Abstract
Understanding the regulation and structure of the eukaryotic ribosome is essential to understanding protein synthesis and its deregulation in disease. Traditionally ribosomes are believed to have a fixed stoichiometry among their core ribosomal proteins (RPs), but recent experiments suggest a more variable composition1–6. Reconciling these views requires direct and precise quantification of RPs. We used mass-spectrometry to directly quantify RPs across monosomes and polysomes of budding yeast and mouse embryonic stem cells (ESC). Our data show that the stoichiometry among core RPs in wild-type yeast cells and ESC depends both on the growth conditions and on the number of ribosomes bound per mRNA. Furthermore, we find that the fitness of cells with a deleted RP-gene is inversely proportional to the enrichment of the corresponding RP in ribosomes bound to multiple mRNAs. Together, these findings support the existence of ribosomes with distinct protein composition and physiological function.
Traditionally eukaryotic ribosomes have been thought to have a fixed composition of 80 core RPs1–3, some of which are represented by several paralogous RPs. Recent studies of eukaryotic ribosomes7–12 have demonstrated that (i) genetic perturbations to the core RPs specifically affect the translation of some mRNAs and not others and (ii) mRNAs coding for core RPs are transcribed, spliced, and translated differentially across physiological conditions13–17. These results suggest the hypothesis4–6 that, depending on the tissue type and the physiological conditions, cells can alter the stoichiometry among the core RPs comprising the ribosomes and thus in turn alter the translational efficiency of distinct mRNAs. However, differential RP–expression can reflect extra ribosomal functions7,18,19. Furthermore, polysomes (multiple ribosomes per mRNA) from different cell–lines have similar core RP stoichiometries20. Thus, the variable RP–stoichiometry in the ribosomes of wild-type cells that is suggested by the ribosome specialization hypothesis remains unproven.
To measure whether the stoichiometry among RPs can change with growth conditions, we used velocity-sedimentation in sucrose-gradients to isolate fractions containing monosomes and polysomes from yeast cells grown in minimal media with either glucose or ethanol as the sole source of carbon and energy21 (Fig. 1a). The proteins from individual fractions were spiked-in with known amounts of universal proteomics standard (UPS2) proteins, digested to peptides, labeled with tandem mass tags (TMT), and quantified on Orbitrap Elite based on the MS2 intensities of the TMT reporter ions21. The accuracy of estimated fold-changes for peptides can be gauged by the good agreement between the measured and the spiked-in fold-changes for UPS2 (Fig. 1b) that were measured simultaneously with the yeast RPs. The measured fold-changes for UPS2 peptides are about 12% smaller than expected from the spiked-in levels, as indicated by the 0.88 slope of the linear fit in Fig. 1c, most likely due to coisolation interference 22.
To quantify RP paralogs independently from one another, the fold-change of each RP was estimated as the median fold-change of its unique peptides, i.e., peptides whose sequences are found only in that RP. Our data have multiple high–confidence unique peptides for most RPs (Extended Data Fig. 1a, b) except for highly homologous paralogs. These quantified unique peptides allow estimating the relative RP levels in monosomes and polysomes collected from yeast grown in glucose or ethanol media (Fig. 1c). We systematically tested whether the variability in the estimated RP levels (Fig. 1c) reflects stoichiometry differences among the RPs or other factors and artifacts, such as noise in the MS measurements, a differential distribution of nascent RP polypeptides among monosomes and polysomes, posttranslational modifications (PTMs) of the RPs, and the presence of 90S ribosomal biogenesis particles23. These factors are unlikely to contribute significantly to the data in Fig. 1c since the majority of RP fold-changes that are estimated from multiple unique peptides differ by 10% or less (Extended Data Fig. 1c, d and Supplementary Discussion) and the 90S particles are less abundant than the mature ribosomes (Extended Data Fig. 2 and Supplementary Discussion).
We find that the relative RP levels depend on two factors: the carbon source in the growth media and the number of ribosomes per mRNA (Fig. 1c). The RP levels that are higher in glucose compared to ethanol also tend to increase with the number of ribosomes per mRNA (Fig. 1c). Importantly, the RP composition of trisomes in ethanol, Eth (iii), is more similar to the composition of monosomes than to tetrasomes. This observation shows that polysomes (3 ribosomes per mRNA) may have similar RP composition to monosomes, suggesting that the RP composition of monosomes is not necessarily indicative of a non–functional state.
To explore the extent to which the RP variability reflects paralogous RPs substituting for each other, we separately plotted the relative levels of RPs without paralogs (Fig. 1d) and of paralogous RP pairs (Fig. 1e). The levels of RPs without paralogs (Fig. 1d) vary similarly to the levels of all RPs (Fig. 1c), indicating that exchange among paralogous RPs is not the sole reason for the variable RP–stoichiometry. Furthermore, while the levels of some paralogs, such as Rpl37ap and Rpl37bp, are anticorrelated and consistent with paralog–exchange, the levels of other paralogs, such as Rpl17ap and Rpl17bp, are positively correlated and inconsistent with paralog–exchange across the analyzed ribosomes (Fig. 1e). The correlations between all paralogous pairs that we can confidently quantify based on unique peptides alone indicate similar distribution between positively and negatively correlated paralogous pairs (Fig. 1f).
Having found variability in the stoichiometry among yeast RPs, we sought to test its generality. We separated the ribosomes of exponentially growing mouse ESC on sucrose gradients (Fig. 2a) and quantified the core RPs, as annotated by Swiss–Prot, across different fractions corresponding to the number of ribosomes per mRNA (Fig. 2b). As in yeast, we observe significant variation in the stoichiometry among the mouse RPs depending on the number of ribosomes bound per mRNA. Furthermore, the ratios between the polysomal and monosomal levels of mouse RPs correlate to the corresponding ratios for their yeast orthologs (Fig. 2c; p-value < 0.03), suggesting that the RP-stoichiometry differences between monosomes and polysomes are conserved across yeast and mouse.
Next, we tested the variability among RPs and its phenotypic consequences by independent fitness measurements. Our observation that the RP stoichiometry depends on the number of ribosomes bound per mRNA parallels the observation that the translational activity per ribosome increases with the number of ribosomes bound per mRNA24,25. We therefore hypothesized that genetic deletions of RPs enriched in the more active ribosomes – as compared to RPs enriched in less active ribosomes – may result in a larger decrease of the translation rate and thus lower fitness. To test this hypothesis, we computed the correlation (Fig. 3a) between the fitness (in ethanol medium) of yeast strains with single RP-gene deletions26 and the corresponding relative RP levels measured in the tetra-ribosomal fraction (4 ribosomes per mRNA). Consistent with our hypothesis, the fitness of strains lacking RP-genes is inversely proportional to the relative levels of the corresponding RPs in the tetra-ribosomes (Fig. 3a). Extending this correlation analysis to the RP-levels in all sucrose fractions from both glucose and ethanol (Fig. 1c) results in a correlation pattern (Fig. 3b) that further supports our hypothesis by showing the opposite dependence for fractions with fewer ribosomes per mRNA: the fitness of strains lacking RP-genes is proportional to the relative levels of the corresponding RPs in fractions with fewer ribosomes per mRNA (Fig. 3b). This correlation pattern holds both for ethanol and for glucose carbon sources. To mitigate possible artifacts in the fitness data26 due to potential chromosome duplications in the deletion strains, we computed the correlations between the levels and the fitness of RP-deletion strains only for RPs without paralogs (thus unlikely to be affected by chromosome duplication) and found much higher magnitudes of the correlations (Fig. 3a, c). This result suggests that the variability in the RP stoichiometry is not limited to paralogous RPs substituting for each other.
We extended our fitness analysis from yeast to mouse using the published depletion data from CRISPR knockouts in human ESC (HUES62)27. We identified the closest mouse orthologs of each human RP with deletion data by BLAST alignment (Fig. 3d) and correlated the fitness of human ESC lacking the human RP orthologs to the RP levels across sucrose fractions that we measured (Fig. 2b). The correlation pattern (Fig. 3e) is similar to the one in yeast (Fig. 3a-c) and highly significant (false discovery rate (FDR) < 0.1%). This pattern indicates that the fitness of ESC lacking RP-genes is directly proportional to the relative RP levels in monosomes and inversely proportional to the relative RP levels in polysomes. The magnitude of this inverse proportionality increases with the number of ribosomes per mRNA (Fig. 3e), consistent with our hypothesis. The fact that the fitness of human ESC lacking RPs correlates significantly to the levels of the corresponding mouse orthologous RPs suggests that the variability of the RP stoichiometry and its biological function are likely conserved across mouse and human. The magnitude of this correlation increases significantly when the correlation is computed based only on orthologs whose sequences are over 80% identical between mouse and human (Fig. 3e), providing further evidence for the conserved fitness consequences of the altered RP stoichiometry.
Pioneering experiments performed in the 1960’s established that the rate of incorporating 14C-leucine in proteins per ribosome increases along the sucrose gradient as the number of ribosomes bound per mRNA increases24,25. The increasing translational activity across sucrose gradients24,25 parallels both our findings of altered RP stoichiometry across sucrose gradients (Fig. 2b) and the correlation of the altered RP stoichiometry to the fitness effects of mammalian RP deletions (Fig. 3e). These parallels can be explained by the expectation that the higher the translational activity of a ribosome, the higher the fitness cost of its perturbation in rapidly growing stem cells. The key factor required by this expectation is the variable ribosomal composition that we measured. The variable RP stoichiometry in the absence of external perturbations suggests that cells use variable RP composition of their ribosomes as a regulatory mechanism of protein translation.
Competing Interests
The authors declare that they have no competing financial interests.
Contributions
N.S., S.S. and B.B. performed experiments and collected data. N.S., A.v.O. and E.A. analysed the data. N.S. and A.v.O. discussed the results and wrote the manuscript.
Acknowledgments
We thank P. Vaidyanathan for help with the sucrose gradients and S. Kryazhimskiy, W. Gilbert, P. Vaidyanathan, G. Frenkel, and A. Murray for discussions and constructive comments. This work was funded by a grant from the National Institutes of Health to A.v.O. (R01-GM068957) and Alfred P Sloan Research Fellowship to E.M.A.