The Importance of Being RNA-est: considering RNA-mediated ribosome plasticity

ABSTRACT

For over 40 years, ribosomes were considered monolithic machines that translate the genetic code indiscriminately. However, over the past two decades, there have been a growing number of studies that suggest ribosomes to have a degree of compositional and functional adaptability in response to tissue type, cell environment and stimuli, cell cycle or development state. In such form, ribosomes themselves take an active part in translation regulation through an intrinsic adaptability provided by evolution, which furnished ribosomes with a dynamic plasticity that confers another layer of gene expression regulation. Yet despite the identification of various sources that give rise to ribosomal heterogeneity both at the protein and RNA level, its functional relevance is still debated, and many questions remain. Here, we will review aspects, including evolutionary ones, of ribosome heterogeneity emerging at the nucleic acid level, and aim to reframe ribosome ‘heterogeneity’ as an adaptive and dynamic process of plasticity.The terms on which this article has been published allow the posting of the Accepted Manuscript in a repository by the author(s) or with their consent.

KEYWORDS:

• rRNA heterogeneity
• rRNA evolution
• rRNA variants
• rRNA expansion segments
• rRNA modifications
• 5.8S rRNA
• heterogeneous ribosomes
• ribosome plasticity

The Ribosome: historical monolith – evolutionary chameleon

The ribosome is generally considered a highly conserved ribonucleoprotein complex translating messenger RNA (mRNA) into proteins, a function that makes it essential across all species and domains of life. Composed of ribosomal RNAs (rRNAs) and proteins (r-proteins), and assembled from a small and large subunit, ribosomes across the phylogenetic tree follow a common architectural blueprint; one that appears to have nevertheless been adapted and modified to different species during evolution in accordance with their cellular and organismic needs. This is especially true for higher eukaryotes, in line with the necessity for more intricate control over translation as cells and organisms become increasingly complex.

While the core of the ribosome, which harbours its catalytic sites, has remained highly conserved throughout evolution, it is its ‘outer shell’, which contains most of the r-proteins, that has increased in size and complexity (Figure 1)[1–6]. While Bacterial and Archaeal ribosomes contain 26 bacteria specific, or 34 archaeo-eukarya specific r-proteins, respectively, eukaryotes have not only added an additional 44 to the 33 universally conserved r-proteins, and have, furthermore, acquired insertions and extensions, and, in rare cases, deletions over time [6–8]. Whole genome duplications, such as the one which occurred in the lower eukaryote S. cerevisiae about 200 million years ago, have also added to this diversification, resulting in an over-representation of r-proteins in this species compared to other eukaryotes, including humans [9–11]. Yet, these paralogs do not simply represent gene duplications, as deletions of individual paralogs were shown to distinctly affect the translation of specific mRNA reporters, and exhibit alterations under certain stress conditions [12,13] similar examples have also been found in higher eukaryotes, such as Drosophila and even humans [10,14,15].

Figure 1. Side view of ribosomes from bacteria, budding yeast, and human. Cryo-EM structures of E. coli, S. cerevisiae and H. sapiens ribosomes are depicted along with their approximate masses and compositions. 23S, 25S and 28S are shown in blue, 5.8S in red, 5S in black, 18S in orange, and r-proteins in green [1]. (PDB identifiers: 4v54 [2], 4v88 [3], and 4ug0 [4].

However, evolutionary changes have also occurred with regard to the number and length of ribosomal RNA (Figure 1) and affected both the ribosome core and ‘outer shell’. Unlike r-proteins, rRNAs exhibit a high degree of homology across kingdoms. Nevertheless, kingdom- or species-specific features have been acquired over time, with S. cerevisiae having evolved an extra 908 nucleotide (nt) and humans 2650 nt in addition to the 4567 nt that make up the conserved rRNA core [1,5,16,17]. Some of these sequences have formed what is known as ‘expansion segments’ (ESs), large rRNA insertions that are specific to eukaryotic ribosomes and vary greatly in length and sequence between species [18]. However, the evolution of rRNA has also included more subtle changes in only a small number of nucleotides in some places as well as new sites and types of rRNA modifications, which were proposed to have contributed to the precise fine-tuning of ribosome function in diverse environments, from single-cell organisms to complex tissues and organisms.

However, over the last two decades, an increasing amount of data has suggested that this evolutionary fine-tuning has not just left us with a static and monolithic macromolecular machine of determined parts. Rather, ample support has emerged for the ribosome as an entity of physiological flexibility whose evolutionary changes – be those in the form of ribosomal protein paralogs, ribosomal protein-like factors, rRNA sequence extensions of changes that enable dynamic association of so-called accessory proteins, or protein or RNA modifications that allow fine-tuning of translation – allow for additional layers of translation regulation in response to tissue-specificity, developmental stage, environmental cues, cellular mRNA expression levels, or even disease, to name just a few [9,10,14,19–21]. The potential role of ribosomal proteins, accessory factors, and post-translational protein modifications as contributors to functional and compositional ribosome heterogeneity has, to date, been discussed at length in numerous reviews [9,10,14,19–21]. This review will instead focus on aspects of ribosome heterogeneity emerging from the nucleic acid level.

Variation in ribosomal DNA sequence and repeats

The high cellular demand for ribosomes necessitates a high abundance of rRNAs. As such, ~60% of active transcription in eukaryotes is dedicated to rRNA [22]. To support these high levels of rRNA production, each cell carries many copies of ribosomal DNA (rDNA), which are arranged in an operon structure on either one or several chromosomes across all kingdoms of life. In eukaryotes, the rDNA operon contains the 18S, 5.8S and 25S/28S rRNA genes, with hundreds of copies distributed in tandem arrays that are transcribed by RNA polymerase I, while 5S rRNA is transcribed separately [23]. Owing to its large cellular quantities and despite its cell physiological importance, high-throughput sequencing approaches systematically exclude rDNA. In addition, the lack of a reference genome assembly due to its high sequence homogeneity and repetitiveness makes alignment and establishment of a true estimate of rDNA copy number and sequence variation difficult, causing both to be highly understudied. In humans and mice, rDNA copy numbers have, however, been shown to vary widely, and a recent bioinformatic analysis identified both inter- and intraindividual variations in 18S, 28S, 5.8S and 5S rRNA sequences [20,24–28], with single nucleotide polymorphism (SNP) as the predominant variant in human rDNA [25]. Across individuals, 128 positions within the rDNA sequence were found to be multiallelic and, consistent with previous reports, sites of insertions (1–12 nt) and deletions (1–9 nt) were also identified. Overall, 19 variants were observed in more than 50% of humans, while on an individual level, variations were only detected in ~ 5% rDNA operons, and, furthermore, allele frequency varied broadly within individuals, from a single rDNA copy to ~ 300 copies present in a given genome. Differentially expressed rRNA variants across different organs and tissues were also identified in mice, and, for the most part, variants were shown to be evolutionarily conserved and incorporated into actively translating ribosomes [20,24–28].

Separated from the rDNA operon and transcribed by RNA polymerase III, eukaryotic 5S rRNA is also distributed across the eukaryotic genome and its regulation is tightly coupled to the expression of the rDNA operon [29–32]. In Xenopus and zebrafish, maternal 5S rRNA is highly expressed during oogenesis at~20,000 and 2,330 copies per genome, respectively, prior to being entirely replaced by its somatically expressed form during embryo development, at much lower levels of ~ 400 and 12 copies, respectively, per genome [33,34]. Such change in expression pattern has also been observed for 18S, 5.8S, and 28S rRNAs during zebrafish development, whereby maternal-type 5.8S, 18S, and 28S rRNAs are replaced by their somatic-types during embryogenesis and which differ substantially in sequence [35,36]. Similar observations have also been made in lower eukaryotes and prokaryotes; in Plasmodium falciparum, sequence-divergent copies of rRNA were found to be differentially expressed between the mosquito and human infection stages, and evidence from E.coli and Vibrio vulnificus has demonstrated the expression and incorporation of variant rRNA alleles into active ribosomes in response to stress [37,38]. Thus, while the mechanisms by which variant rRNA sequences enable selective translation and ribosome function remain currently unknown, their role in providing a source for ribosome heterogeneity has become more evident over the past decade. The long-awaited key to a more integrated assessment of rDNA and rRNA variation may lie in current and future advances in long-read sequencing approaches, including sequence, structural, and copy number variation, and their functional consequences for ribosome function and translation.

Ribosomal RNA expansion segments

An evolutionary increase in rRNA size and complexity has most commonly been observed through the insertion of so-called expansion segments (ESs), GC-rich insertions of varying lengths into the rDNA sequence [18]. Guided by information from both high-resolution 3D and secondary structure modelling, a model has emerged, termed ‘the accretion model’, which implies a general tendency towards molecular growth, i.e., expansion, of rRNA [39]. According to its principles, rRNA expanded iteratively and incrementally in size during evolution without a substantial remodelling of the ribosome’s core or common trunk; as such, ES sites have increased as organisms became more complex. While most commonly observed in eukaryotes, ESs have been identified in the 5S rRNA of 36 bacterial strains and Archaea, which possess ESs with most insertions ranging between~5–20 nt in length at sites that coincide with those of eukaryotic ESs [18,39]. An exception to these short Archaeal ESs, are the Asgard ES39 and ES9, which have been discovered in the LSU rRNA Asgard archaeal phyla and, ranging in length from 50 to 200 nt, are closer to eukaryotic ESs, where ES39S varies between 80 nt in protists and 138 nt in S.cerevisiae to 178 nt in Drosophila and 231 nt in humans [40]. In birds and mammals, ESs often containing long stretches of GC-rich sequences, exhibiting high sequence and length variability and can span hundreds of angstroms in length, prompting their description as ‘tentacle-like’ rRNA segments, dynamic and flexible in structure [39]. This flexibility and the fact that their limited interactions with r-proteins have made it difficult to resolve their structure and gain further insight into their functions [5,16,18,39,41,42].

ESs have been shown to interact with non-ribosomal, or so-called ‘accessory’, proteins, and mRNAs. In budding yeast S.cerevisiae, ES7L and ES27L are two of the largest ribosomal ESs and are essential for ribosome biogenesis [18,43]. ES7L covers an area 20,000 Å [6] in mature, assembled ribosomes, providing a sufficient solvent-exposed surface to interact with many non-ribosomal proteins [44–46]. Recent data suggested ES7L to act as a reactive oxygen species sensor in yeast and undergoes endonucleolytic cleavage during an early response step to reactive oxygen species (ROS) production [47]. Ribosomes containing cleaved ES7L were still able to participate in polysome formation, suggesting an adaptive response of ribosomes to oxidative stress; however, the precise mechanism of this ES-ROS sensor system, or any translational changes as a consequence of cleaved ESL7-ribosome bound mRNAs, has not been yet elucidated [47]. In addition, in vitro rRNA-protein pull-down experiments of ES7L identified a variety of non-ribosomal proteins involved in many biological functions, including large ribosomal subunit biogenesis, protein transport and localization as well as transcription. It has also been proposed that ES7L could additionally serve as a hub for tRNA aminoacylation and improve translation efficiency together with expansion segment ES27L by positioning N-terminal acetyltransferase A below the ribosome exit tunnel via an interaction with both ES7L and ES27L thus aiding removal of the N-terminal initiator methionine from nascent peptides and ribosomes decoding accuracy [41,44]. ES27L, ES24L, and ES39L have also been suggested to function as targeting of ribosomes to the endoplasmic reticulum and the translocation of nascent chains [18,48–50].

Using cryo-electron microscopy (EM), ES27L was also found to interact with the methionine amino peptidase EBP1 at the peptide exit tunnel (PET) in the non-translating 80S ribosome in HeLa cells, whereby the ES was recruited by EBP1 via a consensus sequence and fixed in place [51]. This interaction was shown to create a potential steric hindrance for any nascent chain emerging from the PET as well as prevent any co-translational protein modifications, and the effect of an EBP1-ES27L interaction on translation is still unclear. EBP1 was also identified as a chief component of ribosomes in the development of murine neocortex, binding the PET during active translation, regulating both start codon initiation and N-terminal peptide elongation; yet no interaction between EBP1 and ES27L was identified in this tissue [52]. However, EBP1 abundance in murine neocortex ribosomes was shown to be cell-type and development-stage specific, with EBP1 specifically regulating the morphology of early-born neocortex neurons, suggesting that both the role of accessory proteins and their interaction with selected ES or other rRNA regions in the ribosome may not only be cell-type but also cell-stage specific.

The same may be true for ES:mRNA interactions. A decade ago, a computational study analysing complementarities between mRNA and rRNA in higher eukaryotes was the first to suggest an interaction between ES and 5’ untranslated regions (UTRs) of mRNAs [53], and a more recent in silico analysis showed widespread complementarity between mRNAs and both 18S and 28S rRNA ESs in humans to indeed exist [54]. In zebrafish, where maternal and somatic 5.8S, 18S, and 28S rRNA variants were shown to be selectively expressed in egg or embryogenesis during development, two specific expansion segments, ES6S and ES3S, have been identified in silico and have been suggested to promote differential operation of the maternal- and somatic-type ribosomes. Within 18S rRNA, ES6S may selectively recruit maternal mRNA specifically expressed during embryogenesis, while ES3S was suggested to selectively recruit mRNAs via their 5’UTRs to prevent their translation by sequestration and thus effectively driving the selective ribosome substrate recognition and translation at different developmental stages [35].

This is the first in vivo evidence of a specific ES directly binding an mRNA and the species-specific regulation of an mRNA by a single ES has recently been demonstrated using chimeric ‘“humanized”’ yeast ribosomes. Hox genes are master regulators of head-to-tail axis formation in metazoans and are subject to complex multi-tier regulation during development, and the IRES-like element within the 5’UTR of Homeobox (Hox) 9 mRNA, more specifically its short stem-loop P4, was shown to interact with the human ES9S in a species-dependent manner [55–57]. Moreover, ES9S was also found to associate with hundreds of mouse embryonic mRNAs [56], which, together with its role in Hox gene translation, adds not only another intriguing layer to an already complex mechanism of gene expression regulation but further hints at a potentially important function for direct ES–mRNA interactions in translation regulation in a cell- and species-dependent manner during development [24,58].

rRNA Modifications

Post-transcriptional modifications of ribosomal RNA have been shown to be another source of ribosome heterogeneity and can either target base or ribose moieties. Base modifications include pseudouridylation, methylation, dihydrouridylation, hydroxylation, acetylation, as well as formation of the more complex 1-methyl-3-amino-α-carbonyl-propyl-pseudouridine (m¹acp3Ψ) modification , while ribose modifications include 2’-O-methylation. Some of these modifications can be found across all kingdoms, while others are kingdom- or species-specific. Pseudouridylation and 2’-O-methylation are the most abundant modifications found in eukaryotes, accounting for 90% of all rRNA modifications in human ribosomes [59], while 2’-O-methylation is also the most abundant rRNA modification in Archaea [7,60–62]. In both kingdoms, these modifications are, respectively, mediated by small nucleolar ribonucleoprotein particles (snoRNPs) of the box H/ACA and C/D families that guide the location of the modification through their snoRNA components, while the catalytic reactions are carried out by the proteins Dyskerin (DSK1; yeast Cbp5) or Fibrillarin (FBL; yeast Nop1) for pseudouridylation (Ψ) and 2’-O-methylation, respectively [63]. There are around a hundred combined pseudouridylated and 2’-O-methylated sites in S. cerevisiae, while over a hundred for each modification have been identified in human rRNA [63], and knockdown of FBL and DSK1 were shown to reduce global rRNA 2’-O-methylation and pseudouridylation levels [31,64], resulting in reduced IRES-mediated translation and reduced translational fidelity in yeast, mouse embryonic fibroblasts, and HeLa cells [31,64].

Since the advent of high throughput sequencing approaches and their adaptations to measure different modifications alongside new mass spectrometry methods (e.g., 2’-O-methylation: RibomethSeq; pseudouridylation: HydraPsiSeq; m³C and m⁷G: AlkAniline-Seq; N4-acetylcitidine: ac⁴C-seq; all PTMs: SILNAS [65–68], accurate measurements of modification occupancy at specific sites have suggested that while some sites are close to 100% modified, others show less occupancy, ranging from 70% to less than 10% [9,59,68–70]. Modification clusters on rRNA have been conserved during evolution, but they have also expanded with the accretion of rRNA segments over time. While most modifications occur in conserved and functionally important regions of the ribosome, such as the A-, P-, and E-sites as well as along the peptide exit tunnel and inter-subunit bridges, additional modifications have been identified within ESs in eukaryotes [59,62,63,71]. Modified clusters near functional centres were shown to have structural implications on the active sites and are required to maintain translation rate and accuracy [63,72–76], while different effects for modifications along the inter-subunit bridge have been observed, with modifications in the B2a bridge affecting translation fidelity, while modified sites within the B1a bridge influenced subunit association but had little effect on translation accuracy beside a specific effect on UGA readthrough in the absence of all modifications, suggesting a high degree of functional specificity of modified sites [59,77–80]. Moreover, 2’-O-methylation levels were also shown to vary between different cells and conditions at fractionally methylated positions, while the fully methylated sites were generally invariant [72]. This implies potential regulatory functions for certain fractional modification sites and evidence for ribosome plasticity at the level of rRNA modifications.

Although there have so far only been a limited number of investigations into the role of specific rRNA modification sites on ribosome function, data from yeast and human cells demonstrated that alteration or ablation of 2’-O-methylation or pseudouridylation at distinct sites impacted translation profiles of select mRNAs, in addition to cell viability and proliferation, and was linked to both ribosomopathies and various cancers [81,82]. In prostatic neoplasia and breast cancer, FBL overexpression caused variations in site-specific rather than global 2’-O-methylation leading to altered translation regulation of a small subset of mRNAs, 8% of which were IRES-containing and implicated in tumorigenesis (e.g., p-53 and c-myc), growth factor receptor and apoptosis regulation, while translational fidelity was significantly reduced [83–88]. DSK1 mutations, on the other hand, have been linked to dyskeratosis congenita (DC), and significantly lower pseudouridylation levels at two known 28S rRNA Ψ sites, 4331 and 4966, have been detected in DC patients and are associated with altered IRES-dependent mRNA translation and translation fidelity [89–91]. Overall, while the identification of variations in rRNA 2’-O-methylation and pseudouridylation at specific sites suggests that differential rRNA modification patterns can underlie the functional plasticity of the ribosome, the precise mechanisms by which these modifications may influence ribosome function remain to date unknown.

However, 2′-O-methylation and pseudouridylation are not the only rRNA modifications [7,9,60–63], and others have also been associated with changes in the translation output of mRNA subsets and phenotypes that support the idea of ribosome heterogeneity. One striking example is the single hypermodified site m¹acp3Ψ at nucleotide 1248 within 18S rRNA [92]. Conserved within the decoding centre of eukaryotic ribosomes, hypo- or sub-stoichiometric modification levels of m¹acp3Ψ have been observed in more than 20 distinct cancer types, concomitant with an increase in translation levels of r-protein mRNAs [92]. Deletion of the S.cerevisae methyltransferase Rcm1 and its ortholog NSUN5 in C.elegans and Drosophila, which target and modify 25S:m⁵C₂₂₇₈ and 28S:m⁵C₃₇₈₂, respectively, provided cells with an increased resistance to oxidative stress as well as an increased life span [93,94]. Moreover, in 38% of the human glioma-derived cells, epigenetic inactivation of NSUN5 and unmethylated C3782 within 28S rRNA, resulted in an adaptive translational program for survival under conditions of cellular stress and increased long-term survival [95]; however, the mechanisms underlying this selective translation of certain mRNAs are still unclear.

In C.elegans, N⁶-adenosine methylation (m⁶A) at 18S:1717, catalysed by the METTL-5 methyltransferase, was also shown to be required for the selective translation of the enzyme cyp-29A3, which is involved in the processing of stress signalling molecules [96]. Loss of 18S:1717 m⁶A affected specifically the translation of cyp-29A3 in response to stress via a so far unknown mechanism in metl-5 mutant worms, which exhibited an increased stress resistance similar to cyp-29A3 worms, while m⁶A levels in 18S rRNA were unaffected, thus implicating this rRNA modification in mediating a translational response to stress signals. However, the modification status of two highly conserved m [9]A sites within the 3’ of 18S, A₁₇₈₁ and A₁₇₈₂, was found to change in response to sulphur starvation in yeast and mammalian cells [97,98]. Located in the decoding centre as part of an inter-subunit bridge and modified by the dimethylase Dim1/DIMT1, the constitutively dimethylated 18S:m⁶A_1781–1782 changed to mono-methylation of both adenosines in translating ribosomes under these conditions, and ribosomes carrying 18S:m⁶A_1781–1782 exhibited an increased specificity for the translation of sulphur metabolism genes [97,99].

Another rRNA modification, N⁴-acetylcytidine (ac⁴C), was identified in humans, S.cerevisae, and some Archaea of the Thermococcales species [65]. In Thermococcus kodakarensis, ac⁴C residues located specifically along the subunit interface regions and making direct interactions with the ribosomal substrates were shown to change in a temperature-dependent manner. Further, cryo-EM studies suggest a role for this modification in stabilizing the ribosome at higher temperatures; however, whether this is required for the function of mature ribosomes, or to facilitate rRNA folding and processing at higher temperatures remains to be determined [65].

Finally, there is also evidence suggesting that rRNA modifications are involved in development. In zebrafish, a subset of modified sites was shown to change modification status and stoichiometry during the transition from early maternal to somatic rRNA expression, in correlation with the availability of their cognate snoRNPs [100]. It has been proposed that early maternal rRNA and its methylation status confer ribosomes with a translational specificity for specific maternal mRNAs, although this remains to be confirmed [100]. Variations in site modifications related to development have also been observed in mice, where multiple organs, including the brain, liver, heart, lung, and kidneys, were screened for ribose methylation. While most methylation sites were fully modified in adult mouse tissues, a subset of sites showed lower and two sites,18S:A₅₇₆ and 28S:G₄₅₉₃, higher methylation levels in developing tissues compared to adults. G₄₅₉₃ is predicted to be 2’-O-methylated by snoRNA SNORD78, which is intron-encoded within the lncRNA GAS5 along with eight other box C/D snoRNAs, none of whose targeted sites were modulated during development [101]. Interestingly, the loss of SNORD78 and methylation at its target sites has also been linked to hindbrain-specific malformations in zebrafish [102]. While further work is required to determine the role of the majority of rRNA modifications as well as the underlying mechanisms of their function, there is growing evidence that they indeed represent a source of ribosome heterogeneity and influence ribosome function.

The case of 5.8S rRNA isoforms

In eukaryotes, differential 5’end processing of 5.8S rRNA provides another source of ribosome heterogeneity. Evolutionarily derived from the 5’-end of bacterial 23S, and with a sequence homology of~50% between E.coli to S.cerevisiae, Drosophila, Xenopus, and HeLa cells, the overall length of 5.8S rRNA varies from~80 to 200 nt across 56 eukaryotic organisms for which 3,666 sequences have been genome mapped [103,104]. However, in addition to this length variance, three 5.8S isoforms have been described, which differ in their 5’ end: a long form (5.8S_L), a short form (5.8S_S), and an even shorter 5.8S_C (c for cropped) [105]. The existence of different stable forms of 5.8S rRNA was first described by Gerald Rubin in 1974 in S.cerevisiae [106], and concomitant expression of both a long- and short-form has subsequently been found in all eukaryotes studied so far, although the ratio between the two forms can vary between species [105,107]. In most eukaryotes, the 5.8S-short, which is of very similar in length to S. cerevisiae (158 nt) and H. sapiens (157 nt), constitutes the major form, representing ~90% of the total 5.8S rRNA in cells. The 5.8S-long, which makes up the remaining 10%, harbours a 6 nt or 5 nt extension at its 5’end in S.cerevisiae and higher eukaryotes, respectively [105,107,108]. More recently, active ribosomes and rRNA containing a 5.8S-cropped form, which is 10 nt shorter than 5.8S-short have also been isolated from mouse and human cells [105].

The main maturation pathway for 5’end formation of 5.8S in yeast involves an endonucleolytic cleavage within the internal transcribed spacer 1 (ITS1) of the rRNA precursor, followed by a series of 5’-3’ exonuclease trimming steps producing 5.8S-short rRNA [109]. The minor pathway leading to the maturation of 5.8S-long is thought to involve a so-far unidentified endonuclease cleaving ITS1 at site B1L in yeast [108–113]. In human cells, exonucleolytic trimming by Xrn2 has been suggested to generate both long- and short-forms of 5.8S [114,115], and in mouse, Xrn2 has also been implicated in the 5’end maturation of both 5.8S-long and 5.8S-cropped forms, although the exact mechanisms remain unclear [105]. Additional alternative pathways that bypass the two canonical processing pathways have been identified in S.cerevisiae, where the 35S rRNA precursor is either cleaved at the 3’end of 18S and ITS1 then exonucleolytically processed to produce 5.8S-long [116], or ITS 1 is cleaved at an alternate site, A4, the latter leading to a change in 5.8S-short:long ratio [117]. However, while different forms of 5.8S are incorporated into active ribosomes and may convey functional ribosome heterogeneity, to date, any potential differences in substrates of these differential 5.8S ribosomes remain unknown.

Unlike 18S and 25S/28S rRNAs, 5.8S harbours only very few modification sites. Only two 2’-O-methylation (U14 and G75, with low and high modification levels, respectively) and two pseudouridylated sites (U55 and U69, both ~60% modified) have been identified in human 5.8S rRNA, while S.cerevisiae was shown to contain only one pseudouridylation site (U63, ~78% modified) in its 5.8S sequence, and no sites of methylation have been identified [70,71,118,119]. Finally, despite its short length, differences between maternal, and somatic rRNA sequences have also been found within the 5.8S rRNA in zebrafish, where sequence variances are spread over the central and terminal regions of 5.8S, in addition to 5’-short and -long forms; however, their impact on translation regulation remains currently unknown [35].

Final Remarks

Evolutionary changes in the level of ribosomal RNA, whether in the form of sequence variations, expansion segments, developmental variance in rRNA genes, rRNA expression levels due to differences in rDNA repeat numbers, rRNA processing isoforms, and RNA modifications are as much of a part of the emergence of ribosome heterogeneity as ribosomal proteins and accessory proteins. But it is not just the question of ‘what’ makes ribosomes heterogeneous but the ‘how’ – their actual cellular function and physiological relevance; and, to some extent, the ‘if’, as the term ,ribosome specialization, is still fraught with scepticism. However, some of that scepticism has been assuaged in recent years by a reframing of the notion as ‘ribosome heterogeneity’, which provides a less restrictive and ‘absolute’ image of ribosomes as fixed compositionally different pools of stable ribosomes in cells and, instead, provides a potential image of a more ‘flexible’ ribosome that may be adaptable to certain circumstances. The notion of ribosome heterogeneity first emerged with the ribosome filter hypothesis, which proposed that ribosomal subunits may directly regulate translation through base-pairing with specific mRNAs via complementary sequences, a theory that was demonstrated for the mouse mRNA of the Gtx homeodomain protein, which contains a 9‐nt complementary match within helix 26 of the 18S rRNA [120,121]. Later, Mauro and Edelman widened their hypothesis of differential mRNA translation meditated by ribosome heterogeneity conferred also by ribosomal proteins, their varying post-translational modifications as well as select rRNA modifications, in line with current theories [121]. However, even the idea of a compositionally ‘flexible’ ribosome remains problematic since, i) ribosomes appear to ensure uniform ribosome composition, and ii) are exceptionally stable, as making ribosomes is energetically costly for a cell [22,122]. On the other hand, evolution has provided a certain ‘built-in’ flexibility of ribosome through rRNA modifications, expansion segments, isoforms on one side and ribosomal and -like proteins as well-accessory factors on the other, not to mention complementary rRNA:mRNA sequences. It is, therefore, feasible to consider a generally produced ribosome that may be uniform in its base features and components as well as stable, but which, at the same time, is dynamic insofar as it can respond to physiological challenges through a dynamic exchange of certain proteins, as has been suggested for axonal ribosomes [123], or even – hypothetically speaking – conformational changes within its ES regions to create binding platforms for distinct factors. In such a scenario, a ribosome would become an adaptive machine, a ‘plastic’ ribosome – a macromolecular complex that is more dynamic than static, but does not require a complete compositional overhaul, and thus population change, under such circumstances. Of course, this scenario does not exclude the potential negative consequences of adaptive ribosome plasticity in response to cellular or environmental cues. However, while, organismally speaking, the compositional and functional adaptation of ribosomes and its regulation of the expression of specific mRNAs may have detrimental consequences, on a cellular level, such processes are purely circumstantial and, consequentially, adaptive regardless of the outcome for disease etiology.

The combinatorial effect of rRNA allele variants and ESs along with rRNA modifications and differently processed rRNA isoforms, in addition to ribosomal proteins and accessory factors, provides an enormous repertoire for adaptive ribosome plasticity, of which to date we have only scratched the surface. The ongoing methodological developments including long-read sequencing and biochemical approaches, such as VELCRO-IP RNA-seq and many others will without a doubt assist us in elucidating many of the remaining questions of ribosome plasticity – at the RNA and protein level as well as their functional relevance at the cell and tissue level, in organs and development, and in different species, including pathogens and their hosts.

Acknowledgments

We would like to thank all the current and past lab members for stimulating discussions, and Daniel Zenklusen for critical reading of this manuscript. We sincerely apologize to the authors whose work we could not cite or discuss. M.O. also wishes to thank Renée Schoeder for the opportunity to contribute to and edit this special issue.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

Data sharing is not applicable to this article as no new data were created or analysed in this study.

Additional information

Funding

The Oeffinger lab is supported by the Natural Sciences and Engineering Research Council of Canada [RGPIN-2015-06568] and collaborative funds from the Canadian Institute for Health Research [PJT-425798].