded DNA damage response (Tripathi et al., 2013). MALAT1 depletion increases levels of MDM2 at later time points; however, the role of MDM2 in this aspect is not clear. Despite the possible link between MALAT1-mediated regulation of mitosis and transcription factor B-MYB (Mybl2) (Tripathi et al., 2013), further studies are still needed to determine whether MALAT1 represses p53 through other mechanisms.

MEG3

MEG3 is an imprinted gene; it is expressed in a temporal and spatially regulated manner. Knockout mice with the paternal deletion of MEG3 are normal because there are no changes in expression of either MEG3 and its downstream MEGs or paternally expressed genes compared with wild-type controls (Zhou et al., 2010). In contrast, mice carrying the maternal deletion die perinatally and have major skeletal muscle defects. In mouse, MEG3 is expressed early in embryos in visceral yolk sac and embryonic ectoderm and subsequently in paraxial mesoderm, epithelial ducts, and also in skeletal muscle, cochlea, brain, and eye (Schuster-Gossler et al., 1998). In particular, MEG3 is expressed in adult mouse brain (Miyoshi et al., 2000) and highly expressed in the human pituitary. However, its expression is lost in tumors, whereas ectopic expression of MEG3 inhibits growth in human cancer cells (Zhang et al., 2003). Further studies indicate that the loss of MEG3 expression is prominent in clinically nonfunctioning pituitary adenomas; on the other hand, although MEG3 expression levels appear lower in functioning tumors than in normal pituitaries, this difference is not statistically significant. From examination of human cancer cell lines of various tissue origins for MEG3 expression, it appears that loss of MEG3 expression is a common phenomenon in tumors or cancer cell lines, including those derived from brain, bladder, bone marrow, breast, cervix, colon, liver, lung, meninges, and prostate. In meningiomas, there is a strong association between loss of MEG3 expression and tumor grade (Zhang et al., 2010a). Together, these studies suggest that MEG3 functions as a tumor suppressor, which might involve multiple mechanisms (Zhou et al., 2012).

As a tumor suppressor, MEG3 inhibits tumor cell proliferation possibly through induction of apoptosis. Thus, it is not surprising that MEG3 participates in the p53-mediated gene expression. For example, MEG3 stimulates p53-mediated transactivation in human meningioma cell lines (Zhang et al., 2010a), which seems to be related to its ability to downregulate MDM2 expression (Zhou et al., 2007). However, more work is needed to determine how MEG3 downregulates MDM2. It would be interesting to know whether MEG3 can interact with MDM2 or p53 to disrupt p53–MDM2 interaction or simply facilitate degradation of MDM2. MEG3 RNA can fold into three major motifs, M1, M2, and M3, as predicted by mFold program (Zhang et al., 2010b). Disruption of M2 significantly alters its folding and importantly, abolishes the p53-activating function of MEG3 (Zhou et al., 2007). It remains to be determined how this secondary structure leads to loss of the activity. Interestingly, MEG3-mediated activation of p53 transcription is selective. MEG3 enhances p53-dependent expression of GDF15, an inhibitor of cell proliferation, but has no effect on p21/Cip1 promoter (Zhou et al., 2007). Thus, an intriguing question is whether other lncRNAs such as eRNAs (discussed below) are also involved in this selective regulation.

RoR

As mentioned earlier, RoR is unique in the p53 pathway. On one hand, RoR can suppress p53; on the other hand, RoR is under control of p53. The first indication for RoR involvement of p53 pathway is that knockdown of RoR led to upregulation of genes involved in the p53 response to oxidative stress, DNA damage-inducing agents, as well as cell death pathways (Loewer et al., 2010). Of interest, a recent study suggests that RoR also serves as a sponge for miR-145 (Wang et al., 2013), which has been previously shown to play a role in human iPSCs (Xu et al., 2009) because miR-145 can directly target pluripotency factors OCT4, SOX2, and KLF4. RoR shares microRNA response elements with these core transcription factors and prevents them from microRNA-mediated suppression in self-renewing human ESCs (Wang et al., 2013). These results suggest that RoR forms a feedback loop with core transcription factors and microRNAs to regulate ESC maintenance and differentiation. Since miR-145 is a direct transcriptional target for p53 (Sachdeva et al., 2009), this further supports the role for p53 in apoptosis and stem cell self-renewal.

Using a screening approach for focused lncRNA sets that are selected from lncRNA database (Amaral et al., 2011), we identified two lncRNAs, RoR and loc285194, as p53 targets (Liu et al., 2013; Zhang et al., 2013a). Briefly, we first treated HCT-116 cells with doxorubicin (doxo) and profiled lncRNAs using RT–PCR arrays. These two lncRNAs are significantly induced in HCT-116 cells expressing wild-type p53, but not in HCT-116 p53-null cells. Furthermore, ectopic expression of wild-type p53, but not mutant p53, causes a similar induction, suggesting that this induction is specific to p53.

In silico analysis identified four potential p53 response elements (p53REs) within a 1 kb fragment upstream RoR (Zhang et al., 2013a). p53RE-1 is the most conserved among all four p53REs. Furthermore, luciferase assays with a reporter carrying this 1 kb fragment indicate that this regulation by p53 is at the transcription level and p53RE-1 is critical to the induction of RoR. ChIP assays further confirm that p53 specifically interacts with p53RE-1. Together, these experiments demonstrate that RoR is under control of p53, which is likely through interaction with the putative p53RE. Apparently, since RoR is also able to repress p53, this forms an autoregulatory feedback loop, through which p53 may be delicately kept in check so that the cell is able to better respond to intracellular or extracellular stresses.

Wrap53

Wrap53 is a natural p53 antisense transcript. Natural antisense transcripts potentially constitute a large class of regulatory RNAs that may play a role in regulation of the corresponding sense gene expression. Given that antisense transcription occurs in up to 70% of all mammalian genes (Katayama et al., 2005), it is conceivable that they may play a pivotal role in eukaryotic gene expression, in particular for genes like p53. Identification of Wrap53 supports this notion (Mahmoudi et al., 2009) because the highly conserved Wrap53 plays a critical role in the regulation of p53 at the RNA level.

Wrap53 is located immediately upstream of the p53 gene on the opposite strand; Wrap53 also encodes a WD40 domain protein homologous to members of the large WD40 family of proteins with diverse functions. However, the function of the Wrap53 protein is unknown. Of interest, a similar structure is also detected for the p73 gene, suggesting that it is conserved. Wrap53 is capable of regulating p53 levels in both normal and cancer cell lines (Mahmoudi et al., 2009).

Although Wrap53 is expressed as three isoforms, α, β, and γ, only the α form that contains exon 1a can regulate p53, whereas other two forms lacking this sequence fail to affect p53 levels, supporting the importance of the antisense exon 1a in this regulation. Furthermore, the antisense exon of Wrap53 (exon 1a) alone is sufficient for its effects on p53, suggesting Wrap53 mRNA, rather than the Wrap53 protein, as a regulator of p53 expression (Mahmoudi et al., 2009).

There is only a 227 bp fragment overlap between p53 exon 1 and Wrap53 exon 1a, and antisense oligonucleotide against Wrap53-E1a or p53-E1 causes a specific reduction of p53 expression, demonstrating that this RNA–RNA interaction is required to maintain normal levels of p53 in the cell, which is also supported by RNase protection assays. Furthermore, knockdown of endogenous Wrap53 abrogates p53 induction in response to DNA damage; on the other hand, ectopic expression of Wrap53 potentiates p53-induced apoptosis. However, there is no evidence that de novo synthesis of p53 mRNA is involved. Further studies with luciferase reporter assays, along with the finding that Wrap53 silencing only decreases p53 mRNA but not p53 pre-mRNA levels, suggest that Wrap53a regulates p53 at the posttranscriptional level (Mahmoudi et al., 2009).

p53-specific eRNAs

As mentioned earlier, eRNAs are a class of non-coding RNA molecules transcribed from the DNA sequence of enhancer regions. The functional importance of gene enhancers in regulation of gene expression is well documented (Bulger and Groudine, 2011). By the size, most of eRNAs fall into lncRNA category (Natoli and Andrau, 2012). They were originally identified through genome-wide search in neurons and can bind to the transcription complex to activate the corresponding genes (Kim et al., 2010). However, it appears that eRNAs are a wide-spread phenomenon (Li et al., 2013). The expression of a given eRNA seems to correlate with the activity of its corresponding enhancer in a context-dependent fashion. Increasing evidence suggests that eRNAs actively play a role in transcriptional regulation in cis and in trans. It is likely to have important functions in many regulated programs of gene transcription (Li et al., 2013).

An important feature about eRNA-mediated gene expression is that enhancers control expression of one or multiple genes over distance and in an orientation-independent manner. This is because the enhancer regions can be brought together in close proximity of its target promoters. In support of this notion, a recent study identified several p53-specific eRNAs (p53-eRNAs) including DUSP4, PAPPA, and IER5 (Melo et al., 2013) that are expressed in a p53-dependent manner. Importantly, they contain enhancer activity and interact intrachromosomally with multiple neighboring genes to regulate expression of genes distantly, conveying long-distance p53-dependent transcription regulation (Melo et al., 2013). However, these eRNAs do not seem to directly bind to the genomic DNA at the promoter of their regulated genes, as demonstrated by ChIP-seq; instead, they are able to interact with RNP II complex to activate gene expression. Identification of such p53-eRNAs implies that the number of p53-regulated genes may be much larger than what we previously thought. By this mechanism, p53 may be able to control a specific set of p53 target genes under certain circumstance, while imposing no effect on the rest of p53 target genes.

Other potential p53-associated lncRNAs

In addition to those discussed above, there are several other lncRNAs that might also be involved in regulation of p53. For example, PCGEM1 was identified by differential display analysis of paired normal and prostate cancer tissues (Srikantan et al., 2000). Later, it was shown that PCGEM1 is associated with high-risk prostate cancer patients (Petrovics et al., 2004) and is able to inhibit doxo-induced apoptosis (Fu et al., 2006). Given that DNA damage-induced apoptosis often involves p53, it may imply that PCGEM1 plays a role in the p53 regulatory network because this p53 induction is delayed in LNCaP cells stably overexpressing PCGEM1 compared with control LNCaP cells. During identification of senescence-associated lncRNAs, Abdelmohsen et al. (2013) reported that SAL-RNA1 is able to delay senescence and suppression of SAL-RNA1 increases p53 level although the detailed mechanism remains to be determined yet.

LncRNAs as p53 effectors

Like protein-coding genes or microRNAs, lncRNAs can also serve as p53 effectors. Profiling experiments probably are the best way to identify such lncRNAs. Guttman et al. (2009) have identified 39 lncRNAs that are significantly induced in the wild-type p53 background but not in p53-null cells, demonstrating a strong association in the expression patterns of certain lincRNAs and genes in the p53 pathway. We used RT–PCR array for a focused set of lncRNAs and identified upregulation of several lncRNAs including RoR and loc285194 by p53 (Liu et al., 2013; Zhang et al., 2013a). Therefore, the following lncRNAs participate in the p53 regulatory network by serving as p53 effectors.

LincRNA-p21

Using DNA damage and p53 restoration experiments, Huarte et al. (2010) identified a number of p53-mediated lincRNAs and among them is lincRNA-p21. Further experiments indicate that lincRNA-p21 functions as a repressor in the p53 pathway because knockdown of p53 or lincRNA-p21 individually impacts a large number of genes including those involved in apoptosis and cell cycle arrest. In contrast, enforced expression of lincRNA-p21 impacts expression of the p53-regulated genes. These results suggest that lincRNA-p21 serves as a p53 target and plays a role in triggering apoptosis. In addition, lincRNA-p21 is induced in various tumor models including lung tumor, sarcoma, and lymphoma.

RNA precipitation combined with mass spectrometry demonstrates that lincRNA-p21 interacts with hnRNP K, a member of hnRNP protein family known to have multiple functions. Further studies suggest that this interaction between lincRNA-p21 and hnRNP K is required for proper localization of hnRNP K and transcriptional repression of p53-regulated genes. It is known that hnRNP K is a component of a repressor complex that acts in the p53 pathway (Kim et al., 2008); the binding of lincRNA-p21 to hnRNP K provides a repression selectivity to those genes due to p53 induction through possible interaction with their promoters.

In addition, lincRNA-p21 may function through posttranscriptional regulation. It is known that gene expression can be robustly regulated at the posttranscriptional level by RNA-binding proteins and by non-coding RNAs (ncRNAs). In addition to hnRNP K, lincRNA-p21 can also interact with the RNA-binding protein HuR, through which lincRNA-p21 acts as a modulator of translation (Yoon et al., 2012). For instance, association of HuR with lincRNA-p21 facilitates the recruitment of let-7/Ago2 to lincRNA-p21, leading to lower lincRNA-p21 stability. At a reduced HuR level, lincRNA-p21 is accumulated in cells, increasing its association with JunB and β-catenin mRNAs and selectively lowering their translation. As the level of HuR increases, lincRNA-p21 declines, which in turn derepresses JunB and β-catenin translation and increases the levels of these proteins. Thus, HuR controls translation of a subset of target mRNAs by influencing lincRNA-p21 levels. These findings support a role for lincRNA as a posttranscriptional inhibitor of translation.

PANDA

A number of lncRNAs are expressed in a cell cycle-regulated manner. By using ultrahigh-resolution tiling microarrays, Hung et al. (2011) identified PANDA as one of p53-regulated lncRNAs involved in cell cycle progression and apoptosis. Their survey covers the transcriptional and chromatin landscape around the transcription start sites of 53 cell cycle genes including those encode all known cyclins, CDKs and CDKIs at 5-nt resolution across 25 kb of the 9p21 locus as well as from 10 kb upstream to 2 kb downstream of each transcription start site.

PANDA is a bidirectional transcript from the CDKN1A promoter; DNA damage induces five lncRNAs from the CDKN1A promoter, including PANDA, in a p53-dependent manner. Expression of PANDA and CDKN1A were positively regulated by p53. There is a p53-binding site immediately upstream of the CDKN1A transcription start site, as demonstrated by ChIP–ChIP analysis. PANDA and CDKN1A are diametrically situated 2.5 kb from this intervening p53-binding site, which supports the possibility of p53 co-regulation. Indeed, knockdown of p53 by RNAi before DNA damage inhibits the induction of PANDA from DNA damage. However, CDKN1A does not induce PANDA, suggesting that PANDA is not a linked transcript of CDKN1A, nor its expression is dependent on p21 (Hung et al., 2011).

Lower expression levels of PANDA are seen in human primary breast tumors harboring mutant p53; ectopic expression of wild-type p53 can restore DNA damage-inducible expression of PANDA. In contrast, PANDA has no effect on expression of CDKN1A and p53 (Hung et al., 2011). These findings support the notion that PANDA is a p53 effector in response to DNA damage to suppress apoptosis while CDKN1A induces cell cycle arrest. Suppression of PANDA leads to the doxo-induced apoptosis and at the same time it upregulates genes involved in apoptosis, including the apoptotic activators APAF1, BIK, FAS, and LRDD (Hung et al., 2011). Mechanistically, PANDA RNA interacts with the transcription factor NF-YA (Li et al., 1991); they together are able to suppress expression of pro-apoptotic genes that carry putative NF-YA binding sites in their promoter regions.

Loc285194

Loc285194 is a lncRNA located at osteo3q13.31. Since the osteo3q13.31 locus harbors frequent focal copy number alterations (CNAs) and loss of heterozygosity (LOH) in primary osteosarcoma samples, it implies that loc285194 may function as a potential tumor suppressor. Further studies suggest that loc285194 may regulate genes related to cell cycle progression and apoptosis (Pasic et al., 2010), supporting the role of loc285194 as a tumor suppressor. Focal osteo3q13.31 CNAs and LOH are also common in cell lines from several other cancers in lung, autonomic and central nervous system, blood, endometrium, soft tissue, skin, gastrointestinal tract, urinary tract, and breast. However, little is known as to how loc285194 is regulated in cancer cells.

Similar to RoR, loc285194 is also found to be induced by p53 through interaction with the putative p53RE in the upstream region of loc285194, as demonstrated by promoter analysis and ChIP assays (Liu et al., 2013). As a putative tumor suppressor, loc285194 suppresses cell growth both in vitro and in vivo. In consistent with these findings, loc285194 is downregulated in colon tumor specimens (Liu et al., 2013). A possible mechanism of loc285194 in suppression of colon cancer cell growth may involve its ability to repress the oncogenic miR-211 that has been previously shown to be upregulated in colon cancer and promote tumor cell growth (Cai et al., 2012). In particular, loc285194 and miR-211 form a reciprocal repression feedback loop through a similar microRNA-mediated silencing mechanism involving the RISC complex (Liu et al., 2013). Thus, their interaction may function as a part of ‘CeRNA’ network (Salmena et al., 2011).

H19

Like MEG3, H19 is also an imprinted gene and is maternally expressed. It carries a spliced and polyadenylated RNA that is expressed at a high level in a wide variety of fetal tissues when cell differentiation occurs. Of interest, H19 is imprinted at the IGF2 locus and expressed from the maternal allele, while the neighboring IGF2 gene is transcribed from the paternal allele. H19 has complex functions and its role in cancer is controversial. For example, H19 may function as an oncogene in hepatocellular carcinoma (HCC) and bladder carcinoma (Matouk et al., 2007). However, in murine models of tumorigenesis, the H19 locus controls the size of experimental teratocarcinomas, the number of polyps in the APC murine model of colorectal cancer, and the timing of appearance of SV40-induced hepatocarcinomas (Yoshimizu et al., 2008), clearly displaying a tumor suppressor effect in mice. Therefore, it appears that the role of H19 in cancer may be dependent on cell type or tissue. Relevant to this review, H19 is regulated negatively by p53. Dugimont et al. (1998) showed that antisense p53 increases H19 transcripts in HeLa cells. Moreover, a temperature sensitive (ts) 143 Ala p53 mutant reveals temperature-dependent regulation of H19 expression, which appears to depend on the putative promoter of H19. However, little follow-up has been carried out regarding the underlying mechanism of this p53-mediated repression of H19. Although a recent study suggests that H19 may interact with p53, leading to inactivation of p53 (Yang et al., 2012), further verifications of this interaction in various systems are still needed in order to determine the function of this interaction.

The role of hnRNP proteins in p53 regulatory network

The hnRNP proteins are a large group of proteins that have been shown to interact with pre-mRNAs. Several different hnRNP proteins can bind to the same pre-mRNA to form a ribonucleoprotein complex immediately after the RNA is transcribed. A major function of these hnRNP proteins is to participate in RNA processing, splicing, and mRNA transport. For example, hnRNP A1 protein can influence the choice of splice sites in pre-mRNA, thus playing a role in alternative splicing (David et al., 2010), whereas other hnRNP proteins may be involved in the transport of mRNA to the cytoplasm. In this regard, the hnRNP C protein contains a retention sequence that appears to prevent mRNA from being transported out of the nucleus. However, hnRNP proteins are multiple functional proteins. In particular, they can interact with lncRNAs to regulate gene expression and cellular functions. For example, lncRNA RMST is involved in regulation of genes implicated in neurogenesis by interaction with hnRNP A2/B1 and transcription factor SOX2 (Ng et al., 2013).

Emerging evidence suggests that hnRNP proteins are also involved in the p53 regulatory network through interaction with lncRNAs. As discussed earlier, hnRNP K can interact with lncRNA-p21; similarly, hnRNP K can also interact with MDM2, leading to degradation of MDM2 after DNA damage. DNA damage-induced hnRNP K sumoylation regulates p53 transcriptional activation. SUMO modification plays a crucial role in the control of hnRNP K's function as a p53 co-activator in response to DNA damage by UV (Lee et al., 2012). DNA damage stimulates hnRNP K sumoylation through Pc2 E3 activity, and this modification is required for p53 transcriptional activation (Pelisch et al., 2012). Furthermore, ATM-dependent hnRNP K phosphorylation is required for its stabilization and its function as a p53 transcriptional cofactor in response to DNA damage (Moumen et al., 2013).

In addition to hnRNP K, hnRNP H/F is capable of maintaining p53 pre-mRNA 3′-end processing in response to DNA damage through interaction with a G quadruplex structure located downstream from the p53 cleavage site (Decorsiere et al., 2011). hnRNP Q regulates translation of p53 in normal and stress conditions. Regulation of translation initiation is also crucial for p53 protein accumulation. In this case, hnRNP Q binds to the 5′-untranslated region (UTR) of mouse p53 mRNA and regulates translation efficiency of p53 and apoptosis progression (Kim et al., 2013).

Our studies with RoR suggest that hnRNP I functionally interacts with RoR (Zhang et al., 2013a). Of interest, this interaction regulates DNA damage-induced p53 levels because the ability of RoR to repress p53 is dependent on the interaction of RoR with hnRNP I and suppression of hnRNP I by RNAi substantially decreases the doxo-induced p53. Deletion and site-directed mutagenesis analysis revealed that a 28-base RoR sequence carrying the potential hnRNP I binding motifs is essential and sufficient for this repression, providing further evidence of the involvement of hnRNP I in RoR-mediated p53 repression. hnRNP I is an RNA-binding protein that carries several RNA-binding domains and is well known for its role in mRNA splicing (Oberstrass et al., 2005); it is relatively abundant in the cell. While the majority of hnRNP I is retained in the nucleus; only small fraction is in the cytoplasm. Since mRNA splicing takes place in the nucleus, this function is mostly likely dependent on the nuclear hnRNP I. On the other hand, our study shows that the cytoplasmic hnRNP I is phosphorylated and is responsible for the interaction with RoR because almost all of the hnRNP I that interacts with RoR is phosphorylated. As shown in Figure 2, DNA damage induced by doxo or UV causes subcellular redistribution of hnRNP I because we detect more cytoplasmic hnRNP I in the doxo-treated cells. Moreover, the cytoplasmic hnRNP I is likely phosphorylated by protein kinase A (PKA) because a previous study shows that PKA serves a kinase for hnRNP I phosphorylation (Xie et al., 2003). Importantly, the phosphorylated hnRNP I can interact with the 5′-UTR of p53 mRNAs to stimulate its translation. Alternatively, the phosphorylated hnRNP I can also interact with RoR to suppress its translation. Since RoR is induced by p53 after DNA damage, the elevated RoR is exported to the cytoplasm and competes with p53 mRNA for hnRNPs. Thus, this forms an autoregulatory feedback loop, which provides an additional mechanism to keep p53 levels in check.

Figure 2.

ROR–p53 feedback auto-regulatory system through hnRNP I. RoR is a p53 transcriptional target; meanwhile, RoR can also repress p53 expression in response to DNA damage. This RoR-mediated repression of p53 is likely through the interaction of RoR with the phosphorylated hnRNP I such that RoR suppresses the capability of hnRNP I to interact with the 5′-UTR of p53 mRNA, a process known to stimulate its translation.

Perspective

As a master gene regulator, p53 is strictly regulated because an unwanted induction of p53 could be deleterious to the cell. In order for the cell to maintain p53 at a certain level, lncRNAs are recruited as an additional mechanism into this complicated regulatory system. LncRNAs such as MALAT1, MEG3, RoR, Wrap53, and p53-eRNAs may be just a few of such examples. On the other hand, lncRNAs can serve as p53 effectors. Given the large number of lncRNAs, we expect that more p53 effector lncRNAs will be identified, and thus this will greatly expand the repertoire of p53 targets, providing more flexibility for cells to adapt to dynamic environments.

Despite the progress in the lncRNA field in recent years, overall, lncRNA studies are still at a very early stage. To better understand the lncRNA-p53 regulatory network, we need to address several issues. There is an urgent need for a systematical identification of 53-regulated lncRNA. Current studies are still scattered. A more systematic screen may be able to provide a more comprehensive picture of p53-regulated lncRNAs. More importantly, whether these lncRNAs can functionally interact with p53 remains to be determined. As a transcription factor, p53 activates gene transcription often through binding to p53 responsible elements in the promoter of p53-regulated genes, and thus, it is conceivable that p53 is also able to interact with RNAs, including lncRNAs. To test this possibility, one approach would be to perform a global search by RIP-seq with p53 antibody under various conditions including nonstress and stress conditions. Finally, given the important role of p53 in cancer, future studies may also need to evaluate the utility of p53-associated lncRNAs as cancer biomarkers or therapeutic targets. For instance, targeting the lncRNAs associated with cell cycle regulation or apoptosis, along with conventional chemotherapy, may enhance cancer therapy efficacy.

Funding

This work was in part supported by NIH grant R01 CA154989 (Y.-Y.M.).

Conflict of interest: none declared.