Minireview: Switching on Progesterone Receptor Expression with Duplex RNA

Abstract

It has long been appreciated that gene expression is regulated by protein complexes at promoters. More recently, research has demonstrated that small duplex RNAs such as micro-RNAs and short interfering RNAs complementary to mRNA provide another layer of regulation. Evidence now supports the existence of regulatory pathways that use small duplex RNAs to control transcription. Synthetic RNAs complementary to gene promoters [antigene RNAs (agRNAs)] can either activate or inhibit gene expression. Activity of agRNAs is mediated by argonaute, a protein required for RNA interference. Unlike protein transcription factors, agRNAs do not bind to chromosomal DNA but recognize noncoding transcripts that overlap gene promoters or 3′-gene termini. This review describes recent studies with agRNAs and focuses on the robust and potent agRNA-mediated regulation of progesterone receptor. The ability of small RNAs to alter transcription provides a new layer of potential regulation for gene expression.

Small RNAs complementary to gene promoters can modulate transcription. Promoter-targeted RNAs either silence or enhance progesterone receptor (PR) gene expression in different cell lines.

The traditional view of the genome is that chromosomal DNA encodes mRNA, and translation of mRNA yields proteins that regulate gene expression and other cellular processes. Sequencing of the human genome has revealed that only a small fraction of the genome encodes mRNA, leading to the rest of the genome being termed “junk DNA.”

Recent genomic studies have revealed that most of the genome is transcribed (1,2,3) (Fig. 1). Junk DNA is, therefore, accompanied by vast amounts of Junk RNA. Much of this RNA overlaps the regions of genes that encode mRNA and can extend tens of thousands of bases from one gene to the next. More than 30% of genes have antisense transcripts that overlap their promoters (4), offering a clear challenge to simple explanations for how transcription occurs. A few reports have begun to suggest roles for these overlapping transcripts in gene regulation (5,6,7,8,9,10,11), but our understanding of their potential function is at the earliest stages.

Figure 1.

What is a gene? A, Traditional view of gene expression. B, Emerging transcriptional landscapes showing transcripts that overlap mRNA, promoter, and terminal regions of genes.

One role for noncoding RNA is production of micro-RNAs (miRNAs) that target 3′-untranslated regions (3′-UTRs) and inhibit protein expression (12,13,14,15). miRNAs are synthesized as hairpin precursors that are processed by the enzyme Dicer into short duplexes. These duplexes are partially complementary to target genes and alter gene expression through recognition of 3′-UTRs. The full role of miRNAs in regulating gene expression is not known, but miRNAs play vital roles in some physiological processes.

There have also been suggestions that RNA can mediate recognition of chromosomal DNA. In 1994 a report noted that expression of viroid RNA in plants caused methylation of the corresponding integrated sequence within chromosomal DNA (16). Subsequent studies have revealed that RNA-directed methylation of plant DNA is a widespread mechanism for controlling transcription (17). In yeast, data suggest that RNA contributes to the specificity of heterochromatin formation (18,19). RNA-mediated transcriptional silencing has also been reported in Tetrahymena (20) and Drosophila (21,22,23).

In 2004, two reports suggested that duplex RNA could mediate DNA methylation and inhibit gene expression in human cells (24,25). One of these reports was subsequently retracted (26). Our laboratory had been examining the control of gene expression by synthetic agents that target chromosomal DNA (27). After considering the strengths and weaknesses of the prior literature, we initiated our own examination into the potential of small RNAs to target gene promoters and regulate gene expression.

In this review, we focus on modulation of gene transcription in mammalian cells by small RNAs that target gene promoters and regions downstream from gene termini (Fig. 2). We refer to these RNAs as “antigene RNAs” (agRNAs) to distinguish them from duplex short interfering RNAs (siRNAs) or miRNAs that target mRNA and modulate translation. agRNAs contain 19-bp duplex regions that are complementary to the target sequence and are identical in structure to siRNAs (Fig. 2A).

Figure 2.

agRNAs and their targets. A, Typical agRNA or siRNA design. Nineteen base duplex, two deoxythymidines at the 3′-termini. B, Duplex siRNAs are most often designed to target mRNA and silencing gene expression. C, agRNAs can target gene promoters or regions downstream from gene termini and either silence or activate gene expression.

Unlike RNAs that modulate endocrine responses by forming primary interactions with proteins (see later text), agRNAs recognize target sequences though Watson-Crick hybridization (Table 1). Although agRNAs appear to share some properties with protein transcription factors, rather than bind DNA, evidence suggests that they recognize noncoding RNA transcripts that overlap gene promoters.

Table 1.

Contrasting agRNAs and other agents that control gene expression

Mechanism	Known target	Effect
Transcription factor	DNA	Transcription
DNA methylation	DNA	Transcription
SRA	Protein factor	Transcription
Gas5	DNA-binding domain	Transcription
miRNA	mRNA	Translation
siRNA	mRNA	Translation
Antisense oligomer	mRNA	Translation
AgRNA	Noncoding transcript	Transcription

Research in this area has confronted several questions. Might gene silencing be an off-target effect? If the phenomenon were real, what would be the molecular target of agRNAs? What factors are involved? Finally, if synthetic agRNAs can mediate gene expression, is there also evidence that endogenous agRNAs might help control the activity of gene promoters?

Inhibiting Gene Expression

We chose progesterone receptor (PR) as a model gene for examining agRNA-mediated regulation of gene expression in mammalian cells. We chose PR because 1) the promoter and major transcription start sites were well characterized (28), 2) expression was easily detectable by Western analysis or quantitative PCR (qPCR), 3) PR is an inducible gene, the expression of which varies more than 10-fold among model cell lines, and 4) PR expression is important for understanding both normal physiology and disease (29,30,31).

PR protein has two major isoforms, PR-B and PR-A (28). They are transcribed from distinct promoters within the PR gene. The transcription start site for PR-B is located upstream of PR-A. We designed agRNAs to target promoter sequences upstream from the PR-B transcription start site. We then introduced the agRNAs into T47D breast cancer cells using standard transfection protocols and tested their affect on endogenous PR gene expression. We chose T47D cells because they express high levels of PR protein when grown in standard conditions.

Several anti-PR agRNAs were potent inhibitors of PR expression even though they possessed little or no complementarity to PR mRNA (Fig. 3A) (32). Inhibition was robust and easily reproducible, hinting at the possibility that we were tapping into a natural mechanism for gene silencing.

Figure 3.

Control of PR expression by agRNAs. Western analysis demonstrating A) RNA-mediated inhibition (32) and B) RNA-mediated activation (36) of PR gene expression. MM, Mismatch containing RNA duplex. PRrna1, PRRNA2-RNA duplexes complementary to PR mRNA. Lipid only refers to cationic lipid added to cells in the absence of RNA. The target sites for agRNAs relative to the +1 transcription start site were PR2 (−2/+17), PR9 (−9/+10), PR11 (−11/+8), PR24 (−24/−5), PR26 (−26/−7). The top band is the PR-B isoform, and the lower band is the PR-A isoform.

Further characterization revealed important properties of gene silencing by agRNAs. Silencing by agRNAs was dependent on expression of argonaute 2 (AGO2) (33), a key protein in the RNA interference (RNAi) pathway (34,35). Involvement of AGO2 in the action of agRNA suggested that RNA-mediated recognition of promoter DNA was related to RNAi. Unlike standard RNAi, however, nuclear run-on assays (33) and quantitation of RNA polymerase II (RNAPII) levels at the PR promoter (36) revealed that agRNAs caused inhibition of transcription.

Inhibition of PR expression by agRNAs was sensitive to the exact target sequence. Moving the RNA target only one or two bases was sufficient to abolish gene inhibition. This sensitivity is reminiscent of protein transcription factors, in which small changes in conformation can alter gene expression.

Gene silencing by agRNAs appears to be a general phenomenon. Aside from PR, our laboratory has observed agRNA-mediated inhibition of cyclooxygenase-2, major vault protein, androgen receptor, low-density lipoprotein receptor, and huntingtin. Other laboratories have used promoter-targeted RNAs to inhibit expression of E-cadherin (37), EF1A (24,38), RASSF1 (39), c-myc (40), CCR5 (39), ubiquitin C (41), urokinase plasminogen activator (42), the HIV-1 genome (43), and endothelial nitric oxide synthase (44). In what may be a related phenomenon, duplex RNAs have also been reported to control alternative splicing through a mechanism that involves transcriptional silencing (45).

Activating Gene Expression

While characterizing gene silencing by agRNAs in T47D cells, we observed an intriguing result. Some agRNAs that were not inhibiting expression of PR were causing small, but reproducible, increases in PR expression. Although not definitive, these increases hinted that gene activation might be occurring.

To further examine this hypothesis, we tested the effect of these putative activating agRNAs on PR expression in MCF7 cells. We chose MCF7 cells because they express PR protein at low levels relative to levels in T47D cells. We reasoned that if activation were occurring it would be obvious when compared against a low background of PR expression.

This hypothesis was proven correct. Transfection of agRNAs into MCF7 cells led to substantial increases in PR expression (Fig. 3B) (46) and recruitment of RNAPII to the PR promoter (36). We also demonstrated that the same agRNAs that activated PR expression in MCF7 cells could also activate PR expression in T47D cells when the basal levels of PR were lowered.

PR expression can be lowered in T47D cells by depleting hydrophobic compounds from the media. When agRNAs are added to T47D cells grown in this media, PR levels increase (46). Therefore, in two different cell lines, agRNAs were able to mediate increased gene expression. It is worth noting that the extensive literature characterizing PR along with the availability of cell lines expressing varied levels of PR was essential for the success of these experiments.

As with gene inhibition, gene activation was acutely dependent on the exact sequence of the target site. Shifts of one or two bases were sufficient to convert an activating agRNA into one that could not alter gene expression. Importantly, the inert agRNAs were able to compete with activating agRNAs and block their activity (46). This result suggests that inert and activating agRNAs both bind to the same molecular target and that the exact location of binding is critical in determining whether RNA association will lead to enhanced expression.

It is interesting to speculate that our synthetic agRNAs may be interfering with a repressive effect of the antisense transcript, but we have not yet observed any evidence to support this hypothesis. Activating agRNAs do not induce detectable cleavage of the antisense transcript or reduce its abundance.

Sensitivity to target sequence and location is reminiscent of protein transcription factors that can positively or negatively regulate gene expression by interacting with promoter DNA. Other laboratories have also observed gene activation by promoter-directed duplex RNAs (47,48,49).

Molecular Targets

For traditional RNAi, duplex RNAs have a well known target: mRNA. By definition agRNAs have no mRNA target, raising the important question: what do agRNAs bind? One possibility was that agRNAs were directly recognizing DNA. Whereas binding to chromosomal DNA appears to be a simple solution, it is inconsistent with involvement of AGO, a protein well established to promote RNA-RNA interactions.

Chromosomal DNA is not the only nucleic acid present at gene promoters. Genomic studies have revealed that many genes express noncoding transcripts that overlap their promoters (1,2,3,4,5,6,7,8,9,10,11). Although no transcript was known to exist at the PR promoter when we began our work, given the prevalence of noncoding transcription throughout the genome, we evaluated transcription at the PR promoter.

Strand-specific PCR revealed the existence of an antisense transcript synthesized opposite in orientation relative to PR mRNA and originating within the PR gene. We found that agRNAs bound the antisense transcript at the PR promoter and recruited AGO2 protein to the transcript. Other laboratories have identified noncoding transcripts as targets for small RNAs that silence cMyc expression (40) or control alternative splicing (45).

How might recruitment of AGO2 protein to an antisense transcript affect gene expression? PR mRNA expression is modulated by interactions between proteins, small molecules, and nucleic acids. The explanation that best fits our experimental findings is that interactions between agRNAs and noncoding transcripts can recruit AGO2 protein, and possibly other proteins, to the PR promoter (Fig. 4). The presence of AGO protein near the promoter then tips the delicate balance between high and low PR expression.

Figure 4.

Scheme showing factors that associate with gene promoters during agRNA-mediated gene activation or silencing (36). Solid lines denote factors that associate with promoter DNA upon addition of agRNA. Dashed lines denote factors that dissociate from promoter DNA upon addition of agRNA. Pol2, Polymerase II.

Rationales for Using agRNAs

PR has proven to be an informative target gene, but even the initial characterization of anti-PR agRNAs required a substantial investment in resources. Not all genes are likely to be similarly productive targets, raising the question of how target genes for agRNAs should be chosen.

The first step toward using agRNAs is to prepare a clear rationale for their use over other approaches. What advantages might be provided by agRNAs that cannot be achieved by the much better precedented approach of using duplex RNAs to target mRNA? Potential advantages for using agRNAs include:

1. For some genes it may be difficult or impossible to identify siRNAs that are potent and selective. agRNAs may provide an alternative for silencing these genes.

2. mRNA targets can vary in abundance up to 10,000 or more copies per cell, whereas there are generally only two agRNA targets per cell. In some cases, agRNAs may be more potent.

3. More broadly, using agents that recognize sequences within the gene promoter and alter transcription may yield results that are different from those obtained using siRNAs. Activation of PR expression is a specific example of this. Regulation at gene promoters is complex, and unanticipated results could result from alternate promoter usage or a shift in the major transcription start within the target gene.

4. agRNAs provide a strategy for activating gene expression. There are few alternate approaches for selectively inducing expression of a target gene, and agRNAs may provide an approach for increasing the expression of genes for research and therapy.

Choice of Target Genes

Successfully applying agRNAs to a problem requires careful planning (Table 2). Based on our experience with PR, activation may be most easily observed for genes with relatively low basal gene expression but can be up-regulated in response to stimuli. Inducible genes may be most susceptible to agRNA-mediated regulation because they are already poised to be modulated by environmental stimuli.

Table 2.

Planning an experiment with agRNAs

Procedures
Identify target gene. Justify benefits of RNA-mediated gene silencing or activation relative to other approaches
Identify methods for measuring gene expression: antibodies for Western analysis, primers for qPCR.
Identify a positive control: siRNA that silences expression by targeting mRNA
Identify a cell line. A cell line with low expression may be best for examining gene activation. No evidence yet that agRNAs can activate a gene that has no basal expression.
Research literature/databases for data on transcription start and termination sites
Confirm start/termination sites in chosen cell line by RACE
Identify and characterize the overlapping transcripts
Use positive control RNA to optimize transfection conditions
Design at least six agRNAs

From a practical standpoint, the ability to enhance PR mRNA expression by 5- to 10-fold creates an unambiguous window for examining RNA-mediated activation. Our experience with genes that can be activated only 2- to 3-fold is that distinguishing activation from off-target effects requires a greater investment in resources.

Characterizing Target Genes

Once a gene is chosen, the next step is to characterize where transcription is initiated. Information about where transcription is initiated is important because inaccurate assumptions may lead to the design of duplex RNAs that are complementary to mRNA and act by the standard posttranscriptional silencing mechanism.

Information about transcription start sites can usually be found in databases or in the literature. Although useful, existing information cannot be assumed to be reliable or applicable to the experimental cell lines of interest. For example, the 5′- and 3′-termini for our model gene, PR, were misannotated in GenBank until 2008.

One useful source for experimental information for many genes is the Database of Transcriptional Start Sites (dbtss.hgc.jp) (50). We recommend using 5′-rapid amplification of cDNA ends (RACE) as a complementary experimental method to confirm the 5′-terminus of the gene. A less precise, but faster, supplementary assay is to design qPCR primers to amplify transcripts both upstream and downstream of the predicted transcription start site.

Target Sequences for agRNAs

Before testing agRNAs it is essential to possess a positive control for gene silencing. An siRNA that is complementary to the mRNA of the gene of interest and efficiently silences gene expression demonstrates that efficient transfections are being achieved and provides a benchmark for evaluating the efficiency of agRNAs. agRNAs are transfected into cells in parallel with the positive control siRNA, and expression is evaluated using Western analysis or qPCR.

We typically design at least six agRNAs complementary to sequences within 200 bases upstream of the most 5′-transcription start site. The need to test several duplex RNAs and differences in activities between closely related duplexes is also typical of siRNAs that are complementary to mRNA. We design duplexes to have a balanced representation of C/G and A/T nucleotides and have minimal complementarity to sequences located elsewhere in the genome. We note, however, that target regions may sometimes be A/T or C/G rich, and it may be necessary to test agRNAs that do not possess ideal base composition.

Validating agRNA Activity

One of the concerns when using siRNA or agRNA technology is the potential for off-target effects (Table 3). Off-target effects are phenotypic changes that result from interactions between a nucleic acid and an unintended cellular target. The potential for artifactual results is also a major concern when interpreting gene-silencing experiments with siRNAs. The causes and consequences of RNA-mediated off-target effects have been the focus of two excellent recent reviews (51,52).

Table 3.

Validating potential agRNAs

Validation procedures
Assay candidate duplex RNAs using qPCR and/or Western analysis
If any duplex RNAs are active, design multiple mismatch or scrambled control duplexes
Develop structure/activity relationships by testing additional duplex RNAs to identify additional active compounds
Determine potency of active agRNAs (dose-response analysis)
Examine potential for candidate agRNAs to cause interferon response
Use RIP to investigate recruitment of AGO to overlapping transcript
Use ChIP to examine recruitment of RNA polymerase to promoter
Note phenotype. Active agRNAs should have similar phenotypes.

ChIP, Chromatin immunoprecipitation; RIP, RNA immunoprecipitation. 

We pursue four strategies to test the alternate hypothesis that agRNA-mediated gene modulation might be an off-target effect:

1. We test several mismatch-containing or scrambled (blocks of bases are interchanged within the parent agRNA) duplex RNAs.

2. Because introduction of duplex RNA into some cells can invoke an interferon response and alter gene expression unexpectedly, we evaluate whether the active agRNAs are altering the expression of interferon responsive genes.

3. It is unlikely that compounds whose only similarity is complementarity to the same target gene would have the same off-target effect. Therefore, we attempt to identify multiple active agRNAs that do not overlap in sequence.

4. Finally, we monitor cell growth and toxicity. Some RNA sequences are toxic to cells and will broadly affect gene expression. Toxicity may be a legitimate agRNA phenotype but, if it is observed, caution becomes especially important.

Studying Mechanism

Several assays can both provide mechanistic information and further validate agRNAs (Table 2). The nuclear run-on assay measures the synthesis of nascent transcripts and can be used to evaluate levels of transcription. This assay, however, requires millions of transfected cells per data point and is cumbersome. We find that it is more convenient to use chromatin immunoprecipitation to evaluate recruitment of RNAPII to gene promoters in the presence or absence of active and negative control agRNAs.

Our proposed mechanism of agRNA action involves recruitment of AGO2 protein to the gene promoter through interactions with a noncoding transcript. Acquiring evidence for AGO2 recruitment provides important support for concluding that agRNA-mediated modulation of gene expression is target specific.

We first use PCR primers sets that recognize sequences beyond the known ends of the target gene mRNA to determine levels of overlapping transcripts. We then use 5′- and 3′-RACE to characterize these transcripts, and finally perform RNA immunoprecipitation with an anti-AGO2 antibody to examine the ability of agRNAs to recruit AGO2 protein to the overlapping transcript.

Recognition Beyond the 3′-UTR

Many noncoding transcripts overlap the 3′-termini of genes. The existence of these transcripts led us to hypothesize that agRNAs complementary to sequences beyond the 3′-termini of mRNA might also be able to control gene expression. We chose PR as a model gene to investigate potential activities of 3′-agRNAs (53).

Initially, the 3′-terminus of the PR gene was misannotated in GenBank, and the literature characterization of the 3′-terminus was inexact. We began our studies by characterizing the 3′-terminus of PR by 3′-RACE, Northern analysis, and qPCR. The 3′-region of PR is highly AT rich, but several sequences with adequate C/G representation were identified as potential targets for agRNA.

Our results using 3′-agRNAs were strikingly similar to our results with 5′-agRNAs. In T47D cells, addition of 3′-agRNAs led to a reduction in PR mRNA and protein and decreased RNAPII occupancy at the PR promoter (Fig. 5A). Conversely, in MCF7 cells addition of 3′-agRNAs led to increased levels of PR mRNA and protein, and increased RNAPII occupancy. A noncoding transcript overlaps the 3′-terminus of PR mRNA, and both activating and inhibitory 3′-agRNAs recruit AGO2 to it.

Figure 5.

Effects of agRNAs. A, Inhibition of RNAPII recruitment at the PR promoter by 5′-agRNA PR-9, siRNA PR3593 (targets mRNA), and agRNA PR13580 (targets beyond the 3′-termini of PR mRNA). B, Addition of silencing 3′-agRNA PR13580 reverses PR gene activation after treatment with estradiol. C, Activating 3′-agRNA has additive effect when combining with estradiol (E2) MM, Mismatch-containing RNA control. Assays were performed using cells grown in serum-stripped media to maximize effects of E2 addition. All fold change are relative to cells treated with mismatched duplex. Similar effects in combination with E2 were observed for 5′-agRNAs. RNAP2, RNA polmerase II.

Effect of Physiological Stimuli on agRNA Action

Different physiological stimuli can either activate or repress expression of PR. For example, addition of the hormone 17β-estradiol to MCF7 cells enhances PR expression (54,55). Conversely, PR levels decrease in T47D or MCF7 cells grown in media supplemented with charcoal-treated serum and can be further reduced upon the addition of IL-1β (56), or epidermal growth factor (57).

We investigated how these endogenous stimuli would affect modulation of PR gene expression by agRNAs (53). Addition of inhibitory agRNAs to cells treated with 17β-estradiol counteracted the stimulatory effects of 17β-estradiol and led to lower levels of PR expression (Fig. 5B). Addition of activating agRNAs to cells treated with 17β-estradiol increased PR expression to levels substantially above those observed after treatment with activating agRNA or 17β-estradiol alone (Fig. 5C). Addition of activating agRNAs also counteracted the inhibitory effects of charcoal-treated serum, IL-1β and epidermal growth factor on PR expression.

These data suggest that agRNA action cannot be blocked by physiological stimuli. Rather, agRNAs can enhance or counteract stimuli known to be powerful modulators of PR gene expression.

The PR promoter and the 3′-terminal region of PR gene are separated by approximately 100,000 bases within chromosomal DNA. How could agRNAs modulate transcription over such a great distance? Reports suggest that genes can form higher order structures in which gene promoter and terminator regions are juxtaposed. This phenomenon might place 3′-noncoding RNAs and their associated agRNA/AGO protein complexes much closer to gene promoters than their chromosomal locations would suggest.

Addition of an agRNA complementary to a sequence beyond the 3′-termini of PR followed by immunoprecipitation resulted in recovery of the 5′-noncoding transcript as well as the 3′-noncoding transcript. Similarly, in cells transfected with 5′-agRNAs, the 5′- and 3′-transcripts were immunoprecipitated together. RNA immunoprecipitation involves a cross-linking step that couples nucleic acids and proteins. Therefore, the data do not indicate a direct association of the 5′- and 3′-transcripts. Rather, the data suggest involvement in a common complex.

Chromosome conformation capture analysis is a powerful method for detecting long-distance interactions between sequences within chromosomal DNA (58,59,60). In this technique genomic DNA is cross linked and subsequently purified. The genomic DNA is treated with a restriction enzyme and then ligated. This procedure allows the joining of sequences that are in close spatial proximity even if they are encoded by distant genomic regions. DNA sequences are amplified by PCR to evaluate the proximity of regions within the gene.

Chromosome conformation capture analysis revealed that the PR promoter is in close proximity to the PR terminus in both MCF7 and T47D cells. The relative amount of gene looping was similar in both cell lines despite their large differences in basal PR expression. Looping was unaffected by addition of 17ß-estradiol or IL-1ß. The extent of looping was also unaffected by addition of activating or inhibitory agRNAs.

Our results suggest that gene looping may not be involved in moving from PR expression from an activated to an unactivated state but can provide a short cut that allows 3′-agRNAs to affect gene expression at a distant sequence. The looped structure facilitates interactions between 3′-agRNA/AGO complexes and chromosomal DNA to regulate transcription at the promoter (Fig. 6).

Figure 6.

Gene looping, agRNAs, and recruitment of AGO2.

Endogenous Regulation?

Gene activation and gene silencing by agRNAs is robust. If artificial RNAs can so readily modulate gene expression, it seems reasonable to believe that natural miRNAs might also be candidates for recognition of gene promoters.

Two reports have appeared suggesting that endogenous miRNAs can control gene expression by targeting promoters. Dahiya and co-workers (61) searched for potential miRNA-binding sites within E-cadherin, a gene that they had already reported to be regulated by synthetic RNA duplexes. They identified a potential binding site for miR-373 and showed that a mimic of miR-373 increased E-cadherin expression. Rossi and co-workers (62) searched for perfect complementarity between miRNAs and gene promoters and identified a match between miR-320 at its genomic location and the region new gene JPOLR3D.

Our laboratory has developed computational algorithms to predict potential miRNA targets within either gene promoters or regions that lie beyond the 3′-UTR (63). Promoter sequences or sequences beyond the 3′-UTR from the University of California Santa Cruz genome browser are matched with miRNA sequences within miRBase, a database that contains the sequences of mature and precursor miRNAs. Using these algorithms we identified an enrichment of seed-sequence matches within gene promoters and downstream from the 3′-termini of mRNA. In a large number of cases we observed near-perfect complementarity. These studies suggest a rich potential for gene regulation by endogenous miRNAs.

To more closely examine the potential for miRNAs to regulate gene expression by targeting sequences beyond the 3′-UTR, we used our algorithm to predict matches downstream of the 3′-UTR termini of PR. Several matches preserved seed-sequence complementarity and had high overall complementarity. We obtained the synthetic miRNAs and introduced them into cells. One of these RNAs, miR123b, inhibited PR expression and blocked recruitment of RNAPII at the PR promoter. This result supports the suggestion that miRNAs can regulate gene expression by targeting noncoding RNAs.

More Mechanisms for RNAs

agRNAs are just one example of noncoding RNAs that control transcription. For endocrine pathways, one of he best known RNA controllers is the steroid receptor RNA activator (SRA). SRA was discovered by O’Malley and co-workers (64) in 1999 as a transcriptional coactivator and provides a model for RNA controlling transcription through RNA-protein interactions. Other laboratories demonstrated that SRA can activate several other nuclear hormone receptors (reviewed in Refs. 65 and 66). Although the exact mechanism of SRA action is unknown, many transcription factors possess RNA-binding domains (67), and SRA may affect their function. Specifically, SRA has been shown to interact with several proteins including, SRC-1 (64), SLIRP (68), SF-1 (69), and Dax (69). Recently, gene expression profiling has expanded the range of potential targets for SRA (70).

Recent data have also identified noncoding RNA Gas5 as a repressor of glucocorticoid receptor (71). Gas5 was identified as an abundant RNA in cells subjected to growth arrest. Gas5 interacts with the DNA-binding domain of the glucocorticoid receptor and acts as a decoy, reducing the pool of receptor available for regulating gene expression at promoters. Taken together with data on the action of SRA, these results show that structured noncoding RNAs can act through varied mechanisms as transcriptional activators or repressors.

Conclusions

Modulation of gene expression by agRNAs suggests that RNA has the ability to play a larger role in gene regulation than had been assumed previously. There may be a complex web of regulation involving noncoding transcripts that overlap mRNA, small RNAs, and proteins that regulate gene expression. An ability to exploit RNA-RNA recognition to augment transcription factors would have obvious benefits for fine tuning or evolving the function of genes. Related agRNAs can either activate or inactivate transcription, a functional feature that makes them resemble protein transcription factors.

A recent review has noted that many processes take place cotranscriptionally (72). In this context, association of agRNAs and AGO to noncoding RNAs provides a Watson-Crick-based mechanism for changing the composition of proteins at gene promoters and altering the transcription elongation complex.

On a practical level, useful studies require well-controlled experiments and extensive characterization. Experimenters entering this area should expect that obtaining persuasive results will require the investment of substantial resources. Some genes are likely to be more amenable to regulation by agRNAs than others, and PR has proven to be an ideal model system. We have also found that androgen receptor is a reliable target for agRNA-mediated regulation, and it is possible that other sensitively regulated genes within endocrine pathways will also be productive targets.