1. Cell Biology

One-shot analysis of translated mammalian lncRNAs with AHARIBO

  1. Luca Minati
  2. Claudia Firrito
  3. Alessia Del Piano
  4. Alberto Peretti
  5. Simone Sidoli
  6. Daniele Peroni
  7. Romina Belli
  8. Francesco Gandolfi
  9. Alessandro Romanel
  10. Paola Bernabo
  11. Jacopo Zasso
  12. Alessandro Quattrone
  13. Graziano Guella
  14. Fabio Lauria
  15. Gabriella Viero
  16. Massimiliano Clamer  Is a corresponding author
  1. IMMAGINA BioTechnology, Italy
  2. Department of Biochemistry, Albert Einstein College of Medicine, United States
  3. Mass Spectrometry Facility, Computational and Integrative Biology (CIBIO), University of Trento, Italy
  4. Laboratory of Bioinformatics and Computational Genomics, Department of Cellular, Computational and Integrative Biology (CIBIO), University of Trento, Italy
  5. Laboratory of Translational Genomics, Department of Cellular, Computational and Integrative Biology (CIBIO), University of Trento, Italy
  6. Department of Physics, University of Trento, Italy
  7. Institute of Biophysics, CNR Unit at Trento, Italy
https://doi.org/10.7554/eLife.59303
Share this article
Cite this article
  1. Luca Minati
  2. Claudia Firrito
  3. Alessia Del Piano
  4. Alberto Peretti
  5. Simone Sidoli
  6. Daniele Peroni
  7. Romina Belli
  8. Francesco Gandolfi
  9. Alessandro Romanel
  10. Paola Bernabo
  11. Jacopo Zasso
  12. Alessandro Quattrone
  13. Graziano Guella
  14. Fabio Lauria
  15. Gabriella Viero
  16. Massimiliano Clamer
(2021)
One-shot analysis of translated mammalian lncRNAs with AHARIBO
eLife 10:e59303.
https://doi.org/10.7554/eLife.59303
  2. Download BibTeX
  3. Download .RIS

Abstract

A vast portion of the mammalian genome is transcribed as long non-coding RNAs (lncRNAs) acting in the cytoplasm with largely unknown functions. Surprisingly, lncRNAs have been shown to interact with ribosomes, encode peptides, or act as ribosome sponges. These functions still remain mostly undetected and understudied owing to the lack of efficient tools for genome-wide simultaneous identification of ribosome-associated and peptide-producing lncRNAs. Here, we present AHA-mediated RIBOsome isolation (AHARIBO), a method for the detection of lncRNAs either untranslated, but associated with ribosomes, or encoding small peptides. Using AHARIBO in mouse embryonic stem cells during neuronal differentiation, we isolated ribosome-protected RNA fragments, translated RNAs, and corresponding de novo synthesized peptides. Besides identifying mRNAs under active translation and associated ribosomes, we found and distinguished lncRNAs acting as ribosome sponges or encoding micropeptides, laying the ground for a better functional understanding of hundreds of lncRNAs.

Introduction

An incredibly small fraction of the mammalian genome is protein-coding (<3%), while the number of potentially functional non-coding genes remains unclear (Djebali et al., 2012). Long non-coding RNAs (lncRNAs) are defined as non-coding RNA exceeding 200 nt. They have gained much attention because of their role in a variety of cellular processes, from chromatin architecture (Minajigi et al., 2015) to mRNA turnover (Kleaveland et al., 2018) and translation (Ingolia et al., 2011). Typically, lncRNAs are abundant transcripts (Iyer et al., 2015) that display short and not evolutionarily conserved Open Reading Frames (ORFs with minimal homology to known protein domains (Guttman and Rinn, 2012). The majority of lncRNAs are localized in the cytoplasm (Carlevaro-Fita et al., 2016), where they are supposed to remain untranslated. Ribosome profiling (RIBO-seq), which provides positional information of ribosomes along transcripts (Clamer et al., 2018; Ingolia et al., 2012), identified several ribosome-associated lncRNAs (Bazzini et al., 2014; Ingolia et al., 2011; Lee et al., 2012; Zeng et al., 2018). A handful of lncRNAs have been shown to be involved in translation regulation (Carrieri et al., 2012; Yoon et al., 2012), while others are themselves potentially or partially translated (Anderson et al., 2015; Aspden et al., 2014; Bazin et al., 2017; Ingolia et al., 2011; Nelson et al., 2016; Ruiz-Orera et al., 2014; van Heesch et al., 2019). As coding RNAs, lncRNAs can be associated with actively translating or translationally silent ribosomes (Chandrasekaran et al., 2019; Chen et al., 2020; Jiao and Meyerowitz, 2010; Kapur et al., 2017). Hence, the potential involvement of lncRNAs in translation increases the complexity of the mammalian control of gene expression at the translatome and proteome level. Unfortunately, classical RIBO-seq approaches barely distinguish between lncRNAs producing peptides from those that sequester ribosomes (lncRNA bound to ribosomes without translation) and act as ribosome sponges. Proteomics approaches, such as mass spectrometry, can help to define and quantitatively monitor the production of peptides, but are less sensitive techniques than RNA sequencing (Slavoff et al., 2013; van Heesch et al., 2019). Therefore, proteomics and RIBO-seq alone cannot unravel the wide functional range of cytoplasmic lncRNAs associated with the translation machinery.

To fill this gap, we developed AHA-mediated RIBOsome isolation (AHARIBO), a combination of protocols that simultaneously isolate RNAs and nascent proteins associated with translationally active ribosomes. AHARIBO is based on the isolation of ribosomes trapped with their nascent peptides by incorporating the non-canonical amino acid L-azidohomoalanine (AHA), followed by parallel RNA-seq, ribosome profiling, and proteomics.

We applied AHARIBO to human and mouse cells and showed that it enables to (1) purify translating ribosomes via nascent peptide chains, (2) co-purify RNAs and proteins for transcriptome/de novo proteome-associated studies, and (3) detect the regulatory network of lncRNAs translated or associated with ribosomes.

Results

Nascent peptide labeling and separation of the ribosome complex with AHARIBO-rC

To simultaneously purify ribosomes under active translation, associated RNAs, and corresponding growing peptide chains, we optimized a protocol in HeLa cells (Figure 1A). Briefly, the protocol consists of the following phases: (1) incubation with a methionine-depleted medium, (2) addition of the methionine analog AHA, (3) on-ribosome anchorage of nascent peptide chains with a small molecule, (4) cell lysis and AHA ‘copper-free click reaction’ (Jewett and Bertozzi, 2010) for (5) ribosome capture with magnetic beads. We reasoned that the protocol for isolating ribosomes through AHA can be used to obtain information about nascent peptides, constitutive components of ribosomes, mRNAs, and lncRNAs associated with them. For this reason, we optimized several parameters from washing steps to nuclease treatments (Figure 1A) to isolate (1) the full translational complex (AHARIBO-rC, ribosomal complexes: ribosomes, ribosome-associated proteins, nascent peptides, and RNAs), (2) the de novo synthesized proteome (AHARIBO-nP, nascent proteome), and (3) ribosome-protected fragments (RPFs) (AHARIBO RIBO-seq: RIBOsome profiling by sequencing).

Figure 1 with 3 supplements see all
Download asset Open asset
L-Azidohomoalanine (AHA) labeling of nascent peptide chains and ribosome separation.

(A) Schematic representation of AHA-mediated RIBOsome isolation (AHARIBO) workflow. After methionine depletion, AHA incubation, and sBlock treatment, cell lysates can be processed for (1) AHARIBO-rC: isolation of translational complexes (ribosomes, ribosome-associated proteins, nascent peptides, and RNAs); (2) AHARIBO-nP: isolation of de novo synthesized proteome; and (3) AHARIBO RIBO-seq: for ribosome profiling. (B) Polysomal profiles in HeLa cells. On the right of each profile, example of SDS-PAGE of protein extracts from each fraction of the profile. Staining of the membrane was performed by biotin cycloaddition followed by streptavidin-Horseradish peroxidase (HRP). RPL26 protein was used as a marker of the large ribosome subunit. (C) Box plot showing the AHA signal enrichment in the polysomal fractions of the profiles in cells untreated (NT) and treated with either cycloheximide (CHX) or sBlock. Results are shown as the median (±SE) of three independent experiments. NS: not significant. *p-value=0.05 was obtained through an unpaired t-test. (D) Volcano plots of AHARIBO-rC-isolated proteins. Data are compared with input (AHA-containing lysate, left) or with streptavidin-coated beads without biotin-DBCO (right). DBCO: dibenzocyclooctyne. Red line: t-test p-value<0.05.

Figure 1—source data 1

A table with the relative abundance of AHARIBO-rC-isolated proteins.

Relative abundance of AHARIBO-rC-isolated proteins. AHARIBO: AHA-mediated RIBOsome isolation.

https://cdn.elifesciences.org/articles/59303/elife-59303-fig1-data1-v2.xlsx
Figure 1—source data 2

Gene Ontology analysis data.

https://cdn.elifesciences.org/articles/59303/elife-59303-fig1-data2-v2.xlsx

To minimize the amount of AHA-tagged and fully synthesized proteins released from ribosomes and achieve optimal on-ribosome polypeptide stabilization, we tested multiple incubation times of AHA exposure and compared the effect of two small molecules (namely cycloheximide [CHX] and sBlock, an anisomycin-based reagent). Anisomycin is known to inhibit the activity of eukaryotic ribosomes, while keeping polypeptides bound to translating ribosomes (Garreau de Loubresse et al., 2014; Grollman, 1967; Seedhom et al., 2016).

We observed that 30 min is the optimal incubation time for sufficient AHA incorporation and maximum RNA recovery (Figure 1—figure supplement 1A–C). Next, we compared the efficiency of CHX and sBlock in stabilizing the nascent peptide by co-sedimentation analysis of AHA-tagged polypeptides with ribosomes along the sucrose gradient (Figure 1B). As a control, cells were treated in parallel with puromycin to cause ribosome disassembly and release of the growing peptide chains (Figure 1B; Blobel and Sabatini, 1971; Enam et al., 2020). In agreement with literature, we found that both CHX and sBlock are able to stabilize AHA-peptides on ribosomes and polysomes (Biever et al., 2020; Mathias et al., 1964). The efficiency of anchoring polypeptides on ribosomes in CHX- and sBlock-treated cells was about 50% higher compared to untreated cells, confirming that the treatment effectively stabilizes nascent polypeptides (Figure 1C). The high signal observed in lighter fractions is likely caused by AHA-labeled proteins released from ribosomes. To overcome this problem, it is possible to perform a pre-cleaning of the cell lysate by sucrose cushioning. This step can increase the efficiency of total RNA isolation with AHARIBO compared with the control (no AHA) (Figure 1—figure supplement 1D). As expected, in puromycin-treated samples, the AHA signal was mainly detected in the first two fractions of the gradient, proving that the signal observed in the heavier fractions of CHX- and sBlock-treated cells was not caused by diffusion of AHA-labeled peptides from lighter to heavier fractions. Since sBlock outperformed CHX in anchoring efficiency (Figure 1C), we used this compound in all further experiments.

Prompted by the evidence that nascent peptides can be stably anchored on ribosomes by a small molecule, we isolated RNAs and proteins associated with the translation complex. To this aim, we performed a label-free liquid chromatography-mass spectrometry (LC-MS) analysis of AHARIBO-captured proteins relative to the input, to the background biotin-DBCO- (Figure 1D) or AHA- (Figure 1—figure supplement 2A; Figure 1—source data 1) and to a sample treated with puromycin (AHA+ puromycin) (Figure 1—figure supplement 2B), which causes the release of nascent chains. We observed that ribosomal proteins belonging to both the large and small ribosome subunits are indeed more abundant in AHARIBO-rC samples than in controls. LC-MS results were confirmed by western blot analysis of proteins that are component of the large and small ribosomal subunits (RPS6, RPL26) (Figure 1—figure supplement 2B). Gene ontology (GO) analysis revealed that terms related to translation (biological process), nucleic acid binding (cellular function), and ribonucleoprotein complex (cellular component) are enriched in AHARIBO-rC compared to the control (no AHA), confirming efficient pulldown of translation-related proteins (Figure 1—source data 2).

Then, we used AHARIBO-rC to determine the translational status of cultured cells. To this aim, we downregulated protein synthesis by treating HeLa cells with puromycin, heat shock (HS) (10 min at 42°C, during AHA incubation), or arsenite (Ar) treatment, which induces translational inhibition and stress granules formation (Wang et al., 2016). We observed a reduction of RNA captured in puromycin-, HS-, and Ar-treated cells relative to the control (Figure 1—figure supplement 3A–C). In line with this finding, qRT-PCR analysis showed about 50% reduction in 18S rRNA levels when translation was inhibited (Figure 1—figure supplement 3D).

To further validate AHARIBO-rC, we took advantage of a micropeptide (176 aa) originating from an open reading frame of the TUG1 lncRNA, called TUG1-BOAT (Lewandowski et al., 2020). The wild-type (WT) ORF has a non-canonical start codon and a methionine 75 nt upstream of the stop codon. We ectopically expressed the WT TUG1-BOAT transcript and two mutant constructs (Figure 1—figure supplement 3E): (1) the ΔTUG1-BOAT, without the methionine 75 nt upstream of the stop codon and (2) the +1Met TUG1-BOAT with an ATG (methionine) as start codon. The +1Met TUG1-BOAT has two methionines, one at the N terminal and the other at 25 aa (75 nt) before the C-terminal. Our RT-qPCR analysis performed 24 hr or 48 hr after transfection showed a good efficiency of AHARIBO in capturing the TUG1-BOAT RNA when methionines are present (about 50 times more in +1Met TUG1-BOAT than in ΔMet TUG1-BOAT after 24 hr) (Figure 1—figure supplement 3E), confirming the efficiency of AHARIBO-rC in capturing translated RNA.

AHARIBO-nP: genome-wide portray of the de novo synthesized proteome

Motivated by the evidence that AHARIBO-rC can be used to isolate bona fide active ribosomes, we further tested our method genome-wide in mouse embryonic stem cells (mESCs) under basal condition and after differentiation into early neurons (ENs) (Tebaldi et al., 2018; Figure 2—figure supplement 1A). We analyzed both AHARIBO-rC-isolated RNA and newly synthesized polypeptides associated with actively translating ribosomes by RNA-seq and LC-MS, respectively. The protocol for the isolation of the de novo synthesized polypeptides (named AHARIBO-nP) is based on urea washing to remove all proteins that are not nascent peptides (Figure 2—figure supplement 1B). In parallel, we isolated and analyzed the global translatome by extracting the RNA after 30% sucrose cushioning of cytoplasmatic lysates (Wang et al., 2013), and then analyzed the global proteome by pulsed SILAC (pSILAC) (Schwanhäusser et al., 2009; Figure 2A).

Figure 2 with 1 supplement see all
Download asset Open asset
AHARIBO-nP and pSILAC.

(A) Workflow for parallel AHARIBo-nP and pSILAC. mESCs: mouse embryonic stem cells; EN: mouse embryonic stem cells differentiated in early neurons. (B) Venn diagram representing the number of differentially expressed proteins (EN/mESCs) identified by AHARIBO-nP and pSILAC (p-value<0.05). (C) Volcano plot for each differentially expressed protein (EN/mESC) of AHARIBO-nP proteome versus -log2(p-value). Red broken line indicates p-value<0.05. Orange and purple dots represent upregulated proteins involved in cytoskeleton organization (GO:0007010) and neurogenesis (GO:0022008), respectively. Blue, green, and magenta dots represent downregulated proteins related to RNA processing (GO:0006396), protein synthesis (GO:0006412), and mouse pluripotency (WP1763). Gray dots represent all other proteins. (D) Schematic representation of combined cell treatments for pSILAC and AHARIBO-nP. (E) Volcano plots displaying for each protein the -log2 t-test p-value against the fold changes of protein turnover (heavy/light) in pSILAC proteome (left) and AHARIBO-nP (right) for double-treated mESCs. GO: gene ontology; AHARIBO: AHA-mediated RIBOsome isolation; pSILAC: pulsed SILAC. 

Figure 2—source data 1

A table with the pulsed SILAC (pSILAC) proteomic data.

https://cdn.elifesciences.org/articles/59303/elife-59303-fig2-data1-v2.xlsx
Figure 2—source data 2

A table with AHA-mediated RIBOsome isolation (AHARIBO) differentially expressed proteins.

Proteins are considered differentially expressed when adjusted p-values are smaller than 0.05 AHARIBO-nP differentially expressed proteins.

https://cdn.elifesciences.org/articles/59303/elife-59303-fig2-data2-v2.xlsx

Quantitative proteomic analysis of ENs versus mESCs (EN/mESC) led to the identification of 2654 differentially expressed proteins (Figure 2B, Figure 2—source data 1). As expected, differentiated cells (EN) showed a reduced turnover compared to mESCs (Figure 2—figure supplement 1C). In parallel, EN and mESC cells were analyzed by AHARIBO-nP, which captured 1365 and 2215 proteins, respectively. Of note, 74% of proteins identified through AHARIBO-nP is in common with the pSILAC dataset. The smaller number of proteins identified with AHARIBO-nP compared to pSILAC is most probably related to the shorter time of incubation with AHA (30 min) compared to pSILAC (24 hr) and is consistent with previous observations from similar pulldown enrichment strategies (Bagert et al., 2014; Rothenberg et al., 2018). Differential expression analysis (EN/mESC) identified 573 proteins (p-value<0.05) in AHARIBO-nP (Figure 2B; Figure 2—source data 2). The GO analysis of differentially expressed proteins showed that proteins involved in cytoskeleton organization and neurogenesis were upregulated (Figure 2C), further confirming the reliability of AHARIBO-nP in monitoring de novo protein expression. We focused on proteins captured by AHARIBO-nP during differentiation (Figure 2C, Figure 2—source data 2) and found that several are known to be expressed during early stages of development of the nervous system (e.g., Map1b, Tubb3, and Dync1h1) (Fiorillo et al., 2014; Gonzalez-Billault et al., 2002; Latremoliere et al., 2018). In addition, we performed AHARIBO-nP pulldown in mESCs double-labeled for pSILAC (24 hr) and AHA (30 min) (Figure 2D). Interestingly, we observed high fold changes of heavy amino acids in AHARIBO-nP (Figure 2E) and a significantly higher protein turnover in the AHARIBO-nP compared to the pSILAC proteins (Figure 2—figure supplement 1D), suggesting that AHARIBO-nP is indeed able to capture the de novo synthesized polypeptides.

Collectively, these results show that AHARIBO-nP captures de novo synthesized proteins and produces meaningful descriptions of phenotypic changes occurring upon cell differentiation. Moreover, these results demonstrate that our AHARIBO-nP protocol is suitable to monitor dynamic changes in protein expression by LC-MS analysis.

Combination of AHARIBO-rC and AHARIBO-nP: parallel genome-wide analysis of translated RNAs and de novo synthesized proteome

Prompted by previous results, we asked if mRNAs purified using AHARIBO-rC are a good proxy of protein levels. To this aim, we compared AHARIBO-rC RNA and the global translatome with AHARIBO-nP in mESCs during differentiation.

To exclude any bias related to protein length, we checked whether AHARIBO-nP preferentially captures long or short proteins. We plotted the peptide size against the enrichment resulting from AHARIBO-rC compared with the global transcriptome (Figure 3A). This value represents the extent to which AHARIBO-rC RNA differs from the standard method. Our results confirm that AHARIBO captures transcripts encoding for polypeptides in a wide range of length (Figure 3A). Since in all eukaryotes proteins are initiated with a methionine residue and the average protein size in eukaryotes is about 300 aa (Frith et al., 2006), virtually any protein can be captured as soon as the nascent peptide exits the ribosome (i.e., when it reaches a length of about 35–40 aa). In about 70% of the proteome, the N-terminal methionine is co-translationally cleaved when the peptide is at least 50 aa long by the enzyme methionine aminopeptidase (Wild et al., 2020), while the remaining 30% retains the methionine (Martinez et al., 2008). Therefore, there is a reasonable probability for at least one AHA residue to be available for each peptide when the inhibitor of translation (sBlock) is added to the cell medium, enabling the capture of the polypeptide outside the ribosome exit tunnel.

Figure 3 with 1 supplement see all
Download asset Open asset
AHARIBO-rC RNA versus de novo proteome analysis.

(A) Enrichment of a given transcript obtained with AHA-mediated RIBOsome isolation (AHARIBO) versus global translatome (x-axis) as a function of the theoretical protein length (y-axis) for mouse embryonic stem cells (mESCs) (left) and early neurons (ENs) (right). Each bar represents the number of enriched transcripts with the defined theoretical protein length. (B) Fraction of coding genes expressed above a minimum threshold in EN. The AHARIBO-rC and global translatome group are represented in yellow and cyan, respectively. For each group, the mean (solid line) and SD (shades) of the fractions for a given count per million (CPM) threshold are calculated over all samples (n = 6) in that group. (C) Scatter plot of RNA fold change (global translatome on the left, AHARIBO-rC on the right) compared to protein fold change (AHARIBO-nP) obtained by comparing EN with mESC. N: number of differentially expressed genes (DEGs) with p-value<0.05.

Figure 3—source data 1

A table with differentially expressed genes (DEGs) from RNA-seq data comprising logFC, LogCPM, LogFWER, and LogPval.

Genes are considered differentially expressed when both log fold changes are higher/smaller than 1.5/−1.5 and False Discovery Rate (FDR)-adjusted p-values are smaller than 0.01. DEGs from RNA-seq data.

https://cdn.elifesciences.org/articles/59303/elife-59303-fig3-data1-v2.txt
Figure 3—source data 2

A table with RNA and protein differentially expressed genes (DEGs) from AHARIBO-nP, pSILAC, AHARIBO-rC, and global translatome.

Genes are considered differentially expressed when both log fold changes are higher/smaller than 1.5/−1.5 and FDR-adjusted p-values are smaller than 0.01. Proteins are considered differentially expressed when adjusted p-values are smaller than 0.05. RNA and protein DEGs. AHARIBO: AHA-mediated RIBOsome isolation; pSILAC: pulsed SILAC. 

https://cdn.elifesciences.org/articles/59303/elife-59303-fig3-data2-v2.xlsx

To further prove the reliability of our method, we measured the efficiency of AHARIBO-rC to capture coding transcripts compared to a global translatome analysis. Using increasing abundance thresholds in EN, we observed that AHARIBO-rC efficiency is comparable to the global translatome for low abundant transcripts in EN and for all transcripts in undifferentiated mESCs (Figure 3—figure supplement 1A). Strikingly, AHARIBO captures abundant transcripts in EN with much higher efficiency than the global translatome (Figure 3B).

Finally, we tested whether the RNA isolated with AHARIBO-rC can predict the de novo synthesized proteome. After comparing differentially expressed genes (DEGs) during differentiation to the AHARIBO-nP proteome (Figure 3—source data 1), we observed that AHARIBO-rC RNA is a good proxy of the newly synthesized proteome (Pearson’s correlation r = 0.75, Figure 3C, Figure 3—figure supplement 1B). In particular, we found that AHARIBO-rC RNA presents less uncoupled genes (up-RNA and down-protein or down-RNA and up-protein) than the global translatome (Figure 3—figure supplement 1C), thus faithfully recapitulating proteome changes. The correlation of the global translatome with the global protein turnover measured with pSILAC shows a Pearson’s r = 0.27 (Figure 3—figure supplement 1D, Figure 3—source data 2). This result demonstrates that AHARIBO-nP does reflect the labeling of peptides rather than completely synthesized proteins.

Combined AHARIBO approaches define the functional role of lncRNAs in translation

Based on the evidence that a combination of AHARIBO approaches can simultaneously detect RNAs under active translation and peptides in the process of being produced, we applied our methods to detect ribosome-associated and translated native lncRNAs.

In AHARIBO-rC data, we identified a total of 687 lncRNA genes in mESCs and about 400 differentially expressed (DE) lncRNAs during neuronal differentiation (Figure 4—figure supplement 1A, Figure 4—source data 1). Among the top five DE lncRNAs (fold change >10; p-value<1×10−10), we found Pantr1 and Lhx1os, known to be involved in neuronal development (Biscarini et al., 2018; Carelli et al., 2019). To identify potentially translated lncRNAs, we applied the abundance threshold analysis to the subset of AHARIBO-rC non-coding RNAs in common with a published dataset (n = 270) of lncRNA identified by ribosome profiling data in mESCs (Ingolia et al., 2011; Figure 4—figure supplement 1B). The analysis of 100 lncRNAs in common between the two datasets showed a stronger enrichment of ribosome footprints in the AHARIBO-rC than in the global translatome (Figure 4A, Figure 4—figure supplement 1C). Altogether, these results suggest that a fraction of non-coding transcripts, which is efficiently isolated with AHARIBO-rC, is potentially translated.

Figure 4 with 3 supplements see all
Download asset Open asset
The AHA-mediated RIBOsome isolation (AHARIBO) platform can be used to detect ribosome-interacting long non-coding RNAs (lncRNAs).

(A) Linear plot illustrating the fraction of non-coding genes expressed above a minimum threshold in early neurons (EN). The AHARIBO-rC and the global translatome group are represented in yellow and cyan, respectively. For each group, the mean (solid line) and the SD (shades) of the fractions for a given count per million (CPM) threshold are calculated over all samples (n = 3) in that group. Expression values are indicated as normalized CPM. AHARIBO-rC was performed on the ribosome pellet after sucrose cushioning. (B) Venn diagram of the number of lncRNAs genes with at least 1 CPM identified by RNA-seq, AHARIBO-rC, RIBO-seq, and AHARIBO RIBO-seq. (C) Classification of lncRNAs interacting with ribosomes and relative detection through the multiple AHARIBO and standard approaches. ND: no detection of protein synthesis. (D) (Left) Schematic representation of the number of mouse embryonic stem cell (mESC) lncRNAs in common between AHARIBO RIBO-seq, AHARIBO-rC RNA, and standard RIBO-seq. These lnRNAs were validated by liquid chromatography-mass spectrometry (LC-MS). (Right) Example of an AHARIBO RIBO-seq ribosome occupancy profile of lncRNA 1810058I24Rik displaying the reads distribution along the entire transcript and the accumulation of reads at the known short open reading frame (shadow area and blue arrow on top).

Figure 4—source data 1

A table with the list of long non-coding RNAs (lncRNAs) identified by RNA-seq by RNA-seq in mouse embryonic stem cells (mESCs).

https://cdn.elifesciences.org/articles/59303/elife-59303-fig4-data1-v2.txt
Figure 4—source data 2

A table with the list of long non-coding RNAs (lncRNAs) identified by RIBO-seq in mouse embryonic stem cells (mESCs).

https://cdn.elifesciences.org/articles/59303/elife-59303-fig4-data2-v2.txt
Figure 4—source data 3

A table with the list of matching peptides from AHA-mediated RIBOsome isolation's (AHARIBO) identified long non-coding RNAs (lncRNAs).

https://cdn.elifesciences.org/articles/59303/elife-59303-fig4-data3-v2.xlsx

To understand if and how lncRNAs interact with ribosomes, we performed ribosome profiling experiments after AHARIBO pulldown (named AHARIBO RIBO-seq), with parallel standard RNA-seq (on inputs) analysis in mESCs. For protein-coding genes, both standard and AHARIBO RIBO-seq show an enrichment of RPFs in the coding sequence (Figure 4—figure supplement 2A). The two datasets show high correlation (Figure 4—figure supplement 2B) and the expected codon periodicity in the coding sequence in AHARIBO RIBO-seq (Figure 4—figure supplement 2C). These results further confirm the capability of AHARIBO in capturing ribosomes. With AHARIBO RIBO-seq, we identified a list of lncRNAs covered by ribosome footprints (Figure 4—source data 2). By intersecting our AHARIBO RIBO-seq data with those obtained from standard methods (RIBO-seq and RNA-seq after sucrose cushioning) or AHARIBO-rC, we identified 125 common putative translated lncRNAs (Figure 4B). Some of these lncRNA (n = 19) are known to be translated in mouse tissue (van Heesch et al., 2019). The vast majority of these lncRNAs do not have a known function. Two of the identified lncRNAs (9330151L19Rik and Gm9776) were detected only by standard RIBO-seq and RNA-seq but not with AHARIBO (Figure 4C). This result may be due to the absence of translation events (i.e., transcripts loaded with idle ribosomes). Next we validated the coding potential of lncRNAs that are in common between AHARIBO and standard RIBO-seq (Figure 4D). We translated in silico the transcripts in all frames to find potential ORFs with a canonical start codon (AUG). Translated sequences were semi-trypsin-digested in silico and then manually annotated to find confident matching spectra from the AHARIBO-nP protein dataset. Out of the about 46,000 collected spectra (Figure 4—source data 3), our MS-based proteomics analysis detected peptides with highly corresponding ribosome footprints (e.g., Gm42743, Gm26518, B230354K17Rik, D030068K23Rik, 1810058I24Rik). From the list of 129 lncRNAs that are in common among all AHARIBO protocols and standard RIBO-seq (Figure 4D), we identified by MS analysis a micropeptide (Mm47) of 47 aa (Figure 4D) at a high degree of confidence. This micropeptide derives from a lncRNA expressed in murine macrophages, and recently characterized by an independent group (Bhatta et al., 2020) as a relevant peptide able to modulate the innate immunity in mice. Several other lncRNAs show high confidence of translation events with in silico prediction even if they were not perfectly matching our proteomic spectra (Figure 4—figure supplement 3), paving the way for a better characterization of translatable lncRNA that has not been reported before. These results, combined with (1) AHARIBO’s efficiency in detecting an ectopically expressed micropeptide (TUG1-BOAT) and (2) concordance with recently published data, prove that our approach could be useful to unravel translation events in lncRNAs that are misannotated as non-coding. Altogether, our data confirm that our three diverse and complementary AHARIBO approaches represent a unique method to identify ribosome-associated and translated RNAs.

Discussion

LncRNAs localize in the nucleus or in the cytoplasm. In the nucleus, they modulate transcription, pre-mRNA splicing, or act as scaffold for protein interaction during chromatin organization (Sun et al., 2018). In the cytoplasm, the majority of lncRNAs is associated with polysomes (Carlevaro-Fita et al., 2016), where they either can or cannot produce proteins (Chen et al., 2020; Ingolia et al., 2011). Numerous lncRNAs are misannotated as non-coding but contain short ORFs encoding for micropeptides with biological relevance in cancer (D'Lima et al., 2017; Huang et al., 2017), bone development (Galindo et al., 2007), immunity (van Solingen et al., 2018), metabolism (Magny et al., 2013; Nelson et al., 2016), and DNA repair (Slavoff et al., 2014). Different methodological approaches have been developed to quantify the variations of RNA abundance by sequencing or imaging techniques (Blumberg et al., 2019; Jao and Salic, 2008; Morisaki et al., 2016; Wu et al., 2016), RNA engagement with the translational machinery by RIBO-seq or polysomal profiling (Arava et al., 2003; Clamer et al., 2018; Eden et al., 2011; Taniguchi et al., 2010), and protein synthesis by mass spectrometry or metabolic labeling (Aviner et al., 2013; Dieterich et al., 2006; Schwanhäusser et al., 2009; Yan et al., 2016). Despite these advantages, available technologies hardly capture in a single experiment the dynamics of translation across multiple biological conditions, the translation of unannotated coding transcripts, and translation-related functions of lncRNAs. Now that it is widely accepted that a portion of the genome annotated as non-coding can result in a complex transcriptome partially engaged with ribosomes (Chen et al., 2020; Djebali et al., 2012; Iyer et al., 2015), RNA sequencing and ribosome profiling should include micropeptide detection.

Our data show that AHARIBO serves as a flexible tool to detect translated RNAs, identify lncRNAs bound to elongating ribosomes, and detect de novo synthesized proteins. The intersection of standard RIBO-seq, RNA-seq, and AHARIBO approaches allowed us to identify translated lncRNAs. We demonstrated that AHARIBO is efficient in capturing short translated open reading frames, both native or ectopically expressed. Although LC-MS technologies are not as sensitive as RNA sequencing, we successfully identified a mouse-specific micropeptide reported to originate from a native lncRNA ORF, confirming the effectiveness of AHARIBO. To overcome existing limitations in LC-MS detection, many other translation events on lncRNAs can be predicted combining AHARIBO approaches with in silico translation of the identified leads. This approach would likely allow to selectively validate a list of still uncharacterized lncRNAs. Although the unlabeled background cannot be avoided, a pre-cleaning of the cell lysate with a cushioning step can help to increase the resolution with difficult samples. Moreover, a puromycin treatment instead of sBlock could be added as control in proteomic experiments. A unique feature of AHARIBO is the possibility to simultaneously isolate ribosomes, RNA engaged with ribosomes, and the corresponding proteins produced. Besides the versatility of the method, AHA labeling has the advantage of minimal interference with protein synthesis (Hodas et al., 2012; Tom Dieck et al., 2012).

The most prominent limitation of the method relies on the methionine starvation required for efficient AHA incorporation (Calve et al., 2016; Hodas et al., 2012; Saleh et al., 2019). This step can modify the physiological conditions of the cell and needs to be taken into consideration when planning experiments requiring certain stimuli (e.g., drug treatment) during methionine depletion. The conditions used in the AHARIBO protocol give robust protein labeling, but AHA concentration can be conveniently tuned based on specific cell types or biological questions. Additionally, we observed that there are still challenges for LC-MS verification of putative lncRNA peptides identified with AHARIBO. Of note, a potential contribution from background signal needs to be taken into consideration in LC-MS and Ribo-seq analysis.

With AHARIBO we introduce a strategy for the selective isolation of active ribosomes using the nascent peptide chain as bait for a more comprehensive interrogation of lncRNA biology and proteogenomic studies. Overall, we provide evidence that AHARIBO is a comprehensive and reliable toolkit suitable for downstream parallel RNA-seq, RIBO-seq, and LC-MS analysis, empowering scientists to shed light on the functional complexity of translation.

Materials and methods

Key resources table
Reagent type (species)
or resource		DesignationSource or referenceIdentifiersAdditional information
Cell line (Homo sapiens)Papillomavirus-related endocervical adenocarcinomaATCCRRID:CVCL_0030
Cell line (Mus musculus)46C embryonic stem cellsATCCRRID:CVCL_Y482Quattrone A. Lab. (CIBIO)
AntibodyAnti-β3-tubulin (mouse monoclonal)PromegaCat. #G712A
RRID:AB_430874		(1:2000)
AntibodyAnti-Oct4 (mouse monoclonal)Santa Cruz BiotechnologiesCat. #SC 5279
RRID:AB_628051		(1:2000)
AntibodyAnti-human RPL26 (rabbit polyclonal)AbcamCat. #ab59567
RRID:AB_945306		(1:2000)
AntibodyAnti-human RPS6 (rabbit polyclonal)AbcamCat. #ab40820
RRID:AB_945319		(1:2000)
AntibodyAnti-human beta actin (rabbit polyclonal)AbcamCat. #ab8227 RRID:AB_2305186(1:2000)
Recombinant DNA reagentWT TUG1-BOAT (plasmid)PMID:32894169
Recombinant DNA reagentΔ TUG1-BOAT (plasmid)This paperSee 'Materials and methods section: 'TUG1-BOAT ectopic expression and qPCR’
Recombinant DNA reagent+1Met TUG1-BOAT (plasmid)This paperSee 'Materials and methods' section: 'TUG1-BOAT ectopic expression and qPCR’
Peptide, recombinant proteinPrecision Protein StrepTactin-HRP ConjugateBioRadCat. #1610380(1:5000)
Chemical compound, drugL-Arginine-13C6,15N4 hydrochlorideSigma-AldrichCat. #608033
Chemical compound, drugL-Lysine-13C6,15N2 hydrochlorideSigma-AldrichCat. #608041
Chemical compound, drugL-Azidohomoalanine (Click-IT AHA)InvitrogenCat. #C10102
Chemical compound, drugDibenzocyclooctyne-PEG4-biotin conjugateSigma-AldrichCat. #760749SML1656
Chemical compound, drugsBlockIMMAGINA BioTechnologyCat. #SM8
Chemical compound, drugPuromycinSigma-AldrichCat. #P8833
Chemical compound, drugCycloheximideSigma-Aldrich#C4859
Chemical compound, drugLipofectamine 3000 Transfection ReagentThermo Fisher Scientific.Cat. #L3000001
Chemical compound, drugMag-DBCO beadsIMMAGINA BioTechnologyCat. #MDBCO
Chemical compound, drugeMagSi-cN beadsIMMAGINA BioTechnology#018-eMS-001
commercial assay or kitSMART-Seq Stranded KitTakaraCat. #634443
Commercial assay or kitSuperScript III Reverse TranscriptaseThermo FisherCat. #18080044
Commercial assay or kitKapa Probe Fast Universal qPCR KitKapa Biosystems#KK4702
Software, algorithmImage analysisImageJRRID:SCR_003070
Software, algorithmStatistical packageedgeRRRID:SCR_012802

Cell culturing and treatments

View detailed protocol

For protocol development, optimization, and validation, HeLa cells were used. HeLa cells were maintained on adherent plates in Dulbecco's modified Eagle's medium (DMEM; EuroClone #ECM0728L) supplemented with 10% fetal bovine serum, 2 mM L-glutamine, 100 units/mL penicillin, and 100 µg/mL streptomycin at 37°C, 5% CO2. For passaging, cells were washed with 1× Phosphate-Buffered Saline (PBS), detached using 0.25% trypsin-EDTA, and spun down at 260 × g for 5 min.

For treatments, 250,000–400,000 HeLa cells per well were seeded in six-well plates and grown to 80% confluence. At the time of treatment, culture medium was removed and cells were washed once with warm 1× PBS. Subsequently, cells were incubated with Dulbecco's modified Eagle's limiting medium (DMEM-LM; Thermo Scientific #30030) supplemented with 10% fetal bovine serum and 800 µM L-leucine for 40 min to deplete methionine reserves. Methionine-free medium was then supplemented with L-azidohomoalanine (Click-IT AHA; Invitrogen #C10102) at a final concentration of 250 µM and incubation time (ranging from 10 min to 120 min; 30 min set as incubation time for the protocol). Cells were then treated with 1× sBlock (IMMAGINA BioTechnology, catalog no. #RM8; sBlock is an anisomycin-containing proprietary reagent) for 10 min. Then, six-well plates were placed on ice, medium was removed, and cells were washed once with cold 1× PBS supplemented with 1× sBlock. After removing residual PBS with a pipette, hypotonic lysis buffer (0.01 M NaCl, 0.01 M MgCl2, 0.01 M Tris-HCl, 1% Tx-100, 1× sBlock, 1% sodium deoxycholate, 5 units/mL DNAse I [Thermo Scientific #89836], 200 units/mL RiboLock RNase Inhibitor [Thermo Scientific #EO0381], 1× Protease Inhibitor Cocktail [Cell Signaling Technology #5871S]) was added to each well, and cells were lysed with the aid of a scraper. After hypotonic lysis, nuclei and cellular debris were removed by centrifuging at 18,000 × g, 4°C for 5 min. For quantification of the total absorbance value of cell lysates, the absorbance was measured (260 nm) using a Nanodrop ND1000 UV-VIS Spectrophotometer. Lysates were aliquoted and processed directly or stored at −80°C.

Arsenite pre-treatment was performed by adding sodium arsenite (Sigma-Aldrich #S7400) at a final concentration of 500 µM for 1 hr.

For RNA-seq and proteomics experiments, two biological settings were assessed in triplicate experiments: (1) undifferentiated mouse 46C embryonic stem cells (mESCs) (Ying et al., 2003) and (2) mESCs induced to differentiate into ENs. mESCs were maintained in mESC self-renewal medium composed of Glasgow’s MEM (Thermo Scientific #11710-035) supplemented with 1000 units/mL ESGRO Recombinant Mouse LIF protein (Millipore #ESG1107), 10% fetal bovine serum, 55 μM 2-mercaptoethanol, 1 mM sodium pyruvate (Thermo Scientific #11360070), MEM non-essential amino acids (Thermo Scientific #11140050), GlutaMax (Thermo Scientific #35050061), and penicillin/streptomycin. For passaging, mESCs were washed twice with 1× PBS, detached using 0.02–0.05% trypsin-EDTA, and spun down at 260 × g for 3 min. Pellet was resuspended in fresh medium and plated onto 0.1% gelatin-coated culture vessels.

For treatments, 5 × 105 mESCs/cm2 were seeded in Petri dishes and grown to 60% confluence. For pSILAC proteomics, 24 hr before lysis mESCs were washed twice with 1× PBS and the medium was replaced with SILAC Advanced DMEM/F-12 Flex Medium (Thermo Scientific #A2494301), supplemented with 1000 units/mL ESGRO Recombinant Mouse LIF protein, 10% dialyzed fetal bovine serum, 4500 mg/L glucose, 17.25 mg/L proline, and penicillin/streptomycin. Either light or heavy L-arginine (Sigma-Aldrich #608033) and L-lysine (Sigma-Aldrich #608041) were added at 84 mg/L and 146 mg/L, respectively. For both AHA+ proteomics and RNA-seq experiments, treatments were performed as described above for HeLa cells, with the exception that methionine-free medium was supplemented with 1000 units/mL ESGRO Recombinant Mouse LIF protein and 10% dialyzed fetal bovine serum. After methionine depletion, cells were treated with 250 µM AHA for 30 min. The remaining treatment steps and hypotonic lysis were performed as detailed above.

Neuronal differentiation was performed according to a previously described protocol (Ying et al., 2003). Briefly, 2.000 mESCs/cm2 were seeded on gelatin-coated culture vessels in N2B27 medium. Cells were gently washed with 1× PBS, and medium was renewed every 1–2 days until 15DIV. N2B27 medium is composed of 1:1 mix of DMEM/F-12 (Thermo Scientific #21331020) and Neurobasal Medium (Thermo Scientific #21103049), supplemented with 0.5% N-2 (Thermo Scientific #17502048), 1% B-27 (Thermo Scientific #17504044), GlutaMax, and penicillin/streptomycin.

Upon differentiation, ENs were treated directly in culture vessels. For pSILAC proteomics, 24 hr before lysis ENs were washed once with 1× PBS and the medium was replaced with SILAC Advanced DMEM/F-12 Flex Medium, supplemented with 0.5% N2, 1% B27, 4500 mg/L glucose, 17.25 mg/L proline, and penicillin/streptomycin, 4500 mg/L glucose, 17.25 mg/L proline, and penicillin/streptomycin. Either light or heavy L-arginine and L-lysine were added at 84 mg/L and 146 mg/L, respectively. For both AHA+ proteomics and RNA-seq experiments, ENs were treated as described above for HeLa cells, with 250 µM AHA for 30 min. The remaining treatment steps and hypotonic lysis were performed as detailed above.

Cell lines were purchased directly from ATCC and passaged fewer than 15 times. Mus musculus 46C ES were obtained from Quattrone A. Lab (CIBIO, RRID:CVCL_Y482). All cells tested negative for mycoplasma contamination.

Immunocytochemistry

Request a detailed protocol

For immunofluorescence assay, cells were fixed with 4% paraformaldehyde for 15 min at room temperature, permeabilized using 0.5% Triton X-100 in 1× PBS for 15 min at room temperature, and blocked using 5% fetal bovine serum, 0.3% Triton X-100 in 1× PBS for 2 hr at room temperature. Cultures were then incubated overnight at 4°C with either anti-β3-tubulin (Promega #G712A) or anti-Oct4 (Santa Cruz Biotechnologies #SC-5279) primary antibodies diluted in 2% fetal bovine serum, 0.2% Triton X-100 in 1× PBS. Cells were then washed three times with 1× PBS and incubated with Alexa-555 anti-mouse secondary antibodies for 2 hr. Nuclei were counterstained with Hoechst 33258 before imaging with a Zeiss Axio Observer Z1 inverted microscope equipped with a 2.83 Megapixel AxioCam 503 mono D camera.

AHARIBO-rC/AHARIBO-nP: purification of active ribosomes for RNA/protein isolation

Request a detailed protocol

For RNA-seq experiments, lysates were diluted in W-buffer (10 mM NaCl, 10 mM MgCl2, 10 mM HEPES, 1× sBlock) to a final Nanodrop-measured absorbance (260 nm) of 1–2 a.u./mL, supplemented with 40 U of Superase-In RNase Inhibitor (Thermo Scientific #AM2696) and incubated with dibenzocyclooctyne-PEG4-biotin conjugate (Sigma-Aldrich #760749; 50 µM final concentration) in a reaction volume of 100 µL for 1 hr on a rotator in slow motion (9 rpm) at 4°C. Lysates were then incubated with 50 µL of eMagSi-cN beads (IMMAGINA BioTechnology #018-eMS-001) for 30 min at 4°C on the rotator in slow motion (9 rpm). Subsequently, samples were taken off the rotator and placed on a magnetic rack on ice, and supernatants were discarded. Beads were washed two times with 500 µL of 1× PBS supplemented with 0.1% Triton-X100, 1× sBlock, and 1:10,000 RiboLock RNase Inhibitor (Thermo Scientific #EO0381) on the rotator in slow motion at 4°C, removing supernatants from the tubes sitting on the magnetic rack and gently adding new washing solution each time. After the final wash, beads were resuspended in 200 µL of W-buffer and transferred to a new vial. Then, 20 µL of 10% SDS and 5 µL of Proteinase K (Qiagen #19131) were added to each sample, and samples were incubated at 37°C for 75 min in a water bath. Subsequently, suspensions were transferred to a new vial, and acid phenol:chloroform:isoamyl alcohol RNA extraction was performed. Briefly, an equal volume of acid phenol:chloroform:isoamyl alcohol (pH 4.5) was added, and samples were vortexed and centrifuged at 14,000 × g for 5 min. Aqueous phases were then transferred to new vials, 500 µL of isopropanol and 2 µL of GlycoBlue (Thermo Scientific #AM9516) were added, samples were mixed and incubated at room temperature for 3 min, and then stored overnight at −80°C. The following day samples were centrifuged at 14,000 × g for 30 min, supernatants were removed, 500 µL of 70% ethanol were added to each sample, and samples were then centrifuged at 14,000 × g for 10 min. Finally, pellets were air-dried and resuspended in 10 µL of nuclease-free water. When quality check and quantification was needed, RNA samples were run on a 2100 Bioanalyzer (Agilent) using the Agilent RNA 6000 Nano Reagents kit (Agilent #5067–1511) and assayed on the Qubit fluorometer using the Qubit RNA HS Assay Kit (Thermo Scientific #Q32852). For visualization of total RNA patterns, samples were run on a 1% agarose gel. ImageJ software (v 1.45s) was used for the quantitation of signal intensities of ribosomal RNA bands.

For proteomics experiments, lysates were diluted in W-buffer to a final Nanodrop-measured absorbance (260 nm) of 1–2 a.u./mL in a final volume of 100 µL. Ribosome pulldown was performed using Mag-DBCO beads (IMMAGINA BioTechnology #MDBCO). Lysates were incubated with 50 µL of beads for 1 hr on a rotator in slow motion (9 rpm) at 4°C. Supernatants were discarded after placing samples on the magnetic rack. Beads were washed three times with 500 µL of 200 mM Tris, 4% CHAPS, 1 M NaCl, 8 M urea, and pH 8.0 at room temperature on a shaker at 1000 rpm, using the magnetic rack to replace the washing solution. After the final wash, beads were resuspended in 30 µL of water and transferred to a new vial.

qRT-PCR analysis

Request a detailed protocol

Total RNA was extracted from samples processed through the AHARIBO-rC protocol as described above. Depending on the available input material, RNA was retrotranscribed using either RevertAid First Strand cDNA Synthesis Kit (Thermo Scientific #K1621) or SuperScript III Reverse Transcriptase (Thermo Fisher #18080044), per manufacturer's protocols. qPCR was run on CFX Connect Real-Time PCR Detection System (BioRad) using Kapa Probe Fast Universal qPCR Kit (Kapa Biosystems #KK4702). Reactions were performed in technical duplicates of biological triplicates. The following TaqMan probes were used: Hs99999901_s1 (18S), Hs02800695_m1 (HPRT1).

For the normalization of qRT-PCR results, HPRT1 was used as housekeeping gene. The fold change in normalized 18S RNA levels between untreated (control) and treated (arsenite) samples was calculated. A second normalization to threshold cycles from non-AHA-treated samples was done to account for background signal.

TUG1-BOAT ectopic expression and qPCR

Request a detailed protocol

We ectopically express the putative protein produced by the open reading frame of TUG1, called TUG1-BOAT (Tug1-Bifunctional ORF and Transcript), in HeLa cells. Briefly, construct generation and transfection was performed as in Lewandowski et al., 2020 with some minor changes to adapt the experimental setup to the AHARIBO method. We synthesized three different constructs for human Tug1 ORF1 (Thermo Scientific):

  1. The first (called WT TUG1-BOAT) is the one reported in Lewandowski et al., 2020. It has a non-canonical start codon and a methionine at 75 nt (25 aa) upstream of the stop codon.

  2. The second (called Δ TUG1-BOAT) is deleted by the only methionine of the sequence present at 75 nt from the stop codon. No methionines are present.

  3. The third (called +1Met TUG1-BOAT) has an ATG start codon (methionine) instead of the non-canonical CTG start codon e. Therefore, the third construct has two methionines, one at the N terminal and the other at 25 aa (about 75 nt) upstream of the C-terminal.

We cloned the constructs in the pcDNA3.1(+) plasmid with HindIII and EcoRV restriction enzymes. For transfection of TUG1-BOAT constructs, we seeded HeLa cells in a six-well plate and transfected the cells with 2.5 μg of plasmids (pcDNA3.1(+) containing each of the inserts) using 742 Lipofectamine 3000 Transfection Reagent (Thermo Fisher Scientific). After 24 or 48 hr post transfection, cells were processed with AHARIBO-rC protocol followed by RNA extraction.

We performed qPCR analysis on AHARIBO pulldowns and input for each vector to validate the efficiency in capturing short translated ORF deriving from RNA annotated as lncRNA (TUG1). Briefly, 200 ng of DNase I-treated RNA was used as input to generate cDNA using High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems), according to the manufacturer's protocol. qPCR was run on CFX Connect Real-Time PCR Detection System (BioRad) using Powerup Sybr Master Mix (Applied Biosystems) and a couple of primers design to amplify 150 nt of the CDS of TUG1-BOAT transcript (see below). Reactions were performed in technical duplicates of biological duplicates. For normalization, 18S was used as housekeeping gene. Ct values were analyzed using the ∆∆Ct method (Livak and Schmittgen, 2001).

  • Fw PRIMER: GGCTCTTTCTCCTGCTCTGG

  • Rev PRIMER: CTCCTCGTCGAATCGCAAAC

  • Insert size: 150 nt

TUG1-BOAT sequences are listed below:

Italics: 5′ UTR leader sequence; bold: canonical and non-canonical start codons; red: methionine.

  • >WT TUG1-BOAT

  • GGCCGAGCGACGCAGCCGGGACGGTAGCTGCGGTGCGGACCGGAGGAGCCATCTTGTCTCGTCGCCGGGGAGTCAGCCCCTAAATCGAAGAAGCCCTGGCGCGCCCTCCCCCCCTCCCGGGTCTGGTAGGGCGAAGGAACGGGCGTGCGGTCGATCGAGCGATCGGTTGGCGGCTCTTTCTCCTGCTCTGGCATCCAGCTCTTGGGGCGCAGGCCCGGCCGCCGCGGCGCGCGCCCGGTGGCCGTTGGCGCTCGCGCCGCGTCTTTCTTCTCGTACGCAGAACTCGGGCGGCGGCCTATGCGTTTGCGATTCGACGAGGAGTCGTCCGGGTGGTCGGCGGCGGCGGGCAGCTGCTCCGCCCCGCTCCGGGGGAGGCGGCGGCGGCAGCGGCCGCGGGATTTGGAGCGGCCGGGGAGGCGGGGGTGGCCGGGGCCGGCTTGGAGGCCTGGCGCCACCCTTCGGGGCCTGCAAGGACCCAGTTGGGGGGGCAGGAGGGGGCCGGAGGATGGTTGGTTGTGGGATTTCTACTTTGCCTTTTCCTCCTTATGCCGCCTGACTACAAAGACCATGACGGTGATTATAAAGATCATGACATCGACTACAAGGATGACGATGACAAGTAG

  • >ΔTUG1-BOAT

  • GGCCGAGCGACGCAGCCGGGACGGTAGCTGCGGTGCGGACCGGAGGAGCCATCTTGTCTCGTCGCCGGGGAGTCAGCCCCTAAATCGAAGAAGCCCTGGCGCGCCCTCCCCCCCTCCCGGGTCTGGTAGGGCGAAGGAACGGGCGTGCGGTCGATCGAGCGATCGGTTGGCGGCTCTTTCTCCTGCTCTGGCATCCAGCTCTTGGGGCGCAGGCCCGGCCGCCGCGGCGCGCGCCCGGTGGCCGTTGGCGCTCGCGCCGCGTCTTTCTTCTCGTACGCAGAACTCGGGCGGCGGCCTATGCGTTTGCGATTCGACGAGGAGTCGTCCGGGTGGTCGGCGGCGGCGGGCAGCTGCTCCGCCCCGCTCCGGGGGAGGCGGCGGCGGCAGCGGCCGCGGGATTTGGAGCGGCCGGGGAGGCGGGGGTGGCCGGGGCCGGCTTGGAGGCCTGGCGCCACCCTTCGGGGCCTGCAAGGACCCAGTTGGGGGGGCAGGAGGGGGCCGGAGGATGGTTGGTTGTGGGATTTCTACTTTGCCTTTTCCTCCTTCCGCCTGACTACAAAGACCATGACGGTGATTATAAAGATCATGACATCGACTACAAGGATGACGATGACAAGTAG

  • >+1Met TUG1-BOAT

  • GGCCGAGCGACGCAGCCGGGACGGTAGCTGCGGTGCGGACCGGAGGAGCCATCTTGTCTCGTCGCCGGGGAGTCAGGCCCCTAAATCGAAGAAGCCATGGACTACAAGGATGACGATGACAAGGCGCGCCCTCCCCCCCTCCCGGGTCTGGTAGGGCGAAGGAACGGGCGTGCGGTCGATCGAGCGATCGGTTGGCGGCTCTTTCTCCTGCTCTGGCATCCAGCTCTTGGGGCGCAGGCCCGGCCGCCGCGGCGCGCGCCCGGTGGCCGTTGGCGCTCGCGCCGCGTCTTTCTTCTCGTACGCAGAACTCGGGCGGCGGCCTATGCGTTTGCGATTCGACGAGGAGTCGTCCGGGTGGTCGGCGGCGGCGGGCAGCTGCTCCGCCCCGCTCCGGGGGAGGCGGCGGCGGCAGCGGCCGCGGGATTTGGAGCGGCCGGGGAGGCGGGGGTGGCCGGGGCCGGCTTGGAGGCCTGGCGCCACCCTTCGGGGCCTGCAAGGACCCAGTTGGGGGGGCAGGAGGGGGCCGGAGGATGGTTGGTTGTGGGATTTCTACTTTGCCTTTTCCTCCTTATGCCGCCTGACTACAAAGACCATGACGGTGATTATAAAGATCATGACATCTAG

RNA-seq

Request a detailed protocol

RNA samples were subjected to library preparation for the Illumina platform using the SMART-Seq Stranded Kit (Takara #634443) as per manufacturer's instructions using 5 ng of RNA as starting material. For quality check and quantification, the final libraries were run on a 2100 Bioanalyzer (Agilent) using the Agilent DNA 1000 Kit (Agilent #5067-1504) and assayed on the Qubit fluorometer using the Qubit dsDNA HS Assay Kit (Thermo Scientific #Q32851). Libraries were sequenced on an Illumina HiSeq2500 by the NGS Core Facility (University of Trento).

Polysome profiling

Request a detailed protocol

HeLa cells were treated and lysed as described above, adding one of the following blocking drugs: (1) sBlock (IMMAGINA BioTechnology #SM8, final concentration 1×, 10 min treatment); (2) cycloheximide (Sigma-Aldrich #C4859; final concentration 30 µM, 5 min treatment); (3) puromycin (Sigma-Aldrich #P8833; final concentration 50 µM, 5 min treatment); and (4) no blocking drug. Cleared supernatants obtained from cytoplasmic lysates were loaded on a linear 15–50% sucrose gradient and ultracentrifuged in a SW41Ti rotor (Beckman) for 1 hr and 40 min at 180,000 × g at 4°C in a Beckman Optima LE-80K Ultracentrifuge. After ultracentrifugation, gradients were fractionated in 1 mL volume fractions with continuous absorbance monitoring at 254 nm using an ISCO UA-6 UV detector. Each fraction was flash-frozen in liquid nitrogen and stored at −80°C for subsequent protein extraction.

Polysome profiles were analyzed as follows. The relative intensity of each individual fraction was determined for both on-membrane AHA and RPL26 signals, then the AHA/RPL26 relative intensity ratio was calculated for each fraction. For each profile, the relative intensity ratios of polysome-containing fractions (fractions 8/9–10/11) were averaged and normalized to the relative intensity ratio of the 60S fraction, which was chosen as internal baseline for background signal based on the fact that it should be devoid of translationally active ribosomes. To assess the effect of the different blocking drugs, averaged normalized relative intensity ratios for the profiles obtained from different blocking drugs and from the untreated control sample were compared. ImageJ software (v 1.45s) was used for quantitation of signal intensities of protein bands.

Sucrose cushioning for ribosome enrichment (global translatome)

Request a detailed protocol

HeLa cells were treated in Petri dishes and lysed as described above, adding 1× sBlock as blocking drug. Sucrose cushioning was performed according to a modified version of a previously described protocol (Ingolia et al., 2012). For each sample, a volume of cell lysate corresponding to 1.7 a.u. (based on Nanodrop measurement of absorbance at 260 nm) was layered on top of 900 µL of 30% sucrose cushion (30 mM Tris-HCl pH 7.5, 100 mM NaCl, 10 mM MgCl2, 1 M sucrose in nuclease-free water) supplemented with 1× sBlock. Samples were ultracentrifuged at 95,000 rpm at 4°C for 1 hr and 40 min using a TLA100.2 rotor (Beckman). Pellets were resuspended in 100 µL of nuclease-free water supplemented with 30 mM Tris-HCl, pH 7.5, 100 mM NaCl, 10 mM MgCl2.

Protein extraction from sucrose gradient fractions

View detailed protocol

Polysomal fractions (1 mL) or pellet/supernatant fractions from 30% sucrose cushioning (1/5th of total amount, adjusted to 260 µL volume) were processed for methanol/chloroform protein extraction. Briefly, 600 µL of methanol and 150 µL of chloroform were added to each sample and samples were vortexed. Then, 450 µL of deionized water were added to each sample and samples were vortexed again. Samples were centrifuged at 14,000 × g for 1 min at room temperature, and the resulting aqueous phase was removed without disrupting the underlying white ring (protein interface). Subsequently, 450 µL of methanol were added to each sample, samples were vortexed, and then centrifuged at 14,000 × g for 2 min at room temperature. After centrifugation, supernatants were removed and pellets air-dried. Finally, pellets were resuspended in deionized water supplemented with Pierce Lane Marker Reducing Sample Buffer (Thermo Scientific #39000) to a final volume of 15 µL and either stored at −80°C or heated at 95°C and directly used for SDS-PAGE.

On-membrane click chemistry

Request a detailed protocol

Cell lysate or protein extracts obtained from sucrose gradient fractions were supplemented with Pierce Lane Marker Reducing Sample Buffer (Thermo Scientific #39000), heated at 95°C for 10 min, and separated by SDS-PAGE. Separated proteins were transferred to nitrocellulose membranes, then membranes were blocked overnight at 4°C in 5% milk prepared in 1× Tris-Buffered Saline (TBS) − 0.1% Tween20 supplemented with dibenzocyclooctyne-PEG4-biotin conjugate (Sigma-Aldrich #760749; 50 µM final concentration). Membranes were washed three times in 1× TBS − 0.1% Tween20 for 10 min each, then incubated with Precision Protein StrepTactin-HRP Conjugate (BioRad #1610380; 1:1000 in 5% milk prepared in 1× TBS − 0.1% Tween20) for 1 hr at room temperature, then washed again. Membranes were subsequently developed using Amersham ECL Prime Western Blotting Detection Reagent (GE Healthcare #RPN2236). Images were acquired through the ChemiDoc MP Imaging System. ImageJ software (v 1.45s) was used for quantitation of AHA signal intensities.

Immunoblotting

Request a detailed protocol

Aliquots of 10–20 µL of cell lysate or protein extracts obtained from sucrose gradient fractions were supplemented with Pierce Lane Marker Reducing Sample Buffer (Thermo Scientific #39000), heated at 95°C for 10 min, and separated by SDS-PAGE. Separated proteins were transferred to nitrocellulose membranes, then membranes were blocked for 1 hr at room temperature in 5% milk prepared in 1× TBS − 0.1% Tween20. Membranes were subsequently incubated for 1 hr at room temperature with the following primary antibodies, diluted in 5% milk prepared in 1× TBS − 0.1% Tween20: anti-RPL26 (Abcam #ab59567; 1:2000), anti-RPS6 (Abcam #ab40820; 1:1000), and anti-beta-actin (Abcam #ab8227; 1:2000). Membranes were washed three times in 1× TBS − 0.1% Tween20 for 10 min each, then incubated with the appropriate HRP-conjugated secondary antibodies for 1 hr at room temperature and washed again as before. Membranes were then developed using either Amersham ECL Prime Western Blotting Detection Reagent (GE Healthcare #RPN2236) or SuperSignal West Femto Maximum Sensitivity Substrate (Thermo Scientific #34095), depending on signal intensities. Images were acquired through the ChemiDoc MP Imaging System. ImageJ software (v 1.45s) was used for quantitation of signal intensities of protein bands.

RNA-seq data analysis

Request a detailed protocol

FASTQ files from Illumina HiSeq2500 were first checked for adapters and quality-base distribution using FASTQC tool (http://www.bioinformatics.babraham.ac.uk/projects/fastqc), followed by trimming with Trimmomatic-0.35 (Bolger et al., 2014) with the following settings: ILLUMINACLIP:ADAPTOR_FILE:2:30:8:1 LEADING:3 TRAILING:3 SLIDINGWINDOW:4:15. Prior to sequencing data processing, technical replicates (different sequencing lanes) from the same library were merged together generating a unique FASTQ per sample. Reads were aligned onto mm10 Mouse genome using STAR-2.6.0a (Dobin et al., 2013) with a maximum mismatch of two and default setting for all other parameters. Once uniquely mapped reads were selected, the GRCm38.92 mouse gene annotation from Ensembl (http://www.ensembl.org) was incorporated in the HTSeq-count v0.5.4 (Anders et al., 2015) tool to obtain gene-level counts. Genes with counts per million (CPM) <1 in all samples were considered as not expressed and hence removed from the analysis. Trimmed mean of M values (TMM) (Robinson and Oshlack, 2010) normalization and CPM conversion were then performed to obtain gene expression levels for downstream analyses. For each comparison, differential expression testing was performed using the edgeR-3.20.9 (Robinson et al., 2010) statistical package from Bioconductor. According to the edgeR approach, both common (all genes in all samples) and separate (gene-wise) dispersions were estimated and integrated into a negative binomial generalized linear model to moderate gene variability across samples. Finally, genes having a Log fold change higher/smaller than 1.5/−1.5 and an FDR-corrected p-value of 0.01 (or smaller) were considered as significant.

Proteomics experiments

Request a detailed protocol

Proteomic analysis was performed on samples processed through the pSILAC and AHARIBO workflows, as described above. For pSILAC experiments, cells were prepared as described above (see 'Cell culturing and treatments'). Then, 50 μg of lysates was subjected to acetone precipitation and protein pellets were dissolved in 50 mM ammonium bicarbonate and 6 M urea. For AHARIBO enrichment, the beads used for ribosome pulldown were reconstituted in 100 µL 6 M urea with 50 mM ammonium bicarbonate.

Samples were reduced using 10 mM DTT, 1,4-Dithiothreitol (DTT) for 1 hr at room temperature and alkylated with 20 mM iodoacetamide in the dark for 30 min at room temperature. Subsequently, proteins were digested at room temperature with 0.5 μg Lys-C (Promega, #VA1170) for 4 hr, after which the solution was diluted four times in 50 mM ammonium bicarbonate. Then, 1 μg of trypsin (Promega, #V5111) was added to the samples and proteolysis was carried out overnight. Digestion was interrupted by adding 1% trifluoroacetic acid. Samples were then desalted by C18 stage-tip, lyophilized, and resuspended in 20 μL of buffer A (0.1% formic acid) for LC-MS/MS analysis.

Samples were analyzed using an Easy-nLC 1200 system coupled online with an Orbitrap Fusion Tribrid mass spectrometer (both Thermo Fisher Scientific). Peptides were loaded onto a 25-cm-long Acclaim PepMap RSLC C18 column (Thermo Fisher Scientific, 2 µm particle size, 100 Å pore size, id 75 µm) heated at 40°C. For pSILAC samples, the gradient for peptide elution was set as follows: 5–25% buffer-B (80% acetonitrile, 0.1% formic acid) over 90 min, 25–40% over 15 min, 40–100% over 18 min, and 100% for 17 min at a flow rate of 400 nL/min. For AHARIBO pulldown samples, the same steps for peptide elution were set over a total gradient of 80 min. The instrument was set in a data-dependent acquisition mode. The full MS scan was 350–1100 m/z in the orbitrap with a resolution of 120,000 (200 m/z) and an AGC target of 1 × 106. MS/MS was performed in the ion trap using the top speed mode (3 s), an AGC target of 5 × 103, and an HCD collision energy of 30.

MS raw files were analyzed by using Proteome Discoverer (v2.2, Thermo Scientific). MS/MS spectra were searched by the SEQUEST HT search engine against the human or the mouse UniProt FASTA databases (UniProtKB 11/2018). Trypsin was specified as the digestive enzyme. Cysteine carbamidomethylation (+57.021 Da) was set as fixed modification, and methionine oxidation (+15.995 Da) and N-term acetylation (+42.011 Da) as variable modifications. SILAC labeling (Lys +8.014 Da, Arg +10.008 Da) was used as quantification method for pSILAC samples. All other values were kept as default.

Proteomics data analysis

Request a detailed protocol

Heteroscedastic T-test was used to assess the significant differences in peptide/protein abundance (p-value lower than 0.05) unless stated otherwise. Data distribution was assumed to be normal, but this was not formally tested. GO and Kyoto Encyclopedia of Genes and Genomes pathway analysis were performed using DAVID version 6.8, PANTHER 14.1, and Enrichr (http://amp.pharm.mssm.edu/Enrichr/).

Identification of lncRNA peptides from result spectra

Sequenced non-coding RNAs were in silico translated into amino acid sequences using the EMBOSS Transeq tool from EMBL. Only the three forward frames were translated. Spectra obtained from the AHARIBO enrichment of newly synthesized proteins were searched against a database of typical contaminants like keratins, trypsin, and bovine serum albumin provided by MaxQuant (Cox and Mann, 2008). The software utilized for database searching was Proteome Discoverer (v2.4, Thermo Scientific); the non-fragment filter and the Top N Peaks Filter (with N = 4 per 100 Da) were also used in the workflow to eliminate noise signals from the MS/MS spectra. The spectra not matching with high confidence this database were searched against the human SwissProt database. Those not matching with both databases were used to match the in silico translated database generated by EMBOSS Transeq using semi-specific tryptic cleavage to consider also unexpected translation start sites. We considered only those spectra that passed the 1% FDR threshold and created two distinct groups for those peptides with an AUG ‘in-frame’ versus not in-frame.

Ribosome profiling

Request a detailed protocol

mESCs at 80% confluence were pre-treated with the elongation inhibitor cycloheximide before rapid harvest on ice and cell lysis (lysis buffer, IMMAGINA BioTechnology #RL001-1). Clarified cell lysates (1.7 total a.u., measured by Nanodrop) were treated with 1.3 U of RNase I (Thermo, #AM2295) in W-buffer (IMMAGINA BioTechnology #RL001-4) containing 1× sBlock to digest RNA not protected by ribosomes. Digestion was performed for 45 min at RT and then stopped with Superase-In RNase Inhibitor (Thermo Scientific #AM2696) for 10 min on ice. Samples were then processed differentially according to the specific approaches described below.

Standard RIBO-seq

Request a detailed protocol

80S ribosomes were isolated by centrifuging lysates through a 30% sucrose cushion at 95,000 rpm, for 2 hr at 4°C. The cushion was resuspended in W-buffer and treated with SDS 10% (final 1%) and 5 µL of proteinase K (20 mg/mL), and incubated at 37°C in a water bath for 75 min before acid phenol:chloroform:isoamyl alcohol (pH 4.5) RNA extraction.

AHARIBO RIBO-seq

Request a detailed protocol

The lysates were incubated with dibenzocyclooctyne-PEG4-biotin conjugate (Sigma-Aldrich #760749; 50 µM final concentration) in a reaction volume of 100 µL for 1 hr on a rotator in slow motion (9 rpm) at 4°C. Lysates were then incubated with 50 µL of eMagSi-cN beads (IMMAGINA BioTechnology #018-eMS-001) for 30 min at 4°C on the rotator in slow motion (9 rpm). Subsequently, samples were taken off the rotator and placed on a magnetic rack on ice, and supernatants were discarded. Beads were washed two times with 500 µL of 1× PBS supplemented with 0.1% Triton-X100, 1× sBlock, and 1:10,000 RiboLock RNase Inhibitor (Thermo Scientific #EO0381) on the rotator in slow motion at 4°C, removing supernatants from the tubes sitting on the magnetic rack and gently adding new washing solution each time. After the final wash, beads were resuspended in 200 µL of W-buffer and transferred to a new vial. Then, 20 µL of 10% SDS and 5 µL of Proteinase K (Qiagen #19131) were added to each sample, and samples were incubated at 37°C for 75 min in a water bath. Subsequently, suspensions were transferred to a new vial, and acid phenol:chloroform:isoamyl alcohol (pH 4.5) RNA extraction was performed.

For both approaches, protocol steps starting from RNA extraction were performed as follows. Briefly, an equal volume of phenol:chloroform:isoamyl alcohol was added, and samples were vortexed and centrifuged at 14,000 × g for 5 min. Aqueous phases were then transferred to new vials, 500 µL of isopropanol and 2 µL of GlycoBlue (Thermo Scientific #AM9516) were added, samples were mixed and incubated at room temperature for 3 min, and then stored overnight at −80°C. The following day samples were centrifuged at 14,000 × g for 30 min, supernatants were removed, 500 µL of 70% ethanol were added to each sample, and samples were then centrifuged at 14,000 × g for 10 min. Finally, pellets were air-dried and resuspended in 10 µL of nuclease-free water. Extracted RNA was then resolved by electrophoresis through denaturing TBE-urea gels, and fragments between 25 nt and 35 nt were excised. Libraries were prepared using the RiboLace kit_module 2 (IMMAGINA BioTechnolgy #RL001_mod2) and sequenced on an Illumina HiSeq 2500 sequencer with a single-end 50 bp run.

RIBO-seq data analysis

Request a detailed protocol

Reads were processed by removing 5' adapters, discarding reads shorter than 20 nucleotides, and trimming the first nucleotide of the remaining ones (using Trimmomatic v0.36). Reads mapping on the collection of M. musculus rRNAs (from the SILVA rRNA database, release 119) and tRNAs (from the Genomic tRNA database: gtrnadb.ucsc.edu/) were removed. Remaining reads were mapped on the mouse transcriptome (using the Gencode M17 transcript annotations). Antisense and duplicate reads were removed. All alignments were performed with STAR (v020201) employing default settings.

The identification of the P-site position within the reads was performed using riboWaltz (v1.1.0) computing the P-site offsets from the 3' end of the reads. The percentage of P-sites falling in the three annotated transcript regions (5' UTR, CDS, and 3' UTR) was computed using the function region_psite included in riboWaltz (v1.1.0). Transcript counts were normalized using the TMM method implemented in the edgeR Bioconductor package. Transcripts displaying 1 CPM in at least one replicate were kept for further analyses.

Data availability

All data generated or analysed during this study are included in the manuscript and supporting files. Source data files have been provided. All sequencing data are deposited in public archives and made available upon publication.

The following data sets were generated
    1. Minati L
    2. Romanel A
    3. Peretti A
    4. Gandolfi F
    5. Clamer M
    (2021) ProteomeXchangeConsortium
    ID PXD022679. One-shot analysis of translated mammalian lncRNAs with AHARIBO - Mass-spectrometry proteomics data.
    https://www.ebi.ac.uk/pride/archive/
    1. Minati L
    2. Romanel A
    3. Peretti A
    4. Gandolfi F
    5. Clamer M
    6. Firrito C
    (2021) GEO repository
    ID GSE167865. One-shot analysis of translated mammalian lncRNAs with AHARIBO - Ribo-seq processed data.
    https://www.ncbi.nlm.nih.gov/geo/
    1. Minati L
    2. Romanel A
    3. Peretti A
    4. Gandolfi F
    5. Clamer M
    6. Firrito C
    (2021) NCBI BioProject
    ID PRJNA692822. One-shot analysis of translated mammalian lncRNAs with AHARIBO - NGS processed BAM files.
    https://www.ncbi.nlm.nih.gov/bioproject/
The following previously published data sets were used
    1. Ingolia NT
    2. Lareau LF
    3. Weissman JS
    (2011) GEO
    ID GSE30839. Ribosome Profiling of Mouse Embryonic Stem Cells Reveals the Complexity and Dynamics of Mammalian Proteomes.
    https://www.cell.com/fulltext/S0092-8674%2811%2901192-5#secd1715918e1381

References

    1. Grollman AP
    (1967)
    Inhibitors of protein biosynthesis. II. mode of action of anisomycin
    The Journal of Biological Chemistry 242:226–233.

Decision letter

  1. Howard Y Chang
    Reviewing Editor; Stanford University, United States
  2. James L Manley
    Senior Editor; Columbia University, United States
  3. John L Rinn
    Reviewer; University of Colorado at Boulder, United States

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Acceptance summary:

Using AHA labeling approach, the authors reported a one-shot approach by combining RNA-seq, Ribo-seq, and LC-MS. The authors provide evidence that this approach helps identification of translatable lncRNAs.

Decision letter after peer review:

Thank you for submitting your article "One-shot analysis of translated mammalian lncRNAs with AHARIBO" for consideration by eLife. Your article has been reviewed by three peer reviewers, and the evaluation has been overseen by a Reviewing Editor and James Manley as the Senior Editor. The following individual involved in review of your submission has agreed to reveal their identity: John L Rinn (Reviewer #2).

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

Summary:

Minati et al. reported a method called AHARIBO to detect RNAs with active translation. Using L-azidohomoalanine (AHA) to label nascent polypeptides, the authors purified the translation complex via click chemistry followed by high-throughput analysis including RNA-seq. LC-MS, and Ribo-seq. The authors compared translatome between ESC and differentiated neurons (EN) using AHARIBO and found some long non-coding RNAs (lncRNAs) that could encode peptides.

The overall methodology is straightforward and useful. The authors claim that AHARIBO is able to distinguish ribosome-associated lncRNAs and ribosome-translating lncRNAs. This is indeed an important question in the field since a growing body of evidence suggests that a substantial amount of lncRNAs contain functional open reading frames. Although the title, Abstract, and Introduction primarily focus on this topic, the result section did not serve this goal at all. A revision needs to address the key issues below, as well as temper claims and list the potential limitations and interpretations in a revised Discussion.

Essential revisions:

1) AHARIBO detection of nascent peptide vs. mature protein association with RNA.Nascent peptide capture does seem to offer a new approach to measure translation. The core strategy is not adequately validated, however, and so it is not clear that these techniques are capturing proteins and RNAs through nascent-chain labeling. Further, individual applications of this technique for proteomics or sequencing are not subject to incisive tests that clearly distinguish the proposed nascent chain capture mechanism from alternative explanations. It is in fact unclear if there is a straightforward path to address these central concerns.

a) It is clear that the great majority of AHA label is found in completed free proteins rather than in nascent proteins (Figure 1B). Capture of ribosomal proteins (along with a range of other unspecified proteins) in AHA- controls doesn't exclude the possibility that this (Figure 1D) reflects enrichment of new proteins. In addition to labeling of mature proteins, it seems that only 2- to 3-fold enrichment is achieved in comparison with AHA- samples (Figure 1—figure supplement 1D). This means that a large fraction of captured protein, including captured ribosomes/polysomes - is unlabelled background.

b) Likewise, AHA labeling is often used to measure nascent protein synthesis, much like pSILAC labeling. What is the evidence that the concordance in pSILAC and AHARIBO-nP doesn't simply reflect the labeling of fairly new but completely synthesized protein in AHARIBO-nP lysates?

c) If AHARIBO is capturing nascent peptides, certain strong polarity effects are expected: peptides should be strongly enriched near the N-terminus of proteins relative to the C-terminus. Ribosome footprints should be absent from the first 30 – 40 codons, because these proteins should not expose nascent peptide.

2) Evidence for newly detected lncRNA encoded peptides via epitope tagging. Can the authors express a tagged version (even if ectopically) of some of the newly identified peptides. This would be a complimentary validation to the mass spec performed and provide spatial localization information. For example, the lncRNA TUG1 was recently reported in Cell and Genome Biology to encode a peptide and is highly abundant in both cell types. Cell : DOI:https://doi.org/10.1016/j.cell.2019.05.010) and genome biology: in press, currently: https://www.biorxiv.org/content/10.1101/562066v1. Do the authors find TUG1 translated in these cells (the RNA is abundant in both) that would be another validation of a "validated" new peptide that is larger than 100a.a

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

Thank you for resubmitting your work entitled "One-shot analysis of translated mammalian lncRNAs with AHARIBO" for further consideration by eLife. Your revised article has been evaluated by James Manley (Senior Editor) and a Reviewing Editor.

The manuscript has been improved but there are some remaining issues that need to be addressed before acceptance, as outlined below:

1) The validation of TUG1 peptide strengthens the original conclusion. Since TUG1 cannot be identified from AHARIBO, it will be highly desirable if the authors could show any newly identified peptides from lncRNA, i.e., translatable lincRNA that has not been reported before. At least, a thorough discussion about this potential is needed.

2) The puromycin treatment resulted in ~15-20% reduction in AHARIBO signal rather than the >80% seen in ribosome profiling experiments, which implies that a substantial fraction of the AHARIBO signal comes from nonspecific background. Please incorporate text to explicitly address this point – should puromycin treatment always be done to confirm AHARIBO screen hits?

https://doi.org/10.7554/eLife.59303.sa1

Author response

Essential revisions:

1) AHARIBO detection of nascent peptide vs. mature protein association with RNA.Nascent peptide capture does seem to offer a new approach to measure translation. The core strategy is not adequately validated, however, and so it is not clear that these techniques are capturing proteins and RNAs through nascent-chain labeling. Further, individual applications of this technique for proteomics or sequencing are not subject to incisive tests that clearly distinguish the proposed nascent chain capture mechanism from alternative explanations. It is in fact unclear if there is a straightforward path to address these central concerns.

a) It is clear that the great majority of AHA label is found in completed free proteins rather than in nascent proteins (Figure 1B).

We agree with the reviewer that, to some extent, AHA labeled proteins, fully synthesized and released from ribosomes during the incubation time with AHA, could be captured by AHARIBO. This is also mentioned in our original manuscript: “The high signal observed in light fractions is caused by AHA-labeled proteins released from ribosomes”. Proteins fully synthesized could affect AHARIBO’s efficiency in different ways. Therefore, we approached the problem as follow:

1) We check if a sucrose cushioning of the cell lysate could increase AHARIBO’s efficiency in terms of total RNA recovery. Hence, the separation of free full-synthesized proteins from ribosomes before performing AHARIBO, could help to reduce the background.

2) We performed additional controls to further demonstrate the robustness of the method, even without the need of a sucrose cushioning of the cell lysate

1) Sucrose cushioning of the AHARIBO cell lysate

Starting from a HeLa cell lysate, we compared the efficiency of AHARIBO before and after sucrose cushioning in terms of total RNA capture (calculated as the fold change enrichment between AHA treated and control, AHA+/AHA-). As a matter of fact, a clarification of the cell lysate from full length proteins allows better AHARIBO-rC efficiency expressed as RNA recovery.

The cushioning step could help to improve the AHARIBO efficiency in highly proliferative cells, when protein synthesis rate is maximized and a larger amount of protein can be produced and released from ribosomes during AHA incubation. We included this information in the manuscript and in Figure 1—figure supplement 1. It is important to note that the cushioning step can be relevant for the identification of translated lncRNAs. In fact, the differences between AHARIBO and global translatome for the enrichment in lncRNA are more pronounced after sucrose cushioning (Figure 4A) than without sucrose cushioning (Figure 4—figure supplement 1C).

That being said, since (i) AHARIBO is meant to be a fast protocol with low instrumental needs (e.g. ultracentrifuge) and (ii) we further confirmed that a clarification of the cell lysate is not essential in low proliferative cell lines (primary neurons, Author response image 1); for the sake of clarity, we preferred to not include this step in the standard experimental setup of Figure 1.

We included this sentence to the manuscript: “To overcome this problem it is possible to perform a pre-cleaning of the cell lysate by sucrose cushioning. This step can increase the efficiency of total RNA isolation compared with the control (no AHA) (Figure 1—figure supplement 1D).”

2) Additional controls to prove the robustness of the method (sections A, B and C)

A) We reasoned that AHARIBO’s effectiveness could be affected by free AHA-tagged in the cell lysate, if the AHA-tagged proteins are RNA binding proteins (RBPs). In this case, a possibility of co-isolation of RNAs bound to AHA-tagged RBPs cannot be excluded. If this is the case, we should observe a bias toward specific subtypes of isolated mRNAs. In particular, we should expect that RBPs binding sites are significantly overrepresented in the isolated AHARIBO mRNA when compared with the global translatome. To rule out this hypothesis, we analyzed differentially expressed genes during cell differentiation (mESC/EN), searching for specific enrichment of sequences known to be recognized by RNA binding proteins. We screened 1000 differentially expressed mRNA (P value < 0.005) from AHARIBO-rC and from the global translatome, and we mapped the RBP motives in the untranslated regions (UTRs) of the selected transcript (http://aura.science.unitn.it/about/) by using the Atlas of UTR Regulatory Activity (AURA, mouse database) (Dassi et al., 2014).

Author response table 1
AHARIBO-rC.

UTR regulatory elements enrichment

Regulatory factor/elementRegulated query genesEnrichment p-valueBH-corrected p-value
Tardbp119 ( 11.90 % )< 1.0E-07< 1.0E-07
Nova2175 ( 17.50 % )< 1.0E-07< 1.0E-07
Srsf3201 ( 20.10 % )< 1.0E-07< 1.0E-07
Ezh2159 ( 15.90 % )1.1E-070.00000034
Ptbp222 ( 2.20 % )1.9E-070.00000057
Apobec121 ( 2.10 % )6.3E-070.0000019
Mbnl224 ( 2.40 % )1.26E-050.00003775
Ago21 ( 2.10 % )7.72E-050.00023166
Srsf268 ( 6.80 % )0.0013780.00275579
Elavl42 ( 0.20 % )0.0016550.00331066
Elavl22 ( 0.20 % )0.0032590.00651716
Elavl13 ( 0.30 % )0.0048740.00974701
Rbms31 ( 0.10 % )0.0236890.02368882
Rbm315 ( 1.50 % )0.0364230.03642291
Rbm8a1 ( 0.10 % )0.0468170.04681703
Author response table 2
Global translatome.

UTR regulatory elements enrichment

Regulatory factor/elementRegulated query genesEnrichment p-valueBH-corrected p-value
Tardbp96 ( 9.60 % )< 1.0E-07< 1.0E-07
Nova2139 ( 13.90 % )< 1.0E-07< 1.0E-07
Srsf3168 ( 16.80 % )< 1.0E-07< 1.0E-07
Srsf256 ( 5.60 % )2.95E-061.77E-05
Apobec119 ( 1.90 % )9.13E-063.65E-05
Ptbp219 ( 1.90 % )1.08E-054.33E-05
Srsf159 ( 5.90 % )7.26E-050.0002177
Elavl22 ( 0.20 % )0.003258580.00651716
Elavl13 ( 0.30 % )0.00487350.00974701
Ago15 ( 1.50 % )0.019498880.03899777
Rbms31 ( 0.10 % )0.023688820.02368882
Rbm8a1 ( 0.10 % )0.046817030.04681703

We noticed that AHARIBO and global translatome retrieve similar results for highly represented regulatory elements. Therefore, we can exclude any major bias due to the co-isolation of RNAs through AHA-tagged RBPs. The only significant and robust difference between the two datasets is related to the regulator factor histone-lysine N-methyltransferase, Ezh2. This protein is present in actively dividing cells, widely expressed in early embryos and in central and peripheral nervous system. Since we did not identified this protein in AHARIBO-nP, we can claim that the protein is not AHA-labelled and the related RNA motive is enriched because it probably binds UTRs of differentially expressed transcripts during translation.

In conclusion, our analysis on UTR regulatory elements did not show a strong difference between the two datasets (AHARIBO and global translatome), suggesting that full length AHA-tagged RNA binding proteins do not affect AHARIBO RNA results.

B) A possibility to reduce the free protein content is to reduce the AHA incubation time. Unfortunately, a compromise in the time of incubation is required when cell lines have different protein synthesis rates. We tested AHARIBO on different cellular models (HeLa, mESC) and differentiation state (mESC differentiate into early neurons). Early neurons (EN) have a protein turnover slower than embryonic stem cells (Author response image 1) and proliferative (HeLa) cells. In this work, we empirically defined the optimal time point for RNA enrichment on AHARIBO beads. To be consistent, we set a constant incubation time of 30 min among all experiments. This time is required to maximizing the recovery of both RNA and ribosomal proteins in HeLa, mESCs, and mESC differentiated in early neurons (Author response image 1). This time point is probably more than enough for a proper nascent protein labeling of human immortalized cancer cells (HeLa) and mESCs (where at 30 min a lot of proteins are produced and released) but it is the minimum requirement for a good efficiency in EN without additional step of sucrose cushioning.

Author response image 1
Download asset Open asset
Total protein staining.

Left, total proteins from mESC and early neurons (EN) cell lysates (20 µg total protein for each sample measured by Bradford assay) loaded on a SDS-PAGE and the membrane stained by biotin-cycloaddition followed by streptavidin-HRP. Central and right panel, agarose gel electrophoresis of total RNA extracted from input lysates (1/10 of the total lysate volume) and lysates subjected to AHARIBO pulldown from mESC (left) or EN cells (right) either treated (+) or not (-) treated with AHA.

In conclusion, we provide additional evidences that the enrichment of RNA with AHARIBO is consistent across all biological samples tested.

C) To better characterize AHARIBO for its ability to capture ribosome complexes, we performed the following additional experiments:

i) We inhibit translation with heat shock (10 min at 42°C) during the 30 min of AHA incubation time;

ii) We forced the release of the nascent protein chain from ribosomes by puromycin treatment instead of sBlock.

In both experimental setups we should observe a reduced AHARIBO’s efficiency. In the first case, due to a reduced protein synthesis. In the second case, due to the inhibition of protein synthesis and a release of the nascent chain from the ribosome.

i) Heat shock. First, we checked if the heat shock was effective by performing a RT-qPCR to monitor the RNA level of a member of the 70 kilodalton heat shock proteins (HSPA4 ), a family of proteins known to be involved in the response to heat shock (Richter et al., 2010). We observed that HSPA4 is upregulated in AHARIBO-rC (Author response image 2A) for both HeLa and mESC. Then, we checked if the heat shock affects global protein synthesis by SDS-PAGE (Author response image 2B). As a consequence of the heat shock, we observed a decrease of the AHA-labelled proteins (representing de-novo protein synthesis) compared with untreated cells (no heat shock).

Author response image 2
Download asset Open asset
a) HSPA4 fold change (ΔΔct) measured by qPCR in mouse embryonic stem cells and HeLa with or without heat shock (10 min at 42°C).

For each sample, HSPA5 AHA+/AHA- signal ratios were normalized to the control and to the housekeeping gene. b) Left, quantification of AHA content before (control) and after heat shock. On the right, representative image of a SDS-PAGE reporting the total protein content for heat shock (with and without AHA) and not treated sample (with and without AHA). In each lane a total of 1 ug of protein as measured by Bradford assay was loaded. Staining of the membrane was performed by biotin cycloaddition followed by streptavidin-HRP. Experiments were performed in triplicates.

As reported in the original manuscripts, when protein synthesis is inhibited (e.g. arsenite treatment), AHARIBO-rC captures less ribosomes and associated RNAs. According to that, and after having established that the heat shock condition was effective, we applied AHARIBO-rC on HeLa cells with or without heat shock followed by RNA extraction form AHARIBO-rC. We observed that the total AHARIBO-rC RNA is reduced after heat shock, as expected due to the reduction in protein synthesis.

In conclusion, a down-regulated protein synthesis by heat shock results in a reduction of RNA captured relative to the control, as reported in the main text for the arsenite treatment. We included these results in Figure 1—figure supplement 3.

ii) Puromycin treatment. We performed AHARIBO experiment in HeLa cells treated with 50 μm of puromycin. Puromycin reacts with the nascent chain, terminating translation and causing ribosome disassembling. Therefore, AHA-tagged nascent chains will no longer be available on ribosomes for AHARIBO capture. Strikingly, we observed a strong improvement of the signal-to-background ratio for AHARIBO RNA enrichment as further validated by LC-MS and western blot analysis (see below, answer to reviewer #1b).

In conclusion, the release of the nascent peptide chain from the ribosomes by means of puromycin results in a reduction of RNA captured relative to the control. We included this additional validation of AHARIBO in Figure 1—figure supplement 3.

Capture of ribosomal proteins (along with a range of other unspecified proteins) in AHA- controls doesn't exclude the possibility that this (Figure 1D) reflects enrichment of new proteins. In addition to labeling of mature proteins, it seems that only 2- to 3-fold enrichment is achieved in comparison with AHA- samples (Figure 1—figure supplement 1D). This means that a large fraction of captured protein - including captured ribosomes/polysomes - is unlabelled background.

We thank the reviewer for drawing our attention to the unlabeled background present in AHARIBO. A main source of this background could be generated by unwanted reactions of the biotin-dibenzocyclooctyne (DBCO) reagent in the cell lysate when AHA is not present. In fact, the kinetic of the reaction of DBCO with AHA is fast and efficient, but when AHA is not present (as in the “AHA-” control) the ligand can react more efficiently with other chemical groups (van Geel et al., 2012). To better understand the problem, we performed AHARIBO with two different controls:

control 1: AHARIBO with AHA but without DBCO

control 2: AHARIBO with AHA but with puromycin treatment instead than sBlock treatment.

As suggested by the reviewer, we strengthened and validated our manuscript by adding new set of experiments (data included in Figure 1—source data 1):

i) New LC-MS analysis with control 1

After isolation of the ribosomal complexes with AHARIBO-rC from HeLa cell lysates, we extracted all proteins captured. In parallel the same experiment was performed without biotin-DBCO, therefore no specific pull-down of AHA-tagged nascent peptides and ribosomes is possible. After trypsin digestion and LC-MS analysis on all proteins co-purified with AHARIBO-rC we selectively identified all ribosomal proteins. We observed from a 4-fold (e.g. RPL23A) to 18-fold (e.g. RPL4, RPL7) enrichment is achieved for AHARIBO-rC in comparison with “control 1” samples. This means that a portion of the unlabeled background is DBCO related.

ii) New LC-MS analysis and western blots results with control 2

We reasoned that an additional experiment to account for the DBCO-related background is to include a control where the nascent peptide is released from ribosome instead of being stabilized on it with sBlock. We treated cells with puromycin instead of sBlock and we repeated the LC-MS analysis (in this case AHA is present in both control and puromycin treated sample). Our results show that less background and a stronger enrichment is obtained in LC-MS and in western blot data compared with the AHA- control (no AHA). This result is in agreement with our data on the RNA recovery reported at the reviewer answer 2cii.

iii) Dedicated experiments with an ectopically expressed micropeptide.

See answer to reviewer #2. The ectopic expression of the short peptide Tug1-Boat demonstrates that in the absence of methionine in the nascent peptide chain, the Tug1-Boat transcript is not efficiently co-purified with AHARIBO (50-fold signal-to-background).

Although the DBCO-related background is hardly removable, these additional experiments confirm the robustness of the method. We changed Figure 1 and the main text according to the new results and we added this to Figure 1 and Figure1—figure supplement 2.

b) Likewise, AHA labeling is often used to measure nascent protein synthesis, much like pSILAC labeling. What is the evidence that the concordance in pSILAC and AHARIBO-nP doesn't simply reflect the labeling of fairly new but completely synthesized protein in AHARIBO-nP lysates?

The reviewer is correct in highlighting the relation between pSILAC and AHARIBO-nP. We apologize that this point was not clear enough.

We tackled the question comparing 100 differentially expressed proteins (mean of triplicate sets of data) detected by pSILAC and AHARIBO-nP, whose transcripts were detected by AHARIBOrC RNA as well. Our data show a poor concordance between the two datasets (Author response image 3). Therefore, AHARIBO-nP is not reflecting the total protein turnover as measured by pSILAC, but it is capturing only a fraction of the pSILAC proteome.

Author response image 3
Download asset Open asset
Protein fold change (log2) of differentially expressed proteins during cellular differentiation of mESCs to early neurons.

Comparison between pSILAC(black) and AHARIBO proteome (red).

Additionally, we compared differentially expressed genes detected using the global translatome versus the protein turnover (pSILAC). The analysis revealed a poor concordance between pSILAC and global translatome RNA variations (Pearson’s r = 0.27), differently from what we observed with AHARIBO (main Figure 3).

Therefore, we can conclude that AHARIBO protein does not mostly reflect all pSILAC data, but only a fraction of the them represented by translated mRNA during AHA incubation. We modified the Results section and we add to Figure 3—figure supplement 1.

c) If AHARIBO is capturing nascent peptides, certain strong polarity effects are expected: peptides should be strongly enriched near the N-terminus of proteins relative to the C-terminus. Ribosome footprints should be absent from the first 30 – 40 codons, because these proteins should not expose nascent peptide.

We thank the reviewer for suggesting the possibility of a polarity in our ribosome profiling data. Related to that, other groups reported a reduced coverage of ribosome footprints at the beginning of the transcript in selective ribosome profiling experiments to capture ribosomes engaged with chaperon proteins. For example, the Bakau’s lab (Figure 3A in Oh et al., 2011) reported that the E. coli trigger factor (TF) engages with the polypeptide emerging from the exit channel after the first 100 amino acids are translated. A similar result was observed in yeast (Figure 1B in Döring et al., 2017) where a selective Ribo-seq revealed a reduced read density in the first 50 codons reflecting the interaction of chaperon proteins with the nascent chains beyond a length of 50 aa.

Given that the first 30 aa are buried inside the ribosome exit tunnel and chaperon proteins need a certain number of amino acid to support the proper folding of the newly synthetized polypeptide, it is clear that these proteins engage with nascent chain when at least 20 residues are exposed from the exit tunnel of the ribosome.

To understand if this was the case for AHARIBO as well, we analyzed the differences between the P-site density of AHARIBO Ribo-seq and total RIBO-seq (input).

First, we looked at the ratio between the average P-sites density distribution in AHARIBO RIBO-seq and the average number of P-sites on the whole CDS, in the first 120 nucleotides from initiation. Strikingly, we observed a significant (measured with two-sided Wilcoxon– Mann–Whitney test) lower P-site density for AHARIBO in the first 120 nt, confirming that the first 40 amino acids are less covered by reads (Author response image 4, left).

Author response image 4
Download asset Open asset
Left, P-site density initiation/CDS ratio for total RIBO-seq (input) and AHARIBO Ribo-seq.

Right, metagene profiles showing the ratio between the AHARIBO density and input read density within the first 450 codons from the start codon. Violet arrow, first 25 codons from the start codon. Blue arrow, first 50 codons from the start codon.

Then, we looked at the ratio between AHARIBO and Input metagene profiles, and we observed a reduction of the RPF signal in the first 50 codons, with the strongest reduction in the first 25 codons (Author response image 4, right).

These results confirm that:

1) In agreement with the hypothesis of an enrichment of the signal near the N-terminus, we observed a reduced RPF density in the first 120 nt in AHARIBO when compared to the input.

2) The reduction of signal is not broadly distributed in the first 100 codons as reported for chaperon proteins, but mainly concentrated in the first 25-50 codons. This could be explained by the different type of engagement with the nascent chain. Unlike experiments published for chaperon proteins, AHARIBO captures the nascent chain through a small molecule, i.e. AHA reaction with biotin-DBCO followed by beads pulldown. We cannot exclude that the small biotin-DBCO molecule reacts with the N-terminal AHA-tagged nascent chain just outside the exit tunnel, or partially inside the main exit tunnel. According to that, we speculate that the molecule could bind the nascent chain, much before large chaperon proteins interact with the nascent chain.

In support of this hypothesis, we observed that Joachim Frank’s lab (Gabashvili et al., 2001) and Song’s lab (Dao Duc et al., 2019) reports that the ribosome exit tunnel have a dynamic behavior and multiple exits to facilitate the extrusion of nascent peptides (Figure 1 in Duc DK et al., 2019) from ribosomes. In particular, by comparing 20 different cryo EM and X-ray crystallography structure of the ribosome from all three domains of life, D. Duc. et al. observed that the first 25-30 aa buried inside the ribosome exit tunnel are passing through two main constrictions. A first, at about 30 Å (about 9 aa) from the PTC and made by the uL4 and uL22 protein loops. The second, at about 50 Å (about 15 aa) from the PTC formed by an extended arm of uL4 (Figure 4A in Duc DK et al., 2019). The radius of the constriction is about 4 Å at the first troughs and 3-4 Å at the second (in human). While the radius at the exit is about 7-8 Å.

With a length of 32 Å and a radius of 7 Å, biotin-DBCO can potentially reach the internal part of the tunnel at about 60-80 Å from the PTC. This ribosome architecture at the exit tunnel may influence the interaction of small molecules with the ribosomes. Therefore, we should expect a reduction of the Ribo-seq signal in the first 75 nt from the start codon. This is in agreement with our meta-profile reported in Author response image 4 (right) and with the Tug1-boat micropeptide ectopic expression experiment (Reviewer answer #2).

3) We still observed some signal at the start codon, although less than in the input. This can be related to the background or to a possible capture of the tRNAimet (substituted by tRNAiAHA in AHARIBO) within the 48S translational initiation complex. This complex could generate signal at the start codon as previously reported for selective 48S profiling experiments (Bohlen et al., 2020).

At the present stage, the hypothesis on ribosome profiling data reported above are not supported by additional biochemical and structural data. For the sake of clarity we did not include these information in the manuscript. Nevertheless, we further discussed this point in the answer to reviewers essential revision 2.

We are aware that AHARIBO for ribosome profiling could have a better mechanistic description, but this topic is out of the scope of this work and could result in a separate publication.

2) Evidence for newly detected lncRNA encoded peptides via epitope tagging. Can the authors express a tagged version (even if ectopically) of some of the newly identified peptides. This would be a complimentary validation to the mass spec performed and provide spatial localization information. For example, the lncRNA TUG1 was recently reported in Cell and Genome Biology to encode a peptide and is highly abundant in both cell types. Cell : DOI:https://doi.org/10.1016/j.cell.2019.05.010) and genome biology: in press, currently: https://www.biorxiv.org/content/10.1101/562066v1. Do the authors find TUG1 translated in these cells (the RNA is abundant in both) that would be another validation of a "validated" new peptide that is larger than 100a.a

The suggestion of the reviewer is extremely useful for a better understanding of AHARIBO mechanism and efficiency. Unfortunately, we did not find TUG1 transcripts in our mESC and EN RNA-seq and Ribo-seq datasets. Hence, a proper discussion about the capture TUG1 with AHARIBO in mESC and EN is not possible.

We, therefore, decided to ectopically express the putative protein produced by the open reading frame of TUG1, called TUG1-BOAT (Tug1-Bifunctional ORF and Transcript) in HeLa cells as reported in (Lewandowski et al., 2020).

Construct generation and transfection was performed as in Lewandowski J.P. et al., 2019 with some minor changes (reported in the updated Materials and methods section of the manuscript) to adapt the experimental setup to the AHARIBO method. Briefly, we synthesized three different constructs for human Tug1 ORF1 (Thermo Scientific):

1) The first (called WT TUG1-BOAT) is the one reported in Lewandowski J.P. et al., 2019. It has a non-canonical start codon and a methionine at 75 nt (25 aa) upstream the stop codon.

2) The second (called ΔTUG1-BOAT), is deleted by the only methionine of the sequence present at 75 nt upstream the stop codon

3) The third (called +1Met TUG1-BOAT), has the non-canonical CTG start codon exchange with an ATG codon (methionine). Therefore, the third construct has two methionines, one at the N terminal and a second one at 25 aa before (75 nt) the stop.

We performed target TUG1-BOAT qPCR analysis on AHARIBO pull-down and input for each experiment/construct after 12 and 24 hours of transfection, to validate the efficiency in capturing short translated ORF deriving from a lncRNA (TUG1).

Our RT-qPCR analysis shows a good efficiency (10x – 50x compared with the control) of AHARIBO in capturing TUG1-BOAT RNA when a methionine at the N terminal of the encoded peptide is present (about 50 times more in +1Met TUG1-BOAT than with ΔTUG1-BOAT).

We observed that the WT transcript is captured as well, although the methionine is only 25 aa far from the stop codon. The 25 amino acids are just enough for the polypeptide to protrude from the exit tunnel (25 aa = 75 nt = about 90 Å) This is in agreement with our ribosome profiling data (showing a lower density in the first 25 codons) confirming that AHARIBO can engage with short nascent peptides at the exit of the ribosome tunnel (answer to reviewers essential revision #1c).

The expression of the exogenous small ORF TUG1-BOAT is a valuable demonstration of the AHARIBO efficiency. We described this result in the main text of the manuscript and we included this in Figure 1—figure supplement 3.

Next, to gain more insights into the function of thee lncRNAs in mESCs and early neurons we:

i) Compared our results with a dataset of experimentally validated lncRNA producing short peptides (van Heesch et al., 2019) in rat and mouse heart.

We noted that 19 lncRNAs detected by AHARIBO-rC in mESCs, are present in the dataset reported by Heesch S. et al., 2019, highlighting the fact that AHARIBO can capture translated lncRNA. Since most of these lncRNA are tissue specific (Heesch S et al., 2019), we expected that only a small subset of mESC and EN lncRNAs captured by AHARIBO were in common with the lncRNAs identified by Heesch S et al. in rat and mouse heart.

ii) Analyzed the lncRNA differentially expression level after heat shock in mESCs

We exposed mESCs to a short heat shock (42°C for 10°C), to induce the expression of a subset of genes involved in the adaptive response. First, we confirmed that cells react to heat shock by monitoring the upregulation of the heat shock transcript (HSPA4). Next, we confirmed that AHARIBO can successfully detect changes induced by the stress (see answer reviewer #1).

Then, we performed AHARIBO-rC RNA-seq in mESCs before and after heat shock and we analyzed differentially expressed lncRNAs. We did not observe any significant change in lncRNAs differential expression at the level of the global translatome, while in AHARIBO-rC we detected 11 lncRNA (Table 3) differentially expressed after heat shock. Unfortunately, none of them has been validated by other groups or is known to produce micropeptides. Therefore, we decided to not include these data in the manuscripts, but they further demonstrate the potential of our technique. In fact, this experiments could help further investigation on the effect of lncRNAs in heat shock, as previously reported on other species (Bernabò et al., 2020), since only a few data related to human cells are published so far (Place and Noonan, 2014).

Additional proteomic analysis are needed to better characterized potentially translated lncRNA after heat shock in human cells, which may result in a separate work.

Author response table 3
lncRNA differentially expressed in mESC after heat shock and captured by AHARIBO-rC.
geneSymbollogFClogCPMPValue
Gm285924.045326-0.185441.01E-07
Gm266352.2374280.0550597.37E-08
1110019D14Rik-2.130093.2246594.19E-07
4930467D21Rik-2.461512.5672172.90E-07
Gm26776-2.61450.7522261.83E-07
Malat1-2.6527311.844341.38E-08
AC162181.1-2.815431.3013711.40E-10
Gm30551-3.160130.5891254.86E-09
Gm5432-3.70264-0.112551.06E-07
mt-Rnr2-4.5688415.378532.27E-10
4930440I19Rik-4.611880.0149353.90E-07

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

The manuscript has been improved but there are some remaining issues that need to be addressed before acceptance, as outlined below:

1) The validation of TUG1 peptide strengthens the original conclusion. Since TUG1 cannot be identified from AHARIBO, it will be highly desirable if the authors could show any newly identified peptides from lncRNA, i.e., translatable lincRNA that has not been reported before. At least, a thorough discussion about this potential is needed.

We thank the reviewer for having appreciated our additional experiments and for the constructive comment and thorough suggestion about the TUG1 results.

We addressed the question raised by the reviewer through additional analysis of our AHARIBO RNA-seq, RIBO-seq and proteomic data.

In AHARIBO RIBO-seq, we identified numerous lncRNAs covered by ribosome footprints. By intersecting AHARIBO RIBO-seq data with those obtained from standard methods (RIBO-seq and RNA-seq after sucrose cushioning) and AHARIBO-rC RNA-seq, we identified more than hundred common putative translated lncRNAs (Figure 4B, new version of the manuscript). To validate the coding potential of lncRNAs identified by AHARIBO RNA-seq (Figure 4D, new version of the manuscript) and Ribo-seq, we translated in silico the transcripts in all frames to find potential ORFs. In silico peptides were used to match confident matching spectra obtained from the AHARIBO-nP protein dataset. We found 129 lncRNAs in common between all AHARIBO protocols and standard RIBO-seq (Figure 4D). Among these lncRNAs we identified a micropeptide (Mm47) of 47 aa (Figure 4D, new version of the manuscript) with a high degree of confidence in mass spectroscopy data. According with our hypothesis and results, this micropeptide was recently discovered and characterized by Bhatta and colleagues (Bhatta et al., 2020).

Several other peptides show high confidence of translation events with in silico prediction. Even if they were not perfectly matching our proteomic spectra, these peptides are likely generated from lncRNA not known to be translated. These list of transcripts is reported in Figure 4—source data 3 and listed below (in brackets the unique putative peptide identified).

List of matching peptides from lncRNAs as “high” confidence matching spectra:

ENSMUST00000051089 (NSFVNDIFER), ENSMUST00000181328 (KIDNQINLPK), ENSMUST00000181149 (KINQLQNMVKDNK), ENSMUST00000099446 (NLMNVINVVKLLHFS), ENSMUST00000180524 (MSPSQLLELKRNQ), ENSMUST00000182499 (VCVALIINICHIMI), ENSMUST00000134140 (NGGGLLMSYVIK), ENSMUST00000180432 (ELAEQPSSALKTSNREQ), ENSMUST00000181251 (QLTDNQRVNQKA), ENSMUST00000179344 (KELQLK), ENSMUST00000181443 (KGPNDISLAQSYLPI), ENSMUST00000071101 (KNNPPPQNAKPK), ENSMUST00000180407 (IELRENLQTY), ENSMUST00000180489 (EISASANLELNGAPSQQ), ENSMUST00000188038 (LALEELR), ENSMUST00000149246 (LLLPGVIK).

In addition, six abundant lncRNAs identified in AHARIBO RNA-seq show massive ribosome occupancy in our AHARIBO RIBO-seq data ( > 99th percentile of all lncRNA identified by AHARIBO RIBO-seq). In silico-translation of these peptides results in many potential micropeptids (ENSMUST00000180396 (23), ENSMUST00000181751 (61), ENSMUST00000182010 (43), ENSMUST00000192833 (94), ENSMUST00000200021 (27), ENSMUST00000223012 (86)

To further strengthen our results, we analysed the ribosome occupancy profile of three lncRNAs, which display hallmarks of translation in AHARIBO RIBO-seq and the predicted micropeptides position along the transcript in the three different frames of transation. None of them are annotated as translated lncRNA.

As requested by the reviewer, we add the results shown in a new supplementary figure (Figure 4—figure supplement 3) and modify the manuscript text as follow in the Results paragraph:

“Several lncRNA show high confidence of translation events with in silico prediction, even if they were not perfectly matching our proteomic spectra (Figure 4—figure supplement 3), paving the way for a better characterization of translatable lncRNA that has not been reported before.”

and in the Discussion:

“To overcome existing limitation in LC-MS detection, many other translation events on lncRNAs can be predicted combining AHARIBO approaches with in-silico translation of the identified leads. This approach would likely allow to selectively validate a list of still uncharacterized lncRNAs.”

2) The puromycin treatment resulted in ~15-20% reduction in AHARIBO signal rather than the >80% seen in ribosome profiling experiments, which implies that a substantial fraction of the AHARIBO signal comes from nonspecific background. Please incorporate text to explicitly address this point – should puromycin treatment always be done to confirm AHARIBO screen hits?

We thank the reviewer for his/her suggestion. To better understand how much signal is caused by nonspecific background, we performed a set of experiments with puromycin treatment before applying AHARIBO. In particular:

1) we compared AHARIBO’s efficiency in isolating total RNA with or without puromycin pre-treatment. We observed a 12-fold signal-to-background in puromycin treat sample relative to control (Figure 1—figure supplement 3).

2) we compared AHARIBO’s efficiency in isolating proteins with or without puromycin pre-treatment. By both LC-MS analysis and immunoblotting we identified ribosomal proteins among of all proteins captured by AHARIBO. Our proteomic results show a 1.2- to 2-fold signal-to-background in puromycin treated sample compared with the control; i.e not puromycin treated (AHARIBO-rC/AHA+PURO). Immunoblotting performed to analyse two selected ribosomal protein shows a 2- to 9-fold signal-to-background in puromycin treat sample relative to control (Figure 1—figure supplement 2).

Overall, RNA detection methods (RNA-seq, RIBO-seq, Syber gold gel staining of total RNA) were more sensitive than LC-MS detection. This may explain the better results seen in ribosome profiling experiments mentioned by the reviewer.

We agree that a fraction of the AHARIBO signal is not specific background. By performing a sucrose cushioning step before applying AHARIBO, it is possible to reduce the background caused by peptides not associated to ribosomes, as shown in Figure 1—figure supplement 1D. We discuss about this possible strategy in the Results paragraph as follows:

“To overcome this problem it is possible to perform a pre-cleaning of the cell lysate by sucrose cushioning. This step can increase the efficiency of total RNA isolation compared with the control (no AHA) (Figure 1—figure supplement 1D)”.

We agree with the reviewer that a puromycin control could be useful control to account for the nonspecific background in proteomic analysis. Therefore, we modified the Discussion of the manuscript accordingly to the reviewer suggestion:

Although the unlabelled background cannot be avoided, a pre-cleaning of the cell lysate with a cushioning step can help to increase the resolution with difficult samples.

“Moreover, a puromycin treatment instead of sBlock could be added as control in proteomic experiments.”.

https://doi.org/10.7554/eLife.59303.sa2

Article and author information

Author details

  1. Luca Minati

    IMMAGINA BioTechnology, Trento, Italy
    Contribution
    Conceptualization, Formal analysis, Validation, Investigation, Writing - original draft
    Contributed equally with
    Claudia Firrito, Alessia Del Piano and Alberto Peretti
    Competing interests
    L.M. is an employee of IMMAGINA BioTechnology S.r.l.

  2. Claudia Firrito

    IMMAGINA BioTechnology, Trento, Italy
    Contribution
    Investigation
    Contributed equally with
    Luca Minati, Alessia Del Piano and Alberto Peretti
    Competing interests
    C.F. is an employee of IMMAGINA BioTechnology S.r.l.

  3. Alessia Del Piano

    IMMAGINA BioTechnology, Trento, Italy
    Contribution
    Investigation
    Contributed equally with
    Luca Minati, Claudia Firrito and Alberto Peretti
    Competing interests
    A.D.P. is an employee of IMMAGINA BioTechnology S.r.l.

  4. Alberto Peretti

    IMMAGINA BioTechnology, Trento, Italy
    Contribution
    Investigation, Methodology, Writing - first workflow and methods
    Contributed equally with
    Luca Minati, Claudia Firrito and Alessia Del Piano
    Competing interests
    A.P. is an employee of IMMAGINA BioTechnology S.r.l.

  5. Simone Sidoli

    Department of Biochemistry, Albert Einstein College of Medicine, Bronx, United States
    Contribution
    Data curation, Formal analysis, Methodology
    Competing interests
    No competing interests declared

  6. Daniele Peroni

    Mass Spectrometry Facility, Computational and Integrative Biology (CIBIO), University of Trento, Trento, Italy
    Contribution
    Data curation, Formal analysis, Investigation, Methodology
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-0862-266X

  7. Romina Belli

    Mass Spectrometry Facility, Computational and Integrative Biology (CIBIO), University of Trento, Trento, Italy
    Contribution
    Data curation, Formal analysis, Visualization, Methodology
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-5690-2797

  8. Francesco Gandolfi

    Laboratory of Bioinformatics and Computational Genomics, Department of Cellular, Computational and Integrative Biology (CIBIO), University of Trento, Trento, Italy
    Contribution
    Data curation, Formal analysis
    Competing interests
    No competing interests declared

  9. Alessandro Romanel

    Laboratory of Bioinformatics and Computational Genomics, Department of Cellular, Computational and Integrative Biology (CIBIO), University of Trento, Trento, Italy
    Contribution
    Data curation, Formal analysis, Investigation, Visualization, Methodology
    Competing interests
    No competing interests declared

  10. Paola Bernabo

    IMMAGINA BioTechnology, Trento, Italy
    Contribution
    Data curation, Project administration
    Competing interests
    P.B. is an employee of IMMAGINA BioTechnology S.r.l.

  11. Jacopo Zasso

    Laboratory of Translational Genomics, Department of Cellular, Computational and Integrative Biology (CIBIO), University of Trento, Trento, Italy
    Contribution
    Visualization, Methodology
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-3151-6443

  12. Alessandro Quattrone

    Laboratory of Translational Genomics, Department of Cellular, Computational and Integrative Biology (CIBIO), University of Trento, Trento, Italy
    Contribution
    Methodology
    Competing interests
    A.Q. is a shareholder of IMMAGINA BioTechnology S.r.l

  13. Graziano Guella

    Department of Physics, University of Trento, Trento, Italy
    Contribution
    Writing - review and editing
    Competing interests
    G.G. is shareholders of IMMAGINA BioTechnology S.r.l

  14. Fabio Lauria

    Institute of Biophysics, CNR Unit at Trento, Trento, Italy
    Contribution
    Formal analysis
    Competing interests
    No competing interests declared

  15. Gabriella Viero

    Institute of Biophysics, CNR Unit at Trento, Trento, Italy
    Contribution
    Conceptualization, Data curation, Formal analysis, Supervision, Methodology, Writing - review and editing
    Competing interests
    G.V. is a scientific advisor of IMMAGINA BioTechnology S.r.l.
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-6755-285X

  16. Massimiliano Clamer

    IMMAGINA BioTechnology, Trento, Italy
    Contribution
    Conceptualization, Resources, Data curation, Formal analysis, Supervision, Funding acquisition, Validation, Investigation, Visualization, Methodology, Writing - original draft, Project administration, Writing - review and editing
    For correspondence
    mclamer@immaginabiotech.com
    Competing interests
    M.C. is the founder of, director of, and a shareholder in IMMAGINA BioTechnology S.r.l., a company engaged in the development of new technologies for gene expression analysis at the ribosomal level.
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-8185-059X

Funding

Autonomous Province of Trento (LP6/99)

  • Luca Minati
  • Claudia Firrito
  • Alessia Del Piano
  • Alberto Peretti
  • Paola Bernabo
  • Massimiliano Clamer

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Senior Editor

  1. James L Manley, Columbia University, United States

Reviewing Editor

  1. Howard Y Chang, Stanford University, United States

Reviewer

  1. John L Rinn, University of Colorado at Boulder, United States

Version history

  1. Received: May 25, 2020
  2. Accepted: February 16, 2021
  3. Accepted Manuscript published: February 17, 2021 (version 1)
  4. Version of Record published: March 4, 2021 (version 2)

Copyright

© 2021, Minati et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 2,299
    Page views
  • 399
    Downloads
  • 9
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

  1. Luca Minati
  2. Claudia Firrito
  3. Alessia Del Piano
  4. Alberto Peretti
  5. Simone Sidoli
  6. Daniele Peroni
  7. Romina Belli
  8. Francesco Gandolfi
  9. Alessandro Romanel
  10. Paola Bernabo
  11. Jacopo Zasso
  12. Alessandro Quattrone
  13. Graziano Guella
  14. Fabio Lauria
  15. Gabriella Viero
  16. Massimiliano Clamer
(2021)
One-shot analysis of translated mammalian lncRNAs with AHARIBO
eLife 10:e59303.
https://doi.org/10.7554/eLife.59303

Categories and tags

Research organisms

Further reading

    1. Cell Biology
    2. Genetics and Genomics

    Regulation of defective mitochondrial DNA accumulation and transmission in C. elegans by the programmed cell death and aging pathways

    Sagen Flowers, Rushali Kothari ... Joel H Rothman
    Research Article

    The heteroplasmic state of eukaryotic cells allows for cryptic accumulation of defective mitochondrial genomes (mtDNA). ‘Purifying selection’ mechanisms operate to remove such dysfunctional mtDNAs. We found that activators of programmed cell death (PCD), including the CED-3 and CSP-1 caspases, the BH3-only protein CED-13, and PCD corpse engulfment factors, are required in C. elegans to attenuate germline abundance of a 3.1-kb mtDNA deletion mutation, uaDf5, which is normally stably maintained in heteroplasmy with wildtype mtDNA. In contrast, removal of CED-4/Apaf1 or a mutation in the CED-4-interacting prodomain of CED-3, do not increase accumulation of the defective mtDNA, suggesting induction of a non-canonical germline PCD mechanism or non-apoptotic action of the CED-13/caspase axis. We also found that the abundance of germline mtDNAuaDf5 reproducibly increases with age of the mothers. This effect is transmitted to the offspring of mothers, with only partial intergenerational removal of the defective mtDNA. In mutants with elevated mtDNAuaDf5 levels, this removal is enhanced in older mothers, suggesting an age-dependent mechanism of mtDNA quality control. Indeed, we found that both steady-state and age-dependent accumulation rates of uaDf5 are markedly decreased in long-lived, and increased in short-lived, mutants. These findings reveal that regulators of both PCD and the aging program are required for germline mtDNA quality control and its intergenerational transmission.

    1. Biochemistry and Chemical Biology
    2. Cell Biology

    Genetically engineered mesenchymal stem cells as a nitric oxide reservoir for acute kidney injury therapy

    Haoyan Huang, Meng Qian ... Zongjin Li
    Research Article Updated

    Nitric oxide (NO), as a gaseous therapeutic agent, shows great potential for the treatment of many kinds of diseases. Although various NO delivery systems have emerged, the immunogenicity and long-term toxicity of artificial carriers hinder the potential clinical translation of these gas therapeutics. Mesenchymal stem cells (MSCs), with the capacities of self-renewal, differentiation, and low immunogenicity, have been used as living carriers. However, MSCs as gaseous signaling molecule (GSM) carriers have not been reported. In this study, human MSCs were genetically modified to produce mutant β-galactosidase (β-GALH363A). Furthermore, a new NO prodrug, 6-methyl-galactose-benzyl-oxy NONOate (MGP), was designed. MGP can enter cells and selectively trigger NO release from genetically engineered MSCs (eMSCs) in the presence of β-GALH363A. Moreover, our results revealed that eMSCs can release NO when MGP is systemically administered in a mouse model of acute kidney injury (AKI), which can achieve NO release in a precise spatiotemporal manner and augment the therapeutic efficiency of MSCs. This eMSC and NO prodrug system provides a unique and tunable platform for GSM delivery and holds promise for regenerative therapy by enhancing the therapeutic efficiency of stem cells.

    1. Biochemistry and Chemical Biology
    2. Cell Biology

    The FAM104 proteins VCF1/2 promote the nuclear localization of p97/VCP

    Maria Körner, Susanne R Meyer ... Alexander Buchberger
    Research Article Updated

    The ATPase p97 (also known as VCP, Cdc48) has crucial functions in a variety of important cellular processes such as protein quality control, organellar homeostasis, and DNA damage repair, and its de-regulation is linked to neuromuscular diseases and cancer. p97 is tightly controlled by numerous regulatory cofactors, but the full range and function of the p97–cofactor network is unknown. Here, we identify the hitherto uncharacterized FAM104 proteins as a conserved family of p97 interactors. The two human family members VCP nuclear cofactor family member 1 and 2 (VCF1/2) bind p97 directly via a novel, alpha-helical motif and associate with p97-UFD1-NPL4 and p97-UBXN2B complexes in cells. VCF1/2 localize to the nucleus and promote the nuclear import of p97. Loss of VCF1/2 results in reduced nuclear p97 levels, slow growth, and hypersensitivity to chemical inhibition of p97 in the absence and presence of DNA damage, suggesting that FAM104 proteins are critical regulators of nuclear p97 functions.