LncRNAs regulating stemness in aging

Abstract One of the most outstanding observations from next‐generation sequencing approaches was that only 1.5% of our genes code for proteins. The biggest part is transcribed but give rise to different families of RNAs without coding potential. The functional relevance of these abundant transcripts remains far from elucidated. Among them are the long non‐coding RNAs (lncRNAs), a relatively large and heterogeneous group of RNAs shown to be highly tissue‐specific, indicating a prominent role in processes controlling cellular identity. In particular, lncRNAs have been linked to both stemness properties and detrimental pathways regulating the aging process, being novel players in the intricate network guiding tissue homeostasis. Here, we summarize the up‐to‐date information on the role of lncRNAs that affect stemness and hence impact upon aging, highlighting the likelihood that lncRNAs may represent an unexploited reservoir of potential therapeutic targets for reprogramming applications and aging‐related diseases.

The non-coding transcriptome encloses a variety of RNA species, spanning from small non-coding RNAs, including microRNAs (miR-NAs), Piwi RNAs (piRNAs), and small nucleolar RNAs (snoRNAs) to long non-coding RNAs (lncRNAs), that are >200 bp long, but could be as large as several kilobases and be subdivided into different categories. LncRNAs are transcribed majorly by Pol II and Pol I RNA polymerases and are present throughout the genome, either as antisense of coding genes (natural antisense transcripts-NATs), pseudogenes, or intergenic (long intergenic non-coding RNAs-lincRNAs), additionally they could be bidirectional, arise from trans-splicing or adopt different structural forms which increase their stability ( Figure 1a) Several lncRNAs have been implicated in gene-regulatory networks performing roles such as chromosome dosage compensation, genomic imprinting, epigenetic regulation, cell cycle control, splicing, and cell differentiation (Mercer, Dinger, & Mattick, 2009;Rinn & Chang, 2012). Mutant mouse strains for different lncRNAs (Fendrr,Peril,Mdgt,Brn1b,or Pint) revealed phenotypes ranging from growth defects to abnormalities in the structure of the neocortex (Sauvageau et al., 2013). This study and others were a proof of concept that, similarly to coding genes, lncRNAs might play critical roles in vivo (Li & Chang, 2014). However, considering that lncRNAs account for 10% in mice and 24% in humans of all RNA transcripts (Atianand & Fitzgerald, 2014), the number of lncRNAs with an assigned function is still limited.
Aging is a biological process characterized by a cascade of biochemical changes which result, ultimately, in an observable functional decay (Lopez-Otin, Blasco, Partridge, Serrano, & Kroemer, 2013) caused by the accumulation of senescent cells which are cells with an irreversible proliferative arrest (de Jesus & Blasco, 2012).
Manipulation of the number of senescent cells impacts in the aging progression, demonstrating the feasibility of antisenescence therapies for age-associated syndromes (Baker et al., 2011(Baker et al., , 2016Campisi & d'Adda di Fagagna, 2007;de Jesus & Blasco, 2012;Gil & Withers, 2016;Itahana, Campisi, & Dimri, 2007). Similarly to the lncRNA footprint existing and correlating with the complexity of different tissues, the percentage of expressed lncRNA genes during senescence of primary human fibroblasts better reflects the different cell identities, when compared with the expression of coding genes ( Figure 1b). Previously, a "footprint" of several senescence-associated lncRNAs (SAL-RNAs) has been identified (Abdelmohsen et al., 2013), highlighting a link between lncRNAs and aging. Furthermore, targeting of selected lncRNAs (e.g., SAL-RNA1-XLOC_023166) was found to actually delay senescence, suggesting a direct role for lncRNAs in the acquisition and/or maintenance of senescence features.
Additionally to their sequence-dependent role, lncRNAs may adopt different structures with the same sequence, leading to different biological properties. One example are the circular RNAs. One curious example is the antisense transcript coexisting in the INK4a-ARF locus (a tumor suppressor associated to stemness, aging, and cancer; Li et al., 2009) named ANRIL (Aguilo, Zhou, & Walsh, 2011;Holdt et al., 2016). ANRIL could adopt a linear and/or circular form.

| Impact of aging on adult stem cells
Stem cells have the potential to self-renew and to differentiate into different lineages, being a source of different adult specialized cell types and tissues (Watt & Hogan, 2000). Most adult organs retain a limited regenerative capacity which seems to depend on the stem cells reserves (which maintain self-renewal and pluripotency potential after mobilization signals; Bianco & Robey, 2001;Korbling & Estrov, 2003). Although stem cells have specialized characteristics which protect them from external insults, aging impacts on stem cell homeostasis, resulting in halted stem cell renewal and proliferation (Ermolaeva, Neri, Ori, & Rudolph, 2018;Goodell & Rando, 2015).
Stem cells experienced aging-dependent accumulation of DNA damage and telomere shortening (Flores & Blasco, 2010;Flores et al., 2008), directly impacting on stem cell function and ultimately on lifespan (Ruzankina et al., 2007;Vilas et al., 2018). Interestingly, at least some of the phenotypes of stem cell aging may be partially delayed. An example is the anti-aging effects of caloric restriction (Mazzoccoli, Tevy, Borghesan, Delle Vergini, & Vinciguerra, 2014).
Caloric restriction was shown to prolong the capacity of stem cells to self-renew, proliferate, differentiate, and replace cells in several adult tissues. Whether lncRNAs may be acting directly or indirectly on stem cell homeostasis and be potential novel targets for stem cell resistance to aging-induced processes has been recently come to stage Bernardes de Jesus et al., 2018;Li et al., 2017;Ramos et al., 2013).  Nigro et al., 1991;Rong et al., 2017). circRNAs could be classified into several subtypes depending on their positioning relatively to the parental linear transcript or from the integration of 1 or multiple introns and/or exons (Qu et al., 2017;Westholm et al., 2014;Zhang, Wang, et al., 2014;. (b) Expression profiles of different RNA species during senescence of human skin fibroblasts. Previously released RNA-seq data from human wt and senescent WI-38 human cells (Chen et al., 2012;Marthandan et al., 2015) were analyzed with ISAT2(v2. Adult stem cells are a rare population of undifferentiated cells capable of self-renewal and to differentiate into lineage-specific tissues usually within the niche they reside (Dulak, Szade, Szade, Nowak, & Jozkowicz, 2015). Adult stem cells replace damaged cells due to tissue turnover or injury. High turnover organs are known to be populated by adult stem cells, although it is believed several adult tissues retain populations of adult stem cells even in the absence of detectable proliferation (Dulak et al., 2015). Well-characterized examples of high turnover tissues are the intestine, blood, or muscle.

Intron
Here, adult stem cells play crucial roles in tissue homeostasis (Wagers & Weissman, 2004). During the lifespan of a person, adult stem cells also age, being this concomitant with a decline in their properties (Ahmed, Sheng, Wasnik, Baylink, & Lau, 2017). Aging affects mostly, but not only, high turnover tissues such as the bone marrow-derived mesenchymal stem cells and subsequently the hematopoietic stem cells (HSCs), the skeletal muscle, or the intestine.
Whether lncRNAs play a role in adult stem cell aging remains to be fully demonstrated. Hereafter, we will describe the role of known lncRNAs in adult stem cells and their potential correlation with the aging process in distinctive tissues.

| lncRNAs in adult skeletal muscle stem cells
Adult skeletal muscle retains partial capacity to regenerate ( MyoG (Yu et al., 2017). The lncRNA MUNC targets RNAs such as myogenin and Myh3 involved in myogenic differentiation (Mueller et al., 2015). Lnc-mg is specifically enriched in skeletal muscle and is essential for muscle cell differentiation and skeletal muscle development . Lastly, Dimartino et al show that lnc-31, a lncRNA required for myoblast proliferation, stabilizes the YB-1 factor, allowing its positive effect on Rock1 mRNA translation (Dimartino et al., 2018; see Table 1).
Other muscle-specific lncRNAs include the LincMD1, which controls muscle differentiation by acting as a competitive endogenous RNA (ceRNA) of miR-133 and miR-135 regulating the expression of MAML1 and MEF2C (Cesana et al., 2011). Overexpression of linc-MD1 correlates with the anticipation of the muscle differentiation program. Although they proved involvement in muscle regeneration programs, the correlation of muscle lncRNAs with the aging process is still missing. Recently, a novel lncRNA (Chronos) has been identified in aged muscle (Neppl, Wu, & Walsh, 2017). Chronos is regulating the process leading to the gradual loss of muscle mass occurring with advancing age. Chronos is positively regulated with age. Inhibition of Chronos induces hypertrophy of the muscle through the modulation of Bmp7 signaling (Neppl et al., 2017).  (Challen et al., 2011). They focused on two lncRNAs, LncHSC-1 and LncHSC-2, which are highly expressed in WT HSC, but absent in Dnmt3a KO HSCs. Additionally, they also identified a small subset of lncRNAs (29 out of 159) with altered expression between 4mo and 24mo HSCs. Surprisingly, the lncRNAs whose expression was changed with aging were not characterized (Luo et al., 2015).

| LncRNAs and HSC
Whether the aging-related lncRNAs may play a similar role in increasing colony formation in the context of aging is currently unknown ( Figure 1c). Recently, Delás and colleagues characterized a subset of mouse lncRNAs with potentially relevant expression during hematopoietic differentiation. Among the candidates was identified one lncRNA, Spehd, which silencing lead to myeloid progenitors deficiency in their oxidative phosphorylation pathway (Delás et al., 2018). With the increasing interest in lncRNAs and the advent of novel technologies, we believe the future will bring major findings on the biology of lncRNAs on HSC dynamics during aging.

| Gut
The gut epithelium is a self-renewing tissue dependent on an intricate process including mobilization, proliferation and differentiation of basal stem cells. The fast division and mobilization of novel cells need to be counterbalanced by a well-regulated apoptotic process . This balance is regulated by internal and external cues. Disruption of the gut epithelial may occur in patients with serious diseases, leading to the passage of toxic substances to the blood. Similarly to other genotoxic signals, aging leads to a severe change in the gut homeostasis . In Drosophila, aging results in an increased number and proliferation of dysfunctional stem cells (Moorefield et al., 2017;Tran & Greenwood-Van Meerveld, 2013). In mammals, aging is associated with decreased intestinal barrier function (Tran & Greenwood-Van Meerveld, 2013) and impaired nutrient absorption (Holt, 2007). Xiao and colleagues described the expression patterns of T-UCRs in the intestinal epithelium (Xiao et al., 2018). T-UCRs exhibited distinct dynamics after food starvation. Here, T-UCR uc.173 stimulated growth of the small intestinal mucosa. Due to the conservation observed by this class of transcripts, these findings may provide a venue for therapeutic strategies stimulating the regeneration of the intestinal mucosa such as during aging (Xiao et al., 2018). Other lncRNAs participating in the gut biology are the lncRNA H19 and the lncRNA SPRY4-IT1. H19 is a conserved lncRNA transcribed from the imprinted H19/Igf2 gene cluster. H19 is highly expressed during embryogenesis, but its levels decrease during aging (Fu et al., 2008).
H19 is a molecular sponge or bind to different miRNAs (Kallen et al., 2013). 2007) protects the gut barrier in mice exposed to external stresses.
Interestingly, mucosal SPRY4-IT1 levels decrease in patients diagnosed with increased gut permeability (IGP) comparing to normalmucosal samples from controls . SPRY4-IT1 levels correlate with repressed levels of tight junctions guiding to the potential role for this lncRNA in reverting altered mucosa phenotypes . Manipulation of these lncRNAs may prove beneficial for age-dependent gut loss of homeostasis.

CELLULAR R EPROGRAMMING-A ROLE FOR LNCRNAS?
Several alternatives in vitro methodologies have been optimized for the reprogramming and/or expansion of embryonic-like stem cells  (Li et al., 2010;Samavarchi-Tehrani et al., 2010). Indeed, expression of Zeb2 (Beltran et al., 2008;Wang, Guo, et al., 2013), an EMT factor, is shown to increase with aging and to be a barrier for cellular reprogramming

| LncRNAs as part of the stem cell network
LncRNAs have long been associated with cellular stemness (Loewer et al., 2010) with more than 100 lncRNAs known to bind to pluripotency transcription factors (Sheik Mohamed, Gaughwin, Lim, Robson, & Lipovich, 2010). Several lncRNAs showed direct involvement in the maintenance of pluripotency, regulating directly the levels of transcription factors (TFs), or participating in the reprogramming process (Guttman et al., 2011;Loewer et al., 2010). The synergy between lncRNAs and stemness is further confirmed by the direct association of pluripotency TFs, such as Oct4, Sox2, or Nanog to lncRNAs promoters, suggesting a direct regulation of lncRNAs levels in cell reprogramming and stemness preservation (Loewer et al., 2010). One example is the lncRNA-RoR which was shown to participate in the reprogramming conversion (Wang, Xu, et al., 2013). lncRoR works as a miRNA sponge, protecting pluripotency TFs from miRNA targeting. A pluripotency candidate directly regulated by lncRNAs is the oncogene c-Myc.
Although it was traditionally associated with cancer (Dang, 2012)

| lncRNAs and epigenetic rewiring during reprogramming
Before the discovery of the extensive non-coding transcription across the genome from high-throughput studies, lncRNAs were long known to be players in the epigenetic processes of X-chromosome inactivation (XCI) and genomic imprinting (Lee & Bartolomei, 2013). Genomic imprinting is an epigenetic phenomenon that renders a subset of genes to be mono-allelically expressed according to their parental origin (Barlow & Bartolomei, 2014). These genes are frequently located in the same genomic regions, commonly known as imprinted clusters, an organization implying a common mechanism of imprinting regulation.
Indeed, all imprinted clusters have cis-acting imprinting control regions (ICRs) which are epigenetically differentially marked by DNA CpG methylation in the two parental alleles. Interestingly, most imprinted clusters have at least one lncRNA which is mono-allelically expressed and regulated by CpG DNA methylation. These lncRNAs can be intergenic or antisense to reciprocally imprinted genes. They are believed to regulate imprinted expression of the neighboring genes through the act of transcription itself or by the recruitment of chromatin-modifying complexes, as has been referred to the cases of Airnc, Kncq1ot1, and Meg3 lncRNAs (Kaneko, Son, Bonasio, Shen, & Reinberg, 2014;Latos et al., 2012;Nagano et al., 2008;Terranova et al., 2008). Such studies paved the way for the investigation of the role of many lncRNAs and their link with the epigenetic machinery namely methylating/demethylating enzymes and chromatin-modifying complexes for instance (Quinn et al., 2016). Epigenetically related lncRNAs may be involved in the aging process. For example, Xist lncRNA is known to become downregulated during senescence in vitro (Abdelmohsen et al., 2013).
Recent genome-wide studies clearly pointed out for an epigenetic clock in both mouse and human tissues based on aging-related DNA methylation changes (Hannum et al., 2013;Horvath, 2013;Stubbs et al., 2017;Weidner et al., 2014). Indeed, a DNA methylation signature of aging was uncovered and capable of predicting the chronological age and functional decline of a given tissue (Horvath, 2013;Stubbs et al., 2017). Whether such epigenetic changes are a cause or a consequence of the aging process still needs to be uncovered.
During iPSC reprogramming, a massive epigenetic rewiring of the differentiated program into the stem-like state occurs in a short time window. Aged cells encounter an extra layer of epigenetic rewiring since it requires not only an epigenetic resetting of the donor cell memory, but also of their aging-specific characteristics (Hochedlinger & Plath, 2009;Mertens et al., 2015). This might explain their decreased efficiency in reprogramming, as elucidated clearly from mouse cells studies (Mahmoudi & Brunet, 2012). For instance, while full reversal of aging-specific epigenetic features is believed to occur (Mertens et al., 2015), some might persist (Lo Sardo et al., 2017), which might hinder the pluripotency capacity and quality of iPSCs derived from aged donor cells.
Another aspect is that epigenetic-sensitive loci such as imprinted regions could be deregulated during this process. Indeed, imprinting errors have been documented in both mouse and human iPSCs (Ma et al., 2014;Nazor et al., 2012;Stadtfeld et al., 2010;Sun et al., 2012), giving rise to inappropriate silencing or biallelic expression of imprinted genes, including imprinted lncRNAs. In particular, these errors are recurrent at the Dlk1-Dio3 imprinted cluster, where hypermethylation leads to the loss of expression of several imprinted noncoding transcripts including the Meg3 and Meg8 lncRNAs (Ma et al., 2014;Stadtfeld et al., 2010). As a consequence, these iPSCs lose their pluripotency hallmarks. Indeed, Meg3 OFF mouse hiPSCs contribute poorly in chimeric mice and fail to generate "all-iPSC" mice, the most stringent pluripotent test Liu et al., 2010;Stadtfeld et al., 2010). Likewise, MEG3 OFF human iPSCs fail to differentiate properly down the neuronal lineage (Mo et al., 2015).
These results indicate a major role for Dlk1-Dio3 imprinting in pluripotency and suggest the involvement of imprinted lncRNAs in determining the full developmental potential of iPSCs. Whether this stochastic epigenetic errors affecting imprinting during the inefficient process of iPSC reprogramming of aged cells is exacerbated and whether they can explain, to some extent, their reduced inability to become iPSCs and the role of imprinted lncRNAs on these processes are interesting areas of research to follow.

| MET transition during reprogramming of aged cells
A mesenchymal-to-epithelial transition (MET) is the first important decision that cells undergoing reprogramming need to overcome (Sancho-Martinez & Izpisua Belmonte, 2013), especially if using the favorite mesenchymal-derived dermal fibroblasts as donor cells (Li et al., 2010). Importantly, forced expression of E-cadherin (epithelial marker) (Redmer et al., 2011) or downregulation of Zeb2, which facilitates MET transition, augmentes the efficiency of reprogramming (Wang, Guo, et al., 2013). Whether MET may be delayed during reprogramming of aged iPSCs and act as an aging barrier for reprogramming has been recently unveiled by us (Bernardes de Jesus et al., 2018). Moreover, we identified a lncRNA, called Zeb2-NAT, a natural antisense transcript of Zeb2, as a molecular target to improve reprogramming of aged cells. (Mattick, 2010;Mercer & Mattick, 2013;. NATs are a particular group with very interesting characteristics, in particular due to its antisense transcription with potential regulatory role of the sense protein-coding genes (Beltran et al., 2008;Bernardes de Jesus et al., 2018;Matsui et al., 2008;Wang, Chung, et al., 2014;Zong et al., 2016). This might be a common regulatory module, since according to recent studies, 72% of mice and human genomic loci are transcribed from both sense and antisense strands (Werner, Carlile, & Swan, 2009 | 7 of 14 the functional involvement of antisense transcription in cellular reprogramming and aging. Overall, antisense transcription could act locally, interfering in the functional levels of the sense transcript, or as regulatory hubs responsible for the dispersion of regulatory signals to neighboring genes (Pelechano & Steinmetz, 2013). Whether both sense and antisense transcription may be expressed in the same cell, or at the same time, remains to be elucidated. The importance of divergent transcription, as observed in sense-antisense transcription pairs, has been recently assessed by Lou and colleagues who elegantly linked divergent RNAs to cell lineage commitment (Luo et al., 2016). Divergent lncRNAs are shown to be relatively abundant, to co-localize and to be co-express with developmental and transcription regulator genes and to be associated with epigenetic marks involved in differentiation regulatory networks (Luo et al., 2016). Zeb2-NAT lncRNA is an example of a lncRNA more expressed in aged cells whose modulation of expression can improve iPSC reprogramming from aged cells (Bernardes de Jesus et al., 2018). It is likely that other lncRNAs might exist with similar attributes which might be revealed by highly sensitive transcriptome studies such as the novel native elongating transcript sequencing technology (mNET-seq), which generates single-nucleotide resolution (Nojima et al., 2015) or global run-on sequencing (GRO-seq) (Core, Waterfall, & Lis, 2008) among other high-resolution techniques.

| LNCRNAS AS AN TI-AGING THE RAPIES
As mentioned before, lncRNAs are emerging as potential targets for anti-aging therapies. Their non-coding nature and particularities (such as the conformational complexity, cellular localization, or interactions) need to be taken into consideration in the design of strategies for efficient lncRNA modulation.
Modified oligonucleotides are probably the best characterized and known approach to target lncRNAs. Antisense oligonucleotides have been traditionally used as a research tool to explore function of several lncRNAs in vitro and in vivo. More recently, novel oligonucleotides harboring RNA or DNA recognition and cleavage domains have shown up as potential novel strategies with an increased specificity and stability to target lncRNAs independently of their cellular compartmentalization (Bhartiya et al., 2012;Jadhav, Scaria, & Maiti, 2009;Lennox & Behlke, 2016;Suryawanshi et al., 2012), in particular when including base modifications such as locked nucleic acids (LNA). Regarding in vivo strategies, nowadays, there are still issues at the level of delivery and targeting due to the fact that different oligonucleotides work in a cell and tissue-specific manner. Moreover, the route of delivery is sometimes inefficient and could lead to offtargets. Nevertheless, designed catalytic oligonucleotides harboring base modifications for stability and specificity against lncRNAs are still one of the best strategies to reach satisfactory down-regulated levels of mature lncRNAs avoiding genetic modifications. Examples include the strategy employed to target Angelman syndrome in mice (Meng et al., 2015). Angelman syndrome is caused by maternal deficiency of UBE3A, with the paternal copy of UBE3A being silenced by a lncRNA named UBE3A-ATS (Tan & Bird, 2016). Targeting of the mouse Ube3a-ATS with antisense oligonucleotides (ASOs) ameliorated some cognitive deficits associated with the disease (Meng et al., 2015). Whether the same strategy could be used in humans is still unknown. Another example is SAMMSON, a lncRNA linked to melanogenesis. Targeting SAMMSON through intravenous delivery of ASO in a human xenograft model significantly reduced tumor growth and cell proliferation (Leucci et al., 2016;Matsui & Corey, 2017).
Additionally, modified antisense oligonucleotides have been used to effectively treat human conditions such as hypercholesterolemia and inflammatory bowel disease (Marafini et al., 2015;Toth, 2013). The use of modified antisense oligonucleotides in both neuromuscular and neurodegenerative diseases with a monogenic cause has recently advanced to clinical trials (e.g., Duchene muscular dystrophy; Koo & Wood, 2013;Wilton & Fletcher, 2005).

| Future directions
Rapid advances in genome sequencing have placed long non-coding transcripts as a major player in gene regulation. In this review, we placed the current knowledge on the potential roles of lncRNAs in stemness related to aging. On one hand, we discuss functional roles of lncRNAs in stem cell pools during aging and, secondly, their impact on cellular reprogramming of aged cells. We believe that in the near future, functional tests will undoubtedly uncover anti-aging therapeutic approaches relying on targeting of lncRNAs. We expect many surprises to come, where a complex trait such as aging could be seen at the light of the non-coding transcriptome.

ACKNOWLEDG MENTS
We thank members of the Carmo-Fonseca laboratory for insightful discussions and advice.

CONFLI CT OF INTEREST
None declared.