Ribosome biogenesis and function in development and disease Free

Cell fate decisions are widely governed by the activity of specific signaling pathways and the differential expression of transcription factors. Cell-specific transcription patterns are then re-enforced and maintained by the action(s) of various epigenetic mechanisms (Gökbuget and Blelloch, 2019). Despite the clear importance of differential patterns of transcription during development, the ultimate readout of gene expression is the actual synthesis of the corresponding protein. However, poor correlations between mRNA and protein levels within cells are often observed (Schwanhäusser et al., 2011; Khan et al., 2013), highlighting the importance of translational regulation in molding final patterns of gene expression.

RNA-binding proteins have long been known to bind to elements within the 5′ and 3′ untranslated regions (UTRs) of mRNAs to modulate their engagement with translation machinery (Shi and Barna, 2015). In addition to transcript-specific regulation, emerging evidence highlights the functional significance of global translation capacity within a cell, which depends on the overall number of functionally available ribosomes (Mills and Green, 2017; Saba et al., 2021). Building on our knowledge of ribosome assembly in yeast (Klinge and Woolford, 2019), we have a growing appreciation of the enhanced complexity of ribosome biogenesis (Ribi) in multicellular organisms. In this Review, we briefly outline the basics of ribosome assembly and the regulation of global protein synthesis across species. We review the literature on a group of diseases, collectively known as ribosomopathies, caused by defects in ribosome assembly and mutations in genes encoding ribosome proteins. The study of these disorders provides insights into how ribosomes influence different developmental processes, including translational control in germ cells and early embryos. We discuss emerging models that suggest dynamic changes in Ribi and mRNA translation drive specific cell fate decisions, particularly in stem cell lineages. We end with a review of recent data on specialized ribosomes, including tissue-specific ribosomal protein paralogs. The study of ribosomes, Ribi and mRNA translation is vast, but this Review focuses on new findings and emerging ideas at the intersection between ribosomes, mRNA translation and development.

Much of our current understanding of Ribi in eukaryotic cells stems from genetic and biochemical studies performed using yeast (Baßler and Hurt, 2018; Klinge and Woolford, 2019). Cryogenic electron microscopy work has also provided an unparalleled view of the stepwise progression of this complex process (Cheng et al., 2019; Du et al., 2020; LaPeruta et al., 2022; Sailer et al., 2022), which is summarized in Fig. 1. Briefly, most steps of Ribi occur within the nucleolus (Fig. 1). Ribi starts with Pol I-dependent transcription of ribosomal DNA (rDNA) to produce 47S pre-rRNA. This step, along with other aspects of Ribi, is regulated by various mechanisms and signaling pathways, including MYC and mTOR (reviewed by Mayer and Grummt, 2006; Grummt, 2013; Ni and Buszczak, 2022). This pre-rRNA is subsequently processed to yield 18S, 5.8S and 28S rRNAs. 5S rRNA is independently transcribed by Pol III outside of the nucleolus (Ciganda and Williams, 2011). The rRNAs are then post-transcriptionally modified and processed in various ways and eventually associate with 80 different ribosome proteins. This assembly process occurs either co-transcriptionally or post-transcriptionally to produce a 90S complex (Scull and Schneider, 2019), which is then processed to form pre-40S and pre-60S subunits. Immature subunits are independently transported to the cytoplasm, where they undergo final maturation steps to form functional 40S and 60S subunits (Fig. 1).

Ribi in multicellular organisms has evolved increasing complexity from yeast to humans. Although the steps in the formation of functional 40S and 60S subunits are generally similar between eukaryotic species, the number of factors involved in the process appears to have expanded in mammals. Although the catalog of Ribi proteins in multicellular organisms likely remains incomplete, independent studies have significantly expanded our knowledge in this area. For example, biochemical characterization revealed surprising complexity in the nucleoli in human cells (Andersen et al., 2002), later discovered to include hundreds of proteins participating in Ribi (Wild et al., 2010; Tafforeau et al., 2013). In addition, unbiased RNAi-based screens identified 139 genes that alter the number of nucleoli in human cells (Farley-Barnes et al., 2018). Follow-up biochemical experiments showed that many of these proteins play clear roles in Ribi, many of which do not have clear yeast homologs. For example, in human cells, a four-protein complex regulates recycling of the ribosome maturation factor RSL24D1 late during pre-60S maturation (Ni et al., 2022), whereas a similar step is carried out by only a single protein in yeast (Pertschy et al., 2007; Kappel et al., 2012; Mitterer et al., 2016; Prattes et al., 2019, 2021). The consequences of having more factors involved in specific steps of Ribi remains unclear. Perhaps having additional proteins endows cells with the ability to finely tune each step of Ribi in response to changing developmental and environmental cues. Further work will be needed to test this idea and to gain a comprehensive understanding of this essential process in human cells and other multicellular organisms.

Once made, ribosomes translate mRNAs into proteins. The process of mRNA translation can be divided into three basic steps: initiation, elongation and termination (Fig. 2). In eukaryotes, most differential regulation of mRNA translation occurs at the step of initiation. This regulation can occur at the level of individual transcripts or on a global scale across the entire transcriptome (Wilhelm and Smibert, 2005; Pelletier and Sonenberg, 2019). A common mechanism employed during development involves the binding of transcripts by RNA-binding proteins, which then interfere with cap-dependent translation by mimicking the eIF4E-binding domain of eIF4G. 4E-binding proteins (4EBPs) perform a similar function on a global scale. Of note, one way the mTOR pathway promotes global translation across species is by phosphorylating 4EBPs, which in turn inhibits their ability to interfere with cap-dependent translation initiation (Matsuo et al., 1997; Gingras et al., 1999; Thoreen et al., 2012). Additional ways mTOR influences translation and ribosome production have been extensively reviewed elsewhere (Liu and Sabatini, 2020).

Once formed, a monosome moves down an mRNA and, together with charged tRNAs, decodes successive codons and catalyzes peptidyl bond formation in the process of translation elongation (Fig. 2). Multiple ribosomes can engage with individual mRNAs to form what are known as polysomes. Although a Ribo-seq-based study showed that translation elongation tends to occur at a similar rate across most transcripts (∼5.6±0.5 codons per second) (Ingolia et al., 2011), this process can be regulated in rare cases. For example, elongation across certain transcripts, including ribosome protein mRNAs, differ relative to other cellular mRNAs (Riba et al., 2019). Moreover, phosphorylation of eukaryotic translation elongation factor 2 (eEF2) by eEF2 kinase (eEF2K) causes translation elongation to slow down, which correlates with higher translation fidelity (Xie et al., 2019). Activation of eEF2K has also been shown to positively influence longevity in worms (Xie et al., 2019). Elongation rates can also vary with stress (Guzikowski et al., 2022), and eEF2 phosphorylation states change in response to mTORC1 and ERK signaling (Fig. 2). Ribosome quality control, triggered by stalled ribosomes on mRNAs, plays an important role in cellular responses to stress, and emerging evidence indicates that this process likely also impacts development and disease across different multicellular organisms (Martin et al., 2020; D'Orazio and Green, 2021; Mishra et al., 2021; Saba et al., 2021). Moreover, codon usage and the amino acid sequence of the resulting peptide being synthesized also influence both transcript levels and translation rates (Zhou et al., 2018; Lyu et al., 2021; Yang et al., 2021; Zhao et al., 2021); rare codon usage can impact the tissue-specific patterns of protein expression (Allen et al., 2022). Understanding the underlying mechanisms responsible for these effects represents important work for the future.

Translation termination occurs when a ribosome encounters a stop codon (Fig. 2). The combined action of translation termination factors eRF1 (ETF1), eRF3 [eRF3a (GSPT1) or eRF3b (GSPT2)] and ribosome recycling factor ABCE1 results in the release of the newly formed polypeptide and the recycling of ribosome subunits. One recent study, using sophisticated single-molecule approaches, revealed both the timing and ordering of events during translation termination (Lawson et al., 2021). Moreover, the field has a growing appreciation that stop codon ‘readthrough’, when ribosomes continue translating past the stop codon and extend the sequence length of the resulting protein (Fig. 2), occurs at varying frequencies, often in a developmental and tissue-specific manner. For example, transcripts that encode for the Drosophila Kelch protein exhibit differential stop codon readthrough in the ovaries versus the nervous system (Robinson and Cooley, 1997; Hudson et al., 2021). Variations in stop codon readthrough are also common within Drosophila central nervous system transcripts (Hudson et al., 2021). Although the significance of this is not well understood, one emerging theme is the importance of cis-acting RNA elements that govern readthrough efficiency, as has been shown for both the Drosophila headcase (Steneberg and Samakovlis, 2001) and kelch genes (Hudson et al., 2021). 3′ sequences have also been shown to regulate stop codon usage in yeast and mammals (Namy et al., 2001; Loughran et al., 2014; Cridge et al., 2018; Anzalone et al., 2019; Rajput et al., 2019).

Further understanding of stop codon readthrough has important implications for human health. Suppressor tRNAs and other methods for modulating translation termination are being explored as methods for promoting readthrough of disease-causing nonsense mutations (Keeling et al., 2014; Wang et al., 2022). The observation that suppressor tRNAs do not cause widespread readthrough of native stop codons across the transcriptome reflects the importance of sequence context in translation termination. Additional work will be needed to understand more fully how readthrough efficiencies are controlled in a cell-specific manner during development and how the protein products generated by readthrough impact tissue-specific functions.

Given the universal requirement for ribosomes in cellular protein synthesis, it should come as no surprise that complete disruption of Ribi or ribosome function is incompatible with life. However, perturbations in Ribi factors and ribosome protein genes often result in surprisingly tissue-specific effects. In Drosophila, loss of haplo-insufficient ribosome protein genes typically results in Minute phenotypes characterized by small stature, thin bristles and reduced fertility, caused by reduced protein synthesis and a subsequent decline in cell proliferation and growth (Marygold et al., 2007). Other structures, including the wing, appear morphologically normal in heterozygous Minute flies. However, deleting the Hippo pathway effector yorkie in a Minute background results in severe wing growth defects, marked by increased caspase activity and cell death during development (Wada et al., 2021). This suggests that different signaling pathways, perhaps acting in a compensatory manner, prevent cell death and ensure tissue growth when functional ribosomes drop below a certain threshold.

As expected, mutations in ribosome assembly factors also result in death during development. However, careful examination of rare surviving mutant adults again reveals surprisingly tissue-specific phenotypes. For example, flies that develop to adulthood after whole-body RNAi knockdown of the Ribi factor Nucleostemin display variable head, eye and antennal phenotypes (Rosby et al., 2009). Strikingly, some of these flies appear to have ectopic antennal segments growing out of the eye. Similarly, knockdown of the Pol I transcription factor TAF1B also produces analogous phenotypes (Shalaby et al., 2017). The specific mechanisms that underlie these homeotic-like phenotypes remain unclear, but these observations suggest that certain aspects of tissue patterning may be particularly sensitive to reductions in Ribi and overall levels of functional ribosomes.

Minute mutations have also played a crucial role in our understanding of cell competition; mosaic analysis shows that heterozygous Minute cells induced in an otherwise wild-type background are eliminated by cell death (Morata and Ripoll, 1975; Moreno et al., 2002; Li and Baker, 2007; Kale et al., 2015). Disruption of ribosomes influences cell competition by inducing expression of the transcription factor Xrp1, which drives JNK activity and autophagy induction (Kale et al., 2018; Lee et al., 2018; Kiparaki et al., 2022). Cell competition has also been observed in vertebrates during liver regeneration (Oertel et al., 2006; Menthena et al., 2011), within the immune system (Bondar and Medzhitov, 2010; Marusyk et al., 2010) and during early development (Dejosez et al., 2013).

In humans, mutations in ribosome protein and Ribi genes result in a group of diseases collectively known as ribosomopathies. These include Diamond–Blackfan anemia (DBA), X-linked dyskeratosis and Treacher Collins syndrome (Table 1). These specific ribosomopathies have been extensively reviewed elsewhere (Aspesi and Ellis, 2019; Farley-Barnes et al., 2019; Kampen et al., 2020; Watt et al., 2022). Despite affecting a common process needed in all cells, these diseases often manifest with surprisingly tissue-specific phenotypes. For example, DBA is most often associated with bone marrow failure, short stature and craniofacial defects, whereas Treacher Collins is associated with craniofacial abnormalities and X-linked dyskeratosis with cytopenia, nail dystrophy and skin hyperpigmentation.

Emerging evidence indicates that disruption of specific steps of Ribi can also result in microcephaly, seizures and hearing loss. Ten individuals with microcephaly within the same family carry a homozygous mutation in the ribosomal RNA processing 7 homolog A gene, RRP7A (Farooq et al., 2020). The yeast homolog of RRP7A, RRP7, is an essential gene involved in pre-rRNA processing and ribosome assembly (Baudin-Baillieu et al., 1997). RRP7A exhibits enriched expression in the developing human neocortex and loss of the gene results in neurogenesis defects both in vitro and in vivo (Farooq et al., 2020). Although RRP7A localizes to nucleoli and plays a clear role in ribosome assembly, the protein also localizes to cilia, leaving open the possibility of additional Ribi-independent functions within the cell.

Several studies have also linked allelic variants in the gene encoding the AAA ATPase SPATA5 to neurological phenotypes. SPATA5 is distantly related to the yeast protein DRG1 (Prattes et al., 2021; Ni et al., 2022), which plays a role in pre-60S maturation. SPATA5 interacts with SPATA5L1, cyclin dependent kinase 2 interacting protein (CINP) and C1ORF109 (Ni et al., 2022). Disruption of these proteins results in pre-60S maturation defects and phenotypes, including microcephaly, hearing loss, seizures and intellectual disability (Forstbauer et al., 2012; Tanaka et al., 2015; Karaca et al., 2015; Buchert et al., 2016; Kurata et al., 2016; Szczaluba et al., 2017; Puusepp et al., 2018; Zanus et al., 2020; Richard et al., 2021), strongly suggesting that disruption of Ribi may cause the developmental disorders associated with these genes. SPATA5L1 and CINP were also recently implicated with Ribi in a Perturb-seq screen, which combines single-cell RNA sequencing with pooled genetic perturbations (Replogle et al., 2022). Further work will be needed to test the extent to which all the identified allelic variants of these four genes impact protein synthesis in specific cells within the developing nervous system.

An understanding of why mutations in the Ribi pathway and in different ribosome protein genes result in tissue-specific phenotypes is starting to come into focus. Aspects of human ribosomopathies are recapitulated in human cell culture models (Ebert et al., 2008) and early work had linked perturbations in Ribi with p53-dependent cell-cycle arrest (Pestov et al., 2001). Furthermore, different cell types appear to have ‘threshold sensitivities’ to perturbations in rRNA transcription (Falcon et al., 2022). Multiple in vivo experiments using zebrafish and mouse models revealed that specific aspects of ribosomopathy-like phenotypes can be suppressed by deletion of p53 (Jones et al., 2008; Barlow et al., 2010; Dutt et al., 2011; Jaako et al., 2011; Zhao et al., 2014; Bielczyk-Maczynska et al., 2015; Noack Watt et al., 2016; Watt et al., 2018; Tiu et al., 2021). This work led to a model whereby imbalances in the stoichiometry of ribosome proteins lead to a nucleolar stress response and inhibition of mouse double minute 2 (MDM2), which normally acts to ubiquitiylate p53 (Lohrum et al., 2003; Fumagalli et al., 2009; Narla and Ebert, 2010). Loss of MDM2 activity and stabilization of p53 results in cellular senescence and/or cell death (Lafita-Navarro and Conacci-Sorrell, 2022; Watt et al., 2022). In addition, recent work shows that a HEATR3-dependent nucleolar surveillance pathway, which monitors nucleolar integrity, also participates in the regulation of p53 stability (Hannan et al., 2022). However, in zebrafish models of DBA, loss of p53 rescues morphological phenotypes but not erythroid defects, suggesting that p53-independent pathways also contribute to tissue-specific phenotypes (Boglev et al., 2013; Yadav et al., 2014; Chakraborty et al., 2018).

Cell-specific sensitivities to ribosomal changes may be explained by specific mRNAs showing greater sensitivities to modest ribosome decreases than others (Mills and Green, 2017). Such sensitivity can dramatically affect development, particularly if the mRNAs in question encode for key factors needed for cell specification. One such example provides a partial explanation for the tissue-specific phenotypes observed in DBA. Haploinsufficiency in several ribosome protein genes, including RPS19, account for roughly 50% of DBA cases. Whole-genome sequencing reveals that mutations in GATA1, a transcription factor that promotes erythroid lineage commitment, are also associated with DBA (Ludwig et al., 2014). Haploinsufficiencies in ribosome protein genes result in a notable decrease in the translation of GATA1 mRNA. Moreover, increasing GATA1 expression rescues many of the blood phenotypes observed in DBA patient samples with reduced ribosome function (Ludwig et al., 2014). Of note, ribosome composition remains constant in DBA patient cells (Khajuria et al., 2018). Whether the translation of other cell-specification factors is reduced in different ribosomopathies, such as those that specifically result in nervous system defects, remains an open question.

Although insightful, these previous studies do not fully resolve the question of what accounts for the tissue specificity of ribosomopathies. For example, why do all ribosomopathies not present with anemia if GATA1 translation is especially sensitive to minor perturbations in translational capacity? Why do all cells not exhibit p53-dependent nucleolar stress upon imbalances in ribosome protein stoichiometry? One possible reason is that not all cells share the same translational capacity. Moreover, the dynamics of ribosome assembly likely vary across different cell types (Ni and Buszczak, 2022; Ni et al., 2022). The differential expression of rRNA processing and ribosome assembly factors during development may create distinct cell-specific bottlenecks in Ribi. Such bottlenecks could make certain cell types and tissues more or less sensitive to mutations in genes that act during different steps of Ribi (Fig. 3). Consistent with this model, allelic variants in functionally related groups of ribosome factors have been associated with related disorders (Table 1). Different cell types may tolerate allelic variants of particular ribosome proteins or Ribi factors with suboptimal functionality simply because they naturally express higher levels of these genes or maintain relatively high levels of ribosomes. Perhaps different levels of flux through specific steps in the biogenesis pathway, which likely vary between cell types, may also account for differences in how cells respond to ribosomopathy-linked mutations.

Developing cells exhibit changes in ribosome assembly and protein synthesis. For example, Drosophila female germ cells dynamically regulate rRNA transcription and ribosome assembly within germline stem cells and their early differentiating daughters (Neumüller et al., 2008; Fichelson et al., 2009; Zhang et al., 2014; Sanchez et al., 2016; Martin et al., 2022). Across different species, transcription in oocytes is turned off during the late stages of oogenesis. Regulation of stage-specific gene expression programs leading to the maturation of eggs depends on differential mRNA translation (reviewed by Mercer et al., 2021; Breznak et al., 2022). Thus, oocytes must generate and maintain a large pool of ribosomes during earlier stages of their development to support these later transitions. For instance, recent work has begun to characterize changes in mRNA translation as mouse oocytes transition from late oogenesis through meiosis I (Luong et al., 2020).

Fertilized zygotes re-enter the cell cycle and begins to proliferate. Zygotic transcription does not start immediately after fertilization, although the timing varies between different species. Because Ribi depends on ongoing rRNA transcription, eggs must house enough ribosomes to support early embryogenesis. Indeed, Caenorhabditis elegans eggs have enough maternal ribosomes to support embryonic development through to the first larval stage (Cenik et al., 2019). This is not the case in mammalian embryos, in which zygotic transcription begins at the two-cell stage and active rRNA transcription is needed to transition from a two-cell to four-cell embryo (Yu et al., 2021a). However, this effect likely goes beyond the requirement for new Ribi and may involve the reorganization of the chromatin landscape during this critical transition period. Characterizing the phenotypes of additional ribosome processing and assembly factors will provide a more detailed understanding of exactly when new Ribi is required during mammalian embryogenesis.

Biochemical and molecular studies of early mammalian embryogenesis are often hampered by difficulties in isolating sufficient material for analysis. Since their derivation, mouse and human embryonic stem cells (ESCs) have served as important models for understanding cellular potency and early developmental events in mammals, including human ESC-derived blastoids (Li et al., 2019; Yanagida et al., 2021; Yu et al., 2021b; Kagawa et al., 2022). There is a growing appreciation that ESCs and their differentiating progeny exhibit clear differences in protein synthesis. Although mouse ESCs exhibit lower levels of mRNA translation relative to their differentiated progeny, they counterintuitively display relatively high levels of ribosome protein synthesis (Sampath et al., 2008; Ingolia et al., 2011), which is accompanied by high levels of rRNA transcription (Watanabe-Susaki et al., 2014). This enhanced level of rRNA synthesis promotes ESC proliferation and appears to be controlled by epigenetic mechanisms, including rDNA methylation (Zheng et al., 2012).

Ongoing rRNA transcription contributes to pluripotency of both mouse and human ESCs (Zaidi et al., 2016; Jarzebowski et al., 2018). Accumulating evidence indicates that decreases in rRNA transcription and Ribi during differentiation are functionally significant. For example, disruption of fibrillarin (FBL), a methyltransferase for rRNA that serves a key function in Ribi, compromises stem cell pluripotency, whereas stable expression of this factor has the opposite effect (Watanabe-Susaki et al., 2014). In addition, treatment with the Pol I and Topoisomerase 2 inhibitor CX-5461 results in ectopic differentiation under normal ESC culture conditions (Woolnough et al., 2016). By contrast, HIV-1 Tat-specific factor 1 (HTATSF1) overexpression, which influences rRNA transcription and intron splicing of ribosome protein mRNAs, prevents differentiation (Corsini et al., 2018).

Mammalian adult stem cell lineages also exhibit clear differences in Ribi and protein synthesis during different stages of differentiation. For example, hematopoietic stem cells exhibit low levels of protein synthesis relative to their differentiating progeny (Signer et al., 2014, 2016; Jarzebowski et al., 2018; Magee and Signer, 2021). Other adult stem cell populations, including hair follicle stem cells (Blanco et al., 2016; Liakath-Ali et al., 2018), neural stem cells (Baser et al., 2019) and satellite cells of the muscle (Zismanov et al., 2016), also display similarly low levels of protein synthesis relative to differentiating cells in the same lineage. By contrast, hematopoietic stem cells exhibit robust rRNA transcription levels, indicative of ongoing Ribi (Jarzebowski et al., 2018). rRNA and ribosome protein production declines in multiple lineages as stem cells transition from a multipotent state to a more restricted state, reflecting regulation of Ribi during differentiation (de Klerk et al., 2015; Marcon et al., 2017; Pereira et al., 2019). High levels of Ribi promote stem cell self-renewal and disruption of factors that contribute to ribosome production often lead to loss of stem cell quiescence. For example, disruption of the 60S biogenesis factor Notchless (Nle) results in hematopoietic stem cell and intestinal stem cell loss (Le Bouteiller et al., 2013).

The examples highlighted in the previous subsections provide evidence that changes in Ribi and global protein synthesis levels during cell differentiation has important functional consequences. What regulates these switches in Ribi and protein synthesis remains an open question and represents important work for the future. Members of the field are beginning to explore this area. For example, a recent study describes how the TRIM-NHL domain proteins Mei-P26 and Brat act to decouple ribosome biogenesis and protein synthesis in germline stem cells and neuroblasts in Drosophila (Gui et al., 2022 preprint). Further work along these lines may reveal additional mechanisms that regulate developmentally important Ribi and protein synthesis switches.

Ribosome heterogeneity has long fascinated biologists (for reviews on this subject, see Shi and Barna, 2015; Ferretti and Karbstein, 2019; Li and Wang, 2020). Differences between ribosomes from liver and skeletal muscle cells were first observed in the 1970s, although this variation was largely attributed to sample preparation and other technical considerations (Sherton and Wool, 1974). More recent efforts have shown that ribosome heterogeneity exists in multiple forms, including sequence variation between rRNAs, differential modification of rRNA and overall ribosome composition. This includes incorporation of different ribosome protein paralogs or variability in the stoichiometry of specific ribosome proteins, post-translation modifications and binding of different ribosome-associated factors (reviewed by Li and Wang, 2020). How this variation impacts development and other aspects of the biology of multicellular organisms is under active investigation.

Clearly, the idea of functional specialization amongst subsets of ribosomes is appealing to developmental biologists because it represents another layer of potential regulation that could influence cell-fate specification and global translation programs in a cell-specific manner. Several models have emerged that seek to frame the functional significance of ribosome heterogeneity. These models are not mutually exclusive, but they do weigh the importance of ribosome specialization in driving distinct mRNA translation programs differently. The ribosome filter model (Fig. 4) posits that ribosomes are not uniform translation machines but play a regulatory role to ‘influence or filter the translation of various mRNAs’ (Mauro and Edelman, 2002). More specifically, this model proposes that different ribosomes prefer to interact with specific mRNAs and that there are competitive interactions between mRNA sequences for binding to rRNA and/or ribosomal proteins. Furthermore, it suggests that this filter can be modulated by changing or blocking specific sites on ribosomes (Mauro and Edelman, 2002). By contrast, the ‘ribosome concentration’ model (Fig. 4) posits that the overall number of functional ribosomes within a cell greatly influences the differential translation of specific mRNAs (Mills and Green, 2017). Different transcripts have variable 5′UTR lengths and often contain specific elements, including upstream open reading frames (uORFs), which can either inhibit or promote the translation of the main ORFs (Chen and Tarn, 2019; Zhang et al., 2019). Altering levels of ribosomes and other components of the translation machinery, such as eIF1 and eIF5, greatly influence the rate of translation initiation on mRNAs, particularly those that contain uORFs (Ivanov et al., 2022).

Work performed using yeast provides some of the most compelling evidence that different stresses promote ribosome heterogeneity within the total ribosome pool (Ferretti et al., 2017). These changes appear functionally significant and likely contribute to cellular stress responses, at least in the contexts that have been examined to date. For example, Saccharomyces cerevisiae harbor ribosomes with and without ribosomal protein S26 (RPS26) (Ferretti et al., 2017). RPS26-containing ribosomes recognize Kozak sequences in mRNAs, which help to define the translation initiation site and drive the translation of highly expressed genes. By contrast, exposure to stress results in an accumulation of RPS26-deficient ribosomes, which tend to preferentially translate mRNAs linked with specific response pathways (Ferretti et al., 2017). Strikingly, RPS26 can be removed from ribosomes in times of stress and added back later once cells have recovered in a ribosome protein chaperone TSR2-dependent manner (Yang and Karbstein, 2022). This finding suggests the possibility that the composition of ribosome populations can be widely, and potentially rapidly, altered after the ribosome maturation process is complete.

Ribosome heterogeneity also likely impacts mRNA translation programs in mammals. Quantitative proteomic experiments indicate that the large subunit ribosomal protein RPL10A is not present in all actively translating ribosomes within mouse ESCs (Shi et al., 2017). Ribosomes that contain RPL10A show selectivity in the translation of mRNAs involved in blood vessel development and different metabolic processes (Shi et al., 2017). In addition, haploinsufficiency of the ubiquitous RPS6 protein in the mouse limb causes specific developmental defects (Tiu et al., 2021). Many, but not all, of these defects can be suppressed by loss of p53 or by enhancing global protein synthesis (Tiu et al., 2021). In another example, RPL38 heterozygous mice exhibit homeotic transformation of the axial skeleton (Kondrashov et al., 2011). Further work showed that these phenotypes are accompanied by reduced translation of specific Hox mRNAs (Kondrashov et al., 2011). Here, IRES-like elements within the 5′UTR of different Hox mRNAs, including Hox9a, play important roles in the selective sensitivity of these specific genes to the presence or absence of RPL38 (Xue et al., 2015). Additionally, extension segments within rRNAs appear to foster interactions between ribosomes and stem-loop elements within specific Hox gene 5′UTRs in a species-specific manner (Leppek et al., 2020). However, recent studies suggest there may be errors in the sequence annotations of these genes (Akirtava et al., 2022; Ivanov et al., 2022). For example, cap analysis of gene expression sequencing (CAGE-seq) of mouse somites shows that the 5′UTRs of the bulk of Hox mRNAs in question are much shorter than previously annotated (Ivanov et al., 2022). Moreover, inhibitory upstream uORFs and the context of start AUG codons found within the 5′UTRs of these transcripts sensitize their translation to global levels of 60S subunits (Ivanov et al., 2022). These studies are more consistent with the ribosome concentration model, whereby Rpl38 mutants produce fewer functional ribosomes and that different Hox mRNAs, including Hox9a, are particularly sensitive to the reduced ribosome availability. However, the production of Hox transcripts with longer 5′UTRs, even at low levels, could influence Hox9a protein production in a cell type-specific manner. The differential translation of such transcripts could have subtle, but functionally significant, consequences in the context of development and evolution.

Analysis of ribosome protein paralogs in multicellular species provides further opportunities to examine potential functional differences between heterogeneous ribosomes. Drosophila encodes for several pairs of ribosome protein paralogs that exhibit tissue-specific expression patterns. This results in clearly detectable heterogeneity within ribosome pools (Hopes et al., 2022). One such paralog pair is Drosophila RPS5a and RPS5b (Kong et al., 2019; Jang et al., 2021). These two ribosome proteins have several amino acid substitutions throughout their sequences, with the most striking differences at their N termini. RpS5a displays ubiquitous expression across most cell types during development and in adult flies (Kong et al., 2019). By contrast, expression of RpS5b is largely restricted to the ovaries and testes (Kong et al., 2019). RpS5a mutants exhibit a Minute phenotype, like many other ribosome protein mutants, and whole-body loss of RpS5b results in female sterility marked by egg degeneration during mid-oogenesis (Kong et al., 2019; Jang et al., 2021). These different expression patterns and mutant phenotypes suggest the possibility that they carry out specialized functions. However, GAL4-UAS rescue experiments demonstrate that overexpression of RpS5a can fully rescue the RpS5b sterile mutant phenotype (Kong et al., 2019; Jang et al., 2021). These results suggest that RPS5a and RPS5b may be functionally equivalent. However, caution must be taken in this interpretation. For instance, overexpression of a given ribosomal protein paralog may be able to compensate for differences in the functionality of the alternative paralog. In addition, one must consider that in the lab flies do not encounter the multitude of variable conditions they are exposed to in the wild. The possibility remains that RPS5b-containing ribosomes provide a subtle but specialized function under specific environmental conditions.

Specific alternative modifications to rRNA also contribute to stress responses across species. For example, in yeast, two adenosine residues (A1781 and A1782) within the 18S rRNA were previously thought to be constitutively di-methylated (m⁶₂A) (Van Knippenberg et al., 1984). However, recent work shows that the methyltransferase Dim1p can modify these adenosines with single methyl groups (m⁶A) (Liu et al., 2021). Under normal growth conditions, ribosomes harboring the m⁶A marks are maintained at low levels, but upon sulfur starvation the number of ribosomes containing m⁶A at the A1781 and A1782 positions dramatically increases. Similar switches in the levels of m⁶₂A versus m⁶A ribosomes also occurs in mammalian cells upon sulfur starvation (Liu et al., 2021). Genetic and biochemical analyses suggest that ribosomes carrying m⁶A at the A1781 and A1782 positions preferentially translate transcripts that encode for proteins involved in sulfur metabolism (Liu et al., 2021). Likewise, human 28S rRNA is also regulated by methylation at A4220, mediated by the methyltransferase ZCCHC4 (Ren et al., 2019). ZCCHC4 is frequently overexpressed in tumors and its loss disrupts global mRNA translation and cell proliferation (Ma et al., 2019; Pinto et al., 2020). Furthermore, in C. elegans, N6 methylation of 18S rRNA regulates the differential translation of subsets of mRNAs in response to stress (Liberman et al., 2020). Thus, changes in rRNA modifications may serve as an evolutionarily conserved mechanism that drives the production of proteins that help restore cellular homeostasis in response to changing environmental conditions.

rRNA methylation also plays a role during development. The methyltransferase METTL5 is responsible for m⁶A modification of 18S rRNA at A1832 in mice; loss of Mettl5 in mouse ESCs does not disrupt pluripotency under conditions that foster ESC self-renewal. By contrast, the Mettl5 mutant cells described by Ignatova et al. (2020) exhibit spontaneous loss of pluripotency based on morphological changes, loss of alkaline phosphatase activity and reduced expression of key markers, such as Klf4, Nanog and Sox2. Levels of m⁶A1832 change during early differentiation and both the aforementioned studies provide evidence that loss of METTL5 compromises the overall differentiation potential of ESCs. A fraction of Mettl5 knockout mice survive to birth, indicating that m⁶A1832 is not absolutely required for viability (Ignatova et al., 2020). However, these mutant mice exhibit numerous developmental defects, including craniofacial abnormalities, hearing loss, behavior defects and male sterility. These phenotypes are consistent with human patients with METTL5 allelic variants (Hu et al., 2016; Reuter et al. 2017; Riazuddin et al. 2017). Thus, the differential methylation of 18S plays an important role in the normal development of mammals. Beyond development, emerging evidence indicates that m⁶A1832 also plays a role in different diseases, including cancer and atherosclerosis (Angelova et al., 2018; Dai et al., 2018; Quiles-Jiménez et al., 2020).

Developmental biologists have embraced the idea that ribosome heterogeneity plays important roles during the development of multicellular organisms, often invoking the idea that specific subsets of ribosomes play a regulatory function whereby they translate specific, developmentally important transcripts. Although exciting and provocative, caution should be taken before making such conclusions, as outlined previously (Ferretti and Karbstein, 2019). A recent study using yeast makes the case for both overall translational capacity and specialization being functionally significant. As mentioned above, yeast have many ribosome protein paralogs and the potential combinations of these various 59 paralogs allow for the biogenesis of >10¹⁷ types of distinct ribosomes (Hu et al., 2022). Using reiterative mutagenesis, one group generated a yeast strain that can produce only monotypic 40S subunits (Hu et al., 2022). Strains that carry these ‘homo-40S’ subunits are viable and maintain mRNA translation levels comparable to wild-type strains, but exhibit minor growth defects under various conditions, displaying particular sensitivity to the translation inhibitor paromomycin. Strikingly, doubling the dose of the ribosomal protein genes already present in the homo-40S strain rescues most of these phenotypes, but not entirely back to wild-type levels. These results indicate that ribosome concentration plays a crucial role in supporting yeast viability under different growth conditions, while leaving the door open for a fraction of heterogenous ribosomes having more specialized functions.

Thus, ribosome heterogeneity and regulated alterations in ribosome population diversity may have evolved to maintain the overall robustness of translation and allow for a certain degree of plasticity in the face of everchanging physiological conditions and different developmental contexts. Ribosome heterogeneity may buffer the translational capacity of a cell, so that overall levels of protein synthesis remain within an evolutionarily selected range under variable environmental conditions (Fig. 5).

Ribosomes and the dynamic regulation of mRNA translation programs play essential roles in the development of multicellular organisms. This is a rich field full of unanswered questions. Beyond the dynamic regulation of Ribi, we still do not have a complete catalog of the RNAs and proteins that participate in Ribi, nor do we fully understand how all these factors are coordinated to avoid nucleolar stress. How differentiating cells sense and coordinate ribosome numbers with cell division and function also remains an open question. The dynamics of ribosome production also likely vary between different cell types. Differentiating cells may have distinct molecular bottlenecks in the ribosome production process that make them sensitive to mutations in specific biogenesis factors. Further work will be needed to explore these ideas.

Ribosome heterogeneity exists in many forms across species, but the functional significance of ‘ribosome specialization’ remains less clear and should be carefully considered on a case-by-case basis. Ribosome heterogeneity may have evolved to ensure a particular ‘dose’ of global mRNA translation in different cell populations in different physiological environments. Whether these effects are independent of selectivity, whereby specific subpopulations of ribosomes only (or at least preferentially) translate specific mRNAs, remains to be rigorously tested. In the future, technical advances may allow investigators to examine the extent to which individual mRNAs within polysome fractions engage with a common or heterogenous group of ribosomes.

Ribosome heterogeneity may also represent an adaptive response. Assaying the dynamics of translation initiation, elongation and termination of specific subclasses of heterogeneous ribosomes, both in vitro and in vivo, would begin to address functional differences between distinct types of ribosomes beyond mRNA target specificity. Reciprocally, differential targeting of specific groups of transcripts by subsets of specialized ribosomes could have functionally important consequences during development and through the course of evolution. The ribosome concentration and ribosome filtration models are not mutually exclusive, and both scenarios likely contribute to gene regulation to varying degrees in different cellular and developmental contexts. With new discoveries constantly being made, the study of ribosome production and ribosome function in the context of development promises to be an exciting field for the foreseeable future.

We thank Kathryn O'Donnell, Prash Rangan and Ben Ohlstein for critical reading of the manuscript. The figures were created using BioRender.