Advances in ribosome profiling technologies Open Access

The translation of mRNAs is a fundamental step in gene expression, which orchestrates the production of functional proteins. Translation is increasingly recognized as a dynamic layer of gene regulation that influences how cells respond to development [1,2], stress [3], and disease [4,5]. Since its development, ribosome profiling (or Ribo-seq), a technique based on deep sequencing of ribosome-protected mRNA fragments generated by RNase treatment [6], has opened new avenues for measuring translation across the transcriptome, or the ‘translatome’, in various biological contexts [7-10] (Figure 1A). This technique reveals translational efficiency, identifies new open reading frames (ORFs), and monitors ribosome traversal speed at codon resolution in a genome-wide manner (Figure 1A). Even after 15 years of research, translatomic studies have shown no signs of ending. However, the inherent technical limitations of Ribo-seq that limit its applications have become evident. This review summarizes the drawbacks and recent technological advancements designed to address these issues.

One of the clear challenges in Ribo-seq is applying it to low inputs. Conventional Ribo-seq protocols require ~10⁶ or more cells [11,12] and have difficulties utilizing a small number of cells as starting materials.

Given that step-by-step sample loss is one of the major obstacles for low inputs, a one-pot reaction is an effective option. For this purpose, ligation-free methods have been implemented [13-17]. Typically, ribosome footprints are tailed with A nucleotides by poly(A) polymerase and reverse-transcribed with oligo deoxythymidine (dT) primers. During reverse transcription (RT), another linker sequence is conjugated through a template-switching mechanism. Among these studies, ligation-free, ultralow-input, and enhanced Ribo-seq (termed Ribo-lite) is further optimized by the appropriate concentration of RNase I—a widely used RNase for Ribo-seq, the skipping of rRNA depletion to suppress sample loss, and on-gel size selection without size markers that often contaminate libraries [15] (Figure 1B). Ribo-lite was applied to low-inputs, such as 1,000 HEK293 cells and 100 mouse oocytes, and even ultralow-inputs, such as 50 HEK293 cells and a single oocyte. The combination of Ribo-lite with Smart-seq2 (a low-input RNA-seq protocol) [18] allows simultaneous analysis of the translatome and transcriptome in human oocytes and embryos [termed Ribo-RNA-lite (R2-lite)] [16].

A low-input ribosome profiling (LiRibo-seq) technique [17] also employs ligation-free one-pot library preparation but has a unique method of footprint recovery after RNase digestion. With biotin-conjugated puromycin, which is covalently linked to a nascent peptide chain through a peptidyl transferase reaction by the ribosome, footprint-ribosome-biotinylated puromycin complexes are isolated by streptavidin beads (the method termed RiboLace) [19] (Figure 1C). LiRibo-seq measures the translatome in 5,000 mouse embryonic stem cells and the maternal-to-zygotic transition in mouse embryos [17].

Similarly, serial template-switching reactions during RT allow one-pot reactions. In the ordered two-template relay (OTTR) [20,21] (Figure 1D), ribosome footprints are treated with terminal transferase to add a dideoxynucleotide at the 3′ end and used for template jumping cDNA synthesis by modified reverse transcriptase from eukaryotic retroelements and the first linker. Subsequently, second template switching is induced with the second linker. The difference in hybridizing nucleotide species in the first (3′ A/G at the footprint template with 3′ T/C for the first primer) and second (3′ C for the DNA template with 3′G for the cDNA extension) template-switching reactions avoids concatamerization [20]. This technique is also helpful in reducing the amount of material required for Ribo-seq with limited bias [21].

Another strategy to address limited materials is to amplify nucleic acids at the early step of library preparation. The T7 high-resolution original RNA (Thor) technique [22] (originally described by Lexogen) employs RNA-dependent RNA amplification by T7 RNA polymerase, using an RNA‒DNA chimera as an in vitro transcription template (Figure 1E). Owing to the linear increase in RNA at the beginning rather than before the final PCR step, Thor-Ribo-seq can minimize the risk of material loss during library preparation. Strikingly, Thor-Ribo-seq maintains high data reproducibility for a wide range of inputs from ~10⁶ to ~10³ cultured cells [22]. A successful application has been reported in dissected fly testes [23].

A potential technical limitation of ultralow-input Ribo-Seq may lie in the annotation of novel ORFs. Although Ribo-seq has been used to define ORFs de novo, this analysis generally requires high read coverage along ORFs. Given the restricted RNA molecule complexity in the low-input materials, Ribo-seq data from such samples may offer limited footprint coverage, even with high sequencing depth. Thus, some difficulty in ORF annotation from low-input data could be expected.

Since conventional Ribo-seq is prepared from the lysate of bulk cells, the profiles should present the average translatome for many cells that may intrinsically have variations in translation. Two independent Ribo-seq methods tailored for single-cell-level measurement have been developed [24,25].

The van Oudenaarden group developed single-cell ribosome sequencing (scRibo-seq) [24] (Figure 1F). In this method, individual cells are collected in a 384-well plate by a cell sorter. The single-pot experiment involves cell lysis, RNase digestion by micrococcal nuclease (MNase), two consecutive linker ligations to the 3′ end and the 5′ end of RNA fragments, RT, and PCR amplification with sample barcode addition. All the samples in the plate are pooled, gel-purified, and sequenced. Footprint differences across ORFs among cells are analyzed by well-established tools for single-cell RNA-seq, such as Seurat [26]. The advantage of MNase, which degrades both DNA and RNA, is the stringent control of activity by Ca²⁺ ions; this chelation of the ion completely quenches the activity—otherwise, linker DNAs or cDNAs would also be digested by the enzyme. The disadvantage of MNase is an A/U preference for cleavage [27], hampering the assessment of the boundary of ribosome coverage from the reads. Understanding the edge of ribosome footprints is extremely important for the estimation of the A-site position in the ribosome and thus for codon-wise evaluation of ribosome traversal. To overcome this issue, a random forest classifier is trained to correct for the sequence bias introduced by MNase digestion and is used to assign the A-site location in the reads [24].

Following the development of scRibo-seq, the Cenik group developed a microfluidic isotachophoresis (ITP)-based technique, termed Ribo-ITP [25] (Figure 1G). This system uses a high-yield microfluidic system for RNA purification and footprint enrichment; RNase-treated cell lysate is subjected to ITP together with fluorophore-conjugated marker oligonucleotides, and then, ribosome footprints of the corresponding size are selected. Starting with a sorted single cell, enriched footprints are handled with the ligation-free method [poly(A)-tailing and reverse-transcription with template switching similar to Ribo-lite]. Importantly, Ribo-ITP substantially reduces the sample processing time and the amount of sample needed. Using Ribo-ITP, the authors characterized allele-specific translation of zygotic transcripts during early mouse embryogenesis [25].

Technically, those single-cell techniques have a chance to employ techniques tailored for ultralow-inputs, such as OTTR [21] and Thor [22]. Notably, since both scRibo-seq and Ribo-ITP methods (and Ribo-lite) skip the rRNA depletion step (see below for details), restricted read depth may be one drawback.

Another challenge in Ribo-seq is the quantification of global translation changes. Owing to the nature of deep sequencing-based measurement, the evaluation is always ‘relative’; if translational changes are restricted to a subset of mRNA, the standard relative enrichment/depletion analysis is adequate. However, overall protein synthesis can often be substantially altered; for example, global translation shutoff is associated with stress [28,29] and translation inhibitor treatment [30,31], coupled with mRNA-selective effects. Generally, the accurate track of the global shift in Ribo-seq or, more broadly deep sequencing-based methods, is challenging [32-34].

A straightforward strategy to address this situation is the addition of ‘spike-in’ control. In RNA-seq, this process is relatively easy [35] since the artificial RNAs for this purpose are commercially available (such as the External RNA Controls Consortium spike-in control). However, for Ribo-seq, the preparation of such external standards is difficult.

Several approaches have been developed for spike-ins in Ribo-seq. One example is the addition of molar amount-defined, short synthetic RNA oligonucleotides to samples after RNase digestion [36-40] (Figure 2A). This oligonucleotide-based method should be performed with caution, as it assumes no variance in processes before spike-in addition or during RNase digestion and ribosome footprint recovery. For rigorous evaluations, sequence diversity or the number of RNA species should also be carefully considered.

The alternative method is based on the supplementation of lysates of orthogonal species as spike-in (e.g., yeast lysate spike-in for human Ribo-seq) [41-45] (Figure 2B). Since experiments start with the lysate mixture before RNase digestion and downstream, there is no risk for sample-to-sample variations, as with short RNA spike-ins. Moreover, the high sequence diversity in footprints from orthogonal species is advantageous for suppressing potential sequencing bias for a small subset of RNAs.

Similarly, footprints stemming from mitochondrial ribosomes, which are typically included in lysates in addition to cytosolic ribosomes, could be used for internal spike-ins [14,30,46-49]. Notably, this method is only applied when it is reasonable to assume that translation within the organelle is unaffected under the experimental conditions.

Another spike-in option harnesses purified mRNA‒ribosome complexes. This method, termed Ribo-Calibration [50], uses artificial mRNAs (such as those encoding Renilla luciferase or firefly luciferase) complexed with an explicit number of ribosomes (Figure 2C). After in vitro translation with a rabbit reticulocyte lysate system, the reaction is subjected to a sucrose density gradient to purify complexes with a defined stoichiometric ratio of ribosome-mRNA (e.g., two ribosomes/disomes or three ribosomes/trisomes). These mRNA‒ribosome complexes can be added to the lysate before RNase digestion to minimize unintended variation. With Ribo-Calibration, absolute changes in translation have been measured in various cell types or stress conditions, such as heat shock and aging [50]. Ribo-spike, which is essentially based on the same rationale [51] (Figure 2C), also allows the assessment of global translation changes during the integrated stress response.

In addition to assuming a global change in translation, Ribo-Calibration measures the mean ribosome number loaded across mRNAs; through combination of RNA-seq data from the same samples, the absolute ribosome density on mRNAs is measured, referring to the ribosome numbers on the spike-in mRNAs as 2 (disome):3 (trisome) [50]. Moreover, given the translational elongation rate measured by ribosome run-off assay-coupled Ribo-seq [52], ribosome numbers on ORFs could be converted into translation initiation rates [50]. The stoichiometry and kinetics in HEK293 cells fall into~5 ribosomes every ~270 nt; ~22 sec/event translation initiation; and ~4.1 codon/sec translation elongation as a global average, which are consistent with the values obtained by in-cell nascent chain tracking in real time [53]. Furthermore, given the mRNA half-lives that can be measured by other sequencing-based methods, such as 5′-bromo-uridine immunoprecipitation chase–deep sequencing (BRIC-seq) [54,55], the initiation rates lead to the calculation of lifetime translation rounds of mRNAs (~1,800 times translation before mRNA decay on average).

Technically, one disadvantage of Ribo-seq is the limited sequencing space due to the extensive contamination of rRNAs. For the reduction in rRNA in libraries, the conventional approach is physical subtraction with rRNA-hybridizing oligonucleotides and subsequent immobilization on streptavidin-conjugated beads [11,12,56-59] (Figure 3A) (typically 15–20% usable reads in mammalian tissue culture). Alternatively, RNase-based methods for digesting selected rRNA fragments can be used [60-62] (10–30% usable reads in mammalian tissue culture, depending on the commercial kits or methods) (Figure 3B), although the rRNA pulldown-and-subtraction method is generally suitable for Ribo-seq [58].

While these strategies focus on purified RNA fragments, a recently reported method allows rRNA reduction at an earlier step with an orthogonal rationale. After ribosome‒footprint complex isolation by ultracentrifugation, Ribo-FilterOut [50] splits it into subunits and the footprint by EDTA and then separates the footprint from the subunits through subsequent ultrafiltration (Figure 3C). This process effectively removes fragmented rRNA maintained within subunits. Importantly, the total time required to complete these steps is typically less than 30 minutes, and no expensive kits are needed. The efficacy of Ribo-FilterOut has been demonstrated by a ~16-fold increase in usable reads in combination with commercially available rRNA pulldown-and-subtraction kits (e.g., Illumina's Ribo-Zero and siTOOLs Biotech’s riboPOOL) [50] (80% usable reads in human tissue culture).

Another method focuses on the double-stranded DNA (dsDNA) library after PCR amplification instead of RNAs. Through the use of recombinant Cas9 protein and specifically designed guide RNA to target contaminated rRNA fragments, the undesired portion of dsDNA derived from rRNA is selectively cleaved from the library [12,63-65] (30–35% usable reads in human tissue culture and mouse brain) (Figure 3D).

RNase selection is an important factor in defining the quality of the data. Escherichia coli RNase I is the most common RNase for Ribo-seq due to low nucleotide bias. However, this enzyme nicks rRNAs within ribosomes and affects the integrity of the complex [66]. Ingolia’s group reported that RNase P1, which generally lacks nucleotide specificity for cleavage [67], preserves the integrity of ribosomes better than does RNase I [21] (Figure 1D); this enzyme minimizes the cleavage of rRNAs and strongly reduces rRNA contamination in the final libraries. Thus, RNase P1–Ribo-seq monitors the subpopulation of the ribosomal complex that conventional RNase I–Ribo-seq misses—a class of footprints longer than monosomes but shorter than colliding 80S–80S complexes or disomes [21]. The ‘subdisome’ footprint may represent 80S‒40S complexes or 80S complexes with accessory proteins.

As summarized in this review, recent technical advances are overcoming the limitations of Ribo-seq and expanding its applications to a wide variety of samples with low inputs, single cells, and valuable materials. Measurements of global translation changes, kinetic assessments, and probing of ribosome configurations with increased sequencing space have improved our understanding of protein synthesis. These techniques could be combined with Ribo-seq derivatives for ribosome subsets associated with specific factors [68], scanning preinitiation complexes [69], ribosome collision [12], organelle translation in mitochondria [70] and chloroplasts [71], nascent mRNAs [72], and microbiomes [73].

S.I. is a member of the Scientific Reports editorial board. The remaining authors declare that they have no competing interest.

S.I. was supported by the Ministry of Education, Culture, Sports, Science and Technology (MEXT) (JP24H02307), and RIKEN (Pioneering Projects). K.T. was supported by the Japan Science and Technology Agency (JST) (JPMJBS2418).

K.T.: Writing—Original draft, Visualization, Funding acquisition, Writing—review & editing; S.I.: Writing— Original draft, Supervision, Project administration, Funding acquisition, Writing—review & editing.

We would like to thank all the members of the Iwasaki laboratory for their helpful comments on this manuscript. K.T. was a recipient of a Junior Research Associate Program (JRA) from RIKEN, a World-leading Innovative Graduate Study Program in Proactive Environmental Studies (WINGS-PES) from The University of Tokyo, and Broadening Opportunities for Outstanding young researchers and doctoral students in STrategic areas (BOOST) from JST.

ITP

isotachophoresis

MNase

micrococcal nuclease

OTTR

ordered two-template relay

RT

reverse transcription

dsDNA

double-stranded DNA

scRibo-seq

single-cell ribosome sequencing