The theory of RNA-mediated gene evolution

Abstract

Observations over the last decade suggest that some RNA transcripts, such as non-coding RNAs, function in regulating the transcriptional and epigenetic state of gene expression. DNA methylation appears to be operative in non-coding RNA regulation of gene expression. Interestingly, methylated cytosines undergo deamination to remove the methylation, which if not properly repaired results in the methylated cytosine being recognized by the cell as a thymine. This observation suggests that the process of non-coding RNA-directed epigenetic targeting also has the potential to alter the genomic landscape of the cell by changing cytosines to thymines and ultimately influence the evolution of the cell. This proposed theory of “RNA-mediated gene evolution” might be one possible mechanism of action whereby RNA participates in the natural selective process to drive cellular and possibly organismal evolution.

RNA Functions as Driver of Epigenetic and Transcriptional States

Sequencing of the human genome was one of the greatest milestones in science. Almost immediately following this monumental achievement it became apparent that the vast majority of the genome—roughly 80%—is not involved in generating proteins but rather in producing non-coding transcripts.^1-4 These observations have proven perplexing, because it had previously been assumed that the vast majority of the human genome was functionally inert. For years, various notions have been put forth as to the function of these non-protein-coding regions in the human genome; ranging from acting as genomic spacers involved in partitioning genes to structural entities involved in chromosomal integrity and folding. The rationale for these interpretations was the presumption that the majority of this “junk DNA” was not transcriptionally active.^5-7 This viewpoint, however, has been radically altered as a result of exciting findings generated from the Fantom⁸ and Encode consortiums,^3,4 whereby a paradigm shift has occurred, suggesting that this genomic dark matter—the non-coding or “junk DNA”—is actually transcribed and active.

While it is now evident that the vast majority of the human genome is transcribed into both coding and non-coding transcripts,^4,7,8 the function of these long non-coding RNAs (lncRNAs) has until recently remained largely unknown. Studies carried out over the last decade have demonstrated that many non-coding RNAs are functional in the regulation of gene transcription (reviewed in^9,10). Mechanistic studies in human cells with both small and long forms of non-coding RNAs have found that these transcripts interact with the proteins involved in epigenetic cellular processes of regulating chromatin states and gene accessibility (Fig. 1).

Figure 1.

Non-coding RNA-directed transcriptional silencing. (A and B) Small non-coding RNAs have been observed to target epigenetic silencing complexes to actively transcribed loci, resulting in stable transcriptional gene silencing (reviewed in^10,39). This form of silencing involves the action of DNMT3a interacting directly with the small non-coding RNAs^20,21,44,45 and directing both histone and DNA methylation to the homology containing loci.^17,22 This targeting can ultimately result in stable and heritable epigenetically silent loci.¹⁷ (C and D) Long non-coding RNAs that are antisense to particular loci have also been found to interact with DNMT3a¹⁹ and modulate transcription.^19,46-48

Epigenetics is the study of heritable changes in gene expression caused by mechanisms other than changes in genomic DNA, and focuses predominantly on changes to the chromatin. Chromatin is composed of DNA and proteins, and epigenetics is the modifications to these elements. Changes to the local chromatin architecture can drastically affect the accessibility of a gene to outside influences. Examples abound of epigenetic changes that range from histone modifications to DNA methylation, both of which can result in transcriptional silencing of a particular locus (reviewed in¹¹).

Fascinating studies have found that some phenotypic traits can be passed on to offspring solely by the action of epigenetic changes at particular genetic loci. These observations, termed transgenerational epigenetic inheritance, suggest that our understanding of evolution is incomplete, and that some level of epigenetic heritability is also functional in evolution. Studies carried out in nematode, mouse, and pig model systems all suggest that transgenerational epigenetic inheritance involves non-coding RNAs.^12-14 There is also evidence for this phenomenon in humans, suggesting that some level of epigenetic inheritance is operative in the fabric of our existence.^15,16 Furthermore, work carried out in human cell culture systems has found that small non-coding RNAs direct epigenetic changes that can be passed on to daughter cells in a long-term heritable manner.¹⁷ Collectively, these observations suggest that non-coding RNAs emanating from essentially “junk” DNA, including pseudogenes and natural antisense transcripts (NATs)(reviewed in¹⁸), may actually be involved in not only controlling transcription but also in the natural selection of the genome and the cell via transgenerational epigenetic inheritance. Indeed, we are only now realizing how intricately involved both small and long non-coding RNAs are with regards to controlling gene expression in human cells.⁷

The Theory of “RNA Mediated Gene Evolution”

Based on the emerging realization that some non-coding RNAs are active modulators involved in controlling the transcriptional and epigenetic processes, and that the results of this process appear to be heritable and passed on to offspring, I theorize that an “RNA-mediated form of gene evolution” is endogenous to human cells. This evolutionary process plays out via perturbations to the endogenous non-coding RNA, DNA methylation or DNA deamination pathways. The purported mechanism of action driving this process in human cells is shown schematically (Fig. 2). The notion presented here is that non-coding RNA-associated recruitment of DNA methyltransferase 3a (DNMT3a) and Enhancer of Zeste 2 (EZH2) to targeted loci can result in those cytosines at the target locus undergoing DNA methylation^17,19-23 (Fig. 2A–D). Typically, such methylated cytosines (C) are subjected to the endogenous removal and repair process of deamination, whereby APOBEC3A deaminates the methylated cytosine to a thymine (T),^24-26 which is subsequently followed by repair with thymine DNA glycosylase (TDG)²⁷ (reviewed in²⁸)(Fig. 2E–I). However, if this process is perturbed by some selective condition, and the deaminated nucleotide not properly repaired, then some methylated cytosines will undergo a hydrolysis reaction resulting in the production of ammonia and the retention of the converted methylated cytosine as a thymine in the DNA sequence. Following DNA replication the cell will recognize this deaminated cytosine as a thymine, which will become fixed in one of the 2 daughter strands^29,30 (Fig. 2J and K). While this cytosine to thymine (C${\displaystyle \to }$T) conversion has been considered by some to be random, the spontaneous deamination of methylated cytosines has been found to be ∼2-fold faster than non-methylated cytosines,³⁰ suggesting a bias toward cytosine methylation at CpG regions in the deamination process. In light of these eventualities I envision a cellular process involving RNA and in particular non-coding transcripts that under certain conditions—such as increased lncRNA expression or decreased deamination repair—result in the alteration of the nucleotide content at a particular locus (shown schematically Fig. 2). Note, this entire process involves both RNA and proteins; the RNAs providing target recognition capabilities and the protein components providing the epigenetic modifying and DNA repair components (Fig. 2).

Figure 2.

Non-coding RNA-directed gene editing. (A) Antisense long non-coding RNAs function in trans¹⁹ (and possibly also in cis—not shown) to target epigenetic regulatory complexes to homology-containing loci. (B) This targeting has been found to require DNMT3a and possibly several other protein components known to interact directly with DNMT3a or be involved in epigenetic silencing. (C) The epigenetic silencing is carried out resulting in (D) histone 3 lysine 9 di-methylation (H3K9me2) or histone 3 lysine 27 trimethylation (H3K27me3) and DNA methylation at the homology-containing target loci, chromatin compaction, and, ultimately, gene silencing (reviewed in^18,49). (E) The methylated cytosines at particular CpG sites of the RNA-targeted locus undergoes deamination by the actions of APOBEC3A and (F and G) a thymine becomes incorporated in place of the methylated cytosine. (H and I) This thymine is typically repaired by TDG, MBD4, and SMUG, but (J) if this process is disrupted and/or replication occurs before repair of the locus is completed then the thymine may become fixed in the DNA such that (K) those transcripts emanating from the affected locus are altered and the cytosine becomes replaced with a thymine. These changes might inhibit future trans non-coding RNA targeting as well as affect the translated product or promoter activity and/or transcription factor binding at the RNA-targeted deaminated locus.

Conditions for such occurrences in the genome could be: (1) the result of robust non-coding RNA targeting of a particular locus leading to increased or overt DNA methylation or; (2) when DNA replication occurs without proper repair to epigenetically modified nucleotides. Such conditions could lead to some methylated cytosines undergoing accidental deamination resulting in stable C${\displaystyle \to }$T transitions in the genomic code. The result of such C${\displaystyle \to }$T transitions are essentially gene editing at non-coding RNA targeted sites (Fig. 2). This mechanism of action is perhaps an inherent aspect of using DNA cytosine methylation to regulate gene expression. Indeed, the methylation of cytosine in CpG residues is recognized as an important avenue of epigenetic regulation,³¹ cytosine to thymine changes represent the most common single nucleotide mutation found;^32,33 furthermore, this mutation is also found to be involved in the evolution of transcription factor binding sites.³⁴ With regard to human cells, it is difficult to resist the temptation to speculate that the methylation of CpG residues acts as a selective force driving the evolution of the genome and also in human disease.³³ While such a pathway (Fig. 2) is surmised here to exist, there is limited direct experimental validation for this notion that RNA acts to shape the genomic content of the genome. However, there are several observations over the last decade that suggest that non-coding RNAs function as mediators of selection involved in the evolution of the cell, and there are clear experimental avenues to testing this theory (see below).

Experimental Validation of the Theory of “RNA-Mediated Gene Evolution”

Experimental observations in human cells suggest that methylated cytosines are repaired by the cell using a process of deamination, where APOBEC3A deaminates the methylated cytosine to a thymine,^24-26 which is subsequently followed by repair with TDG. Other glycosylase proteins may also be involved in this repair pathway, including methyl-CpG (mCpG) binding domain protein 4 (MBD4),^35,36 SMUG,³⁷ and uracil DNA glycosylase (UDG),²⁷ which can repair those unmodified cytosines that have been deaminated by APOBEC3A to a uracil (reviewed in³⁸)(Fig. 2). It is noteworthy that UDG has also been observed to interact directly with HDAC-1, which interestingly appears to be a required component in non-coding RNA directed epigenetic regulation (reviewed in^10,39)(Fig. 1). This observation also suggests an intersection between both pathways of interest: RNA-directed epigenetic gene silencing and repair of deaminated cytosines.

Collectively, data generated over the last decade suggest that: (1) non-coding RNAs direct epigenetic gene silencing and DNA methylation to those CpGs in the RNA targeted locus^17,22,40 (Fig. 1); (2) methylated cytosines that do not undergo proper deamination and repair can become incorporated into the genome as thymines (Fig. 2); and (3) these 2 processes have overlapping protein components, namely DNMT3a and HDAC-1. This overlap in proteins provides for a starting point to experimentally validate the theory of RNA-directed gene evolution. One experimental approach is to use a stable and inducible cell lines expressing small¹⁷ or long non-coding RNA¹⁹ to induce RNA-targeted epigenetic silencing (as shown in Fig. 1) in the absence of the endogenous deamination repair pathway (Fig. 2). This works by suppressing APOBEC3A, TDG, SMUG, UDG, and MBD4 via RNA interference or CRISPR/Cas9.⁴¹ These cultures could then be assessed for the genesis of C${\displaystyle \to }$T transitions using either a surveyor assay,⁴² a quantitative RT-PCR small nucleotide polymorphism assay, or deep-sequencing of the RNA-targeted locus in the various cultures.

Conclusions

Although the notions presented here have yet to be experimentally substantiated, one can envision a model whereby RNA-driven genetic variation, followed by selection, could work (Fig. 2). One possible area in which such an evolutionary mechanism could be operative is in the evolution of transcription factor binding sites.³⁴ Cytosine-to-thymine transitions seem to be involved in the evolution of transcription factor binding sites; moreover, promoter elements—where transcription factor binding sites are ubiquitous—are often bona fide targets for antisense lncRNAs (¹⁹ and reviewed in⁴³).

Paradoxically, the ability of this system (Fig. 2) to constitutively target and specifically control a locus is limited. For instance, the greater frequency of cytosine to thymine changes occurring at a particular RNA-targeted locus could result in an overall loss of sequence complementarity between the effector RNA and its targeted locus-specific transcript (Fig. 2K). This eventuality could lead to a loss in the capability of the RNA to target a particular locus, while simultaneously permitting the targeted transcript to fold into a different conformation. As such, changes in the non-coding RNA transcript might lead to a loss of its targeting abilities and/or its susceptibility to different interactions with other lncRNAs and/or proteins (Fig. 2). Such a notion would play out with trans targeted RNAs while a different scenario might play out in the cis functional RNAs, possibly resulting in more stabilized RNA:RNA interacting regions of the genome. Indeed, the possibilities become infinitely complex, an eventuality that may be advantageous in the evolution of biological systems. No doubt as data emerge in this exciting area of research, so will additional layers of regulation of the central dogma of molecular biology and our understanding of the role of RNA in basic evolutionary processes.

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.

Funding

This project is funded by 1P01AI099783–01 to KVM and Australian Research Council Future Fellowship FT130100572.