RNA-Directed Transcriptional Gene Silencing and Activation in Human Cells

Abstract

The overt loss or uncontrolled gain of gene expression is found at some level in virtually every malady afflicting humans. From cancer to HIV-1, the uncontrolled expression or loss of gene expression is prevalent in human diseases. Approaches toward the specific control of gene expression at the transcriptional level could have the potential to revert or reduce disease pathologies. Over the last several years, researchers have developed methodologies that utilize small antisense non-coding RNAs to specifically silence transcription. Only recently has the endogenous molecular pathway usurped by the introduction of these small RNAs to regulate transcription in human cells been defined. Observations suggest that long antisense non-coding RNAs function as the endogenous epigenetic regulators of transcription in human cells, thus explaining why small antisense RNAs were observed early on to silence transcription via directed epigenetic changes at the target loci. The mechanism of action whereby small regulatory RNAs can either turn gene transcription on or off will be discussed as evidence that one day it may be possible to develop therapeutics to regulate gene transcription and ameliorate particular disease conditions.

Small RNA-Directed Transcriptional Gene Silencing

The most basic central dogma in molecular biology, DNA encodes RNA that encodes proteins, teaches that proteins are responsible for the regulation of transcription. This concept, while providing a simplistic paradigm for teaching basic science, has proven to have numerous exceptions. One such exception that is slowly gaining acceptance is the observations that non-coding RNAs can regulate transcription. First observed in doubly transformed tobacco plants, small double-stranded RNAs were shown to direct epigenetic changes such as DNA methylation to loci containing homology to the small RNA (Matzke et al., 1989). The phenomenon was termed small RNA-directed transcriptional gene silencing (TGS). TGS was later shown in Arabidopsis to require the action of RNA-dependent DNA methylation (Wassenegger et al., 1994; Mette et al., 2000) and members of the Argonaute protein family (Lippman et al., 2003). TGS is mechanistically distinct from the abundantly studied post-transcriptional silencing pathway, which requires Argonaute 2 and results in slicing of the target mRNAs. Notably, TGS results in long-term stable epigenetic modifications to gene expression that can be passed on to daughter cells [reviewed in (MORRIS, 2009a)]. To date, small RNAs have been shown to modulate TGS in plants (Arabidopsis), yeast (Schizosaccharomyces pombe), flies (Drosophila), worms (Caenorhabditis elegans), and in human cells [reviewed in (MORRIS, 2005)]. While there are variations in the mechanism of TGS among the studied organisms, the general theme of RNA-based transcriptional regulation resonates.

RNA-Mediated TGS in Human Cells

Observations of RNA-mediated TGS in human cells lagged behind work done in Arabidopsis, S. pombe, and C. elegans (Morris et al., 2004). Several studies have now however validated the initial observation of siRNA-directed TGS in human cells as well as provided insight on the underlying mechanism of action whereby antisense non-coding RNAs regulate gene transcription (Table 1).

Table 1.

Reported Observations of Transcriptional Gene Silencing (TGS) in Human Cells

Observation(s)	References
Observed a long antisense non-coding RNA capable of directing DNA methylation in human genetic diseases	Tufarelli et al., 2003
siRNAs targeted to EF1-alpha promoter direct TGS in human cells requires histone and DNA methylation	Morris et al., 2004
Defined a miRNA (miR-N367) capable of directing TGS to HIV-1 LTR	Omoto et al., 2004, 2005
siRNAs targeted to Huntington gene silence in absence of DNA methylation	Park et al., 2004
TGS is involved in silencing nonsense codon-containing immunoglobulin minigenes	Buhler et al., 2005
TGS of CDH1 gene in the absence of DNA methylation	Ting et al., 2005
E-cadherin promoter-targeted shRNAs can direct DNA methylation to E-cadherin promoter	Castanotto et al., 2005
Antisense RNA is converted into 21 bp RNAs and found to be associated with regional histone H3 lysine 9 (H3-K9) methylation and HP 1 gamma recruitment	Cho et al., 2005
siRNAs target TGS to HIV-1 LTR, first to observe that only small antisense RNAs are required for TGS and DNMT3a is involved in TGS	Weinberg et al., 2005
TGS occurs when targeting RNAPII-binding sites	Janowski et al., 2005
Herpes viral latency-associated transcript gene promotes assembly of heterochromatin on viral lytic-gene promoter	Wang et al., 2005
TGS of CCR5 requires Ago-1 and demonstration of endogenous miRNA-directing TGS	Kim et al., 2006, 2008
Promoter-targeted methylated oligonucleotides can direct DNA methylation to inhibit Bcl-2 transcription	Hoffman and Hu, 2006
HIV-1 TAR RNA acts as a miRNA to direct heterochromatin formation and histone methylation	Klase et al., 2007
Duplex RNAs can activate gene expression	Janowski et al., 2007
Target p 16 promoter with siRNAs observe heterochromatin changes	Wang et al., 2007
TGFB receptor II promoter-targeted shRNA directs methylation of and silencing of the target gene in rat hepatic stellate cells	Kim et al., 2007
TGS requires pRNAs, first observation that promoters in human cells are transcribed	Han et al., 2007
siRNAs-directed TGS to inhibit prostate tumor growth and metastasis in human cells	Pulukuri and Rao, 2007
siRNAs guide heterochromatin formation in human cells, miR17-5p and miR20a involved in TGS	Gonzalez et al., 2008
TAR miRNAs generated in HIV-1 infection processed by Dicer	Ouellet et al., 2008
SIV-1 LTR-targeted siRNAs direct TGS and requires DNA and histone methylation	Lim et al., 2008

The majority of studies carried out to assess RNA-mediated TGS in human cells have relied on using synthetic promoter-targeted small interfering RNAs (siRNAs). Recently, however, observations have emerged demonstrating that some microRNAs (miRNAs) function to direct TGS in human cells (Omoto and Fujii, 2005; Klase et al., 2007; Kim et al., 2008; Tan et al., 2009). From these studies, it has been revealed that RNA-mediated TGS is operative through RNA-directed methylation of histone 3 lysines 9 and 27 (H3K9 and H3K27, respectively) and DNA methylation at the targeted promoter (Morris et al., 2004; Castanotto et al., 2005; Janowski et al., 2005; Weinberg et al., 2005; Kim et al., 2006; Han et al., 2007). These targeted epigenetic changes appear specifically at the RNA target site and are not found at distal un-targeted regions suggesting a level of specificity. However, the role of DNA methylation in TGS in human cells is not as clearly understood as in plants, as some investigators find DNA methylation at the targeted promoter while others do not (Table 1).

The current mechanistic understanding of TGS is that within the first 24 hours following small RNA treatment there is a robust increase in Argonaute 1 (Ago-1) at the targeted promoter followed shortly thereafter by increasing concentrations of H3K9 dimethylation and H3K27 trimethylation (Fig. 1A) (Kim et al., 2006), suggesting that the small RNA guides an epigenetic remodeling complex to the particular target loci. When the small RNA targeting is sustained for ~3–4 days, DNA methylation begins to appear and correlates with the observation of long-term stable gene silencing (Hawkins et al., 2009). Not all groups have observed DNA methylation at the small RNA-targeted promoter in human cells (Ting et al., 2005). These discordant observations may be the result of as yet unknown differences in the small RNA target sites or targeted genes that may affect the mechanism of TGS. The possibility also exists that additional unknown proteins may also play a role in TGS and may include additional histone and histone-modifying proteins, DNA-modifying proteins, and RNA-binding proteins. The observed long-term silencing is presumably the result of silent state epigenetic changes being directed to the targeted RNA polymerase II (RNAPII) promoter. These silent state epigenetic changes are conceptualized, based on the acquired data to date, to change the targeted loci and local nucleosomes in such a manner that there is an obstruction in RNAPII binding (Fig. 1A). These epigenetic changes suggest that we may need to revise our concept of RNA, as RNA-directed TGS can functionally passage information, using this perceived pathway, and affect gene expression to daughter cells in the form of epigenetic memory (MORRIS, 2009b).

FIG. 1.

Model for small RNA-directed transcriptional gene silencing (TGS) in human cells. (A) Small non-coding antisense RNAs targeted to gene promoters, ~1–2 nucleosomes upstream of the TATAA, can function in an Ago-1-dependent manner to recruit an epigenetic silencing complex consisting of DNMT3a, HDAC-1, and possibly Ezh2 to the targeted promoter. The recognition of the target loci is the result of the antisense non-coding RNA associating with the homology containing pRNA. (B) Small non-coding antisense RNAs targeted to gene promoters, directly at or overlapping the TATA and subsequent RNAPII-binding site, can function in an Ago-2-dependent manner to obstruct transcription. The recognition of the target loci is the result of the antisense non-coding RNA associating with the homology containing pRNA.

Not all RNA-directed TGS in human cells appears to function mechanistically in the same manner. Recent observations suggest that when small RNAs are targeted directly at the RNAPII TATA box, the region where RNAPII initiates transcription, that silencing can occur in an Argonaute 2 (Ago-2)-dependent manner and in the absence of silent state epigenetic changes (Janowski et al., 2005, 2006; Napoli et al., 2009) (Fig. 1B). This form of small RNA-directed TGS appears to be functional through the targeted obstruction of RNAPII binding to TATA, thus it is uncertain as to the duration of the observed silencing (Fig. 1B) and the ability for this form of TGS to be passed on via epigenetic memory to daughter cells.

Recently, RNA-directed TGS has been shown to impact mRNA splicing (Allo et al., 2009). Mechanistically, exon/intron splice junction-targeted siRNAs were shown to epigenetically modify their corresponding target loci in a manner similar to promoter-targeted siRNAs (Fig. 1A). The result of these targeted epigenetic changes was a reduction in RNAPII transcriptional fidelity at the splice site and a variation in splicing (Allo et al., 2009). This body of work suggests that TGS not only impacts transcription but can also modulate splicing. This also raises the question as to whether existing splice sites are the result of extant RNA-directed epigenetic modifications as well as to what extent non-coding RNAs direct evolutionary changes in splice site variation.

Mechanistically however, both forms of RNA-targeted TGS of promoters, be it upstream regions (Fig. 1A) or at RNAPII binding to TATA (Fig. 1B), require promoter-associated RNAs (pRNAs) to find their respective target site in the promoter (Han et al., 2007; Napoli et al., 2009). Since the initial discovery that promoters are transcribed and that these pRNAs are functional target sites for small RNA-directed TGS (Han et al., 2007), several others have reported pervasive transcription to be located at, upstream, or even overlapping 5′ regions of protein-coding genes (Kapranov et al., 2007; Taft et al., 2007; Core et al., 2008; He et al., 2008; Preker et al., 2008; Seila et al., 2008; Napoli et al., 2009). Taken together, these observations suggest that the original concept of RNAPII-mediated transcription initiating at distinct loci needs to be revisited.

An interesting insight into the requirement of pRNAs for small RNA-directed TGS was obtained in studies where it was shown that only the 21 bp antisense (“guide”) strand of the promoter-targeted siRNA was required to modulate TGS (Weinberg et al., 2005; Kim et al., 2006). This body of work indicated that in human cells there might be an endogenous mechanism whereby antisense RNAs direct epigenetic silencing complexes to target loci. Only recently have experiments begun to tease apart this mechanism of action (discussed later). Nonetheless, siRNAs have been found to be an excellent and cost-effective tool for determining those sites in gene promoters that are susceptible to TGS.

The amalgamation of all the current studies on RNA-directed TGS in human cells has provided a working model for the mechanism of action (Fig. 1A). This model is based on RNAs-targeted upstream of the RNAPII-binding site and TATAA. In this model, the initiation of TGS requires an abundance of the small antisense RNAs and the protein cofactors Ago-1, DNA methyltransferase 3a (DNMT3a), and histone deacetylase 1 (HDAC-1) (Suzuki et al., 2008; Hawkins et al., 2009; Turner et al., 2009), whereas the long-term stable silencing or maintenance of silencing requires DNMT3a and DNA methyltransferase 1 (DNMT1) (Hawkins et al., 2009). While these mechanistic insights have proven useful with regards to the generation of promoter-targeted small antisense or siRNAs to direct TGS of genes involved in disease, it has remained unknown until recently whether the observed mechanism of RNA-directed TGS was endogenously operative in human cells or whether the observed mechanism was some vestigial pathway being usurped by the synthetic generated promoter-targeted small RNAs.

RNA-Mediated Transcriptional Activation in Human Cells

Studies carried out to define the ability of siRNAs targeted to AT-rich regions in RNAPII promoters to modulate TGS demonstrated that some siRNAs were capable of modulating transcriptional derepression (Li et al., 2006; Janowski et al., 2007; Place et al., 2008). Argonaute 2 (Ago-2) appeared to be required for the observed RNA-mediated activation. Mechanistic studies to determine exactly how these siRNAs were modulating transcriptional activation indicated that the AT-rich-targeted siRNAs were actually targeting a long antisense non-coding RNA. The loss of this antisense non-coding RNA resulted in a loss of epigenetic regulation for the sense/mRNA partner that ultimately resulted in a marked increase in transcription (Morris et al., 2008). In essence, the activating siRNAs were functioning to suppress the suppressor long antisense non-coding RNA that resulted in derepression or activation of gene expression. Importantly, this study indicated that long antisense noncoding RNAs were functioning as the heretofore unknown endogenous RNA-based transcriptional regulators in human cells. Furthermore, these long antisense non-coding RNAs were found to utilize the same previously described TGS mechanism, namely Ago-1 to regulate transcription and the recruitment of H3K27 trimethylation to the targeted promoter (Morris et al., 2008) (Fig. 2A). This observation (Morris et al., 2008) is supported by other observations in human cells whereby long antisense non-coding RNAs regulate transcription through targeting epigenetic silencing complexes to homology containing loci (Tufarelli et al., 2003; Yu et al., 2008; Cliffe et al., 2009; Mahmoudi et al., 2009).

FIG. 2.

Model for both small and long antisense non-coding RNA-directed transcriptional regulation in human cells. (A) Long antisense non-coding RNAs can be expressed at bidirectionally transcribed genes. It is not known if this can occur in cis or trans. These antisense RNAs may (B) fold into secondary structured non-coding RNAs that have regions containing homology and can (C) interact with particular sites in the promoter of sense/mRNA. This homology containing region could then influence the recruitment of Ago-1, DNMT3a, and HDAC-1 to this target site in the sense/mRNA promoter. (D) Small synthetic antisense non-coding RNAs can be designed to take advantage of this endogenous mechanism and thus utilize the same pathway to transcriptionally silence gene expression. (E) The end result of either small or long antisense non-coding RNA transcriptional silencing is the targeted epigenetic remodeling of the particular RNA-targeted loci that makes interactions between the promoter and RNA polymerase II as well as transcription factors more difficult.

Long and Short Antisense Non-Coding RNAs in TGS

Antisense non-coding RNAs, both short and long, appear functional in regulating TGS in human cells (Fig. 2). Interestingly, both forms of short and long antisense noncoding RNA-directed TGS appears to involve Argonaute 1 (Ago-1) (Janowski et al., 2006; Kim et al., 2006; Morris et al., 2008) and the enrichment of silent state histone 3 lysine 27 trimethylation at their respective target loci (Weinberg et al., 2005; Morris et al., 2008; Hawkins et al., 2009; Turner et al., 2009). This mechanism of action emerges when these RNAs are targeted upstream of the dominant RNAPII promoter (Fig. 1A). However, when the non-coding RNAs are targeted at the RNAPII-binding site or TATA box or a splice junction site in the coding RNA, there appears to be a requirement for Ago-2 (Janowski et al., 2006; Allo et al., 2009). Notably, RNA-directed TGS upstream or downstream of the RNAPII promoter results in silent state epigenetic changes (Suzuki et al., 2008; Allo et al., 2009; Hawkins et al., 2009; Turner et al., 2009) (Fig. 1A) whereas RNA targeting directly at the RNAPII promoter or TATA does not appear to result in silent state epigenetic changes (Janowski et al., 2005; Napoli et al., 2009) (Fig. 1B). Taken together these observations suggest that RNA-directed TGS and the mode of action are variable depending on the RNA-targeted locus and possibly also the amount of ongoing transcription at the particular target site.

Small RNA-Guided Off-Target Effects

With the advent of siRNA-targeted post-transcriptional gene silencing (PTGS) came the realization of new off-target potential. In fact whenever we modulate a system with outside influence, such as treating cultures with small RNAs, we most certainly invite off-target changes. For instance, off-target effects that have been proposed resulting from siRNA-directed PTGS include: dsRNA-induced interferon responses, targeting of unintended mRNAs, and the saturation of the miRNA pathway. The recent data indicating a nuclear component of non-coding RNA-based regulation suggests that siRNAs targeted to mRNAs, with the intention of acting in a PTGS-based manner, could potentially have secondary effects within the context of the nucleus. Based on the TGS models (Fig. 2), it is not too difficult to assume that siRNAs designed to target a particular mRNA are in fact also modulating epigenetic changes at the targeted loci in the context of the chromatin. The effects of these off-target epigenetic changes have manifest experimentally as (1) the spreading of silent state epigenetic changes into neighboring gene promoters [data supporting this can be found in (Hawkins et al., 2009)] and also (2) siRNA-targeted changes in mRNA splicing (Allo et al., 2009).

Small RNA-directed TGS is also capable of inviting unintended off-target effects. Some off-target effects from utilizing small antisense RNA-directed TGS could be envisioned to manifest as (1) stable epigenetic changes at the targeted loci that are heritable (MORRIS, 2009b), (2) potential targeting of secondary targets as a result of partial complementarity to other transcribed loci, and (3) diminished endogenous long non-coding RNA-based regulation (MORRIS, 2009a) resulting from usurping the endogenous transcriptional silencing pathway with small single-stranded antisense RNAs targeted a particular loci involved in disease. TGS, however, is likely to elude dsRNA-induced interferon responses as single-stranded antisense RNAs appear to be the main modulators of the endogenous TGS pathway in human cells [reviewed in (MORRIS, 2009a)]. Though it is possible that hybridization of promoter-targeted small antisense RNAs to their targeted pRNA might induce an interferon response, this is an eventuality that has yet to be determined.

RNA-Directed TGS as a Therapeutic

The implications of utilizing the recently described long non-coding RNA-mediated TGS pathway (Fig. 2) to specifically regulate gene transcription are overwhelming. The ability to capitalize on knowledge of this molecular mechanism whereby antisense non-coding RNAs, be them small or long, regulate transcription for therapeutic benefit is infinite. But why utilize TGS and not the well-studied PTGS-based pathway? TGS has the advantage that a short duration of targeting can result in long-term stable changes that are passed on via epigenetic changes to daughter cells. Thus, one would only have to target the disease-related gene promoter for a short duration to establish a long-term effect. Notably, small single-stranded RNAs can be designed to specifically target a particular gene promoter of interest and synthesized at a relatively low cost. The single-stranded nature of these molecules allows for direct targeting in a strand-specific manner, a feature not found with siRNA-directed PTGS-based silencing (Wei et al., 2009), as well as an aversion of an unintended interferon responses. These single-stranded RNAs can also be modified to contain various stabilizing backbones that are resistant to degradation. Though studies to define backbone modifications functional in stabilizing single-stranded molecules while retaining TGS potential are only now beginning, these studies have the potential to identify optimal synthetic parameters for using small RNAs as therapeutics. Theoretically, any candidate gene that is constitutively overexpressed could be targeted and transcriptionally silenced.

The same paradigm of gene-specific targeting can be utilized to induce derepression of under-expressed genes and turn up gene expression. One could surmise that the reason a gene is epigenetically silenced, say a tumor suppressor gene in a particular cancer, could be the result of uncontrolled long antisense non-coding RNA-directed TGS of the tumor suppressor gene promoter. In fact, several tumor suppressor genes have been shown to exhibit bidirectional transcription (Yu et al., 2008). While further research must be undertaken, the hypothesis that these tumor suppressor genes are typically silenced in cancer is the result of an overexpression of the antisense non-coding RNA is plausible. If this proves to be the case, then one could envision designing single-stranded backbone-modified RNAs, or antisense phosphorothiate oligonucleotides to target the degradation of the regulatory long antisense non-coding RNAs. The loss of these regulatory long antisense non-coding RNAs could be expected to manifest in a loss of the epigenetic brake on the promoter expressing the sense strand and ultimately results in gene activation (Morris et al., 2008; Schwartz et al., 2008). However, gene derepression/activation would probably require sustained exposure of the small RNA targeted to the antisense non-coding RNAs. Although it may prove that high-level increases in sense/mRNA expression following activation results in stable long-term gene activation via downstream effects incurred in the antisense non-coding RNA promoter, data delineating this eventuality are not yet present and thus this notion is merely speculation.

In summary, the endogenous long non-coding RNA-based molecular mechanism (Fig. 2) can be utilized to specifically control transcription and either turn a gene on or off as desired. The advantage to turning genes off using the TGS pathway is that the silencing can be long-lasting and requires a relatively short duration of exposure to the non-coding RNA (Hawkins et al., 2009; Turner et al., 2009). The therapeutic potential to utilizing the TGS mechanism to exert transcriptional control has only recently been realized with examples of transcriptional suppression demonstrated for HIV-1 (Suzuki et al., 2005, 2008; Turner et al., 2009), CCR5 (Kim et al., 2006), c-Myc (Napoli et al., 2009), E-cadherin (Ting et al., 2005; Morris et al., 2008), prostate cancer (Pulukuri and Rao, 2007), and progesterone receptor (Janowski et al., 2005). Clearly, the potential for utilizing this emerging pathway of RNA-based transcriptional regulation to exert therapeutic control over gene expression and avert disease progression is in its infancy with great potential to result in a new class of future therapeutics.

Acknowledgments

This project is funded by R01 HL083473-02 to K.V.M. I thank Paula J. Morris at www.sciencegraphics.biz for the generation of figures and Anne-Marie Turner for critically reviewing the manuscript.