Role of lncRNAs in Cellular Aging

Aging is a universal, intrinsic, and time-dependent biological decay that is linked to intricate cellular processes including cellular senescence, telomere shortening, stem cell exhaustion, mitochondrial dysfunction, and deregulated metabolism. Cellular senescence is accepted as one of the core processes of aging at the organism level. Understanding the molecular mechanism underlying senescence could facilitate the development of potential therapeutics for aging and age-related diseases. Recently, the discovery of long non-coding RNAs (lncRNA) provided insights into a novel regulatory layer that can intervene with cellular senescence. Increasing evidence indicates that targeting lncRNAs may impact on senescence pathways. In this review, we will focus on lncRNAs involved in mechanistic pathways governing cellular senescence.

phenotype (SASP) (3)(4)(5)(6)(7)(8)(9). SASP can explain how relatively low number of cells can cause problems in organs and overall body performance. Third, cellular senescence could intervene with mitochondrial and cellular metabolism through stress activation that results in disturbed energy homeostasis. In contrasts, cellular senescence seems to have a protective function against tumor formation by limiting cellular replication that is consistent with the interception of senescence pathway with tumor response pathways.
Hayflick and Moorehead challenged the cellular immortality idea with their findings on the limit of cellular replication after extensive passaging (10,11). Later, this senescence type has been linked to telomere attrition, a process that causes chromosomal instability and activates DNA damage response (DDR), which is now known as replicative senescence (4). The DDR pathway is a signal transduction pathway comprised of multiple interacting components. It is initiated by ATM and ATM-Rad3-related (ATR). Downstream of these proteins are checkpoint kinases (CHK1 and CHK2) that phosphorylate and activate downstream effectors such as p53 and p21. Besides telomere attrition, other factors such as UV, oncogenic insult, and chemotherapeutics can also activate DDR and initiate cellular senescence (11)(12)(13)(14)(15)(16)(17)(18). In addition, loss of PTEN (tumor suppressor) and overexpression of E2F3 (the S-phase transcription factor) can initiate p53-p21 pathway (19)(20)(21).
In vitro senescent cells have distinctive characteristics such as increased cell size, accumulation of senescence-associated β-galactosidase (SA-β-GAL), activation of survival genes to avoid apoptosis, and the production of the SASP (7,(31)(32)(33)(34). The discovery of the SASP provided clues to how senescent cells could disrupt the overall homeostasis of the body albeit their relatively low number. The presence of SASP in mammalian cells appeared to be conserved, and some of the proteins in secretome have been repeatedly detected in multiple cell type. Another senescence-linked phenotype is the formation of senescenceassociated heterochromatin foci (SAHFs), which is mainly mediated by p16-pRB pathway. These foci often exist in the genomic loci harboring proliferative genes, thereby inhibiting their expression and activity, but it is important to keep in mind that SHAF does not exist in all senescent cells (35). In senescent cells lacking SAHFs, p16-pRB pathway silences proliferative genes on the epigenetic level as compensatory pathway (36). Currently, measurement of multiple traits mentioned above is the best way to identify senescent cells since each cell type can develop the phenotype with varying order and composition (Figure 1). Figure 1 shows the generic process for cellular senescence while it does not imply the existence of each component in all senescent cells (37)(38)(39). Table 1 also indicates specific lncRNAs and their targets. With protein-coding genes making up to only 2% of the human genome, the other 98% have been considered "junk DNA" till the encyclopedia of DNA elements (ENCODE) project and the functional annotation of the mammalian genome (FANTOM) consortium made their early release. They have challenged this concept by identifying a large number of novel transcripts, namely non-coding RNAs (ncRNAs), from the "junk" regions.
Based on their transcript lengths, ncRNAs were further divided into small ncRNAs and long ncRNAs (lncRNAs). lncRNAs are a major class of heterogeneous ncRNAs with the lengths more than 200 nucleotides. They are now being recognized as novel regulators in multiple cellular processes, including development, differentiation, chromosome remodeling, imprinting, and cell cycle control (40)(41)(42)(43)(44). Research also shows that lncRNAs are tightly regulated and highly tissue specific (43). Moreover, dysregulation of lncRNAs has been associated with human diseases, such as cancer or neurodegenerative disorders (45). However, the functions of most lncRNAs remain uncharacterized (46,47). Long non-coding RNAs often contain a small open reading frame similar to mRNAs, which was a challenge on their distinction (48). Ribosome-profiling experiments have been employed to address this issue. Initial results of ribosome profiling showed that many lncRNAs were, in fact, ribosome-bound while high sensitivity mass spectrometry failed to identify those small peptides from supposed "lncRNA ORFs" (49). With the help of bioinformatics, ribosome release score has been calculated for protein-coding and non-coding RNAs. Ribosome release score is produced by using ribosome dissociation and association values of given part of RNA. Using this score, most lncRNAs fall into the non-coding category, supporting the results of previous mass spectrometry experiments (44,50).
Regulatory role of lncRNAs in gene expression has been studied. According to these research, their molecular mechanism is quite versatile and there are multiple modes of action (40,51). lncRNAs may regulate gene expression at transcriptional, RNA processing, translational or posttranslational levels by interacting with DNA, RNA, or protein molecules. The subcellular localization of lncRNAs may also provide additional complexity to their function (52).
Most of the lncRNAs are enriched in the nucleus (43,53), and their localization supports their suggested role in the epigenetic regulation of the chromatin (54,55). They function in cis to regulate the expression of nearby genes or in trans to regulate genes from distant genomic locations. Nuclear lncRNAs often bind to heterogeneous ribonucleoprotein complexes (hnRNPs) and polycomb repressor complex (PRC), consistent with their involvement in modulating epigenetic markers and interacting with transcription factors of target genes. One of the well-known nuclear lncRNAs, Xist, is transcribed from the inactive X chromosome, and it controls dosage compensation of chromosome X in human females (56). Xist recruits polycomb repressive complex 2 (PRC2) (56) that switches histone markers to repressive state, and it inhibits transcription of target genes by blocking the access of RNA polymerase II (RNAPII) to inactivated X chromosome (57,58). PRCs are one of the major binding partners of lncRNAs that has been identified so far. This affinity is explained by intrinsic promiscuity of PRCs and lncRNAs ability to guide PRCs.
Although lncRNAs are generally enriched in the nucleus, many lncRNAs are also detectable in the cytosol. lncRNAs can control mRNA stability by acting as "sponge" for miRNAs in the cytosol. Sponge lncRNAs compete for the miRNA binding site, which results in rescued expression of mRNAs (59). lncRNAs can also associate with a coding gene by overlapping exons to increase the stability of the target mRNA, e.g., BACE1-AS. In contrast, lncRNAs may also be involved in mRNA degradation through Staufen1 (STAU1)-mediated mRNA decay (SMD). This process recognizes double-stranded RNA for degradation. It has been shown that lncRNAs can form RNA-RNA duplex with 3′-UTRs of target mRNA at Alu elements and cause SMD (60). lncRNA-mRNA complex (bound by Alu element) can be the result of imperfect base pairing, and this double-stranded RNA complex activates SMD, which happens through recruitment of UPF1 (regulator of nonsense transcript-1) that shortens the poly-A tail. Shortened poly-A translates to the shorter lifespan for mRNA (61). Another function is that H19 is transported to the cytosol, where it is used as a source of microRNA (miR-675) to suppress growth (62)(63)(64)(65)(66)(67)(68)(69)(70)(71).
Several lncRNAs, as discussed below, have been reported to modulate expression of p21 in different ways. Although the involvement of these p21-regulating lncRNAs in aging has not been directly demonstrated, given the key role of p21 in cellular senescence, it warrants further investigation on how these lncRNAs may participate in cellular senescence regulation through p21.
Focally Amplified lncRNA on Chromosome 1 Hu et al. focused on somatic copy number alterations (SCNAs) of lncRNAs. By analyzing a genome-wide survey on SCNAs of lncRNA in 2394 tumor specimens from 12 types of cancer, they identified a correlation between copy number of focally amplified lncRNA on chromosome 1 (FAL1) and progression of ovarian cancer. They found out that FAL1 interacts and stabilizes the epigenetic repressor BMI1 (the PRC1 core protein) so as to repress p21 expression. Silencing FAL1 with siRNAs significantly blocked tumor growth in vivo. The interaction between FAL1 and BMI1 highlights how lncRNAs can take a central role in epigenetic regulation of p21 (80,81).

BRAF-Activated Non-Coding RNA
BRAF-activated non-coding RNA (BANCR) is a lncRNA required for cell migration in different types of cancer (82-84). Shi et al. investigated its contribution to colorectal cancer. They showed that BANCR expression was significantly decreased in colorectal cancer, and when BANCR is overexpressed, it slowed down the growth of colorectal cancer. They also suggested that pcDNA-BANCR-mediated proliferation was linked to activation of G0/G1 cell cycle arrest through regulation of p21. Effects of BANCR were probably post-transcriptional since protein expression levels of p21 increased while that of mRNA remained unchanged. In brief, this article suggests that downregulation of BANCR contributes to the translational regulation of p21 protein (85).

LincRNA-P21
LincRNA-p21 plays a role in regulating p21 levels. Upon DNA damage, p53 activates lincRNA-p21 that is located next to the p21 gene. In the nucleus, lincRNA-p21 activates transcription of p21 by recruiting hnRNP-K to the promoter region of p21 (86). Using conditional lincRNA-p21 knockout mouse model, Dimitrova et al. showed that loss of lincRNA-p21 could block p21 expression and it can activate mouse embryonic fibroblasts (MEFs) expansion. In the cytoplasm, lincRNA-p21 bound HuR recruits the let-7/Ago2 complex to destabilize lincRNA-p21. LincRNA-p21 also associates with its target mRNAs by base pairing and suppresses their translation with the Rck RNA helicase (86,87).

Metastasis-Associated Lung Adenocarcinoma Transcript 1
The transcript of metastasis-associated lung adenocarcinoma transcript 1 (MALAT1) gene was first identified in 1997 (88), and it is highly expressed in several type of cancers (89). In 2010, MALAT1 was reported as a highly abundant nuclear transcript localized to the nuclear speckles (a region which is enriched in pre-mRNA splicing factors) (90,91). In 2013, Tripathi et al. performed genome-wide transcriptome analyses in human fibroblasts. Their result indicated that MALAT1 controlled the expression of cell cycle genes and was required for cell cycle progression (92). They also showed that depletion of MALAT1 leads to activation of the p53 pathway and p53 is required for MALAT1 dependent cell cycle arrest. Moreover, MALAT1-depleted cells have low levels of B-MYB (Mybl2, an oncogenic transcription factor), leading to the discovery of aberrant alternative splicing of B-MYB pre-mRNA. However, the exact mechanism of how MALAT1 impacts p53 signaling needs further investigation (91,92).

7SL
In 1991, Sakamoto et al. cloned 7SL sequence partially (93). They observed the 7SL would block proliferation of HeLa cells. Recently, Abdelmohsen et al. showed that 7SL forms a partial hybrid with the 3′-untranslated region (UTR) of p53 mRNA. The association of 7SL with p53 mRNA reduced p53 translation. When 7SL were silenced, it increased the interaction between HuR and TP53 mRNA, stabilizing TP53 mRNA and increasing p53 translation. The antagonist activity between 7SL and HuR for TP53 3′ UTR contributes to the fine tuning of p53 translation (94).

lncRNAs ReGULATiNG CDKN2A
Very long intergenic non-coding RNA-vlincRNA antisense to DDAH1 (VAD) and ANRIL have been identified from senescent cells and found to have an impact on cellular senescence while MIR31HG and UCA1 have been identified through oncogenic pathways.
very Long intergenic Non-Coding RNA-vlincRNA Antisense to DDAH1

Senescence-Associated lncRNAs
Abdelmohsen et al. identified differentially expressed novel lncRNAs between young versus old human diploid WI-38 fibroblasts populations. These populations are 32 days apart from each other in doubling time. Among those novel lncRNAs, they selected three lncRNAs to further study their involvement in cellular senescence and proliferation. Decreasing expression of those lncRNAs resulted in increased p53 expression, thereby affecting cell survival and senescence (102). However, exact mechanism requires further study.

Senescence-Associated Long Non-Coding RNA
Using human fibroblast cells, Wu et al. performed genome-wide screening of lncRNA expression in cellular senescence. They have identified lncRNAs with differential expressions in senescent cells in contrast with young cells. Within these lncRNAs, they further selected a specific senescence-associated lncRNA (SALNR) that has low expression in senescent cells. When they overexpressed the SALNR, cellular senescence was delayed. Furthermore, they found that SALNR physically interacts with NF90 (nuclear factor of activated T-cells, 90 kDa), which is able to suppress biogenesis of senescence-associated miRNAs, such as miR-22 and miR-181a. Inhibiting NF90 resulted in premature senescence since NF90 is the suppressor of senescence miRNAs.
Moreover, when cells were exposed to Ras-induced stress (activates senescence), NF90 gets translocated to nucleolus and NF90 can no longer suppress senescence-associated miRNA biogenesis, which can be rescued by SALNR overexpression. Thus, SALNR antagonizes NF90 translocation into nucleolus and rescues its inhibitory activity on senescenceassociated miRNA expression. Their data suggest that lncRNA SALNR controls cellular senescence through modulating NF90 localization (103).

Telomeric Repeat-Containing RNA
First, telomeric repeat-containing RNA has been identified in yeast and it has a specific hotspot in the subject of senescence. It forms DNA-RNA pairing with telomere region at R-loop formation that contains G-quadruplex, TERRA, and TERRAbinding proteins. TERRA also increases upon DNA damage at telomere region to involve in recombination and DNA repair. Arnoult et al. showed that TERRA expression decreases upon telomere elongation in multiple cancer cell lines (104). Elongated telomeres have less TERRA expression due to tightly packed telomeric repeats with HP1α and H3K9me3, which make chromosome inaccessible to the transcription machinery.
In parallel, increased levels of TERRA was also found in several primary human tumors containing short telomeres (105). In other words, TERRA is reversely correlated with telomere length.
On a different research, TERRA was shown to maintain the telomere length in the lack of telomerase enzyme as explained above. RNA-DNA complex formed between TERRA and telomeres postpones the senescence by activating homologous recombination at telomeres (106). These suggest multiple functions for TERRA on telomere length and maintenance; however, further research is required to unravel the detailed mechanism of how TERRA functions in telomere and telomerase during aging (107)(108)(109)

P21-Associated ncRNA DNA Damage Activated
With the help of super high-resolution tiling array analysis in CDKN1 region, Puvvula et al. identified P21-associated ncRNA DNA damage activated (PANDA) as a lncRNA, which is induced upon DNA damage. P21-associated ncRNA DNA damage activated plays a dual role in proliferating cells versus senescent cells. In proliferating cells, scaffold-attachment-factor A/hnRNPU (SAFA) interacts with lncRNA PANDA to block senescence genes by recruiting PRC1 and PRC2 to senescence activating genes loci including CDKN1. Silencing SAFA and PANDA induces cell cycle arrest, and it also causes senescence in proliferating cells by allowing transcription of senescence activating genes.
In senescent cells, SAFA-PANDA-PRC interactions are blocked therefore, pro-senescence genes are transcribed. PANDA inhibits the expression of proliferative genes by sequestering the transcription factor NF-YA from occupying its target promoters in senescent cells. Depletion of PANDA in the senescent cells reactivates proliferative genes and allows cells to replicate (117).

CONCLUSiON AND PeRSPeCTiveS
As concluding remarks, with the recent advancement in technology, our understanding on the function of lncRNAs has expanded vastly. Here, we have summarized an emerging role of lncRNAs as regulators of cellular senescence and age-related diseases. Albeit of these progresses, lncRNAs in senescence field is still in its infancy so as lncRNAs in general. Most of these studies remain at the cellular level, but the evidence that lncRNAs can regulate aging process in vivo has yet to be established.
Understanding lncRNAs function may help us fill some knowledge gap in genetic causes for human phenotypes. Many SNPs are not associated with coding genes but with non-coding region where lncRNAs may be transcribed (118). For example, a recent study on Asian centurions revealed a novel SNP rs2440012 associated with longevity (119). Given the growing evidence of the functional role lncRNAs in physiological and pathological processes, we believe it is just a matter of time to illustrate role of lncRNAs in aging.

AUTHOR CONTRiBUTiONS
UD wrote the manuscript. SL reviewed and edited the manuscript.

ACKNOwLeDGMeNTS
This is a short text to acknowledge the contributions of specific colleagues, institutions, or agencies that aided the efforts of the authors.

FUNDiNG
This work was supported by Singapore NRF fellowship (NRF-2011NRF-NRFF 001-025) to SL. This research is also supported by the Singapore National Research Foundation under its CBRG grant (NMRC/CBRG/0070/2014 and NMRC/CBRG/0101/2016) and administrated by the Singapore Ministry of Health's National Medical Research Council. This work is also supported by A*STAR SINGA scholarship to UD.