Analyzing ribosome remodeling in health and disease
AAnalyzing ribosome remodeling in health and disease
Aleksandra A. Petelski, , , & Nikolai Slavov , , , (cid:0) Department of Bioengineering, Northeastern University, Boston, MA 02115, USA Barnett Institute, Northeastern University, Boston, MA 02115, USA Department of Biology, Northeastern University, Boston, MA 02115, USA (cid:0)
Correspondence: [email protected] or nslavov@nor theaster n.edu Increasing evidence suggests that ribosomes actively regulate protein synthesis. However,much of this evidence is indirect, leaving this layer of gene regulation largely unexplored,in part due to methodological limitations. Indeed, we review evidence demonstrating thatcommonly used methods, such as transcriptomics, are inadequate because the variability inmRNAs coding for ribosomal proteins (RP) does not necessarily correspond to RP variability.Thus protein remodeling of ribosomes should be investigated by methods that allow directquantification of RPs, ideally of isolated ribosomes. We review such methods, focusing onmass spectrometry and emphasizing method-specific biases and approaches to control thesebiases. We argue that using multiple complementary methods can help reduce the danger ofinterpreting reproducible systematic biases as evidence for ribosome remodeling.
Introduction
The control of gene expression is crucial for all biological processes, from developmental stagesand homeostasis maintenance to regeneration processes. This regulation occurs at multiple layers,both transcriptional and post-transcriptional levels. Historically, transcription has been studiedmore extensively than translation, in large part because of the accessibility of technologies fornucleic acid analysis. However, gene regulation via translation was appreciated as early as the late1960’s. For example, the production of insulin was linked to the increased number of polysomesand protein synthesis [1]. Similarly, the synthesis of globin (hemoglobin polypeptides) was foundto depend on heme, the oxygen-carrying iron-rich molecule in the blood [2]. In the early 1980s,different aspects of early embryonic development, such as blastocyst formation [3] and disparatepatterns of protein synthesis [4] were linked to post-transcriptional regulation. Key developmentalgenes, such as Oscar, are regulated at the level of translation and spatial localization [5, 6]. Suchpost-transcriptional regulation is mediated at least in part by translational control mechanismsincluding RNA binding proteins, translation factors, and micro RNAs [7, 8]. a r X i v : . [ q - b i o . Q M ] A ug n addition to these well-established mechanisms, increasing evidence suggests that ribosomesmay also regulate translation of mRNAs [9–14]. Ribosomes have long been viewed as passiveplayers in translation, with a fundamental role of catalyzing peptide-bond formation but exertingregulatory effects only based on their availability [15, 16] and translational control being exerted bya teamwork of cis- and trans-regulators in which the interaction of mRNA structures and sequencesworked with translation factors, RNA binding proteins, and microRNAs; the ribosome complexwas simply viewed as an effector of translation. However, ribosome-mediated regulation throughthe alteration of ribosomal RNA and ribosomal proteins were both hypothesized [9, 11, 17] andsupported by mostly indirect evidence as recently reviewed by Emmott et al. [14]. Quantificationof ribosomal proteins from isolated ribosomes has begun to provide more direct evidence, as mam-malian ribosomes have been found to exhibit differential protein stoichiometry that depends on thegrowth conditions and on the number of ribosomes per mRNA[18]. Such observations suggestthat different ribosome complexes may exist in order to fulfill disparate functions, which can con-sequentially have regulatory effects on translation. This model, termed ribosome specialization,challenges the notion that ribosomes are static enzymes and instead introduces them as active par-ticipants in post-transcriptional regulation. In this model, ribosomes could be specific to distinctcell population and can affect the translation of mRNAs. The hypothesis of ribosome-mediatedtranslational regulation has been further supported by observations of differences in ribosomalprotein composition occurring under stress [19–21] and cell differentiation[22–25]. Localizedsynthesized RPs in the axon can also contribute to ribosome remodeling [26, 27], suggesting pos-sible roles of ribosome-mediated translational control in neuronal functions. The possibility ofribosome specialization is further supported by the observations that mutations of specific riboso-mal proteins selectively affect the synthesis of specific proteins and are strongly associated withdistinct phenotypes such as cancer and aging [28, 29]. Such selectivity suggests that ribosomescan regulate gene expression [10, 11, 14, 30]. Interestingly, some RPs, such as RACK1, havebeen observed to dynamically associate and dissociate from ribosomes and specifically affect thetranslation of short mRNAs [31–33].Specialized ribosomes have been suggested to hold specific functional roles, especially in thecontext of immunology and cancer. The idea of ribosomes driving cancer progression could betied to observations of disease states characterized by dysfunctional ribosomes, disorders collec-tively known as ribosomopathies [34, 35]. Patients with such disorders showed increased risk fordiseases of uncontrolled cell growth, such as cancer, later in life [36]. More direct evidence hasshown the association of RP gene mutations with numerous cancers, raising the prospect of the2xistence of oncoribosomes [37, 38]. Ribosomes have been also implicated in the immunosurveil-lance of cancer and other types of pathogenic cells. MHC class I molecules, which are importantin alerting the immune system when cells are virally infected, are believed to be derived fromDRiPs (defective ribosomal products), unstable molecules that degrade much more quickly thanfunctional proteins that are in stable conformation [39]. In order to rapidly produce such prod-ucts, a subset of ribosomes could be assigned to synthesize DRiPS that have enhanced antigenpresentation [40, 41]. Immunoribosomes may also serve as an efficient source of peptides thatcan stimulate antibody production upon the invasion of pathogenic molecules. These models offunctionally specialized immunoribosomes and oncoribosomes remain insufficiently tested, andfurther testing would benefit from the approaches discussed in this review.Ribosomes in eukaryotes are made of about 80 core ribosomal proteins and four ribosomalRNAs (rRNAs) [42]. The modifications of both molecular groups are important in the functionalityof the ribosome complex and are potential sources of ribosome heterogeneity and specialization[43, 44]. Modifications of rRNA, such as methylation and pseudouridylation, which together spana total of 7000 nucleotides, are known to stabilize the ribosome structure. Variant rRNA allelesand rRNA methylation may contribute to ribosomal specialization. In both mice and humans,several rRNA sequence variants were identified and shown to exhibit tissue-specific expression;furthermore, at least 23 percent of rRNA nucleotides are estimated to exhibit variant alleles withinthe general human population [45]. On average, 32 variants were found to be expressed in singleindividuals, while those sequence variants were found to significantly overlap with sequences thatare functionally important to ribosome function. The diversity of rRNA variants are thus suggestedto to have a biologically important role. Indeed, the methylation of the 16S rRNA at a specificguanosine nucleotide revealed that this modification plays a role in controlling mistranslation andcould explain streptomycin-resistant phenotypes of M. tuberculosis [46]. Both rRNA and proteinmodifications may contribute to ribosome specialization; The role of rRNA has been reviewed byMauro and Matsuda [47], and in this review we will focus on methods for investigating ribosomespecialization via modifications of the core ribosomal proteins.
Post-transcriptional regulation of ribosomal proteins
While suggestive evidence for ribosome remodeling has originated from indirect methods (suchas measuring transcripts coding for RPs [50, 51]), such data remain inconclusive because RP syn-thesis and degradation are extensively regulated [48, 49, 52, 53]. RP molecules that are not incor-3
RNA Ribo-seq Proteins
Individuals R i bo s o m a l p r o t e i n s -101 m R N A R i b o - s e q P r o t e i n s -2-1012 F o l d c hange , l og a − l og r P T R r a t i o P−Value
Rac GTPase binding × −5 tricarboxylic acid cycle × −9 respiratory electron chain < − translational elongation × −7 small ribosomal subunit × −7 − Kidney
P−Value translational termination × −7 ribosome × −5 respiratory electron chain × −5 tRNA aminoacylation × −6 Golgi transport complex × −9 Stomach b Figure 1 | Ribosomal proteins are under post-transcriptional control ( a ) A heatmap showing the levels ofribosomal genes at the levels of RNA, ribosome density, and protein in lymphablastoid cell lines based ondata from Battle et al. [48]. The data points for each gene are displayed as log fold-changes relative totheir mean. The corresponding distributions of fold changes for all genes across all cell lines are shownto the right. To control for variable input amounts from different cell lines, the data from each cell linewere normalized to the same total amount of RP gene products. ( b ) Protein to mRNA ratios (PTRs) werequantified across different human tissues, and gene sets with statistically significant shifts in the tissue-typespecific relative protein to RNA ratios (rPTR) are highlighted. Of interest, ribosomal proteins have highrPTRs in the kidney and low rPTRs the stomach, indicating significant post-transcriptional regulation. Thispanel is based on the analysis by Franks et al. [49]. porated into a complex are rapidly degraded[52]. Because of this post-transcriptional regulation,analysis of ribosome remodeling in health and disease should rely on direct protein measurements.While the abundance of RP transcripts are usually the most accessible data, these measures areindirect. Consider, for example, the variability of RP transcripts, ribosome density, and ribosomalproteins across a panel of lymphoblastoid cell lines shown in Fig. 1a. The data indicate substantialtranscript variability, which diminishes at the level of ribosome density and is almost absent forthe ribosomal proteins, Fig. 1a. Thus, variable RP transcripts do not necessarily indicate variableabundance of RPs, Fig. 1a. More generally, the ratio between RPs and their corresponding tran-scripts can vary substantially as shown in Fig. 1b for two tissues and observed in many other cases[49, 54]. Therefore, the levels of mRNAs coding for RPs are rather indirect evidence to supportconclusions about the protein composition of ribosomes.To obtain more direct evidence for protein remodeling of ribosomes, one should directly mea-sure RP abundances as a more direct approach for evaluating ribosome remodeling across differentconditions, Fig. 1a. Such analyses in cell lysates have suggested changes in the protein compo-sition of ribosomes as budding yeast undergoes the diauxic shift [55], during aging [56, 57], andupon LPS-stimulation of mouse dendritic cells[58]. In addition, RP analysis from cell lysates hasshown that ribosomal transcripts exhibit slower elongation rates with decreased protein production4elative to other transcripts with similar ribosome densities[59].Although RPs are known to degrade very quickly when not incorporated into complexes[52],RPs quantified in total cell lysates may originate in part from other extraribosomal complexes[17, 60–64]. Excluding the influence of such extraribosomal complexes requires the analysis ofisolated ribosomes. The use of isolation methods, such as sucrose gradient fractionation or affin-ity purification[65], prior to protein measurements provides an even more direct way to assessribosome remodeling in different conditions [18, 66, 67].In some studies, the measurement noise is comparable to the variability of RPs across condi-tions [68]. In such cases, we may conclude that ribosomes do not remodel across the examinedconditions or the degree of remodeling it too small to be detected by the methodology used. Thesmaller the measurement errors, the more confident we may be that ribosomes do not remodelacross the set of studied conditions. For example, the degree of ribosome remodeling (if any at all)is very small during the aging of mouse brains [68]. As observed RP changes can be small, theymay be comparable to or smaller than the measurement noise. RP mol/cell, GFP R P m o l / c e ll , W e s t e r n B l o t s Ribosomal Protein (RP) abundance = 0.25
Figure 2 | Different methods provide different estimates of ribosomal protein abundance.
A scatter plot com-paring estimated abundance of ribosomal proteins from Western blots [69] and from flow cytometry analysisof RPs tagged by green fluorescent protein (GFP)[70]. The Pearson correlation between the two estimatesis modest, ρ = 0 . , despite the fact that the measurements by each method are highly reproducible. Theweak correlation many reflect shared biases from the construction of the tagged proteins or biological dif-ferences in protein abundance. The first option seem much more likely, but it remains inconclusive withoutadditional data and analysis. recise and reproducible measurements may not be accurate Since each method comes with its own potential for biases and systematic artifacts, the characteri-zation of ribosome remodeling also calls for the use of complementary methods. A highly precisemethod can offer consistently reproducible measurements that are also consistently biased. Repro-ducibility does not necessarily correspond to accuracy. For example, RP levels estimated by GFP-tagging and by Western blots differ significantly (Fig. 2a) despite the fact that replicates withineach method are reproducible [69, 70]. Generally, measurements can be affected by systematic bi-ases, leading to technically reproducible but inaccurate measurements. Such data may consistentlysupport an incorrect representation of the studied biological systems. The effect of biases is espe-cially important to recognize when studying ribosome remodeling, since the changes of ribosomalprotein stoichiometry are often relatively small as observed in previous studies [18, 25, 71]. R p l R p s R p s R p s R p s x R p l R p l R p l R p s R p l R p s R p l p2 R p l R p l R p s R p l R p l R p l p1 R p l R p l R p l R p s R p l R p s R p s R p l R p l R p s R p s R p s R p l R p l R p l R p l R p s R p l R p s R p l R p l R p l R p l R p l R p s R p s a R p l R p s R p l R p l R p s R p l R p l R p l R p l R p l R p s R p l R p s R p s R p l R p l R p s R p s R p s R p l R p l p0 R p l R p s R p l R p s R p s R p s Rpl8Rps3aRps7Rps3Rps4xRpl6Rpl13Rpl19Rps2Rpl37aRps23Rplp2Rpl4Rpl21Rps8Rpl11Rpl12Rplp1Rpl34Rpl29Rpl3Rps20Rpl17Rps21Rps13Rpl31Rpl35aRps16Rps18Rps9Rpl18Rpl28Rpl27Rpl23Rps25Rpl7aRps14Rpl32Rpl15Rpl10aRpl30Rpl26Rps24RpsaRpl36aRps27Rpl35Rpl7Rps15Rpl10Rpl13aRpl18aRpl14Rpl23aRps10Rpl22Rps27aRps19Rpl27aRpl24Rps15aRps29Rps12Rpl38Rplp0Rpl9Rps26Rpl5Rps11Rps28Rps6 -1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1 C C C l u s t e r s o f r i bo s o m a l p r o t e i n s Figure 3 | Covariation of ribosomal proteins across diverse single cells
Pairwise Pearson correlations betweenRPs computed from protein levels measured by SCoPE-MS in individual differentiating stem cells [72, 73].The correlation matrix was clustered based on the cosine between the correlation vectors. This panel isreproduced from Budnik et al. [72]. nalyzing ribosome remolding in individual cells Ribosome remodeling is likely to contribute to the specialized proteomes of the diverse cell typesthat arise during development and diverge to form different tissues [14]. Furthermore, cyclic tran-scription of mRNAs coding for RPs suggests that ribosome biogenesis is highly temporarily or-ganized and metabolically coordinated during the cellular life cycle [74–76]. While these possi-bilities are of considerable interest, their investigation poses particular challenges for direct RPquantification by complementary methods [77, 78]. These challenges stem from the difficulty ofquantifying proteins in tissues comprised of heterogeneous cells. Single-cell proteomics has ap-proached this question in the context of embryonic stem cells differentiating to epiblast lineagesin embryoid bodies [73]. In this system, the correlations among RPs (Fig. 3) revealed one largeand one small cluster, suggesting that the RPs from the small cluster covary in a cell-type specificmanner. However, these data have not been cross-validated by independent methods and thus re-main inconclusive. Nonetheless, advances in single-cell proteomics hold the potential to increasethe reliability of single-cell protein analysis and to enable cross-validation of such measurements[77–80]. Thus these technological advances may soon enable direct examination of ribosome spe-cialization within the diverse cell types that comprise different tissues.
The need for complementary methods
The technical biases of each method type are challenging to overcome, but the use of complemen-tary methods with divergent weaknesses can help guard against the potential influence of system-atic artifacts on the final results. Thus, the use of complementary techniques allows for both theattainment of strong evidence for ribosome remodeling and the cross-validation of results that cansupport more confident conclusions. As an example, RPs from isolated ribosomes may be quanti-fied separately by both mass spectrometry and Western blots and then the results can be comparedas shown in Fig. 4. The measured RP differences obtained from these two disparate methods arehighly similar, indicating that the results are likely to be driven by biological effects rather than theinherent biases attributed to each method. More broadly, if biases were the main force behind theresults, the outcomes would have been presumably different; thus, the more different the methods,the less likely they are to share biases, and the more beneficial they are in being used together ascomplementary methods.Direct quantification of RPs with complementary methods (as shown in Fig. 4) can revealchanges in the ribosome complex that may contribute to post-transcriptional regulation of gene7 ass–Spec Western Blots
Rpl11 Rps29 Rps14 −1.5−1−0.500.511.522.5 P o l ys o m es / M ono s o m es R a t i o s , l og a Rps14Rps29Rpl11
Mono- somes
Poly- somes
Rpl32
Rpl11 Rps29 Rps14 P o l ys o m es / M ono s o m es R a t i o s b Figure 4 | Quantifying ribosomal proteins with complementary methods
Monosomal and polysomal ribosomesfrom embryonic stem cells were isolated by sucrose gradients and analyzed by mass-spectrometry and by Westernblots [18]. The results exemplify qualitative agreement for the polysomal enrichment of Rps29 and Rps14 as well asrelatively small differences across the two methods. The differences may be due both to the different cell lines usedor to the quantification biases of each method [18]. ( a ) The Polysomal enrichment of RPs was quantified by mass-spectrometry using isobaric tags. The analysis was performed by digesting the RPs either with trypsin or with Lys-Cand the results provided consistent estimates that are combined in these boxplots [18]. ( b ) Polysomal enrichment ofRPs quantified by Western blots [18]. RPs were quantified by Western blots in monosomes and polysomes from E14mouse ESCs. Rpl32 was used as a loading control and the boxplots summarize data from 9 ratios for each quantifiedRP. The panels are reproduced from Slavov et al. [18]. expression [18]. Mass spectrometry (MS) offers an array of such complementary approachesfor discovering and validating ribosome remodeling. These powerful methods directly measureprotein or peptide abundance, either through relative or absolute quantification, and can provideinformation about the stoichiometry of the ribosome complex along with modifications of the ri-bosomal proteins. These results can be enhanced and cross-validated by structural methods, suchas X-ray crystallography and cryogenic electron microscopy (cryo-EM). These latter methods candirectly probe the arrangement of ribosomal proteins within the complex. In this review, we dis-cuss the types of biases, limitations, and advantages associated with each approach type, focusingparticularly on the investigation of ribosome remodeling.8 uantifying ribosomal proteins via mass spectrometry The levels of ribosomal protein expression can be directly measured using mass spectrometryproteomics. Whole proteins, or even entire ribosome complexes can be analyzed by top-down massspectrometry [81, 82]. However, because MS analysis of full length proteins is technologicallychallenging, it is more common to first digest proteins into peptides and then analyze the peptidesby MS. This latter approach is known as bottom-up mass spectrometry, which involves the caveatof inferring protein levels from digested peptides. These two branches (top-down and bottom-up)of mass spectrometry can be implemented by many methodologies, each of which has systematicbiases and can directly quantify RP stoichiometry across biological conditions.Top-down and bottom-up proteomics methods provide complementary measurements, as theyquantify proteins in inherently different ways and thus can help detect and mitigate biases. Techni-cal biases that can affect the ultimate biological interpretation of any experiment can be potentiallyintroduced at any step of the proteomics pipeline, from cell lysis and sample preparation to MSanalysis. These distinct biases can be controlled by using different methods. Ultimately reliableresults require cross-validating the results from different methods that share as few biases as pos-sible.
Biases in bottom-up mass spectrometry
Bottom-up proteomics can be implemented by many methods, all of which have caveats, specif-ically for quantifying ribosomal proteins. The amount of ribosomes in most cells is generallyample, about 10 million in a typical mammalian cell. Yet, the sequences of ribosomal proteins arerather short, with a median length of about 133 amino acids in human cells, in stark contrast tothe overall proteome sequence length, which is 375 amino acids. Thus, ribosomal proteins tendto produce fewer peptides during digestion. In addition, ribosomal proteins are likely to produceshorter peptides when using trypsin, the most commonly used protease, due to a large number ofarginine (R) and lysine (K) amino acids present in the protein sequences [83]. The over representa-tion of K and R may contribute to a higher miscleavage rate, which can complicate analyses. Thesedownsides can be taken into account by the use of multiple alternative proteases such as Lys-C andGlu-C in order to cross-validate results with the commonly used protease of trypsin [84, 85]. Thesebiases can be very significant when estimating absolute protein abundances (due to differences inpeptide flyability, a collective term describing efficiency of ionization and detection), and thesebiases may cancel out in relative protein quantification [86, 87].9 uantifying protein stoichiometry by bottom-up mass spectrometry
The differences in peptide flyability and other peptide specific biases poise a major challenge toquantifying stoichiometry (ratios) between different proteins and their proteoforms [86, 88]. In-deed, a peptide might be more intense because it is delivered more efficiently to the MS analyzerrather than because it originates from a more abundant protein. These peptide specific biases cancelout when performing relative quantification of a protein across different samples. This idea of can-celing out biases can be extended by first-principle models, such as HIquant, to allow quantifyingstoichiometry between different proteins independent from peptide specific biases [87]. Such anapproach is likely to be particularly fruitful for quantifying RP preforms originating from differentprologues, alternative splicing events or post-translational modifications [14, 87, 89].In bottom-up proteomics, cells can be lysed and proteins extracted by a variety of methods[90–92]. Then proteins are digested to peptides, which are separated via liquid chromatography(LC) and ionized through electrospray ionization (ESI) or matrix-assisted desorption/ionization(MALDI). Then, these peptides are introduced into the mass spectrometer as precursor ions, whichcan be used for direct quantification or can be further isolated and disintegrated into fragmentions that are then used for sequence identification and quantification. Bottom-up MS may eithersimultaneously isolate and fragment multiple peptides in parallel (known as DIA; data independentacquisition), or analyze a single peptide at a time (known as DDA; data dependent acquisition).DDA methods were introduced in 1990s and have been widely used for decades [93, 94]. DIA wasintroduced later (2004) by Yates and colleagues [95] and has matured into methods that offer theadvantage of parallel analysis and afford identifying and quantifying many thousands of peptidesin a single run [96, 97].Quantification of peptides can be based directly on precursor ions (MS1-based) or fragmentions (MS2-based). Both of these peptide quantification approaches can be implemented by mul-tiple methods. These methods differ in the way and extent to which they control technical biasesstemming from the proteomics pipeline, from sample preparation procedures to LC-MS analysis.Wet-lab procedures encompass the experimental steps of isolating ribosomes from cell lysates,protein digestion into peptides, and various labeling techniques. In LC-MS analysis, biases canbe introduced due to differential peptide separation, ionization, and isolation for MS2 analysis.The available bottom-up methods control for these technical biases at different stages. Due to thedisparities in biases, different bottom-up approaches can be viewed as highly complementary tech-niques that allow us to gain more confidence in results showing biological changes, especially inribosome remodeling where the changes may be small.10 ontrol of biases at the MS1-Level of Peptide Quantification
Methods that use the MS1 level in order to quantify peptides can be further classified either aslabeled or label-free approaches. The label-free approaches are attractive for their simplicity andfewer experimental steps. In these methods, each sample undergoes the sample preparation andmass spectrometry analysis separately. In contrast, labeled techniques allow analyzing multiplesamples in parallel and thus more opportunities to control for biases that occur during the par-allel stages. Labeling approaches for MS1-based quantification include: SILAC (Stable IsotopeLabeling of Amino Acids in Cell Culture), dimethyl labeling, and mTRAQ. SILAC introducessample labeling to living cells during protein synthesis through the metabolic labeling of newlysynthesized proteins[98]. More specifically, cells are incubated in cell culture medium that con-tains stable-isotope enriched amino acids, usually arginine and lysine [99, 100], which are thenincorporated into new proteins. Dimethyl labeling, on the other hand, introduces sample labelingafter digestion. This chemical labeling approach relies on the reaction between formaldehyde andsodium cyanoborohydride with lysine side chains and N-terminal primary amines in order to formdimethylamines [101, 102]. mTRAQ quantification, which also introduces labels after digestion.It was designed specifically for targeted MS, in which specific peptides found in previous shot-gun runs could be probed [103, 104]. All of these methods quantify peptides at the level of theprecursor ions.The introduction of biases can be controlled through mixing the samples as early as possiblewithin the experimental pipeline. The earlier the samples are combined together, the sooner thesamples can be exposed to identical conditions and the sooner biases can be controlled. Among theavailable MS methods, SILAC allows mixing of samples earliest, even prior to ribosome isolationand MS procedures, thus enabling the control of all biases associated with these experimentaland MS steps. This is especially advantageous when studying subcellular fractions, such as theribosome complex. Labeling methods, such as dimethyl labeling and mTRAQ, allows the mixingof samples prior to LC separation and MS analysis, which controls for LC- and MS-related biases,while label-free methods will not control for these biases, as each sample is analyzed individually.The processes of peptide separation via liquid chromatography and subsequent MS analysiscan introduce technical biases that MS1-based labeling methods can control but label-free cannot.Each run in the mass spectrometer is subject to a host of variable factors that might influence mea-sured ion intensities, including variability in peptide separation, ionization, and instrumentation.This can cause different runs to experience drifts in retention time and m/z, leading to complicatedanalyses. Ion suppression, a problematic phenomenon that affects the final amount of charged ions11hat ultimately reaches the detector, can also undermine quantitative accuracy. In the cases of la-beling, digested peptides from each sample are separated, ionized, and analyzed together, allowingfor the samples to experience the same nuances associated with each of those steps. Label-free ap-proaches in particular suffer from the fact that large portions of peptides are not detected in everysample, which is termed as the missing value problem[105]. This problem makes it more difficultto compare RP abundance across different conditions. Advances relying on matching peptide in-tensity readouts, i.e. MaxLFQ[106], and on enhanced peptide identifications via DIA [95–97] orDDA methods incorporating retention time information [78, 107, 108] can mitigate the missingvalue problem when using label-free approaches.While MS-1 based labeling methods allow controlling for biases during peptide separationand ionization, labeling itself can introduce biases, limitations, and artifacts. SILAC introducesheavy isotopes into live cells and animals, which can induce unintended growth changes[109, 110]and may even affect behavioral characteristics of mice[111, 112]. Also, depending on the modelsystem used, the potential conversion of arginine to proline may lead to underestimation of heavy-labeled peptides[113]. An additional limitation when using SILAC is that the time to fully label abiological system of interest can take days to weeks. Dimethyl labeling, on the other hand, takesaround 5 minutes and is highly cost-effective, as the reagents are inexpensive [114]. However,dimethyl labeling has been associated with a loss of hydrophillic peptides[115], which can lead tofewer overall peptide identifications. Overall, any type of labeling approach should be assessed forcompleteness of labeling to allow for confident quantitation of RP remodeling.MS1-based quantification techniques share some common limitations. A disadvantage of MS1-based labeling methods is that the number of multiplexed samples is usually limited to 2 or 3. Asthe number of isotopically labeled samples increases, so does the number of precursor ions in MS1scans. The high density of ions may lead to interference between coeluting ions with very similarm/z ratios. The use of NeuCode metabolic labeling offers higher SILAC multiplicity through theuse of labels that differ in mass on the scale of milli-Daltons[116, 117]; thus, this technique requireshigher resolving power of the mass analyzers [116, 117]. Because of the inherent limitations thatMS1-level methods present, it is important to consider using complementary methods rather thanmultiple technical replicates of just one method type. MS1-based methods can be co-validatedwith MS2-based approaches, which still come with their own biases; however, the set of biasesderived from both quantification approaches are different, allowing for increased confidence inquantifying even small RP changes. 12 ontrol of biases at the MS2-Level of Peptide Quantification
As opposed to MS-1 based methods, precursor ions are isolated and then fragmented, and the re-sulting (fragment) ions analyzed by another (MS2) scan. In most cases, MS2-based quantificationis used with isobaric tags, such as tandem mass tags (TMT)[118] or Isobaric Tags for Relativeand Absolute Quantification (iTRAQ)[119, 120]. DIA analysis can also benefit from incorporatingMS2-level data in order to improve the accuracy of label-free quantification [121].With isobaric labeling, peptides of each sample are covalently labeled with a sample-specificmass tag. Mass tags are comprised of a reporter ion, a balance group and a reactive group, whichis usually an amine reactive group (NHS group) that binds to primary amines from the peptides.After labeling, the samples are mixed together, allowing the analysis multiple samples (up to 16with TMT Pro) in a single mass spectrometry run. Thus, the multiplexed samples undergo ion-ization and ion selection for MS2 analysis together, greatly mitigating the missing value problem.Additionally, the multiplexing permits analyzing more samples per unit time. Another advantageis that the number of precursor ions detected during MS1 scans does not increase with the numberof samples, as each isobaric tag has an identical mass which allows a particular peptide from differ-ent samples to appear as a single feature in the ion map defined by retention times and m/z ratios.Peptides from different samples are thus indistinguishable at the MS1 level. Upon isolation andfragmentation of precursor ions, they release distinct reporter ions and peptide fragments, someof which remain bound to the balance group. Since the reporter ions and the balance groups havesample-specific number of heavy isotopes, they allow for the ratiometric quantification of samples.Since all peptide from a sample release the same reporter ion upon fragmentation, the accuracyof quantifying a peptide from its reporter ions is dependent on isolating only its precursor ion forsubsequent fragmentation and MS2 analysis. In practice, the MS2 scan typically also containscoisolated peptides. The degree of coisolation can be estimated and used to remove peptides withunacceptably high coisolation. Narrower isolation windows [79] and sampling elution peaks attheir apexes [122] reduce coisolation but may not completely eliminate it. When multiple peptideprecursors are coisolated for fragmentation, the measured reporter ion intensities are proportionalto the superposition of peptide abundances, which results in inaccurate quantification and generallycompressed peptide ratios. However, ratio compression is not necessarily due to coisolation sinceit may be caused by other factors, such as unintentional cross-labeling or sample carry-over onthe LC column. Coislation can be reduced by subjecting fragment ions to further isolation andfragmentation steps by taking MS3 or even higher scans[123]. This MS3 approach reduces thenumber of analyzed ions and thus diminishes the sensitivity and the throughput of the analysis. A13econd option for reducing biased due to coisolation is to use the peptide fragments with attachedbalance groups, which are termed complement ions[124]. These complement ions are producedduring the fragmentation step as a result of the mass balance group remaining attached to thepeptide or its fragments, and thus they can be specific to the analyzed peptide.Abundant peptides, such as those originating from RPs, tend to be less affected by coisolationsince the majority of the reporter ions used for quantification will be derived from the abundantpeptides. However, coisolation can still contribute significant bias to RP quantification and thusnecessitates quality controls. One way to benchmark data quality is through the calculation of theagreement of different peptides originating from the same protein. Ribosomal peptides that origi-nate from a RP should indicate consistent quantification of the protein; the degree of consistencycan be quantified by measures such as reliability or coefficients of variation [49].The methods of quantifying peptides at either the MS1 or MS2 levels come with inherent bi-ases that might complicate analyses and may present reproducible results that are actually artifactsinduced from systematic biases. However, the set of biases for each method type is different, andthus different methods may complement each other. Instead of choosing just one mass spectrom-etry method to identify changes of ribosomal proteins, methods that differ as much as possible(such as MS1 and MS2 based quantification) or non mass-spec methods should be used in parallel(as shown in Fig. 4) so that the results can be co-validated.
Top-down mass spectrometry
With top-down proteomics[81, 82], whole ribosomal proteins or even whole ribosomes can beanalyzed, offering an more intact picture of the ribosome and potential characterization of prote-oforms and post-translational modifications (PTMs). These methods identify the full amino acidsequence of a protein including modifications and thus do not require protein inference from pep-tides. However, these techniques are more more challenging because compared to peptides, pro-teins are more difficult to solubilize, separate, ionize and sequence [125]. These challenges resultin lower sensitivity and throughput of the analysis, as well as higher technical requirements fromthe instruments for high resolving power at high m/z ratios [126–128]. Nonetheless, these chal-lenges are rapidly being addressed by innovative methods for protein separation (such as capillaryzone electrophoresis [129]), sensitive methods allowing the detection of individual ions [130, 131],and by community standards [132].The biases associated with either bottom-up or top-down proteomics are very different, whichpresents opportunities for combining them synergistically. The hybrid combination of mass spec-14rometry techniques has been used to characterize whole ribosome complexes by quantifying RPlevels along with proteoforms [127]. Such an approach has offered novel observations about cys-teine modification of RPS27, and ribosome assembly sites through the characterization of ribo-somes originating from human, plant, and bacterial cells [133]. Additionally, the elusive ribosomalprotein SRA was successfully quantified and found to have heterogeneous stoichiometry in
E. coli ribosomes [127].
Complementary structural biology techniques
We have focused our review on the higher throughput MS methods, but many structural biologymethods can provide more detailed information about ribosome modifications and structure albeitat lower throughput. These include chemical cross linking of proteins [134, 135], X-ray diffraction,and cryo-EM. The analysis of both protein sequence in tandem with spatial information allows notonly to co-validate results ribosome remodeling observations but also to start revealing its func-tional significance. Such a combined approach has been used to study other complexes, such asthe human nuclear pore complex scaffold [136], bacteriophage portal complexes[137], 26S pro-teasome complex[138] and COPII vesicles that transport proteins to the Golgi apparatus from theendoplasmic reticulum [139]. In fact, such combination of structural biology and mass spectrom-etry has been used to study a particular subcomplex involved in ribosome biogenesis [140, 141].The synergy between these methods can help identify specialized ribosomal structures and confir-mations regulating protein synthesis.X-ray diffraction requires the crystallization of the sample of interest prior to analysis, whilecryo-EM eliminates this need. The process of crystallization can demand much optimization,which can involve the purification of many ribosomes. The obviation of crystallization allowsfor the study of more kinds of ribosomes without the limitation of crystallizing them. The flash-freezing of samples allows the observation of molecules in a ”near-native” state; since moleculesare not constrained to a crystal, they have more degrees of freedom, thus allowing for more con-formational states to be studied. However, cryo-EM studies can require anywhere from hundredsto thousands single particle images, which involves long hours of image acquisition at the mi-croscope. Furthermore, many of these images are discarded due to the phenomenon known asbeam-induced motion, which produces blurred images [142]. Despite limitations, both techniquetypes continue to reveal the ribosome protein structure. Recently, X-ray diffraction was used todiscover the role of potassium ions within the ribosome on the stabilization of the protein complex15n both the initiation and elongation states of translation [143]. Cyro-EM has been nearing thesensitivity and resolution of X-ray diffraction, as seen through the report of the entire bacterialribosome resolved at two angstroms [144].The synergy of mass spectrometry and structural biology techniques can empower novel ob-servations of the ribosome complexes. These method types are different in terms of sample prepa-ration and measurement acquisition, yet are complementary in the type of information that is pro-vided. Combining protein composition information with conformational changes can help revealdifferent roles of ribosomes within protein synthesis, and ultimately gene expression. Such knowl-edge can help decipher the degree of ribosome remodeling during normal development and phys-iology (e.g., immunoribosomes) and during diseases (e.g., oncoribosomes). The combination ofmass spectrometry with structural biology has been used to elucidate several facets of the ribo-some complex subunits, including ribosome assembly in bacteria[140, 145, 146] and the effect ofdimerization on ribosomes when nutrients are scarce [147]. In addition, ribosome remodeling hasbeen found to occur in the bacterial ribosome through the use of MS and X-ray crystallography[148].This paper emphasized the need to use multiple complementary approaches for quantifyingribosome remodeling and briefly reviewed sources of systematic biases as well as approaches formitigating their influence. These approaches may afford reliable quantification of ribosome remod-eling, which is a starting point for investigating its functional roles in regulating mRNA translation.Identifying ribosomes associated with specific conditions – such as disease states, developmentalstages, or metabolic conditions – can start to reveal different populations of ribosomes. Theseobservations will serve as a starting basis for characterizing new principles in the regulation ofRNA translation that may reshape our understanding of one of the most fundamental biologicalprocesses. In the long term, this new understanding can enable the design of therapies that specif-ically target translation for disease and regenerative treatments.
Acknowledgments:
This work was funded by a New Innovator Award from the NIGMS from theNational Institutes of Health to N.S. under Award Number DP2GM123497.
References
1. Wool I, Stirewalt WS, Kurihara K, Low RB, Bailey P, et al. (1968) Mode of action ofinsulin in the regulation of protein biosynthesis in muscle. Recent progress in hormoneresearch 24: 139–213. 16. Levere RD, Granick S (1965) Control of hemoglobin synthesis in the cultured chick blas-toderm by delta-aminolevulinic acid synthetase: increase in the rate of hemoglobin forma-tion with delta-aminolevulinic acid. Proceedings of the National Academy of Sciences ofthe United States of America 54: 134.3. Braude PR (1979) Control of protein synthesis during blastocyst formation in the mouse.Developmental biology 68: 440–452.4. Rosenthal ET, Hunt T, Ruderman JV (1980) Selective translation of mrna controls thepattern of protein synthesis during early development of the surf clam, spisula solidissima.Cell 20: 487–494.5. Markussen FH, Michon AM, Breitwieser W, Ephrussi A (1995) Translational control ofoskar generates short OSK, the isoform that induces pole plasma assembly. Development121: 3723–3732.6. Martin KC, Ephrussi A (2009) mRNA localization: gene expression in the spatial dimen-sion. Cell 136: 719–730.7. Castello A, Fischer B, Eichelbaum K, Horos R, Beckmann BM, et al. (2012) Insights intoRNA biology from an atlas of mammalian mRNA–binding proteins. Cell 149: 1393–1406.8. Bartel DP (2009) Micrornas: target recognition and regulatory functions. cell 136: 215–233.9. Mauro VP, Edelman GM (2002) The ribosome filter hypothesis. Proceedings of the Na-tional Academy of Sciences 99: 12031–12036.10. Komili S, Farny NG, Roth FP, Silver PA (2007) Functional specificity among ribosomalproteins regulates gene expression. Cell 131: 557–571.11. Gilbert WV (2011) Functional specialization of ribosomes? Trends in biochemical sci-ences 36: 127–132.12. Slavov N (2015) From differential transcription of ribosomal proteins to differential struc-ture of ribosomes doi:10.7287/peerj.preprints.1504v1.13. Preiss T (2015) All ribosomes are created equal. really? Trends in biochemical sciences41: 10.1016/j.tibs.2015.11.009.14. Emmott E, Jovanovic M, Slavov N (2018) Ribosome stoichiometry: from form to function.Trends in biochemical sciences 44: 95–109. doi:10.1016/j.tibs.2018.10.009.15. Lodish HF (1974) Model for the regulation of mrna translation applied to haemoglobinsynthesis. Nature 251: 385–388.16. Mills EW, Green R (2017) Ribosomopathies: Theres strength in numbers. Science 358.177. Naora H, Naora H (1999) Involvement of ribosomal proteins in regulating cell growth andapoptosis: translational modulation or recruitment for extraribosomal activity? Immunol-ogy and cell biology 77: 197–205.18. Slavov N, Semrau S, Airoldi E, Budnik B, van Oudenaarden A (2015) Differentialstoichiometry among core ribosomal proteins. Cell Reports 13: 865 – 873. doi:10.1016/j.celrep.2015.09.056.19. Ferretti MB, Ghalei H, Ward EA, Potts EL, Karbstein K (2017) Rps26 directs mRNA-specific translation by recognition of kozak sequence elements. Nat Struct Mol Biol 24:700–707.20. Samir P, Browne CM, Rahul, Sun M, Shen B, et al. (2018) Identification of changingribosome protein compositions using mass spectrometry.21. Ghulam MM, Catala M, Abou Elela S (2020) Differential expression of duplicated ribo-somal protein genes modifies ribosome composition in response to stress. Nucleic AcidsResearch 48: 1954–1968.22. Kawasaki H, Shigenaga A, Uda M, Baba T, Ogawa H, et al. (2011) Nitration of tryptophanin ribosomal proteins is a novel post-translational modification of differentiated and na¨ıvepc12 cells. Nitric Oxide 25: 176–182.23. Akanuma G, Nanamiya H, Natori Y, Yano K, Suzuki S, et al. (2012) Inactivation of ribo-somal protein genes in bacillus subtilis reveals importance of each ribosomal protein forcell proliferation and cell differentiation. Journal of bacteriology 194: 6282–6291.24. Khajuria RK, Munschauer M, Ulirsch JC, Fiorini C, Ludwig LS, et al. (2018) Ribosomelevels selectively regulate translation and lineage commitment in human hematopoiesis.Cell 173: 90–103.25. Shi Z, Fujii K, Kovary KM, Genuth NR, R¨ost HL, et al. (2017) Heterogeneous ribosomespreferentially translate distinct subpools of mRNAs genome-wide. Mol Cell 67: 71–83.e7.26. Cagnetta R, Frese CK, Shigeoka T, Krijgsveld J, Holt CE (2018) Rapid Cue-Specific re-modeling of the nascent axonal proteome. Neuron 99: 29–46.e4.27. Shigeoka T, Koppers M, Wong HHW, Lin JQ, Cagnetta R, et al. (2019) On-Site ribo-some remodeling by locally synthesized ribosomal proteins in axons. Cell Rep 29: 3605–3619.e10.28. Steffen KK, MacKay VL, Kerr EO, Tsuchiya M, Hu D, et al. (2008) Yeast life span exten-sion by depletion of 60S ribosomal subunits is mediated by gcn4. Cell 133: 292–302.29. Lawrence MS, Stojanov P, Mermel CH, Robinson JT, Garraway LA, et al. (2014) Discov-ery and saturation analysis of cancer genes across 21 tumour types. Nature 505: 495–501.180. Xue S, Barna M (2012) Specialized ribosomes: a new frontier in gene regulation andorganismal biology. Nature Reviews Molecular Cell Biology 13: 355–369.31. Gerbasi VR, Weaver CM, Hill S, Friedman DB, Link AJ (2004) Yeast Asc1p and mam-malian RACK1 are functionally orthologous core 40S ribosomal proteins that repress geneexpression. Molecular and cellular biology 24: 8276–8287.32. Thompson MK, Rojas-Duran MF, Gangaramani P, Gilbert WV (2016) The ribosomal pro-tein asc1/rack1 is required for efficient translation of short mrnas. Elife 5: e11154.33. Johnson AG, Lapointe CP, Wang J, Corsepius NC, Choi J, et al. (2019) Rack1 on and offthe ribosome. RNA 25: 881–895.34. Draptchinskaia N, Gustavsson P, Andersson B, Pettersson M, Willig TN, et al. (1999) Thegene encoding ribosomal protein s19 is mutated in diamond-blackfan anaemia. Naturegenetics 21: 169–175.35. Liu JM, Ellis SR (2006) Ribosomes and marrow failure: coincidental association or molec-ular paradigm? Blood 107: 4583–4588.36. De Keersmaecker K, Sulima SO, Dinman JD (2015) Ribosomopathies and the paradox ofcellular hypo-to hyperproliferation. Blood 125: 1377–1382.37. Sulima SO, Hofman IJ, De Keersmaecker K, Dinman JD (2017) How ribosomes translatecancer. Cancer discovery 7: 1069–1087.38. Kampen KR, Sulima SO, De Keersmaecker K (2018) Rise of the specialized onco-ribosomes. Oncotarget 9: 35205.39. Yewdell JW, Ant ´on LC, Bennink JR (1996) Defective ribosomal products (drips): a majorsource of antigenic peptides for mhc class i molecules? The Journal of Immunology 157:1823–1826.40. Yewdell JW, Nicchitta CV (2006) The drip hypothesis decennial: support, controversy,refinement and extension. Trends in immunology 27: 368–373.41. Wei J, Yewdell JW (2019) Immunoribosomes: Wheres theres fire, theres fire. Molecularimmunology 113: 38–42.42. Warner JR (1999) The economics of ribosome biosynthesis in yeast. Trends in biochemicalsciences 24: 437–440.43. Sloan KE, Warda AS, Sharma S, Entian KD, Lafontaine DL, et al. (2017) Tuning the ribo-some: The influence of rrna modification on eukaryotic ribosome biogenesis and function.RNA biology 14: 1138–1152.44. Penzo M, Galbiati A, Trer´e D, Montanaro L (2016) The importance of being (slightly)modified: The role of rrna editing on gene expression control and its connections withcancer. Biochimica et Biophysica Acta (BBA)-Reviews on Cancer 1866: 330–338.195. Parks MM, Kurylo CM, Dass RA, Bojmar L, Lyden D, et al. (2018) Variant ribosomal rnaalleles are conserved and exhibit tissue-specific expression. Science advances 4: eaao0665.46. Wong SY, Javid B, Addepalli B, Piszczek G, Strader MB, et al. (2013) Functional roleof methylation of g518 of the 16s rrna 530 loop by gidb in mycobacterium tuberculosis.Antimicrobial agents and chemotherapy 57: 6311–6318.47. Mauro VP, Matsuda D (2016) Translation regulation by ribosomes: Increased complexityand expanded scope. RNA biology 13: 748–755.48. Battle A, Khan Z, Wang SH, Mitrano A, Ford MJ, et al. (2015) Impact of regulatoryvariation from RNA to protein. Science 347: 664–667.49. Franks A, Airoldi E, Slavov N (2017) Post-transcriptional regulation across human tissues.PLoS computational biology 13: e1005535.50. Kondrashov N, Pusic A, Stumpf CR, Shimizu K, Hsieh AC, et al. (2011) Ribosome-mediated specificity in Hox mRNA translation and vertebrate tissue patterning. Cell 145:383–397.51. Slavov N, Botstein D (2011) Coupling among growth rate response, metabolic cycle, andcell division cycle in yeast. Molecular Biology of the Cell 22: 1997 – 2009. doi:10.1091/mbc.E11-02-0132.52. Sung MK, Reitsma JM, Sweredoski MJ, Hess S, Deshaies RJ (2016) Ribosomal proteinsproduced in excess are degraded by the ubiquitin–proteasome system. Molecular biologyof the cell 27: 2642–2652.53. Eisenberg AR, Higdon A, Keskin A, Hodapp S, Jovanovic M, et al. (2018) Precise post-translational tuning occurs for most protein complex components during meiosis. Cell Rep25: 3603–3617.e2.54. Zhang B, Wang J, Wang X, Zhu J, Liu Q, et al. (2014) Proteogenomic characterization ofhuman colon and rectal cancer. Nature 513: 382–387.55. Slavov N, Budnik B, Schwab D, Airoldi E, van Oudenaarden A (2014) Constant GrowthRate Can Be Supported by Decreasing Energy Flux and Increasing Aerobic Glycolysis.Cell Reports 7: 705 – 714. doi:10.1016/j.celrep.2014.03.057.56. Janssens GE, Meinema AC, Gonzalez J, Wolters JC, Schmidt A, et al. (2015) Proteinbiogenesis machinery is a driver of replicative aging in yeast. Elife 4: e08527.57. Kelmer Sacramento E, Kirkpatrick JM, Mazzetto M, Baumgart M, Bartolome A, et al.(2020) Reduced proteasome activity in the aging brain results in ribosome stoichiometryloss and aggregation. Mol Syst Biol 16: e9596.58. Jovanovic M, Rooney MS, Mertins P, Przybylski D, Chevrier N, et al. (2015) Dynamicprofiling of the protein life cycle in response to pathogens. Science 347.209. Riba A, Di Nanni N, Mittal N, Arhn´e E, Schmidt A, et al. (2019) Protein synthesis ratesand ribosome occupancies reveal determinants of translation elongation rates. Proceedingsof the National Academy of Sciences 116: 15023–15032.60. Warner JR, McIntosh KB (2009) How common are extraribosomal functions of ribosomalproteins? Molecular cell 34: 3–11.61. Statello L, Maugeri M, Garre E, Nawaz M, Wahlgren J, et al. (2018) Identification ofrna-binding proteins in exosomes capable of interacting with different types of rna: Rbp-facilitated transport of rnas into exosomes. PloS one 13.62. Weisberg RA (2008) Transcription by moonlight: structural basis of an extraribosomalactivity of ribosomal protein s10. Molecular cell 32: 747–748.63. Bhavsar RB, Makley LN, Tsonis PA (2010) The other lives of ribosomal proteins. Humangenomics 4: 327.64. Zhou X, Liao WJ, Liao JM, Liao P, Lu H (2015) Ribosomal proteins: functions beyondthe ribosome. Journal of molecular cell biology 7: 92–104.65. Dougherty JD (2017) The expanding toolkit of translating ribosome affinity purification.Journal of Neuroscience 37: 12079–12087.66. Shigeoka T, Koppers M, Wong HHW, Lin JQ, Cagnetta R, et al. (2019) On-site ribosomeremodeling by locally synthesized ribosomal proteins in axons. Cell reports 29: 3605–3619.67. Hummel M, Cordewener JH, de Groot JC, Smeekens S, America AH, et al. (2012) Dy-namic protein composition of a rabidopsis thaliana cytosolic ribosomes in response tosucrose feeding as revealed by label free ms e proteomics. Proteomics 12: 1024–1038.68. Amirbeigiarab S, Kiani P, Sanchez AV, Krisp C, Kazantsev A, et al. (2019) Invariablestoichiometry of ribosomal proteins in mouse brain tissues with aging. Proceedings of theNational Academy of Sciences 116: 22567–22572.69. Ghaemmaghami S, Huh WK, Bower K, Howson RW, Belle A, et al. (2003) Global analysisof protein expression in yeast. Nature 425: 737–741.70. Newman JRS, Ghaemmaghami S, Ihmels J, Breslow DK, Noble M, et al. (2006) Single-cell proteomic analysis of s. cerevisiaes. cerevisiae