|Home | About | Journals | Submit | Contact Us | Français|
Over the last few years it has become increasingly apparent that RNA is involved in various forms of gene regulation. While much emphasis has been placed on the role of small non-coding RNAs in post-transcriptional modes of gene regulation it has become apparent that a far more complex scenario exists. Recent observations insinuate a paradigm whereby non-coding RNAs are operative effector molecules in the transcriptional regulation of endogenous gene expression. These observations support a route for how epigenetic gene silencing is directed, maintained and passed on as epigenetic memory in human cells. This perspective will highlight the endogenous effector RNAs and mechanism of action whereby non-coding RNAs transcriptionally regulate gene expression in human cells and discuss these recent observations in the context of human evolution.
Epigenetics is the study of how factors not coded for in the DNA are effective in the control of meiotically and mitotically heritable changes in gene expression.1 Epigenetic marks can be observed as DNA or histone methylation and can correlate with either gene activation or suppression. Evidence from animal studies has suggested that certain environmental stimuli can have an influence on particular epigenetic marks and that these epigenetic marks and their subsequent effects can be passed on to offspring.2 The result is that a particular epigenetic state can have long-term consequences with regards to phenotype.3 Taken together these observations suggest in animals that particular patterns of gene expression can be retained and passed on to progeny. However, evidence for this epigenetic memory and how it is directed in humans has been lacking.
During the winter of 1944–1945 there was a famine in the Netherlands. This “Dutch hunger winter” resulted from the German imposed suppression of food rations in the western area of the Netherlands. As a result of these actions several pregnant women became exposed to this imposed famine.4 The regulation of fetal growth and development is governed in part by the action of the epigenetically regulated insulin-like growth factor II (IGF2) protein. IGF2 is a growth promoting protein that is expressed from the paternal allele.5 IGF2 is regulated by IGF2r, which functions as a growth inhibitory protein that is expressed from the maternal allele.5 IGF2 is also under the regulatory control of the maternally expressed H19 gene which expresses an IGF2 regulatory non-coding RNA.6 The regulation of both H19 and IGF2 is maintained in part by the action of DNA methylation specifically at a differentially methylated region (DMR). Hypomethylation of the DMRs for IGF2 induces biallelic expression of IGF2, resulting in increased growth of the fetus.7
A monumental study carried out on humans who have descended from progeny during the Dutch hunger winter has shown that significant epigenetic changes, in the form of DNA methylaton at the DMRs, are present in offspring six decades later.4 Heijmans et al.4 showed that the level of DNA methylation observed at the IGF2 imprinted gene is significantly less than unexposed, same-sex siblings. This observation suggests that during the Dutch hunger winter, the famished mothers passed on an epigenetic state that resulted from the imposed selection for IGF2 to become less methylated at the DNA level. This could be in principal the result of a need for IGF2 to be expressed at greater levels for proper growth and development. From this study, numerous questions arise, such as how epigenetic memory can be maintained, what drives epigenetic memory, and can such a mechanism ultimately drive the selection of epigenetically modified individuals and result in altered phenotypes? Recent evidence implicates antisense non-coding RNAs as driver of epigenetic memory in humans.8,9
The majority of investigators think of RNA interference (RNAi) as a mechanism whereby small non-coding RNAs can suppress the translation of a genes mRNA. With the discovery of RNAi came the explosive discovery of several new small RNA effector molecules such as microRNAs (miRNAs) and piwi associated RNAs (piRNAs) (reviewed in ref. 10). The observation and characterization of these small RNAs has significantly changed the perception of how gene expression is controlled and definitively demonstrated that small non-coding RNAs can alter gene expression.11 While significant work has been carried out on the role of small non-coding RNAs such as miRNAs and piRNAs in post-transcriptional modes of gene regulation, far less is known regarding the role of longer non-coding RNAs and antisense non-coding RNAs in gene regulation and the extent to which genes may be regulated by these non-coding RNAs.
Non-coding RNAs are pervasive throughout the transcriptome of eukaryotes.12 Longer forms of non-coding RNAs appear to be ubiquitously expressed in human cells.13 It has been estimated that roughly 25% of non-coding RNA transcripts are antisense to known protein coding genes.14 Non-coding RNAs have been shown to be involved in X inactivation, dosage compensation, imprinting and polycomb mediated silencing.15 It is intriguing that non-coding RNAs do not appear to be universally conserved between various organisms.16 One interesting observation that may speak to the relative requirement for non-coding RNAs in higher organisms is that the ratio of non-coding DNA increases on an evolutionary scale, a pattern that is not evident in protein coding genomic regions.17 This observation implies a role for non-coding RNAs in complexity that could be envisioned to manifest via non-coding RNA guided epigenetic based modes of gene regulation. However, for non-coding RNAs to function in a regulatory manner, the epigenetic changes would need to be recruited to genomic regions that are operative in gene expression and fundamentally susceptible to epigenetic modifications, such as promoters. Supporting this notion is the observation that several non-coding RNAs have been found to be located at, upstream, or even overlapping 5' regions of protein coding genes.18-22 Many of these reported non-coding RNAs have been detected in the antisense orientation relative to the transcriptional orientation of the coding gene23 and (reviewed in ref. 24) and have been shown to contain some homology to known promoters.22,25 Recent observations in human cells by our group8 and others9 suggest that, mechanistically, antisense non-coding RNAs are active in the epigenetic regulation of gene expression. Furthermore, others have found miRNAs that can modulate gene transcription in a similar manner.26,27 Suppression of the antisense non-coding RNA or miRNA targeting can result in a relief of targeted silent state epigenetic modifications at the promoter, whereas the increased expression of the non-coding RNA or promoter targeted miRNA can result in increased epigenetic silencing at the targeted promoter.8,9,26,27 These observations when taken together support earlier notions that sequence specific non-coding antisense RNAs are involved in the epigenetic regulation of gene promoter activity.28
In a weird twist of fate, mainly as a result of the discovery of RNAi as a tool to study gene function, came the notion that RNAs could regulate transcription in humans.29 In plants such as Arabidopsis, in lower eukaryotes such as S. pombe, and in C. elegans, small double stranded, non-coding interfering RNAs (siRNAs) were shown to function in transcriptional gene silencing by the targeted recruitment of silent state epigenetic marks such as histone and DNA methylation to regions of the genome required for transcription (reviewed in ref. 30). In human cells however the ability of siRNAs to modulate gene transcription was less clear. This was presumably the result of human cells lacking an RNA dependant RNA polymerase (RDRP). Nevertheless, one can avoid the requirement of RDRP by simply generating siRNAs de novo and experimentally introducing them into the cell. When siRNAs were generated to target a genes promoter in human cells, transcriptional gene silencing of the targeted promoter resulted.29,31-40 These observations suggested two hypotheses concerning TGS in human cells; (1) that the machinery for RNA directed transcriptional gene silencing is endogenous and active in human cells, or (2) that the machinery for RNA directed transcriptional gene silencing is endogenous and vestigial in human cells. As these two hypotheses were the result of the small non-coding RNA being exogenous administered to the cell, whether an endogenous effector molecule existed and was driving this machinery remained unknown.
A relatively important insight into one of the possible effector molecules directing the epigenetic silencing pathway in humans was determined by experiments that showed single stranded antisense RNAs (asRNAs) could drive transcriptional silencing in human cells.37 This body of work was also the first observation that the small antisense non-coding RNAs interacted with DNA methyltransferase 3A,37 an enzyme involved in de novo DNA methylation.41 Mechanistically, the exogenously introduced small non-coding RNAs, either siRNAs or small antisense RNAs (asRNAs), direct the enzymatic machineries for methylation of histone 3 at lysines 9 and 27 (H3K9 and H3K27) (reviewed in ref. 42). Some targeted genes also present with methylation of DNA specifically at those promoters containing sequence homology to the non-coding RNA.29,31-39 Thus, the non-coding RNAs were presumed to guide a protein complex to the targeted promoter. This protein complex has now been shown to contain Argonaute 1(Ago-1) which is essential for the RNA-mediated transcriptional silencing in human cells.36,38,43 Recent experimental evidence suggests that histone deacetylase 1 (HDAC-1) and DNA methyltransferase 3a (DNMT3a) are also required for the induction of transcriptional gene silencing in human cells.37,40,44 The non-coding RNAs are able to recognize the targeted promoter by interactions with a low-copy promoter-associated RNA (pRNA) that spans the promoter during the act of transcription.19,45 This pRNA can essentially act as a scaffold for the recruitment of a transcriptional silencing complex (TSC) containing Ago-1, HDAC-1, DNMT3a and possibly several other currently unknown factors39,40,45 (Fig. 1).
Similar to the question of “junk DNA”, is TGS useful or simply retained information from our evolutionary past? While much of the mechanism has been worked out by previous studies (reviewed in ref. 42), whether this mechanism and the endogenous trigger are active or vestigial in humans had remained unknown until recently. This question appears to be on the pathway to resolution. Experimental evidence recently demonstrated that antisense non-coding RNAs are functional and active endogenous regulators of epigenetic gene regulation and gene transcription in human cells.8,9 Importantly, these antisense non-coding RNAs required the action of Ago-1 to direct epigenetic states of gene regulation,8 supporting the notion that the endogenous mechanism involved in siRNA or asRNA mediated forms of transcriptional gene regulation are in fact the same pathway utilized by endogenous antisense non-coding RNAs (Fig. 1). When all of these observations are taken together, a paradigm is beginning to emerge suggesting that antisense non-coding RNAs, which are estimated to represent roughly 25% of human gene transcripts,14 may in fact be actively utilizing the previously described transcriptional silencing pathway to epigenetically regulate gene expression.8,9
Our current understanding of how non-coding RNAs, be them siRNAs, asRNAs, miRNAs or longer forms of antisense non-coding RNAs, regulate gene expression suggests that RNA can function to direct protein effector complexes to target genomic loci. Mechanistically all of these forms of non-coding RNAs appear to be utilizing the same pathway that endogenous antisense non-coding RNAs utilize to target silent state epigenetic marks to the corresponding target genomic loci (Fig. 1). As such, the suppression of the antisense non-coding RNAs by RNAi can result in the loss of the antisense non-coding RNA suppressor and the subsequent loss of the ability to guide epigenetic silencing complexes to a particular locus.8,9 The result is the observation of gene activation of the targeted locus.8 But what might be the endogenous regulator of the antisense non-coding RNAs? Clues as to the regulator of bidirectionally transcribed genes have begun to emerge based on a relatively small subset of observations.
Several of the antisense non-coding RNAs observed to date have been found in bidirectionally transcribed genes, the majority of which appear, to be tumor suppressor genes.9 Much of the sequence overlap between the sense/mRNA and antisense non-coding RNA in those genes reported to exhibit bidirectional transcription appears in the 3' UTR of the coding/mRNA. However, evidence of these antisense non-coding RNAs spanning downstream regions encompassing the sense/mRNA promoter have also been reported.8 Of interest is the notion that the 3' UTR of coding/mRNAs is an area where a majority of the known miRNAs have been shown to interact to regulate a genes translational expression.46 Taken together these observations suggest that we may be overlooking one aspect of miRNAs—that they might play a role as differential regulators of bidirectionally transcribed genes. Theoretically, one would assume that as the surplus of sense/mRNA transcripts for a particular gene that exhibits bidirectional transcription increases, that the cell, no longer requiring such high levels of this mRNA, could modulate the expression by the miRNA binding the sense/mRNA (Fig. 2A). The miRNA binding to the sense/mRNA could result in unimpeded antisense non-coding RNA expression (Fig. 2B) and subsequent antisense non-coding RNA directed epigenetic changes to the sense promoter leading to reduced levels of gene transcription (Fig. 2C). Conversely, when the antisense is in relative abundance and the sense/mRNA promoter is epigenetically targeted (Fig. 2C), the miRNA would potentially bind the surplus antisense non-coding RNA (Fig. 2D). The result of the miRNA targeting the antisense non-coding RNA would be a release of the suppressive capabilities placed on the sense/mRNA promoter by the action of the antisense non-coding RNA (Fig. 2E) and an increase in sense/mRNA expression (Fig. 2F). This may offer an explanation for why miRNAs have so much flexibility in their binding abilities.46 The flexibility in binding targets might be a need to utilize the miRNA to differentially regulate bidirectional gene expression. Of note the majority of bidirectionally transcribed genes reported so far appear to be tumor suppressor genes.9 Several of these tumor suppressor genes have been reported to be epigenetically silenced in human cancers as well as to be genes that contain miRNA target sites in their respective 3' UTRs. Might this progression to oncogenesis to some extent be the result of uncontrolled antisense non-coding RNA directed silencing? Evidence supporting this supposition (Fig. 2) can be found in the tumor suppressor gene E-cadherin, where miR-373 can bind both the sense/mRNA coding 3' UTR or antisense non-coding RNAs.8 The loss of miR-373 could result in an imbalance of regulation whereby the antisense non-coding RNA for E-cadherin uncontrollably targets epigenetic silencing complexes to the E-cadherin promoter. Such uncontrolled targeting would be expected to result in stable epigenetic based silencing of the gene. Unfortunately, to date little is known concerning miRNA regulation of antisense transcripts.
The discovery that antisense non-coding RNAs are operative in endogenous forms of RNA based epigenetic gene regulation in human cells8,9 indicates that it might now be possible to exert superlative control over gene expression. For instance, a small antisense non-coding RNA can be designed to specifically target virtually any promoter. The result of this targeting would be epigenetic gene silencing that appears first in the form of histone methylation in the target locus followed thereafter by DNA methylation.40,47 Such targeting would ultimately lead to stable epigenetic silencing of the targeted locus. Alternatively, if a particular gene exhibits bidirectional transcription, then it is possible to design siRNAs to the antisense non-coding RNA. The suppression of the antisense non-coding RNA would result in the suppression of the suppressor and ultimately a loss of epigenetic silencing at the sense/mRNA promoter8 (Fig. 2). Such a scenario can result in target specific gene activation. As such it is now theoretically possible to pick your gene of interest and turn it on or off. Two conditions apply however, as turning the gene on appears to require the gene to be bidirectionally transcribed8 while turning the gene off requires that there be some level of transcription across the genes promoter.45
Efficient, targeted control of gene expression represents a potentially powerful scientific and medical tool which can also raise ethical questions and concerns. As evident by the Dutch hunger winter, epigenetic changes can be passed from parents to their offspring, suggesting exerting exogenous control over endogenous non-coding RNAs could have the potential to impart epigenetic changes on daughter cells as well. We must therefore consider how the development of RNA based therapeutics will affect current epigenetic patterns and how these changes might impact human evolution. Recent evidence suggests that naked siRNAs administered in vivo are found to localize to the spleen, prostate, lung, liver and testis.48 Widespread localization of naked siRNAs targeted in epigenetic manners could result in unintended off-target effects in various tissues, potentially altering the epigenetic profile more than originally intended. The risk exists that we inadvertently silence a gene not for one person but for future generations, altering the natural evolution of that gene. However, we cannot yet fully understand the potential for these changes to be passed on to future offspring by parents undergoing RNA based therapies, as our understanding of epigenetics remains in its infancy. Targeted gene control has the potential to be a powerful therapeutic, however use should be limited to post-reproductive individuals to minimize the potential risks of RNA directed epigenetic modifications for future generations.
This project is funded by R01 HL083473 to K.V.M. I thank Paula J. Morris at Seainsite http://seainsite.com/index.html for the generation of figures and Anne-Marie Turner for critical evaluation of this manuscript.