The discovery of RNAa came as a surprise. In early 2004, our group was interested in how aberrant DNA methylation of promoter sequences was regulated in cancer cells. It was speculated that ncRNA could induce sequence-specific DNA methylation, a phenomenon that had been known to occur in plants for over 10 years (12
). At the time, our laboratory was investigating epigenetic mechanisms of gene silencing, including DNA hypermethylation in gene promoters. One particular gene of interest was E-cadherin, a tumor suppressor gene silenced in several types of cancers. The E-cadherin promoter contains a typical CpG island surrounding the transcription start site that, upon methylation, silences E-cadherin expression (13
). We sought to examine whether DNA methylation could be induced at the E-cadherin promoter by exposing cells to synthetic small dsRNAs, the known trigger for RNAi. By implementing rational siRNA (small interfering RNA) design rules (15
) against the E-cadherin promoter sequence, two high-scoring targets at sites just outside the CpG island were selected for testing. When dsRNA for either target was transfected into the prostate cancer PC-3 cell line, E-cadherin expression was surprisingly robustly upregulated instead of being downregulated (7
). This sparked subsequent identification of additional examples of RNAa (e.g. p21 and VEGF genes) (7
) (). Shortly thereafter, another group reported activation of the progesterone receptor (PR) and major vault protein (MVP) genes by dsRNA (8
), implying that the phenomenon could be a general mechanism of gene regulation.
1.1. Definition, observations, and features of RNAa
The term RNAa was originally used to describe gene activation at the transcriptional level mediated by small dsRNA designed to target gene promoter sequences (7
). These dsRNAs, termed ‘small activating RNAs’ (saRNAs) to distinguish them from siRNAs, exert an effect opposite to that of RNAi. Similar to the term RNAi, whose definition has been expanded to describe both transcriptional and posttranscriptional gene silencing mechanisms, RNAa may also be used to describe other related mechanisms by which small RNAs positively regulate gene expression and the epigenome, including transcriptional activation via
targeting the 3′ flanking regions of genes with small dsRNAs (16
), piwi-interacting RNA (piRNA)-mediated epigenetic activation (17
), miRNA-mediated translational activation (18
), and additional gene activation mechanisms yet to be discovered.
1.1.2. Promoter-targeting RNAa
In our initial studies, by testing 6 genes and designing 1 or 2 saRNA targets on the promoter of each gene, 3 of the tested genes (E-cadherin, p21, and VEGF) were activated by their respective saRNAs (7
). These saRNAs have a target size of 19 nt with 3′ dTdT overhangs – an identical structure to standard siRNAs. They targeted locations on gene promoters ranging from −200 to −700 relative to the transcription start site. Soon after, the Corey group reported that synthetic dsRNAs, termed antigene RNAs (agRNAs) to distinguish them from siRNAs that target mRNA, targeting the promoter regions of PR and MVP activated the expression of their respective target genes. These agRNAs also have a target size of 19 nt but they target sequences located on or near the transcription start site, ranging from −56 to −2. Interestingly, in the example of the PR gene, the agRNA PR11, which targets sequence −11/+8, robustly induced PR gene expression in the MCF7 breast cancer cell line, but suppressed PR expression in the T47D cell line which has higher basal expression of PR compared to MCF7 cells (8
). In these studies, RNAa was observed when the saRNA/agRNA was introduced into the cells at concentrations ranging from 5 to 50 nM (7
). Since these initial reports, additional examples of promoter-targeting RNAa have been identified ().
1.1.3. Non-promoter-targeting RNAa
Targeting genomic regions outside promoter sequences may also affect gene transcription both negatively and positively. Yue et al.
found that agRNAs targeting the 3′ terminal region of the PR gene caused either transcriptional gene silencing or activation by interacting with an overlapping noncoding sense transcript (16
). It has been proposed that the agRNAs trigger a looping mechanism bringing the 3′ terminus and promoter into close proximity to allow recruitment of additional proteins and modulation of promoter activity (16
). Likewise, targeting enhancer sequences may also exert some regulatory effect on gene activity as enhancers use a similar looping mechanism to activate transcription (20
1.1.4. Distinct kinetics of RNAa
An interesting feature associated with RNAa is its kinetics, which diverges from traditional RNAi. It is well known that RNAi can be induced within hours in mammalian cells, lasting for a period of ~5–7 days and subsequently disappearing when the exogenous siRNA is exhausted (21
). In contrast, RNAa takes a very different time course. Gene activation by transfecting promoter-targeting saRNA has been reported to be delayed by ~24–48 hours in comparison to RNAi. For example, in a side-by-side comparison, p21 activation by saRNA did not emerge until ~48 hours following saRNA transfection, while knockdown of two genes (MOF or E2F1) by siRNA was detectable as early as 6 hours (22
). Perhaps, the delay in RNAa activity reflects a more complex mechanism with additional rate-limiting steps. Because RNAa is a nuclear process acting on gene transcription, acquiring access to the nucleus may be one of the rate-limiting steps. Changes in chromatin structure associated with RNAa may also contribute to the delayed kinetics.
Another peculiar feature of RNAa is its prolonged effect. After a single saRNA transfection, gene activation can last for nearly two weeks in certain cell lines (7
). The duration of PR gene activation demonstrated by the Corey group (8
) and of E-cadherin and p21 induction observed by our group (7
) is remarkably similar, suggesting prolonged activation is a general feature of RNAa. It is reasonable to speculate that such long-lasting gene activation is due to epigenetic changes, which may persist well beyond the life of saRNA molecules. In support, several types of epigenetic changes have been associated with RNAa (see section 1.3.3).
1.1.5. Complementarity requirement of guide-target sequences for RNAa
Classic RNAi mediated by siRNA requires perfect complementarity between the siRNA and its cognate mRNA sequence. This complementarity is necessary for the RNA-induced silencing complex (RISC) to cleave the mRNA transcript at the siRNA recognition site. In contrast, translational repression mediated by miRNA can tolerate mismatches between non-seed sequence and its cognate mRNA (23
). Interestingly, promoter-targeting saRNAs share similar recognition features as miRNA: mismatches in regions outside the seed sequence are tolerable and retain partial RNAa activity (7
). Based on this observation, it is possible that RNAa may be triggered by endogenous miRNA with imperfect matches to promoter sequences.
1.2. Conservation of RNAa
RNAa was initially discovered in human cells (7
). By targeting the mouse VEGF promoter using short hairpin RNAs (shRNAs) via
lentiviral-mediated overexpresion, Turunen et al.
demonstrated RNAa in mouse cells in vivo
and in vitro
). RNAa has also been reported in several other mammalian species, including rat and non-human primates (11
). Because non-human primates share almost identical genome sequences with human, most saRNA targets designed on the promoters of human genes are well-conserved in non-human primates (11
) and these saRNAs can also activate their cognate genes in non-human primate cells (11
). The conservation of saRNA targets in primates suggests the possibility of validating RNAa-based therapeutics in non-human primate disease models, whereas human and rodent promoter sequences diverge significantly, saRNAs that are functional in rodents may not be active in human cells.
1.3. Mechanism of RNAa
Although the mechanism of RNAa remains largely elusive, recent studies have provided insight into several important aspects of RNAa.
1.3.1. Target molecules of RNAa
Unlike targeting single-stranded mRNA sequences, saRNAs/agRNAs designed for promoter regions may have different modes of strand preference: sense or antisense. At present, a universal model for RNAa is lacking and different mechanisms may exist in context with different promoters.
Sense or antisense target?
Whereas the trigger for RNAa is small duplex RNAs, evidence suggests that only one strand is required to guide activity. In a manner seemingly analogous to strand selection in the case of siRNAs, duplex saRNA is recruited by endogenous machinery in which one strand is discarded and the remaining strand facilitates target recognition via
complementary base-pairing to initiate RNAa. The question remains: which strand guides RNAa activity when targeting noncoding sequence? It has been demonstrated that activation of the PR and low-density lipoprotein receptor (LDLR) genes by targeting promoter sequences with agRNA is facilitated through interacting with a noncoding antisense transcript (i.e. synthesized in the opposite direction as target gene transcription), which runs through the promoter (24
). In these cases, the antisense transcript serves as a docking site for target recognition, implying that RNAa activity is mediated by the sense strand in the agRNA duplex. In another system, in which PR expression was activated by targeting the 3′ terminal region with agRNA, it was revealed that a noncoding sense transcript (i.e. synthesized in the same direction as target gene transcription) recruits agRNA (16
). Such evidence would imply that the antisense strand in this agRNA duplex guides gene activation. Overlapping noncoding RNAs transcribed in the sense and antisense orientations have already been shown to serve as docking sites for transcriptional gene silencing mediated by small duplex RNAs (26
). Likewise, such sense and antisense transcripts may serve as the targets for RNAa as well.
Though PR gene activation by targeting promoter sequence was facilitated by the sense strand of its agRNA duplex, other target genes may not have the same requirement. For instance, using chemical modifications to inactivate strand function selectively, we showed that the antisense strand in saRNA duplexes was responsible for RNAa activity in two examples targeting the p21 and E-cadherin promoters (22
). It is likely that strand function in saRNA/agRNA duplexes is dependent on the context of the gene and/or orientation of the targeted molecule. Utilizing modified saRNA/agRNAs with inactivated strands can define strand activity and assist in determining orientation of putative target noncoding transcripts.
RNA or DNA as target?
Recent studies have revealed that both sense and antisense noncoding transcripts are pervasive in the human genome (29
). At promoter sites, transcripts are also abundant, with some as short as ~20 nucleotides (31
). Some, referred to as nascent transcripts, have been found to be usually tethered to gene promoters (31
). Long noncoding RNAs (lncRNAs), as large as a few thousand nucleotides, can also overlap promoter regions. These promoter-overlapping transcripts may serve as binding sites of Ago-loaded agRNAs. For example, long noncoding antisense transcripts overlapping the PR or LDLR promoter are considered to be the target of agRNAs (24
) because their depletion by an RNase H-based mechanism renders the agRNAs nonfunctional (25
). The transcript–agRNA–Ago complex is thought to act as a scaffold for recruiting other proteins [e.g. hnRNP-k (heterogenous nuclear ribonucleoprotein-k), HP1 (heterochromatin protein 1γ), etc.] to effect gene regulation by enhancing or reducing association of RNA polymerase II (RNAP II) (25
). Interestingly, the abundance of these promoter-overlapping transcripts is not affected by agRNA targeting, implying that an RNAi cleavage mechanism is not involved (34
RNA has long been known to hybridize to double-stranded DNA (35
) with high stability comparable to that of protein-DNA complexes (37
). It has been hypothesized that RNA hybridization with the non-template strand of the promoter could expose the template strand for RNAP II binding (36
). In plants, RNA-directed DNA methylation (RdDM) only occurs to the cytosines along RNA-DNA duplexes, indicating a direct RNA-DNA interaction that provides a strong and specific signal for de novo
DNA methylation (38
). Thus, it is possible that, with the help of an Ago protein, saRNA interacts directly with its DNA target without the need of a third ncRNA as a docking molecule. This viewpoint is further supported by recent observations that many ncRNAs can guide chromatin-modifying complexes to specific genomic sites in mammalian cells (40
1.3.2. Dependence of RNAa on Argonaute proteins
Argonaute (Ago) proteins belong to a large family of proteins containing the PIWI and PAZ domains. Humans have a total of four closely related Ago proteins (Ago1-4) that function in small RNA-mediated gene regulation. During conventional RNAi, the Ago2 protein recruits siRNAs to form the RISC complex, which cleaves target mRNA by virtue of Ago2’s catalytic activity (42
). Both Ago1 and Ago2 have been implicated in transcriptional RNAi, frequently referred to as transcriptional gene silencing (TGS) (43
), although the details of the mechanism remain elusive. The requirement of Ago proteins in RNAa has also been examined by RNAi knockdown and chromatin immunoprecipitation (ChIP) experiments. In knockdown experiments, RNAa-mediated induction of E-cadherin and p21 was abolished by Ago2 depletion, whereas knockdown of the other Ago members (Ago1, 3, and 4) did not significantly impair RNAa (7
). Using ChIP assays, an agRNA-dependent association of Ago2 has been demonstrated with the targeted antisense transcripts in the PR and LDLR promoter (24
). Turunen et al.
also showed that Ago2 is recruited to the mouse VEGF promoter by activator shRNAs (10
). The dependence of RNAa on Ago2 is surprising because target cleavage is not observed in RNAa (34
). A reasonable explanation is that Ago2 is required for initial processing of saRNA/agRNA duplexes in a manner similar to siRNA maturation. The endonuclease activity of Ago2 cleaves and discards the passenger strand to form an active RNA-Ago complex capable of recognizing complementary sequences (45
). This view is supported by the observation that 2′-OMe modification at the passenger-strand cleavage site inhibits RNAa activity (22
). In addition, Ago proteins were found to be as abundant in the nucleus as in the cytoplasm in a number of studies (47
). However, it is unclear whether Ago2 is directly involved in subsequent transcriptional activation at the targeted gene or if additional proteins are essential for RNAa. Moreover, other Ago proteins may participate in RNAa when it is triggered by small RNA with an imperfect duplex structure (i.e. miRNA), in a manner similar to conventional RNAi.
1.3.3. Transcriptional activation is associated with epigenetic changes
Most examples of RNAa were demonstrated by an increase in target gene expression at the mRNA and/or protein level. Enhanced mRNA stability may also lead to increased mRNA and protein levels and represents an alternative mechanism for RNAa. However, this possibility does not appear to be true when assessing mRNA stability after saRNA transfection. Determination of both p21 and E-cadherin mRNA levels across a period of time after actinomycin D or α-amanitin treatment revealed that saRNA treatment does not increase the stability of their corresponding mRNAs compared to control treatments (Portnoy et al., unpublished data). Instead, several studies have shown an increase in RNAP II association with the core promoter, indicating RNAa is a transcriptional mechanism (9
Various types of chromatin modifications have also been identified at different promoters following RNAa. For example, loss of di- and tri-methylation at histone H3K9 has been associated with RNAa at the E-cadherin promoter in PC-3 cells (7
). In another study, reduced acetylation at H3K9 and H3K4 together with increased di- and tri-methylation at H4K4 has been associated with RNAa at the PR promoter (8
). Also, activation of the mouse VEGF promoter by shRNAs has been linked to increased H3K4me2 and H3K4me3 levels, as well as decreased H3K9me2, leading to a more accessible chromatin structure (10
). It is plausible to hypothesize that Ago2 guides saRNA molecules to complementary target sequence where they serve as a scaffold to recruit histone-modifying enzymes and activate transcription. The diversity in chromatin modifications may reflect gene-specific epigenetic changes associated with RNAa (8
). Furthermore, the question still remains as to whether the changes in chromatin are the cause or consequence of RNAa. Regardless, successful mapping of epigenetic changes at target-gene promoters has caught RNAa in action and indirectly provides evidence toward target specificity.