|Home | About | Journals | Submit | Contact Us | Français|
Non-coding RNAs have been found to regulate many cellular processes and thus expand the functional genetic repertoire contained within the genome. With the recent advent of genomic tools, it is now evident that these RNA molecules play central regulatory roles in many transcriptional programs. Here we discuss how they are targeted to promoters in several cases and how they operate at specific points in the transcription cycle to precisely control gene expression.
The role of RNA in gene expression has seen a paradigm shift in recent years. Beyond serving as a passive messenger of the genetic material, RNA itself can act as a non-protein coding (ncRNA) regulatory molecule. In addition to their established functions in splicing and translation, it is now clear that ncRNAs are central players in transcriptional control. Over the past few years, we have come to understand that at least half of the mammalian genome is transcribed, but only about 1–2% encodes proteins, and that up to 70% of protein-coding genes are thought to be transcribed bidirectionally,1-5 a phenomenon known as pervasive transcription.6,7 Previously regarded as “junk DNA,” many of these loci generate regulatory ncRNAs, including a novel class of long (> 150 nt and up to 100 kb) non-coding RNAs (lncRNAs), distinct from the well-characterized small (19–23-nt) siRNAs and microRNAs involved in RNA interference and post-transcriptional regulatory pathways.8-10 Due to their common intergenic location, many were also referred to as lincRNAs for large intergenic ncRNAs that regulate local chromatin activity.11 For the purpose of discussion here, the term lncRNA will be used to describe examples of long regulatory RNAs and their function in the transcription cycle irrespective of their genomic location.
The wide diversity of these lncRNAs is commensurate with their diverse roles in gene regulation, ranging from the control of metabolic pathways to cell fate determination and development.12-14 These lncRNAs serve as molecular scaffolds to guide the recruitment or regulate the activity of RNA-protein complexes to control transcription circuits. They can silence or activate gene expression locally (in cis) by acting on proximally transcribed protein-coding genes, or globally (in trans) acting at long-range and affecting a locus at a distance.12,15-17 Here we discuss selected examples of cis- and trans-acting mammalian lncRNAs that function at defined steps of the transcription cycle, including chromatin modification and remodeling, DNA looping and regulation of higher-order structures, assembly of transcription complexes at the promoter, transcription initiation, and elongation (Fig. 1). We particularly focus on chromatin-associated ncRNAs and what is known about their targeting to transcription complexes. Other examples of control by lncRNAs, including ones that do not act on transcription, such as imprinting of inactive X chromosomes by XIST RNA, are described in recent reviews.13,14,18-23 In addition to the newly identified lncRNAs found by genomic methods, other previously known ncRNAs have now been shown to regulate transcription (e.g., U1, Alu, B2, and 7SK) and will be discussed.
The state of chromatin is a key determinant of transcriptional activity, including nucleosome structure and the ensemble of histone modifications that mark regions of the genome as active or silent. Histone post-translational modifications regulate biological processes either by altering chromatin structure (by loosening or compacting the DNA-histone interactions) or by contributing to the recruitment of regulatory factors. For example, genes actively transcribed by RNA polymerase (RNAP) II are marked by H3 lysine 4 methylation (H3K4me2/3) at their promoters and H3 lysine 36 trimethylation (H3K36me3) throughout the gene body, while repressed genes are marked by H3 lysine 27 methylation (H3K27me2/3) or H3 lysine 9 methylation (H3K9me2/3) at promoters.11,24-27 H3K9me2/3 and H3K27me2/3 marks result in compact chromatin around the transcription start site (TSS) of transcriptionally inactive genes, whereas H3K4me3 marks result in a more open, transcriptionally active chromatin conformation. Different histone modifications in combination can form complex networks, which are essential for regulating gene expression in a spatial and temporal manner. Recent studies highlight several cases in which lncRNAs influence the status of chromatin.
Transcription control is essential during development because it requires coordinated expression of neighboring genes through a process termed locus control. In mice and humans, 39 Hox genes encoding homeo domain transcription factors are clustered in four loci (A, B, C, and D) and expressed in a complex pattern. Seminal studies by the Chang lab28 characterized the transcriptional pattern of the four human HOX loci revealing that most of the transcribed region maps to intergenic, and not exonic, domains. This analysis led to the identification of 231 lncRNAs, some as long as 30 kb. The expression of these HOX lncRNAs demarcates broad chromosomal domains of differential histone methylation patterns and RNAP II occupancy. One of these RNAs transcribed in the HOXC locus is HOTAIR (HOX Antisense Intergenic RNA), a 2.2 kb lncRNA that demarcates active and silent chromatin domains and represses transcription in trans across the 40 kb HOXD locus (Fig. 1A). Hox genes are silenced by the Polycomb group (PcG) proteins, which regulate chromatin structure, in part through post-translational histone modifications. The core Polycomb Repressive Complex 2 (PRC2) comprises four components: the histone methyltransferase EZH1/2, SUZ12, EED and RbAp46/48 (or RBBP7/4), and also several other polypeptides: AEBP2, PCLs and JARID2.29 The EZH1 and EZH2 enzymatic subunits of PRC2 methylate H3K27 (H3K27me2/3), which is essential for long-term gene silencing. HOTAIR binds EZH2 and is expressed at the border of an adjacent repressive chromatin domain, which is enriched for H3K27me3 and SUZ12. Interestingly, knockdown of HOTAIR revealed that it maintains the repressed domain at the HOXD locus, but not the A, B, or C loci (Fig. 1A). It remains to be determined if one or all PRC2 subunits are required for HOTAIR activity, and if gene-specific complexes are formed at other locations.30 H3K27 methylation by PRC2 also signals PRC1 recruitment and H2A monoubiquitination (H2AK119ub), which represses transcription by interfering with RNAP II elongation,31 but it is unclear if this is a general mechanism since some genes are targeted by PRC1 and not PRC2. It is clear, however, that both PRC complexes are required to maintain gene repression.
More than 3000 lncRNAs were found to interact physically with either PRC2 or another repressive chromatin-modifying complex termed CoREST.32 CoREST was identified initially as a co-repressor for the silencing factor REST (also called NRSF), which represses neuronal genes in non-neuronal cells. The CoREST complex also contains HDACs 1 and 2, providing a mechanism by which CoREST can mediate silencing through histone deacetylation. Interestingly, PRC2 and CoREST share 40% of interacting lncRNAs, including HOTAIR, suggesting that they may act together to precisely regulate gene expression at specific genomic loci.32 HOTAIR possesses extensive secondary structure33 and acts as a scaffold for EZH2/PRC2, which binds to a 5′ region of the RNA, and LSD1, an H3K4me2 demethylase component of CoREST that binds to a 3′ region.33 Thus, HOTAIR functions to target these two histone modification complexes to chromatin to remove an active H3K4me2 mark, while methylating H3K27 to favor formation of a repressive environment (Fig. 1A). Interestingly, recent chromatin isolation by RNA purification (ChIRP) analyses revealed that HOTAIR occupancy occurs independently of EZH2, suggesting that RNA can guide PRC2 recruitment and specify formation and spread of a repressive environment.30 Additionally, sites of HOTAIR occupancy are significantly enriched for genes, which become de-repressed upon endogenous HOTAIR knockdown.30,33,34
HOTTIP (HOXA transcript at the distal tip) is a 3764-nt lncRNA transcribed at the 5′ end of the HOXA locus, which coordinates activation of 11 HoxA genes (Fig. 1A). HOTTIP appears to act in cis, based on its proximity to its target genes and the distance-dependence of HOXA target gene activation.17 Unlike PRC2-mediated recruitment by HOTAIR, HOTTIP recruits WDR5, a subunit of the MLL1 H3K4 methylation complex, to promoters. Chromosome conformation capture carbon copy (5C) experiments, which define physical chromatin interactions, revealed that actively expressed HOXA regions have compact chromosomal loops that facilitate interactions between these loci.17 HOTTIP knockdown does not affect this structure but is required for MLL1 occupancy and H3K4 methylation. Because HOTTIP remains close to its site of synthesis, it is able to transmit information from these higher order chromatin structures into epigenetic modifications to coordinate long-range gene expression.17 The examples of HOTTIP and HOTAIR show that each regulates HOX loci through chromatin-remodeling complexes but how they are specifically recruited to their respective DNA loci is currently unclear.
The mouse HOXA locus expresses a 798-nt unspliced, polyadenylated lncRNA between the Hoxa6 and Hoxa7 genes called Mistral, which was discovered by RNA-chromatin immunoprecipitation of MLL1.35 In vitro analysis revealed a direct interaction between the MLL1 SET domain and an RNA hairpin loop in the 3′ region of Mistral. RNAi-mediated knockdown of Mistral significantly diminished MLL1 enrichment in the Hoxa6/a7 region as assessed by ChIP and also inhibited the transcription of Mistral, Hoxa6, and Hoxa7 but no other genes in the HOXA locus. Thus, it appears that localized recruitment of Mistral activates transcription of its adjacent genes through the recruitment of the same histone methyltransferase complex as HOTTIP (Fig. 1A) and presumably through changes in chromatin conformation. Like the previous cases, it remains unclear how Mistral is specifically targeted to its DNA loci.
In an attempt to uncover cellular factors that modulate the histone acetyltransferase (HAT) activity of CREB-binding protein (CBP)/p300/KAT2B, the RNA-binding protein TLS (translocated in liposarcoma) was found associated with CBP by mass spectrometry and shown to be an RNA-dependent HAT inhibitor.36 The N-terminal portion of TLS, but not the full-length protein, is a potent inhibitor. Interestingly, addition of RNA oligonucleotides stimulated inhibition by full-length TLS and also enhanced its proteolysis, suggesting that RNA allosterically activates the HAT inhibitory activity. One specific target gene of TLS-mediated inhibition of CBP is the cyclin D1 gene (CCND1), which has several 200–330 nt RNAs transcribed upstream, referred to as CCND1 lncRNAs (Fig. 1A). These regulatory RNAs were found to bind TLS and all are targeted to the CCND1 promoter. The direct requirement of these lncRNAs at the CCND1 locus was confirmed by knockdown experiments, showing decreased TLS occupancy at the promoter and increased levels of H3 acetylation (H3K9-K14Ac), a target of CBP/p300, and of CCND1 transcripts. Thus, it appears that these locally synthesized lncRNAs are induced in response to DNA damage and are tethered to 5′ regulatory regions of the CCND1 promoter to recruit TLS, which controls CBP/p300 HAT activity to cause gene-specific silencing.36
The localized recruitment of lncRNAs to promoters along with chromatin-modifying complexes is emerging as a common theme. Air and Kcnq1ot1 are two additional examples.37-39 Air, a 108 kb lncRNA, regulates genomic imprinting of a cluster of autosomal genes on mouse chromosome 17, and is required for allele-specific silencing of the cis-linked Slc22a3, Slc22a2 and Igf2r genes in mouse placenta. Mechanistically, Air represses Slc22a3 transcription by interacting with its promoter and recruiting the H3K9 histone methyltransferase G9a to dictate localized H3K9 methylation (H3K9me3).38 Likewise, Kcnq1ot1, a 91 kb lncRNA that controls bidirectional silencing of genes in the Kcnq1 domain in placenta, also interacts with G9a to direct H3K9 methylation and, in addition, interacts with PRC2 to direct H3K27 methylation.39 Similarly to HOTAIR (Fig. 1A), Kcnq1ot1 scaffolds two distinct chromatin-modifying complexes to direct transcriptional repression across a locus. Again, the basis for target specificity remains to be determined.
Oct4 is a transcription factor essential for maintaining the pluripotent state, and the epigenetic silencing of its gene is an important step in differentiation. In particular, the methylation state of H3K9 at the Oct4 promoter, mediated by G9a, is a driving factor in differentiation and reprogramming. A recent report describes an antisense non-polyadenylated Oct4 lncRNA, named asOct4, which originates from a pseudogene.40 The model proposes that asOct4 functions in cis by associating with the Oct4 promoter and recruiting the regulatory complexes PRC2 (EZH2 subunit) and G9a to prevent Oct4 transcription, thus pointing to a role in regulating pluripotency. The specificity of recruitment of this antisense lncRNA remains unknown but it is possible that it forms a DNA-RNA hybrid with the Oct4 promoter. Interestingly, this repressive mechanism is counteracted by the nucleic acid-binding protein PURA (purine rich element binding protein A), which interacts with asOct4 and prevents its promoter localization, ultimately upregulating Oct4 gene expression. Because many lncRNAs are generated antisense to protein-coding genes,1,2,41,42 it will be intriguing whether they, too, function to regulate gene transcription by directing epigenetic remodeling complexes to particular loci.
Some lncRNAs are located further than 1 kb from known protein-coding genes and may function as enhancers in cis to regulate neighboring genes.15,16,43 Such lncRNAs are marked with chromatin signatures characteristic of transcribed genes, such as H3K4me3 at the 5′-end and H3K36me3 downstream, and not marks that characterize enhancer elements, such as H3K4 monomethylation.11,44,45 Knockout of these lncRNAs decreases neighboring gene expression,15 where they may function as co-activators to recruit positive-acting factors or form higher order chromatin structures through DNA looping.46-48
Chromatin insulators are DNA elements that can protect a gene from outside influences, which might lead to either inappropriate activation or silencing of the gene. CTCF (CCCTC-binding factor) is a zinc-finger protein required for transcriptional insulation and accomplishes this via long-range physical interactions with CTCF sites and cohesin.49 CTCF-binding sites are widely distributed throughout the genome and function via CTCF and cohesin to either separate or bring together distant regulatory elements.24,49,50 For example, cohesin enables CTCF to insulate promoters from distant enhancers and control transcription at the IGF2/H19 (insulin-like growth factor 2) imprinted control region (ICR) (Fig. 1B). A recent report shows a role for a ~700-nt lncRNA, SRA (steroid receptor RNA activator), in CTCF-mediated insulation.51 SRA, which is reported to act as both a transcriptional co-activator and co-repressor,52-54 is a chromatin-associated lncRNA found in a complex with the DEAD-box RNA helicase p68 (DDX5) and CTCF (Fig. 1B). p68/DDX5 was detected at CTCF sites in the IGF2/H19 locus, and depletion of p68 or SRA reduced CTCF-mediated insulator activity which ultimately increased levels of IGF2.51 Although p68 and SRA depletion does not affect CTCF recruitment to its genomic sites, it does reduce cohesin binding, implicating the p68/SRA protein-RNA complex in stabilizing the interaction of cohesin with CTCF. This stabilization may function as a regulated step in transcription insulation. Thus, lncRNAs are also required for proper insulator function to shield a locus from the effects of flanking chromatin domains.
In eukaryotes, the general transcription factors (GTFs) TFIIA, TFIIB, TFIID, TFIIE, TFIIF, and TFIIH, and RNAP II assemble at promoters into pre-initiation complexes (PICs) to specify the TSS.55 PIC formation usually begins with TFIID binding to the TATA box, initiator, or downstream promoter element found in most core promoters, followed by the entry of other GTFs and RNAP II through either sequential assembly or a preassembled RNAP II holoenzyme pathway. For activator-dependent (or regulated) transcription, additional cofactors typically are required to transmit regulatory signals between gene-specific activators and the general transcription machinery.55,56 In addition to protein components, a number of lncRNAs are known to influence these early steps in transcription.
Alternative promoters for the same gene are a common phenomenon in gene regulation. The human DHFR gene contains two promoters, minor and major, with the major promoter producing 99% of the transcribed RNA. Seminal studies have shown that the transcript generated from the upstream minor promoter can impede the formation of PICs on the major promoter in vitro (Fig. 1C).57 In quiescent cells, the cell-cycle-regulated DHFR gene is repressed in a manner that correlates with the expression of an ~400-nt lncRNA, referred to as DHFR lncRNA, from the upstream minor promoter, which shuts off transcription from the major promoter. This lncRNA was found to directly bind TFIIB in vitro and reduce its occupancy on the major promoter in vivo,58 suggesting a model whereby the DHFR lncRNA represses transcription in cis by preventing PIC formation (Fig. 1C). Interestingly, DHFR lncRNA can inhibit transcription only when its sequence extends into the major promoter sequence, where it has been shown to form a triplex DNA-DNA-RNA structure in vitro. This mechanism is similar to the formation of a DNA:RNA triplex by pRNA (promoter-associated RNA) at the rDNA promoter which signals the recruitment of the DNA methyltransferase DNMT3b to silence the transcription of rRNA genes.59 The pervasive transcription observed near annotated genes7 might act similarly to generate lncRNAs locally that regulate the assembly of nearby transcription complexes.
A number of short-interspersed elements (SINEs) are transcriptionally upregulated during heat shock, such as the 178-nt mouse B2 RNA and the ~350-nt human Alu RNA. The majority of RNAP II transcription is coordinately downregulated during heat shock, and thus it was postulated that these RNAP III-derived SINE transcripts might be responsible for broadly repressing transcription.60 Upon heat shock, B2 and Alu RNAs were found to specifically occupy the promoters of repressed genes in vivo along with RNAP II and GTFs (Fig. 1C).60 After binding to DNA, RNAP II switches from a closed to an open complex and this change involves the separation of the DNA strands to form an unwound section of DNA of approximately 13 bp, referred to as the transcription bubble. Both Alu and B2 RNAs were found to tightly bind RNAP II in vitro, and biochemical experiments showed that they prevent RNAP II from establishing contacts with the promoter both upstream and downstream of the TATA box during open complex formation (Fig. 1C). The results are consistent with a model in which their repression domains bind in the DNA cleft of RNAP II61,62 and repress TFIIH-mediated phosphorylation of RNAP II.60,63,64 A 50-nt fragment of B2 RNA (nts 81–131) is sufficient for its activity and includes an 18-nt single-stranded region flanked by two hairpins, both of which are required for repression.62,65 Altogether the model posits that these lncRNAs mediate transcriptional repression on the DNA template, and presumably are targeted to promoters as a consequence of RNAP II recruitment, but the mechanistic basis remains to be established.
In searching for ncRNAs that regulate transcription initiation, the ~160-nt U1 snRNA, a core splicing component, was found to associate with TFIIH.66 Stem-loop 2 of U1 RNA, and not the entire U1 snRNP, mediates an interaction with the cyclin H subunit of TFIIH.67 In contrast to the repressive action of the B2 and Alu RNAs, U1 enhances TFIIH-dependent transcription initiation in vitro from a RNAP II promoter (Fig. 1C). Reconstituted transcription in vitro demonstrates an increase in the rate of formation of the first phosphodiester bond by RNAP II in the presence of U1 RNA, suggesting that it regulates transcription in addition to its well-established role in RNA processing. While the mechanism is not yet clear, one model posits that U1 RNA is a component of a re-initiation scaffold anchored by proximal 5′ splice sites, potentially linking its roles in transcription control and splicing. It remains to be determined if U1 RNA binding to TFIIH and its subsequent effect on transcription is a regulated or constitutive process and which are the in vivo targets of this lncRNA.
Until recently, it was widely believed that recruitment of RNAP II and the early steps of initiation were rate-limiting for transcription at most promoters. This notion has been challenged by the finding that many cellular genes exhibit RNAP II occupancy downstream of the TSS, a phenomenon known as transcriptional pausing in which RNAP II is unable to escape into productive elongation.68-72 P-TEFb, the positive transcription elongation factor b (composed of cyclin T1 and Cdk9 subunits), has been studied extensively due to its role in facilitating RNAP II escape from this paused state globally.73,74 When recruited to promoters, P-TEFb phosphorylates the C-terminal domain (CTD) of RNAP II allowing escape into productive elongation.74,75 P-TEFb catalytic activity is tightly controlled in cells by regulating the equilibrium between two states: the active P-TEFb form, which can be recruited to chromatin by interacting with Brd4 and other factors, and an inactive ribonucleoprotein form, referred to as 7SK snRNP, containing a 331-nt RNA known as 7SK snRNA.76,77 RNase footprinting and mutagenesis indicate that 7SK contains a high degree of secondary structure, with stem-loops at both the 5′ and 3′ ends.78-80 5′ stem-loop binds P-TEFb as well as the Hexim1 protein, which acts to inhibit the kinase activity, while the 3′ stem-loop binds the Larp7/PIP7S protein, which, in addition to a methylphosphate capping enzyme (Mepce), stabilizes the RNA.76,79,81-86 Even though 7SK does not seem to tightly bind chromatin, recent findings link this RNA with the transcription machinery and localized regulation at promoters, implying a more dynamic role in its recruitment to target genes.
The HIV promoter is well known to rely on P-TEFb for transcription elongation. The virally encoded Tat protein, which binds to a 5′ RNA stem-loop in HIV transcripts known as TAR (trans-activation response element) recruits P-TEFb to the viral promoter to activate the switch from transcription initiation to elongation. Another function of Tat is to dismantle the 7SK snRNP to increase the pool of P-TEFb and does so by binding to cyclin T1 and competing with Hexim1.87 Tat also binds to the 5′ stem-loop of 7SK, apparently triggering a conformational change in the RNA, but it is unclear if RNA-binding is required for P-TEFb displacement or if it occurs after P-TEFb and Hexim1 are removed from the 7SK snRNP.83,88-91 A prominent model is that Tat captures active P-TEFb from a large pool of inactive 7SK-bound P-TEFb complexes via this competition mechanism and then delivers it to the HIV promoter by binding to TAR.87 Recent ChIP analyses on an integrated HIV promoter indicate that the 7SK snRNP proteins Larp7 and Hexim1, along with cyclin T1 and Cdk9, directly occupy the promoter surrounding the TSS, suggesting that inactive P-TEFb complexes are assembled on the DNA template (Fig. 1D).91 The enrichment of the snRNP proteins is lost downstream of the TSS in a Tat- and TAR-dependent manner, suggesting a model in which the protein-RNA interaction fully displaces the inhibitory 7SK snRNP, thereby timing the transition into elongation. It is currently unclear if the prevailing mechanism of P-TEFb recruitment is in its 7SK snRNP-inhibited form or in its pre-activated state, or whether P-TEFb is also recruited without 7SK as part of another complex. Also, it is unknown if there is a hierarchical order of 7SK snRNP remodeling by Tat or if multiple independent Tat complexes exist with P-TEFb or with the 7SK snRNA, and how this relates to TAR binding. Different pools of Tat could extract P-TEFb from the pre-formed 7SK snRNP complex to transfer it to TAR, and could also bind to 7SK RNA to prevent transcription elongation shut off and favor multiple rounds of transcription elongation.
7SK snRNA is enriched in nuclear speckles, a subnuclear domain rich in pre-mRNA processing factors, from which it can be recruited to sites of active transcription. In situ hybridization experiments revealed that 7SK transiently associates with a stably integrated reporter gene within minutes of inducing transcriptional repression and displaces P-TEFb from the locus. An interesting model is that the 7SK snRNP is dynamically recruited from nuclear speckles to cellular promoters, where it can sequester or inhibit P-TEFb and thereby induce transcriptional elongation shut off. These results raise the possibility that 7SK snRNA may be either bound to a subset of cellular genes along with the transcription machinery or be rapidly recruited to promoters to shut off transcription in response to environmental cues. Recent reports suggest that 7SK snRNA may also be associated with other factors that in some cases are found at promoters, including a diverse set of hnRNP proteins (A1, R, Q, and K), cold shock domain protein A, which has an established RNA-binding activity, and an ~120 kDa protein of unknown function annotated as C9orf10.83,92-94 7SK can also function cooperatively with the chromatin regulator HMGA1 to control transcription in both P-TEFb-dependent and -independent modes.95,96 These cases exemplify a vast range of new mechanistic possibilities by which a regulatory RNA may control transcription by interacting with different protein complexes. It is possible that 7SK dynamically switches between chromatin-bound and -unbound states to regulate the genesis or termination of elongation programs. This would differentiate 7SK from other stably chromatin-bound lncRNAs such as HOTAIR that direct the establishment of silent environments by spreading repressive chromatin marks. Since P-TEFb globally regulates transcription elongation, it is reasonable to envisage that 7SK snRNA also functions to shut off transcription elongation programs genome-wide.
While our understanding of the mechanistic details is still at an early stage, it seems that lncRNAs can already be considered another class of transcription factor, along with chromatin modifiers, DNA-binding regulators, and other cofactors. Even though the model of lncRNAs acting as molecular scaffolds is tempting, only a few examples have been described. Additionally, the step of the transcription cycle targeted by certain lncRNAs is currently unclear. An example is the recent discovery of lincRNA-p21 and PANDA which function by assembling with hnRNPK and the heterotrimeric NF-Y factor, respectively, to regulate p53-dependent transcriptional responses.44,97 A common theme among the lncRNAs discussed here is that they function as chromatin-bound regulators. There are many examples of cis-acting lncRNAs, although recent discoveries point to others that act in trans, suggesting both local and global regulatory roles. However, little is known about how lncRNAs are targeted in trans to their loci. Many mechanisms are possible including targeting to specific sequences by Watson-Crick base pairing, the formation of DNA-DNA-RNA triple helical structures as in the case of DHFR and pRNA, or by contacting factors already bound to chromatin. HOTAIR exemplifies that the RNA itself could recruit and spread the chromatin-remodeling complexes to establish the epigenetic state at its target loci. A systematic and comprehensive definition of the assembly of protein factors on selected lncRNAs will be required to define whether these RNAs can form alternative complexes on or off their genomic targets to regulate gene expression programs. These early glimpses reveal that lncRNAs operate throughout the transcription cycle. Even more, they demonstrate ways in which RNA and protein can be functionally interchangeable and how the RNA world remains with us today.
We thank Cheng-Ming Chiang and Nicholas Conrad for their invaluable comments and suggestions. We apologize for the non-comprehensive nature of this review and to colleagues whose work could not be cited due to space constraints. This work was supported by NIH grants R00AI083087 (I.D.) and GM082250 (A.D.F.).
Previously published online: www.landesbioscience.com/journals/transcription/article/19349