The Persistent Contributions of RNA to Eukaryotic Gen(om)e Architecture and Cellular Function

Abstract

Currently, the best scenario for earliest forms of life is based on RNA molecules as they have the proven ability to catalyze enzymatic reactions and harbor genetic information. Evolutionary principles valid today become apparent in such models already. Furthermore, many features of eukaryotic genome architecture might have their origins in an RNA or RNA/protein (RNP) world, including the onset of a further transition, when DNA replaced RNA as the genetic bookkeeper of the cell. Chromosome maintenance, splicing, and regulatory function via RNA may be deeply rooted in the RNA/RNP worlds. Mostly in eukaryotes, conversion from RNA to DNA is still ongoing, which greatly impacts the plasticity of extant genomes. Raw material for novel genes encoding protein or RNA, or parts of genes including regulatory elements that selection can act on, continues to enter the evolutionary lottery.

Many features of eukaryotic genomes (e.g., the modular rearrangement of genes) may have their origins in an RNA or RNA/protein world.

Everything has been said already, but not yet by everyone.

—Karl Valentin

Sturgeon's Revelation: Ninety percent of science fiction is crud, but then, ninety percent of everything is crud.

—Theodore Sturgeon

They think that intelligence is about noticing things that are relevant (detecting patterns); in a complex world, intelligence consists in ignoring things that are irrelevant (avoiding false patterns).

—Nassim Nicholas Taleb (Taleb 2010)

Of all extant cellular macromolecules, RNA is the most ancient, persisting as much as 4 × 10⁹ years in our planet’s life-forms. The ability to combine genotype with phenotype such as catalytic activity (Noller and Chaires 1972; Kruger et al. 1982; Guerrier-Takada et al. 1983; Noller et al. 1992) leveled a major hurdle in understanding the origin of life. The salient discoveries eliminated the virtually impossible prerequisite for two to three different classes of macromolecules to converge as an evolving unit. At the same time, RNA provides a required continuity in the path of evolution (Yarus 2011) during various genetic takeovers or evolutionary transitions (Cairns-Smith 1982; Szathmáry and Smith 1995). In a remarkably insightful article dating back half a century, Alex Rich foresaw much of what now is becoming mainstream, for example, that RNA was ancestral to protein and DNA (Rich 1962). This landmark publication received little attention over the years; even early proponents of an RNA world did not refer to this article (Woese 1967; Crick 1968; Orgel 1968; Gilbert 1986), although at least one of the investigators must have had knowledge about the article, as it was cited in a different context concerning the stereochemical possibility of six distinct base pairs (Crick 1968). The origin of the DNA genome from RNA and that “DNA may be regarded as a derivative molecule which has evolved in the form that it only carries out part of the primitive nucleic acid function” is another correct prediction (Rich 1962). Furthermore, the investigator presaged mechanisms such as antisense RNA control of gene expression, short interfering RNAs (siRNAs), and perhaps microRNAs (miRNAs): “If both strands are active, then the DNA would produce two RNA strands which are complementary to each other. Only one of these might be active in protein synthesis, and the other strand might be a component of the control or regulatory signal” (Rich 1962).

In this article, I shall present the rise and persistence of RNA from the dawn of an RNA world and discuss current evolutionary principles already apparent in an RNA world. In comparison to Archaea and Bacteria, the eukaryotic genome is a better vantage point, as archaeal and bacterial genomes are more derived and, thus, lost many of the RNA signatures that eukaryotes still show. It is likely that eukaryotic DNA genomes not only kept much more of their RNA/RNP world heritage than previously anticipated, but also continue to evolve novel RNAs in various functional roles.

WHEN DOES LIFE BEGIN?

An excellent treatise of possible scenarios leading to and continuing in an RNA world to the last universal common ancestor (LUCA) is available (Atkins et al. 2011). Can the beginning of life be defined along the transitions from physicochemical to biological reactions? Like almost everything in biology, clear boundaries are difficult to demarcate and thus the definition of the first life-form rather occupies a bandwidth on a continuum. One of several possible thresholds to consider would be the fortuitous generation of one or two molecules that could replicate themselves or each other. Should the threshold be set at the transition when the molecules involved could change during replication and the variants are subjected to selection—the initial Darwinian ancestor (IDA) (Szathmáry 2006; Yarus 2011)? The first self-replicating macromolecules must not necessarily have been RNA. Derivatives of RNA, especially with altered backbones, have been suggested as predecessors of RNA owing to more favorable chemistries/stabilities for spontaneous generation and persistence of oligomerization at the presumed planetary conditions (Joyce et al. 1987; Schöning et al. 2000; Zhang et al. 2005; Powner et al. 2009; Robertson and Joyce 2012; Neveu et al. 2013).

An interesting question is, if it is that “simple,” why did life not evolve multiple times? There are a number of explanations. The early environment of the planet with conditions favoring the necessary chemical reactions differed from the more temperate conditions now (Robertson and Joyce 2012). Perhaps life did evolve before LUCA numerous times independently, but the descendants of LUCA are the only survivors. Perhaps primitive forms of life, for example, in the form of IDAs, still do arise, but we are not aware of them, in part, because we have not searched for such simple and different life-forms. Another reason is that a nascent form of life would easily be outcompeted by the established ones, as the latter had a great chronological advantage adapting to current conditions. New forms of life might have a chance only if their metabolism is sufficiently different and, thus, not useful prey to LUCA-related life-forms or if they happened to evolve in an unoccupied niche so as not to succumb to immediate predation by the fitter “incumbents.”¹

RETRACING THE PATH

In any event, by applying in vitro synthesis and selection procedures (Ellington and Szostak 1990; Tuerk and Gold 1990; Gold et al. 2012), several laboratories are making great strides toward generating RNA molecules with the ability to self-replicate (Doudna and Szostak 1989; Johnston et al. 2001; Zaher and Unrau 2007; Lincoln and Joyce 2009; Shechner et al. 2009; Wochner et al. 2011; Attwater et al. 2013; Mast et al. 2013), although these RNA polymerase ribozymes still fail to completely self-replicate (Deamer 2005). Cooperation of two or more RNA enzymes in hypercycles (Eigen and Schuster 1977) may be a solution to this problem (Vaidya et al. 2012) (see also below).

Once a self-replicating ribozyme (mono- or multimeric) arose with the further potential to evolve into a replicator not restricted to only self-copy but to copy other RNA templates as well, a prerequisite for a metabolically self-sufficient RNA conglomerate, further challenges are apparent. First, the replicator indiscriminately copying any “junk RNA” in the mix hardly would be able to persist. Second, if further RNA molecules would arise by copying with errors—just like in extant organisms, new genes still arise by duplication and variation—to eventually take over metabolic functions other than replication (e.g., activated compounds, including nucleotides), these associations would be fleeting, at best, because of diffusion.

SEQUESTRA, AMPLIFICA, DIVIDE ET IMPERA!

A compartmentalization of cellular constituents in droplets, as suggested by Carl Woese (1979), stabilized by simple fatty acids (Szostak et al. 2001) was an early evolutionary transition (Maynard Smith and Szathmáry 1995) crucial for the continuation of life. Recent experiments showed that such bilayered partitions were sufficiently permeable for uptake of small molecules from the environment and copying nucleic acids in such vesicles is possible (Sacerdote and Szostak 2005; Mansy et al. 2008; Adamala and Szostak 2013). This sequestration was a prerequisite for cellular life and evolution as we know it today, with many evolutionary principles in place already (Brosius 2003c). Vesicles could grow along with their RNA contents, divide, fuse while shuffling their contents and divide again, in other words, performing sexual acts. Another “forecast” of the mechanisms generating genomic diversity would be recombination not only between cells, but also between different RNA molecules, a mechanism that actually had been observed in a two-component ribozyme system (Lincoln and Joyce 2009). The origin of viruses could date to this early stage of cellular evolution as well. RNA molecules could, perhaps protected by a lipid envelope, move from cell to cell, blurring the line between infection and horizontal transfer.

MOST EVOLUTIONARY PRINCIPLES ARE AT LEAST AS ANCIENT AS THE RNA WORLD

Lessons from the RNA world apply remarkably well to extant organisms and their genomes. In a primitive RNA cell, conflict and cooperation, selfishness and altruism had to coexist and establish a fine balance. Importantly, the success of individual ribozymes also depended to a large degree on functional interactions with other cellular RNAs (today: gene products), namely, the (genetic) background of the protocell (Brosius 2003c). In contemporary biology and medicine, considerations of interactions of various alleles within the genomic background, as well as individual variability of gene expression levels including complete gene depletions, are beginning to gain wider acceptance (Williams and Nesse 1991; Sibilia and Wagner 1995; Brosius 2003b; Ganten and Nesse 2012; Nesse et al. 2012).

Balances between selfishness and cooperation had to evolve early; this is recently documented by in vitro selection research that revealed in an experiment to generate Tetrahymena group I ribozymes with improved DNA cleavage capability that one of the two RNA aptamers, itself being catalytically inactive, participated in a productive intermolecular interaction with an active ribozyme evolving in parallel, thus, ensuring the survival of both RNAs in the nucleic acid population (Hanczyc and Dorit 1998). Other studies also show that parasites can cause the evolution of further complexity (Takeuchi and Hogeweg 2008). In the case of the bacterially derived AzoΔ ribozyme, a clonal preparation showed no activity toward the phosphorothioate substrate. Presumably, this sequence alone fails to show a functional fold, but could form an active complex in an intermolecular partnership with other RNA molecules (Hayden et al. 2011). Recently, it was shown that cooperative cycles of replication involving three or four participating RNAs have a selective advantage over selfish replication cycles (Vaidya et al. 2012), as formulated earlier in the hypercycle principle of natural self-organization (Eigen and Schuster 1978) and placed into context with extant organisms (Brosius 2003c). The participation of several small RNAs also has the potential to significantly increase the complexity of ribozymes.

CHROMOSOMAL RNA

The dispensation of essential RNA molecules between daughter cells was initially stochastic, probably facilitated by the availability of sufficient copies of each RNA species in the parent cell, such that each daughter would have a reasonable chance to end up with at least one complete set of RNAs. If initially a few different RNAs were strung together as “mini RNA chromosomes,” the advantage would be achievement of a more balanced distribution of RNAs. Perhaps they were distributed and replicated as such, with some of them cut up into functional RNAs, conceivably the birth of RNA processing. Processing signals might have been placed, in part, between the fragments corresponding to mature RNAs, presumably the origin of intergenic or even intronic space (Brosius 2003c). With the advent of templated protein biosynthesis, a major evolutionary transition included co-option of existing functional RNAs as well as longer “chromosomal RNAs” by RNA cutting and pasting to stitch together messenger RNAs (mRNAs) with open and increasingly longer reading frames, which might suggest a very early origin of RNA splicing in a ribonucleoprotein world (RNP world) (Reanney 1974; Darnell and Doolittle 1986). Alternative splicing and other rearrangements would be one of the mechanisms to enhance variation for generating translation products out of a limited repertoire of functional RNAs also doubling as templates for translation (Brosius 2001).

RNA SIGNATURES WRITTEN ALL OVER EXTANT DNA GENOMES

Linear arrangement of RNA genes, as well as the transition to the RNP world evolving an ancestor of the reverse transcriptase enzyme, happened to constitute a useful precondition for a next major evolutionary transition, namely, the conversion of RNA to DNA, the latter merely serving as bookkeeper (Darnell and Doolittle 1986; Gould 2002). For this and other reasons, the central dogma of biology could be revisited or supplemented by a different graphic account, both in appreciation of the major significance of RNA in the cell and the chronological order of the major transitions (Fig. 1) (Maizels and Weiner 1987; Brosius 2003a; Cech 2012). Remarkably, the process of reverse transcription of any RNA (Brosius 1999b) and more or less random integration into DNA genomes still persists and contributes much to their landscapes in a number of ways. The modular arrangement of functional with nonfunctional DNA and its plasticity in extant eukaryotic genomes might be a vestige of early evolutionary transitions (Brosius 2009).

Figure 1.

Alternative to the central dogma of molecular biology. The left part depicts the original central dogma of molecular biology with several adjustments incorporated, for example, the discovery of reverse transcription (thin upward arrow) (Crick 1958, 1970). The grouping on the right better reflects the evolutionary transitions and primacy of RNA. RNA could be replicated directly (semicircular arrow), albeit in extant organisms (e.g., plants, viruses) only with the catalytic activity of protein (RNA-dependent RNA polymerase). In this scheme, catalysis is indicated by (RNA) and or (protein) in parentheses. The major significance of RNAs for peptidyl transferase activity during translation (Noller 2012) is represented by the larger font for (RNA) compared to (Protein). Execution includes structural, catalytic, and regulatory tasks in the cell. The evolutionary developments underscore Stephen Jay Gould’s view that DNA merely is the agent of bookkeeping (Gould 2002). RNAs used to be bookkeepers as well, but remained agents of causality (Brosius 2005a).

Reverse transcriptase not only played an important role in generating the DNA genome but continues to be essential for chromosome maintenance via the action of telomerase. The extant enzyme synthesizes telomeres that serve as protective caps of chromosome ends, thus counteracting their shortening during replication cycles. Telomerase is a complex between RNA and protein in which the RNA serves as template and protein as reverse transcriptase (Greider and Blackburn 1989; Blackburn et al. 2006; Blackburn and Collins 2011). Group II introns, when transcribed as RNA are self-splicing with the activity residing on the RNA at high salt conditions in vitro (van der Veen et al. 1986). For in vivo activity and integration at new genomic sites in a process termed reverse splicing, a reverse transcriptase encoded in the intron is necessary (Lambowitz and Zimmerly 2011). The ribozyme has been proposed to be ancestral to non-LTR (long terminal repeat) retroposons as well as the spliceosome attributable to similarities of their reverse transcriptase and some small nuclear RNA (snRNA) components of the spliceosome, including aspects of the splice mechanism itself, respectively (Xiong and Eickbush 1990; Guthrie 1991; Sharp 1991; Fica et al. 2013). The rapid spread of introns in eukaryotes has been ascribed to group II introns after they were imported by endosymbionts (Cavalier-Smith 1991), perhaps necessitating the separation of slow mRNA production in the nucleus from the fast translation in the cytoplasm, one of the hallmarks of the eukaryotic cell (Martin and Koonin 2006).²

Importantly, the process of continuously converting RNA to DNA and its genomic integration has the potential to grossly inflate genomes with neutrally evolving material and, in conjunction with larger deletion of segments by recombination, leads to a high turnover rate of sequences on an evolutionary time scale, once more calling into question the ENCODE claim that ∼80% of the human genome is functional (Doolittle 2013; Graur et al. 2013; Niu and Jiang 2013). Generally, specific retroposons³ become active in certain lineages, and reverse transcripts from one or several master gene transcripts, such as LINEs (long interspersed elements), autonomous as they harbor gene encoding the retroposition machinery such as reverse transcriptase and nonautonomous SINEs (short interspersed elements), such as Alu or B1 elements. Nonautonomous retroposons rely on the machinery of the autonomous retroposons. Most retrocopies are not active as they are transcriptionally silent because of the lack of internal (e.g., truncation of LINEs) or external promoters in SINEs. Intact and transcribed master genes can spawn, over a few tens or hundreds of million years, more than one million copies, as is the case for Alu elements (Weiner 2006). With a few exceptions (see below), there is no selective pressure on such elements and, accordingly, they deviate over time from the consensus sequence. The human genome has been estimated to contain ∼43% retroposons, a relatively small amount of DNA transposons (∼3%), and as little as 5% conserved sequences (Lindblad-Toh et al. 2011; Ward and Kellis 2012) that potentially have function (Fig. 2).⁴ As argued before, conversion of RNA to DNA is an ancient process and it has been suggested that most DNA in the human genome has been derived by a virtually unabated bombardment of retroposons (Brosius 1999a). Over time, these elements are blurring into oblivion as randomized sequences by incessant changes, such as point mutations and small indels. Recently, it was confirmed by more sensitive computational strategies involving P-clouds, that almost 70% of the human genome may harbor repetitive elements (de Koning et al. 2011). Extrapolating back to or forward from the origin of RNA → DNA genome transition, there is no reason why this figure, except for the generally lesser contributions of DNA transposons, should not approach almost 100% (Brosius 1999a). The fact that essential sequences are interspersed with nonessential sequences as landing pads for transposed elements lessening their detrimental impact genome-wide could be chalked up to the ENCODE project (Bernstein et al. 2012), as well as adherents to intelligent design, creationists, and the like as “functional,” which would cleanse our species from the blemish of living with ∼80%–90% junk DNA in our genomes (Doolittle 2013; Graur et al. 2013; Niu and Jiang 2013).

Figure 2.

Exaptation of a new gene module at any stage of transposed element (TE) deterioration. In this example, a SINE, such as Alu, B1, identifier (ID), or mammalian-wide interspersed repeat (MIR), is retroposed into the first intron of a gene (introns depicted by thin black lines and the SINE as wide yellow bar). The gene harbors three exons (in blue). Open reading frames (ORFs) are shown as wide bars, whereas the 5′-UTR (untranslated region) and 3′-UTR are shown as narrower blue bars, in the terminal exons, respectively. The transcription promoter is depicted in red. The four top representations indicate the gradual decay of the SINE element by continuous mutation fading from yellow to white; thus, blending into the other anonymous sequence of the intron. At any of those stages, part of the TE can be exapted as novel exon (framed in black). Sometimes part of a discernible TE and adjacent anonymous intron can be exapted as novel gene module (bottom).

YESTERDAY’S JUNK COULD BECOME TOMORROW’S NOVEL GENE MODULE, IF ONLY TEMPORARILY SO

The lack of purifying selection concerning nonharmful TEs, including mRNA-derived retrocopies, should not divert from the fact that, occasionally, such sequences can be exapted as genetic novelties (Brosius 1991; Brosius and Gould 1992). Gene duplication as a means of generating novel genes has been realized for a long time (Haldane 1933; Muller 1935; Bridges 1936; Lewis 1951; Stephens 1951; Nei 1969). Gradual change from a duplicated gene, via those encoding isoforms, up to the acquisition of novel functions including subfunctionalization (Lynch and Force 2000) is now well documented (Roth et al. 2007; Kaessmann 2010; Chen et al. 2013). Gene amplification can occur via segmental duplication (Bailey and Eichler 2006) or retroposition (Brosius 1999a; Babushok et al. 2007) with a lower “success rate” for the latter. Retroposition of a copy of the mature mRNA requires the fortuitous presence of a promoter element upstream, which, as a potential benefit, could immediately alter regulation of the retrogene in comparison to the parent gene (Brosius 1999b). Genes, chiefly those encoded by endogenous retroviruses, were independently exapted or domesticated numerous times into novel functions, some of them are meanwhile essential for procreation or survival (Volff 2006, 2009). Also, DNA copies of nonretroviral RNA genomes, or parts thereof, can be integrated into genomes (Koonin 2010). In fungi, exaptations of such genes were reported (Taylor and Bruenn 2009), underscoring once more the notion that any RNA can be a template for the retroposition machinery (Brosius 1999b).

Can novel genes arise de novo from previously gene-free neutrally evolving genome regions? Despite some false positives (Monte et al. 1997; Kriegs et al. 2005), recruitment of entire protein-coding genes out of neutrally evolving sequences does occur (Long et al. 2003, 2013; Heinen et al. 2009; Kaessmann 2010; Carvunis et al. 2012; Murphy and McLysaght 2012; Neme and Tautz 2013). Recently, it has been emphasized that many long non-protein-coding RNAs originated from TEs (Kapusta et al. 2013). This RNA class is discussed in more detail below. In any event, it is much more common that novel gene modules are being added to existing genes. Such co-opted or exapted modules can be derived from inter- or intragenic space and constitute novel protein-coding exons and regulatory regions such as promoters and enhancers (Brosius 2005b, 2009; Baertsch et al. 2008; Rebollo et al. 2012).⁵ A prominent case is the exonization of parts of Alu elements, usually as alternatively spliced exons (Makalowski et al. 1994; Nekrutenko and Li 2001; Sorek et al. 2002; Lev-Maor et al. 2003; Krull et al. 2005; Shen et al. 2011). Many such events are, over evolutionary time, not stable. That is, they persist for a certain time if they are not or only slightly detrimental, especially when the novel splice product only constitutes a fraction of the functional RNA. Should, over time, the novel splice variant happen to become beneficial, it will be under purifying selection and, by point mutations in and around splice signals, its ratio, in comparison to the canonical splice form, might change in its favor. In a phylogenetic study on primates, it has been shown that such exons derived from Alu elements are lost at a high rate in the trial and error mode, typical for the evolution of novelties (Krull et al. 2005). A similar study involving all mammals and the older MIRs revealed that many of the events were already fixed and currently are under negative selection. The study also shows, in the case of a relatively recent exonization of an ancient MIR element, that exaptation can occur at any stage of TE decay (Fig. 3), and consequently also with any nondescript randomized DNA (Krull et al. 2007). Recently, a mechanism was described by which a snoRNA (small nucleolar RNA) extended its RNA-coding region by alternative processing of its 3′ end by introducing a single-point mutation near the site important for processing, generating both the canonical snoRNA and an extended variant (Mo et al. 2013). This event, detected in a cluster of rat Snord115 snoRNA genes in the imprinted Prader-Willi syndrome locus, but not in mouse, must be relatively recent. The additional gene product, L-Snord115, has most likely not yet acquired a function. The odds are that this snoRNA variant will not survive another dozen or so million years.

Figure 3.

The functional versus nonfunctional human genome. The pies represent the various segments of the genome applying more conservative estimates. mRNAs (∼1.3% accounting for ORFs) including their 5′- and 3′-UTRs (∼0.8%) occupy slightly >2% of the human genome. All other RNAs (non-protein-coding RNAs) are estimated to cover about 5% (green), maximally 10%. This figure is under debate. Likewise, there is a conservative estimate for regulatory elements of ∼8%. Other sources argue up to 20% (Bernstein et al. 2012; Shen et al. 2012). Because conserved non-protein-coding DNA elements (CNEs, ∼2%, purple) often encode cis-regulatory elements (Hiller et al. 2012), the other regulatory elements are shown as 6% only (blue). TEs derived from RNA intermediates contribute ∼43% (yellow) and DNA transposons only 3% (brown). The remainder (39%) corresponds to anonymous, scrambled sequences (white). The estimated 26% introns (Bernstein et al. 2012) occupy mostly the yellow and white segments; intronic regulatory elements and encoded RNAs (e.g., snoRNAs, miRNAs) are accounted for in the respective segments. The pie on the right is identical, except that the retroposons (yellow) and, to a lesser extent, the DNA transposons (brown), blend into the anonymous sequence (white) to reflect TE ancestry of the nondescript sequences as well.

ALL WIRED UP ON REWIRING

Examples for TE contributions to gene expression by providing promoter or enhancer modules have been known for some time and are ample (Jordan et al. 2003; van de Lagemaat et al. 2003; Medstrand et al. 2005). Interestingly, TEs that were conserved at unusually high levels over hundreds of millions of years were reported to act as enhancers (Bejerano et al. 2006; Santangelo et al. 2007; Sasaki et al. 2008; Lindblad-Toh et al. 2011; Lowe and Haussler 2012). Recently, a number of publications proposed that the spreading of copies from active TE classes can lead to rapid rewiring, affecting hundreds of genes whose expression is being altered as a consequence, and is caused by the action of additional transcription factors binding to those enhancers (Wang et al. 2007; Bourque et al. 2008; Feschotte 2008; Xie et al. 2010; Lynch et al. 2011; Rebollo et al. 2012). This might even involve Alu elements in the form of Alu-derived miRNAs or non-miRNAs (Du et al. 2013; Hoffman et al. 2013; Liang and Yeh 2013; Mandal et al. 2013; Spengler et al. 2014).

Nevertheless, caution should be applied for the following reasons:

(1) Four ultraconserved elements were deleted in mice. Although there is no indication that these elements show immediate ancestry to TEs, the lesson equally applies to TE-derived enhancers. In this study, enhancer elements adjacent to genes whose protein-coding exons, when deleted, showed clear phenotypes in mice were chosen for deletion and the elements also were known to function as enhancers. In all four cases, the mice were not only viable and fertile, but also failed to reveal any obvious phenotype among the many parameters tested in the laboratory (Ahituv et al. 2007). Experimentally, it appears to be much easier to obtain a positive expression result with reporter constructs than a phenotype by enhancer ablation (Nelson and Wardle 2013).

(2) Should a newly acquired or activated TE class with “ready-to-use” enhancer activity populate a genome in high numbers, it might not smoothly rewire expression of a set of genes in a functionally viable manner, but simply wreak havoc in a genome by juxtaposition of new enhancers to resident genes. Such eruptions of TEs would be expected to have at least as many detrimental effects than beneficial ones. A less radical scenario would be the following: TEs carry sequences resembling transcription factor binding sites (TFBS) or other regulatory regions that pending minor mutations have the potential to become functional.

(3) Enhancers usually harbor clusters of homotypic or heterotypic TFBS and therefore it is less likely that most TEs are instantly functional. Enhancers usually act in a modular fashion and rather than changing expression patterns, they add a cell-type/tissue or a developmental window to existing expression patterns. After a period of testing (just as in the case of exonization; see above), this or other modules might get lost (adding or exchanging a few mains, yes; rewiring, hardly). For a recent comprehensive assessment of these problems, see de Souza et al. (2013).

TE FUNCTIONS: TO THE MOON!

TEs are insertional mutagens and often enough, integration near or into functional modules of the genome is disadvantageous and, for example, can cause disease (Chen et al. 2005; Callinan and Batzer 2006; Iskow et al. 2010; Hancks and Kazazian 2012). Because of affordable high-throughput sequencing technologies, the search for somatic de novo insertions of TEs becomes feasible. Numerous such events could be detected in tumor tissues and cells (Miki et al. 1992; Lee et al. 2012), some of which might be causal. Somatic integration even was proposed to contribute to neuropsychiatric disease, such as schizophrenia (Bundo et al. 2014). Apart from being detrimental or neutral, a minority of events has the potential to turn out beneficial.

TEs of a certain class might harbor not only TFBS, but also numerous other functional sequences in their consensus sequences, and, perhaps, mainly because these elements are defined by their designation, proposals with highly diverse functions are being published, even involving TEs with very narrow phylogenetic distributions (Allen et al. 2004; Espinoza et al. 2004, 2007; Lunyak et al. 2007; Mariner et al. 2008; Gong and Maquat 2011; Yakovchuk et al. 2011; Carrieri et al. 2012; Jady et al. 2012; Holdt et al. 2013; Ponicsan et al. 2013; Wang et al. 2013). Somatic LINE element integrations even have been implied in the development of the brain (Muotri et al. 2005; Coufal et al. 2009; Faulkner et al. 2009; Singer et al. 2010; Baillie et al. 2011; Upton et al. 2011; Perrat et al. 2013; Reilly et al. 2013). Either TEs are chock full of regulatory motifs and control elements, which can be argued in case of promoters, for example, LTRs (Feuchter and Mager 1990), or it is the fact that TEs are defined and designated nucleic acid sequences (nuons) (Brosius and Gould 1992) and, therefore, receive more attention instead of randomized and anonymous sequences in attempts to investigate their functions. Perhaps one or the other of these exhilarating findings will share the fate of ID repetitive elements and TEs of the SINE class, in a development that unfolded about three decades ago. A lot of excitement was generated by reports that ID elements regulate brain-specific gene expression (Milner et al. 1984; Sutcliffe et al. 1984a,b; McKinnon et al. 1986). Unfortunately, these claims did not stand the test of time (Owens et al. 1985; Sapienza and St-Jacques 1986; Goldman et al. 2014).⁶ For sure, any seemingly insignificant novelty could have far-reaching consequences for future lineages (Martignetti and Brosius 1993; Kapitonov and Jurka 2005), but in their infancy, the functional significance, if not potential of novelties, is not easy to assess and often might be transitory. And what is true for regulatory elements and protein-coding genes and their exonic modules should also apply to functions of non-protein-coding RNAs.

FUNCTIONAL RNA: ALIVE AND KICKING

Despite the growing recognition of RNA’s functional significance and versatility, as well as its preeminence in the evolution of life, up to ∼15 years ago, most scholars in the life sciences still deemed RNA molecules as fossils or remnants from bygone eras. The most complex cytoplasmic RNA class, messengers between the genetic information on DNA and ribosomes, organelles in which structural and functional macromolecules (proteins) of a cell are being assembled, did not generate much excitement any longer after the genetic code had been cracked (Nirenberg et al. 1965; Söll et al. 1965).

A minority of investigators sensed that the previously known and rather abundant nonprotein-coding RNAs, such as transfer and ribosomal RNAs, were only the tip of the iceberg (Prestayko and Busch 1968; Zieve and Penman 1976; Lerner et al. 1980; Brown and Fournier 1984; Lee et al. 1985; Erdmann and Wolters 1987; Mattick 1994; Brosius 1996) and that even novel functional RNAs could evolve (DeChiara and Brosius 1987; Brosius 1991). The floodgates began to open when sequencing of copy DNA generated from non-mRNA fractions revealed a plethora of novel RNAs and RNA classes including miRNAs (Filipowicz 2000; Huttenhofer et al. 2001; Lagos-Quintana et al. 2001; Couzin 2002). The development of deep and ultradeep sequencing methods for cellular, organ-specific, and whole transcriptomes of organisms greatly accelerated the deluge of data (Wang et al. 2009). The past 15 years have surprised most of the scientific community with the incessant discovery of thousands of novel RNAs, some extending RNA species from already known classes such as snoRNAs, some establishing novel RNA classes (e.g., miRNAs and siRNAs), and some unclassified (Fire et al. 1998; Ambros 2001; Lagos-Quintana et al. 2001; Lau et al. 2001; Moss 2001; Ruvkun 2001; Couzin 2002; Carninci et al. 2005; Derrien et al. 2012; Djebali et al. 2012; Kapranov et al. 2012; Ross et al. 2014). Even extracellular RNAs are receiving renewed attention (Benner 1988; Wu et al. 2002; Dunoyer et al. 2007; Leslie 2013a). Major RNA classes are summarized in Table 1.⁷

Table 1.

RNA classification

RNA class	Approximate size (nt)	Function
Tiny RNAs, <50 nt
microRNA, miRNA	21–24	Gene regulation, for example, fine-tuning at the translational level
Short interfering RNA, siRNA	20–25	RNA interference, defense
PIWI-interacting RNA, piRNA	26–31	Epigenetic and posttranscriptional gene silencing of TEs
Small RNAs, ∼50–500 nt
Extracellular RNA, exRNA	Wide range	Intercellular communication
Small nuclear RNA (snRNA)
Spliceosomal RNAs, U1, U2, U4, U5, etc. RNAs	100–200	Splicing of mRNA out of primary transcripts
U7 RNA	50	3′ processing of histone mRNA precursors
Ribonuclease P RNA, RNase P RNA or H1 RNA	340	tRNA processing
7SK RNA	330	Regulation of transcription
Y RNA	90–100	DNA replication
Telomerase RNA component, TERC	450	Maintenance of telomeres
Small nucleolar RNA, snoRNA	60–300	Guide RNAs for nucleotide modification
C/D box snoRNA		2′ O-methylation of the ribose moiety of rRNA
H/ACA box snoRNA		rRNA pseudouridinylation
Cajal body–specific RNA, scaRNA		Composite C/D and H/ACA guide RNA, modification of snRNAs
U3, U8, U14, U17, and U22 snoRNAs		Regulation of preribosomal RNA (pre-rRNA) folding, mediation of correct nucleolytic processing (maturation)
“Orphan” snoRNAs		Target of modification (if any) and function unknown, examples are SNORD115 and SNORD116 in the PWS (Prader-Willi syndrome) locus
Small cytoplasmic RNA, scRNA
5S and 5.8S small ribosomal RNAs	120, 150	Translation
Transfer RNAs, tRNAs	73–94	Adapter molecules in translation
Signal recognition particle RNA, 7SL RNA or SRP RNA	300	Targets protein to the endoplasmic reticulum
Vault RNAs	90–100	Components of the vault organelle, function not clear yet
Ribonuclease MRP, mitochondrial RNA processing RNA	290	Mitochondrial DNA replication and rRNA processing in the nucleus
BC1 RNA	150	Neuronal cytoplasmic RNA (some expression in testes) in soma, but also transported to dendrites, possibly involved in regulation of translation; phylogenetically restricted to rodents, originated from tRNA retroposition
BC200 RNA	200	Neuronal cytoplasmic RNA (some expression in testes) in soma, but also transported to dendrites, possibly involved in regulation of translation; phylogenetically restricted to anthropoid primates; not homologous to BC1 RNA, originated from a monomeric Alu element
Long RNAs, >500 nt
18S and 28S large ribosomal RNA, rRNA	1900, 5000	Translation
Messenger RNAs, mRNAs	Various	Templates for protein biosynthesis
Long intervening non-protein-coding RNAs, lincRNAs	Various	Various suggested functions, not many are certain; regulation of gene expression; gene/chromosome silencing/imprinting by interaction with chromatin
Enhancer RNA, eRNA	>1000
Multiexonic poly(A)+ RNA, meRNA	Various
Long antisense RNAs, asRNA, or aRNA or natural antisense RNA, NAS RNA	Various	Possibly regulation of gene expression
Pseudogene transcripts	Various	Decoy or sink for sequestration of RNA or protein
Circular RNA		Decoy or sink for sequestration of RNA or protein

It is difficult to classify RNAs because they are so diverse. This table attempts to categorize according to size; other groupings take their subcellular locations or functions into account. In any event, the distinctions are blurry and apply only to one species or lineage. In yeast, for example, the U-type snRNAs are significantly larger than in vertebrates. More recently, an arbitrary line between large and small RNAs is being drawn at 200 nt, the latter including tiny RNAs, such as miRNAs. The three size classes here reflect the historical division between small and large RNA at ∼500 nt to accommodate SRP RNA, 7SK RNA, or mitochondrial RNA processing (MRP) RNA that are well >200 nt in length, but have been known since their discoveries as small RNAs. Large or long RNAs are defined as being in the size range of mRNAs, sometimes even displaying mRNA-like attributes (e.g., processing, polyadenylation), but devoid of a functional open reading frame. Initially, miRNAs were designated as tiny RNAs to distinguish them from conventional small RNAs (Ruvkun 2001). This tripartite categorization is kept in the table. There is a flurry of attempts to identify additional RNA classes. Their functions are mostly unknown and the RNAs might correspond to spurious or aborted transcription or other by-products of gene expression (see the text). These include promoter-associated short and long RNA (PASR, PALR), termini-associated RNA (TASR), promoter upstream RNA (PROMPT) transcription initiated RNA (tiRNA), transcription start site antisense RNA (TSSa), antisense termini-associated short RNA (aTASR), retrotransposon derived RNA, satellite region RNA, telomere repeat region antisense RNA (TERRA), etc. For further information and original references see Saxena and Carninci (2011a). PIWI, P-element-induced wimpy testis.

It appears that the mining of novel tiny and small RNAs is reaching a point of deceleration (Saxena and Carninci 2011b) and, as a result, long RNAs are receiving more attention (Jacquier 2009; Ponting et al. 2009; Carninci 2010; Derrien et al. 2012; Hu et al. 2012; St Laurent et al. 2012; Cloutier et al. 2013; Di Ruscio et al. 2013; Geisler and Coller 2013; Kung et al. 2013; Orom and Shiekhattar 2013; Sabin et al. 2013; Ulitsky and Bartel 2013; Fatica and Bozzoni 2014; Yang et al. 2014). Oligonucleotide array sequence analysis, conventional RNA sequencing, and RNA-sequencing technologies (FANTOM) (Carninci et al. 2005; Cheng et al. 2005; Birney et al. 2007; Mercer et al. 2012), even in combination with biocomputation, taking phylogenetic considerations and/or the potential for forming secondary structures into account, mostly come up with at least ten thousand long non-protein-coding RNA candidates (Washietl et al. 2007; Managadze et al. 2013; Necsulea et al. 2014; Nielsen et al. 2014).

DID RNA ENTER BUBBLE TERRITORY?

The percentage of the human genome that is being transcribed appears to be approaching the maximum asymptotically. A simple explanation is the increased coverage with ultradeep sequencing methods and coverage of any cell type including tumors, many developmental stages, as well as (tumor) cell lines numerous times over (Carninci 2009; Ritz et al. 2011; Haas et al. 2012; Ho et al. 2012; Shah et al. 2012; An et al. 2013; Eswaran et al. 2013; Hu et al. 2013; Schonberg et al. 2013; Yoshihara et al. 2013). With all this overabundance, there is a debate raging between those that claim almost any identified transcript and snippet of RNA to be functional (Mattick 2003; Lee et al. 2009; Carninci 2010; Kishore et al. 2010; Clark et al. 2011; Bernstein et al. 2012; Kapranov and St Laurent 2012; Khayrullina et al. 2012; Gebetsberger and Polacek 2013; Mattick and Dinger 2013) and more conservative voices (Brosius 2005c; Huttenhofer et al. 2005; Robinson 2010; van Bakel et al. 2010; Graur et al. 2013). The writer concurs with the argumentation for spurious and stochastic transcripts (Kowalczyk et al. 2012; Jensen et al. 2013; Mudge et al. 2013), especially as many transcription promoters are bidirectional (Seila et al. 2008; Neil et al. 2009; Xu et al. 2009; Wei et al. 2011; Uesaka et al. 2014). Even the yeast Saccharomyces cerevisiae featuring a compact genome abounds with transcriptional noise and erroneous initiation of transcription by RNA polymerase II (Struhl 2007). Furthermore, some transcription might be involved in gene regulation without the transcribed RNA being functional (Tisseur et al. 2011). Examples include regulation of transcription by upstream promoters, previously known as promoter occlusion or transcriptional interference (Adhya and Gottesman 1982; Cullen et al. 1984). In yeast, for example, it has been shown that the SER3 gene is repressed by the act of transcription from the upstream SRG1 gene (Martens et al. 2004). In contrast, the PHO5 gene is activated by intergenic transcription in the opposite orientation presumably influencing chromatin remodeling (Uhler et al. 2007). In other words, the act of transcription exerts the function whereas the RNA is merely a by-product, and this mechanism regulates gene expression in multicellular eukaryotes as well (Kornienko et al. 2013). Another relatively passive role for some RNAs could be the following: Many proteins can bind RNA and proteins can associate with each other. Perhaps an RNA molecule could have the sole function to broker the interaction of two or more proteins that otherwise would not be able to interact with each other. However, if the proteins bind directly or even via other RNA binding proteins to the RNA, close proximity might facilitate functional or structural interactions. In addition, such an RNP complex might even serve as a shuttle into subcellular domains or environments that one or the other component, for lack of the appropriate signals, would not be able to reach by itself (Brosius 2005b).

It has been argued that many of the newly discovered long RNAs represent 3′-UTRs extending beyond the annotated 3′-ends of the mRNAs by using alternative distal poly(A) addition signals. If there are large introns, 5′-UTRs with alternative upstream promoters also can be located far from the ORF. For that reason, some investigators took measures to stay clear of transcripts that are located in a certain proximity to annotated protein-coding genes (Managadze et al. 2013). Nevertheless, splicing does occur in 3′-UTRs as well (Bicknell et al. 2012; Camacho-Vanegas et al. 2012; Zhernakova et al. 2013) and, hence, the corresponding exons could map to corresponding protein-coding genes at a much greater distance to the ORFs. Furthermore, apart from spurious initiation of transcription anywhere in the genome, a certain level of readthrough into gene distal regions could account for a significant portion of these long RNA candidates (Finta and Zaphiropoulos 2000; Akiva et al. 2006; Parra et al. 2006; Richard and Manley 2009).

RNAs overlapping with annotated genes or transcribed in antisense orientation, albeit showing regulatory potential, are generally being avoided for now as their structures, regulation, and functions are more difficult to investigate (Ulitsky and Bartel 2013). Hence, the focus has narrowed to investigating long intergenic noncoding RNAs (lincRNAs).⁸ Another point is being overlooked frequently. For the most part, the arbitrary cutoff for an open reading frame is at 100 amino acids in length. If humans were not pentadactyls, but hexa- or tetradactyls, the cutoff would probably have been chosen at 144 or 64 amino acids, respectively. There are numerous mRNAs that have even shorter ORFs, namely, those encoding peptides (Rudd et al. 1998; Sousa et al. 2001; Frith et al. 2006a; Kastenmayer et al. 2006; Savard et al. 2006; Galindo et al. 2007; Ghabrial 2007; Hanada et al. 2007; Hashimoto et al. 2008; Magny et al. 2013). Perhaps a significant fraction of mRNA-like long intervening non-protein-coding RNAs are mRNAs after all, encoding peptides and short proteins. The question on how many of the non-protein-coding RNAs are associated with polysomes for productive translation is currently under debate (Ingolia et al. 2011; Guttman et al. 2013; Ingolia et al. 2013; Ulitsky and Bartel 2013; van Heesch et al. 2014).

NOVEL RNAs OUT OF THE BLUE

De novo arisen long (intergenic or better intervening) non-protein-coding RNA genes out of neutrally evolving DNA including transposed elements also are receiving increased attention (Guttman et al. 2009; Marques and Ponting 2009; Esteller 2011; Hadjiargyrou and Delihas 2013; Kapusta et al. 2013; Managadze et al. 2013; Ulitsky and Bartel 2013). Their abundance as a class is not surprising, however, because parts of TEs, especially LINEs and lone LTRs, can harbor active transcription promoters. As a consequence, a large proportion of long intervening non-protein-coding RNAs show sequence similarities to TEs.

Most investigators agree that among the tens of thousands of long RNA candidates, there will be hundreds if not thousands of bona fide functional RNAs. One article estimates the number to be ∼10% of all suggested candidates (Ulitsky and Bartel 2013). In addition, quite a few of these spurious and currently nonfunctional RNAs might one day be exapted into novel functions and, as discussed above, as protein-coding exons or regulatory regions. The majority will be rendered inactive again and a few, if beneficial, will eventually persist under selective pressure (Brosius 2005c; Siepel 2009; Polev 2012). In an analogy to Alex Rich’s predictions with respect to an RNA world and the persisting significance of RNA, Henry Harris envisaged the potential of spurious transcripts as raw material for evolution about half a century ago: “Only a small proportion of the RNA made in the nucleus of animal and higher plant cells serves as a template for the synthesis of protein. … Most of the nuclear RNA, however, is made on parts of the DNA, which do not contain information for the synthesis of specific proteins. … It plays no role in the synthesis of a cell protein, but serves as a background on which mutation and selection may operate to produce new templates for protein synthesis” (Harris 1965, 2013). This also applies to new genes encoding functional RNAs.

The process of new genes arising out of nongenic regions might not appear as extraordinary if one considers the presumably more trivial reverse process, namely, formation and gradual decay of pseudogenes. A single-point mutation or small deletion can turn an mRNA into a non-protein-coding RNA. Further mutations can silence its transcription and, after enough time, the remnants of a gene cannot be recognized as such any longer.

COMPETITIVE RNAs

Interestingly, transcribed pseudogenes, including those that arose via retroposition, reportedly can be functional as RNAs or mRNAs encoding truncated proteins (Balakirev and Ayala 2003; Duret et al. 2006; Frith et al. 2006b; Pink et al. 2011; Wen et al. 2012; Sen and Ghosh 2013). For example, retropseudogene PTENP1 derived from tumor suppressor gene PTEN is transcribed. Like the canonical mRNA expressed from the parent gene, it harbors binding sites for several miRNAs in what used to be the 3′-UTR. These miRNA target sites are competing for the corresponding miRNAs, thus, ameliorating their inhibitory effect on the intact mRNA, resulting in higher levels of the tumor suppressor protein PTEN (Poliseno et al. 2010). Even two isoforms of an antisense RNA generated from the PTENP1 pseudogene have been reported to be involved in gene regulation (Johnsson et al. 2013). In addition, a plethora of naturally occurring circular RNAs have been discovered, which may regulate gene expression by sponging up complementary RNAs, such as miRNAs (Salzman et al. 2012; Hansen et al. 2013; Memczak et al. 2013; Tay et al. 2014). RNAs not only can be decoys or sinks for other RNAs but also for proteins. A long RNA processing product located between two snoRNAs has been reported to act as a sink for Fox2 splicing factor and, as a consequence, alter the relative abundance of alternative splice products of a number of genes (Yin et al. 2012).

THE (NUCLEIC) ACID TEST FOR FUNCTION

As one of the early advocates of a vibrant RNA componentry in extant organisms (DeChiara and Brosius 1987; Brosius 1991, 1996; Petherick 2008), the writer never doubted that the number of functional non-protein-coding RNAs could easily match those for mRNAs in line with more conservative estimates being in the 3% range (Clamp et al. 2007; Church et al. 2009; Cabili et al. 2011; Managadze et al. 2013).⁹ Considering genes encoding RNAs smaller than long RNAs as well and leaving generous space for potentially functional RNAs overlapping other genes in sense or antisense orientation, the figure might top 5%. It does not matter whether the numbers will end up even at 10% of the total genome, they will not be in the range of 75% (Djebali et al. 2012). There is a tremendous task ahead of us to determine which of the detected transcripts are bona fide RNAs, and what their individual functions and contributions to the cell are. Phylogenetic conservation helps, but its absence is not a perfect criterion for lack of function. Not many investigators would deny the functional significance of the UTRs of mRNAs. These regions, flanking ORFs, can encode other, for example, regulatory functions such as structures to modulate turnover or translation efficiencies, including sequence complementarities to regulatory miRNAs. Yet, for the most part, these sequences show much less phylogenetic conservation in comparison to ORFs. This agrees with findings about Xist, an ∼17-kb-long non-protein-coding RNA in humans that, at best, is conserved only in small patches between mammals (Duret et al. 2006). Inactivation of Xist RNA expression on the paternal X chromosome leads to early embryonic death in a mouse knockout model (Marahrens et al. 1997). Even the best test for functional significance, gene inactivation in knockout model systems, does not always provide immediate and simple answers. In yeast, very few knockouts of snoRNAs showed a clear phenotype as was the case for U14 (Li et al. 1990), but not, for example, snR189 (Thompson et al. 1988), or only over time or under certain conditions with respect to other snoRNAs (Badis et al. 2003; King et al. 2003; Esguerra et al. 2008). Bacterial genomes like those of Escherichia coli do not have much space to waste and everything without a selective advantage would not remain in the genome for long. In agreement, the knockout of the 4.5S RNA gene had been shown to be essential (Brown and Fournier 1984). Surprisingly, ablation of the gene encoding 6S RNA with a wide phylogenetic distribution and structural conservation in bacteria (Barrick et al. 2005) had no effect on immediate viability (Lee et al. 1985). Later, 6S RNA was shown to regulate transcription and enhance long-term survival (Wassarman and Storz 2000; Trotochaud and Wassarman 2004).

In mammalian genomes, not every inactivation of a protein-coding gene results in a discernible phenotype. The same is to be expected from genes encoding RNA. Snora35 is located within the second intron (almost 100 kb in length) of the serotonin 2c receptor gene (Htr2c) and is highly conserved between placental mammals. As a consequence, the brain-specific Snora35 (earlier termed MBI-36 snoRNA) is cotranscribed with and processed out of the primary Htr2c heterogeneous nuclear RNA (hnRNA), for example, in the choroid plexus (Cavaille et al. 2000). Snora35 gene-depleted mice do not appear to display a phenotype different from their wild-type siblings, thus far (BV Skryabin and J Brosius, unpubl.). In another example of an RNA knockout, it took several years of work to tease out a reduced exploratory behavior in mice when small cytoplasmic BC1 RNA, phylogenetically restricted to rodents and expressed in neurons where it is delivered to dendritic processes, was deleted (Tiedge et al. 1991; Martignetti and Brosius 1993; Lewejohann et al. 2004). In a further mouse snoRNA knockout, the entire cluster encoding Snord116 RNA isoforms (earlier termed MBII-85 snoRNA) was deleted from a locus associated with Prader-Willi syndrome, a neurodevelopmental disorder in humans. Two independent studies revealed that the mice showed failure to thrive and growth retardation, but not all symptoms described for humans (Skryabin et al. 2007; Ding et al. 2008). In many multicellular organisms, snoRNAs are cotranscribed with introns of protein-coding genes or non-protein-coding genes. Snord116 is paternally imprinted and expressed in the brain. The host RNA is processed into a set of abundant long non-protein-coding RNAs by splicing out the introns (Runte et al. 2001). Attempts to compensate for the loss of Snord116 with constructs coexpressing some copies of the respective snoRNA in introns of a different host transgene independently failed thus far (Ding et al. 2008; BV Skryabin and J Brosius, unpubl.). The following could explain the findings: (1) The number of Snord116 RNA isoforms from the original transcription unit is not sufficient in the transgenic construct; (2) spatial, temporal expression and/or its levels are not appropriate; and (3) truncation of the long host hnRNA-derived, processed non-protein-coding RNA is the underlying cause for the mouse phenotype and, by extension, human disease.

An interesting case involves the Hotair long RNA that had been shown ex vivo to regulate HoxD genes (Rinn et al. 2007); deletion of the HoxC cluster harboring the Hotair gene was devoid of a phenotype (Schorderet and Duboule 2011), whereas targeted disruption of the Hotair locus alone led to homeotic transformation of the spine and malformation of metacarpal and carpal bones (Li et al. 2013).

A great stride forward has been made through a comprehensive gene knockout project involving 18 large RNA candidates (Sauvageau et al. 2013). The corresponding genes were selected not to overlap with any other gene or gene module, and their coding regions were replaced with the lacZ and neo^R marker genes. At the current level of investigation, five of the 18 mouse lines revealed a phenotype. In three lines peri- and postnatal lethality was reported, two of which had some survivors, but they had growth defects. For the two remaining lines with a phenotype, growth defects were also reported. The most lethal knockout involved Fendrr, a gene with six exons encoding a nuclear RNA of about 2.4 kb in length. An independent knockout only involving the first exon by inserting a transcriptional stop cassette was lethal as well, albeit showing different organ defects (Grote et al. 2013). The <30%, but >10% “success” rate of 18 knockouts could mean a number of things, for example, that the RNA gene candidates were carefully selected in favor of potential functionality and phenotypic consequences (Sauvageau et al. 2013). Still, two-thirds of the gene ablations did not show a noticeable impact under the laboratory conditions tested, but, nevertheless, could be beneficial for the long-term survival of mice in natural, ever-changing environments or that some of the RNAs have no function per se. Also, it remains valid that, in exceptional cases, such transitory “protogenes” could become exaptations encoding novel functional RNAs, even proteins, or predominantly disappear again into the noise of the neutrally evolving genomic mass. In any event, major points in an article written to balance some of the extreme views are confirmed: “[t]here is no doubt that a number of these non-protein-coding RNAs have important regulatory functions in the cell” and “… aberrant transcripts or processing products embody evolutionary potential and provide novel RNAs that natural selection can act on” (Brosius 2005c).

CONCLUDING REMARKS

All scenarios for the beginning of life are far from perfect. The most plausible hypothesis is based on RNA as a primordial (auto)catalytic macromolecule, leading over a variety of transitions (Cairns-Smith 1982; Szathmáry and Smith 1995), of which one of the first was acquisition of a simple lipid enclosure (Szostak et al. 2001; Mansy et al. 2008) to extant forms of life. In the RNA world, pheno- and genotype were united in the same macromolecule. This union began to separate during the major transition to the RNP world with a “division of labor” between RNA and protein. Even before, linkage of RNA molecules, confinement (of sets of RNAs), and discrimination, on one hand, versus escape (reminiscent of viruses and horizontal gene transfer) and more expansive exchange by reshuffling of the RNA componentry (sex), are apparent. This resembles a tug of war, constantly in need of balancing the forces and not allowing for a victory of either side. Up to this day, at the cellular, organismal, and societal level, the predicament of discrimination and exchange as well as the quandary of selfishness and cooperation remains an essence of evolution and life.¹⁰

Perhaps the modular arrangement of genes and chromosomal DNA has its origins in the RNA or RNP worlds as well, reflecting the hypothetical structure of several linked RNA molecules. This also would mean that RNA processing and trimming had very early origins. Components of the extant telomerase as well as group II introns, ancestral to the spliceosome and non-LTR retroposons, might date back as far as the RNP or RNA worlds. The continued assault of extant eukaryotic genomes with retroposons (somewhat countered and alleviated by deletions via recombination) clearly had its origin in the transition period when RNA was replaced by DNA for the task of storing and replicating genetic material. This process still can shape genomes quite drastically in relatively short evolutionary time frames. Importantly, out of this sea of change with islands of rather conserved and fixed gene modules, new modules can be generated fortuitously and “tested” for added or modified function by the forces of selection. Most disappear again with a few eventually exapted as novel gene modules and, occasionally, even as novel genes encoding protein or functional RNA.

After a long lag phase on the sidelines, functional RNA currently is in the spotlight of biology, even medicine, as RNomics shows promise to detect additional disease genes, greatly develop the diagnostic toolbox, and revolutionize therapeutic possibilities (Esteller 2011; Cech 2012; Erdmann and Barciszewski 2012; Gold et al. 2012). However, the development from only two decades ago when the mere mention of RNA generally exposed grant proposals to monkey hammering resulting in poorer scores and the current situation in which a feeding frenzy of RNA discovery fueled by ultradeep RNA-sequencing technologies endorses almost any detected transcript or degradation product as functional RNA borders on the grotesque. Clearly, the pendulum has swung to the other extreme.¹¹ In any event, the renewed interest in RNA and advances in understanding its evolution and biology will continue to fascinate us. RNA, this ancient macromolecule, will grant us exciting new insights into the works of life past, present, and possibly future.