Transcription Impacts the Efficiency of mRNA Translation via Co-transcriptional N6-adenosine Methylation

A Reporter Vector System to Examine the Transcription-Translation Relationship

To examine the relationship between transcription and translation, we set out to determine the effect of different human promoters on the translation of a reporter gene (Renilla luciferase [Rluc]). For this purpose, we defined promoters as 0.5- to 2.5-Kb-long regions characterized by high H3K4Me3 and low H3K4Me1 epigenetic marks upstream of transcriptional start sites (TSSs), supported by RNA sequencing (RNA-seq) data of MCF7 cells (Loayza-Puch et al., 2013). To make our screen versatile and diverse, we cloned promoters from genes connected to stress response, autophagy, ER metabolism, and metastasis, as well as multiple transcription factors. While cloning the promoter sequences, we avoided regions extending downstream the respective TSSs, as these might result in the inclusion of additional 5′ UTRs into the reporter transcripts, potentially complicating the interpretation of results. Because each of the cloned promoters drives the expression of the same reporter gene, we affiliated every promoter with a unique 10 nt barcode (cloned in the 3′ UTR of Rluc) in order to follow its expression in a pool of Rluc mRNAs (Figure S1A). Following these guidelines, we cloned 135 human promoters (Table S1) in order to create a library named Pro-Lib, where most of the promoters were associated with two or more different barcodes to provide higher experimental confidence. As expected, cloned promoter regions substantially induced Rluc expression (Figure S1B). For normalization of expression, Pro-Lib included an additional reporter gene, Firefly luciferase (Fluc), used as an inner control (Figure S1A). Last, we employed an Flp-FRT recombination system and used competent Flp-In MCF7 cells to stably integrate a single copy of a Pro-Lib vector per cell in the identical genetic locus in order to avoid any possible influence of different chromatin neighborhoods on the reporter gene expression.

After establishing the library in a stable population, we performed a polysomal profiling experiment by using sucrose gradients, a classical method to separate mRNAs according to the amount of bound ribosomes. Since we examined only the barcoded Rluc mRNAs, their relative segregation in the gradient indicates ribosome density and TE estimation. We named the whole procedure barcoded polysomal profiling, or BPP (Figure 1A). Relative enrichment of Pro-Lib barcodes in the various fractions of the gradient showed that most Rluc mRNAs are localized in the initial polysomal fractions (i.e., 9–12; Figure 1B). Control total RNA segregation showed a characteristic pattern of polysomes and EDTA sensitivity, in line with the known dependence of polysomes on the availability of Mg²⁺ ions (Figures S1C and S1D).

Next, we tested whether BPP can detect changes in TE. For this purpose, we employed Rluc mRNA containing 5′-terminal oligopyrimidine tracts (5′ TOP) derived from RPL37a. Translation of mRNAs possessing 5′ TOP is highly dependent on mTOR activity, resulting in a rapid translational arrest after mTOR inhibition, as compared to other mRNAs (Thoreen et al., 2012). Indeed, inhibition of mTORC1 resulted in a global moderate shift of the Rluc mRNAs to fractions 9–11 (e.g., SPTBN1 promoter), whereas a 5′ TOP-containing transcript was depleted from the translated fractions of the gradient (Figure 1C). Interestingly, we also identified several additional Rluc transcripts exhibiting similar hypersensitivity toward the inhibition of mTORC1 (Figure S1E). Further analysis of the sequences adjacent to their TSSs (Figure S1F) identified stretches of pyrimidines that could serve as 5′ TOP signals, providing a probable explanation for their dramatic response. To further test detection capabilities of BPP, we examined the effect of splicing on TE. Indeed, we observed that spliced Rluc mRNA shifts to heavier polysomal fractions and yields more protein (Figure 1D), as expected from the known positive effect of splicing on translation (Nott et al., 2004). Altogether, our control experiments demonstrate that BPP is capable of examining multiple Rluc mRNAs in bulk and detecting changes in TE of individual transcripts.

TATA Box Confers Higher TE

Inspecting the segregation of the barcoded Rluc mRNAs on sucrose gradients, we identified 12 promoters that caused a shift of the reporter transcript toward higher ribosomal occupancy fractions in at least two independent experiments. Intriguingly, four of these promoters contained a TATA-box element (Figure 2A) and four others had TA-rich sequences that could potentially serve as non-canonical TATA boxes (Figure S2A), indicating that this promoter element could positively influence translation. To test this possibility, we supplied several TATA-less promoters with an artificial TATA element (consisting of the TATA sequence followed by the short downstream sequence derived from human ACTB). In all cases, this manipulation resulted in Rluc transcripts occupying denser fractions of the gradient (i.e., 11–13; Figures 2B and S2B), supporting the previous observations. Taking the ASNSD1 promoter as a model, we observed a positive effect of TATA addition on TE under various conditions (Figure S2C), thus indicating a robust phenomenon.

Because artificial introduction of TATA may alter the TSS, possibly impacting translational capacity (Rojas-Duran and Gilbert, 2012), we investigated in detail the 5′ UTRs produced from the TATA-containing and TATA-less promoter pairs. By northern blotting, we observed that most of the tested promoters resulted in reporter transcripts of a similar length, with a noticeable enhancement of mRNA levels in TATA-containing promoters (Figure 2C). To establish precisely the 5′ ends of the transcripts, we performed 5′ rapid amplification of cDNA ends (RACE) analyses, which showed a very narrow peak ∼25 nt downstream of the inserted TATA element (Figures S3A and S3B). To test whether these alterations in the 5′ UTRs could explain the observed changes in translation, we in vitro synthesized Rluc transcripts bearing the different 5′ UTRs and examined their relative TE (Figure S3C). Given that the 5′ UTR generated after the insertion of TATA sequence did not confer higher TE, we conclude that the observed positive effect of TATA on translation is unlikely to stem from the differences in the 5′ UTRs of Rluc mRNAs.

So far, we have inferred TE from ribosome occupancy measurements reflected by migration within sucrose gradients. Next, we tested TE changes by measuring separately the levels of the reporter mRNA and protein in the promoter pairs. As expected, we observed significantly more mRNA produced from TATA-containing promoters (7- to 9-fold, Figure 2D), consistent with the levels measured by northern blotting (Figure 2C). In contrast, the increase in the level of protein activity was significantly higher (>20-fold), supporting the connection between the presence of TATA in promoter and enhanced protein production. To further test the role of the TATA element in translation, we mutagenized it within the SV40 promoter. Remarkably, loss of TATA reduced the reporter mRNA expression by ∼14-fold and protein activity by ∼80-fold (Figure 2E), suggesting reduced TE. Indeed, mRNA produced by a TATA-mutated promoter was enriched in the lighter polysomal fractions (Figure 2F), further supporting this indication. We observed similar results upon mutagenesis of TATA in other promoters (Figures S2D and S2E), suggesting that this effect is not restricted to any particular promoter.

Next, we tested whether the positive effect of TATA on TE also applies to the endogenous mammalian gene expression. For this purpose, we chose c-Myc, a ubiquitously expressed gene with a single active TATA-positive promoter in MCF7 cells (Figure S2F) and used the CRISPR-Cas9 system to alter its TATA sequence (Figure 2G). This strategy yielded isolated cell clones with disrupted c-Myc TATA on most of their alleles (Figure 2G and Figure S2G). Examination of c-Myc expression in these clones revealed a reduction of ∼25% in c-Myc mRNA and ∼50% in both protein activity and expression (Figures 2H and 2I and S2I and S2J). Moreover, polysomal profiling of c-Myc transcripts showed reduced ribosomal occupancy upon mutation of the TATA element (Figures 2J and S2H), whereas a control transcript displayed very similar profiles in both clones (Figure S2K), indicating a c-Myc-specific effect. Thus, we conclude that the presence of the TATA element in a promoter enhances the efficiency of mRNA translation and note the validity of this observation to multiple promoters and genes.

General Association, but Not Causal Link, between mRNA Levels and TE

In all the cases we studied, presence of TATA in a promoter stimulated both mRNA expression levels and TE. To examine whether TE positively correlates with mRNA expression levels, we clustered the Pro-Lib BPP data to separate the transcripts of our library into two relative groups, one with lower TE (Figures S4A and S4B) and another with higher TE (Figures S4C and S4D), and compared the expression levels between these two groups. Indeed, we found that transcripts with higher TE tend to be more abundant (Figure 3A). To test whether this coupling is TATA dependent, we repeated this analysis while omitting the promoters with artificially added TATA elements. Notably, this analysis yielded similar albeit less significant results (Figure 3B), thus suggesting a positive correlation between the abundance of Rluc transcripts and their TE, with no dependency on TATA. To further test this observation, we employed cells expressing an inducible version of the barcoded reporter gene (TRex-Rluc), in which the levels of Rluc mRNA are stimulated ∼17-fold after induction (Figure 3C). Importantly, induction of this gene yielded a significantly greater enhancement of protein activity (∼60-fold, Figure 3C), suggesting a translational boost. Indeed, subsequent BPP analysis revealed a strong increase in the ribosome occupancy of the induced reporter mRNA (Figure 3D and Figure S5A), supporting the notion that increased expression results in higher TE. Notably, we performed 5′ RACE analysis of both induced and non-induced TRex-Rluc mRNAs and found the scattering of TSSs to be very similar (Figure S3D). Altogether, we conclude that the observed link between the expression levels of mRNAs and their TE is not restricted to the TATA element and might therefore represent a more general phenomenon.

To test whether genome-wide data support a global relationship between mRNA level and TE, we analyzed pairs of RNA-seq and Ribo-seq (ribosomal footprinting sequencing) datasets from multiple human cell lines. Intriguingly, we observed a positive global correlation between mRNA expression level and TE, which was rather weak but consistent and statistically significant (Figure 3E). Moreover, this correlation between mRNA levels and TE was also apparent in various stress conditions (Figures S5B and S5C). These results indicate that there is a positive correlation between expression levels of mRNAs and their TE in mammalian genomes as well.

To examine whether mRNA levels could directly regulate TE, we transfected different amounts of in vitro transcribed and purified Rluc mRNA into MCF7 cells and monitored the activity of the produced Rluc protein. We anticipated that if levels of mRNA positively stimulated TE, transfecting increasing amounts of mRNA would result in an exponential increase of protein production. However, we observed a clear linear dependency between the two parameters (Figure 3F), suggesting indifference of the protein production rates to mRNA abundance. To test this conclusion further, we employed lentiviral-mediated stable integration of a transcriptional unit including a promoter and Rluc gene and generated two stable populations with either nearly single or multiple integrations of the same transcriptional unit per cell, resulting in an ∼9-fold difference in Rluc mRNA levels (Figure 3G). Also here, increase in protein activity was similar to the increase in mRNA levels, indicating comparable TEs. Supporting this conclusion, Rluc mRNAs derived from both cell populations exhibited similar segregation patterns in sucrose gradients (Figure 3H). Altogether, these results suggest that while mRNA expression levels positively correlate with TE in multiple cellular models, mRNA abundance does not directly regulate it.

Rate of Transcription Positively Affects TE

Next we searched for the causal origin of the observed correlation between mRNA levels and TE. We considered the rate of transcription to be the most likely candidate since it directly regulates the expression levels of mRNAs. To test this possibility, we correlated transcription rates, as estimated by global run-on sequencing (GRO-seq) (Core et al., 2008), and TE, as determined by Ribo-seq and RNA-seq data, in BJ and MCF7 cell lines. In both cell types, we observed a significant positive association between the rates of transcription and TE of genes that do not possess upstream ORFs (Figure 4A and Figures S6A and S6B). This correlation is notable given that this analysis integrates distinct datasets independently obtained from three genomic techniques in two different cellular systems.

To further investigate the relationship between rates of transcription and TE, we employed camptothecin (CPT), a chemical compound that inhibits topoisomerase I and, in mild concentrations, slows down the progression of transcribing RNA polymerase II (RNAPII) (Dujardin et al., 2014). Indeed, when CPT was applied in parallel to the induction of TRex-Rluc gene, we observed a reduction of ∼60% of Rluc mRNA, but a more prominent reduction of Rluc activity of ∼80% (Figure 4B). This difference could be explained by the reduced ribosome occupancy, as reflected by polysomal profiling analyses (Figure 4C and Figure S5D). Thus, impediment of RNAPII progression does not only reduce the transcriptional rate of TRex-Rluc but also attenuates its TE. To further test the role of RNAPII dynamics on translation, we assessed TE of multiple genes in CPT-treated cells. Overall, CPT treatment resulted in TE reduction of approximately 700 genes by more than 1.5-fold in two independent experiments (Table S2). Given that highly transcribed genes showed higher TE (Figure 4A), we speculated that CPT treatment would cause a more pronounced repressive effect on the translation of these genes. Indeed, intersection with GRO-seq data demonstrated that impediment of RNAPII caused a significant reduction in TE of genes with a relatively high transcription rate (Figure 4D). We further examined the polysomal segregation of several mRNAs, suggested by the genome-wide experiments, and validated that these mRNAs indeed displayed reduced TE coupled with reduced expression, as compared to ALG8 control mRNA (Figure 4E). To eliminate the possibility that the transcripts exhibiting translational shifts are truncated due to the CPT treatment, we re-examined these and other mRNAs by using primers annealing to 3′ UTRs only. Also in this case, the shift of mRNAs was apparent (Figure S6C). Taken together, we conclude that transcription rate is an important positive determinant of TE.

Transcription-Dependent Translation Regulation Requires Canonical Translation Initiation

To further examine the mechanism by which transcription affects TE, we introduced an internal ribosome entry site (IRES) sequence into the inducible TRex-Rluc cassette (TRex-IRES-Rluc, see schematics in Figure 4F) to bypass the canonical 5′ cap-mediated initiation of translation. Interestingly, induction of TRex-IRES-Rluc resulted in Rluc activity that mirrored the changes in the mRNA levels (i.e., ∼12- to 15-fold induction at both levels, Figure 4F), which contrasted the previously observed induction of the Rluc protein activity (>60-fold, Figure 3C). Moreover, BPP analyses revealed similar segregation of both induced and non-induced IRES-Rluc mRNAs (Figure 4G), further suggesting lack of translational induction. Furthermore, introduction of IRES into constructs containing the artificial TATA element (e.g., SZT2 promoter) abolished the previously observed enhancement of TE (Figure 4H, compare with Figure 2D) and resembled the effect of TATA mutagenesis (Figure S2D). Thus, we conclude that canonical translation initiation is required for transcription-dependent enhancement of translation.

m⁶A Modification Mediates Transcription-Responsive Translation

Next, we looked for a molecular event that could explain the link between transcription and translation. Since changes in transcription affected translation of multiple genes (Table S2), we reasoned that this molecular event should be robust and widespread. m⁶A is an abundant nuclear dynamic mRNA modification identified in thousands of human transcripts (Dominissini et al., 2012, Meyer et al., 2012), recently implicated in translational regulation (Wang et al., 2015). Therefore, we asked whether m⁶A could bridge transcription and translation.

First, we examined whether m⁶A levels could be affected by different rates of transcription. We therefore performed immunoprecipitations of m⁶A-modified RNAs (MeRIP) from cells with either induced or non-induced TRex-Rluc and compared the relative recoveries of the different Rluc transcripts. Strikingly, we observed very significant differences between the two Rluc populations; the non-induced mRNA was isolated much more efficiently in all tested mRNA regions (Figure 5A). Importantly, a control mRNA (AHNAK) showed similar recovery in both samples, suggesting a specific enrichment of the non-induced Rluc transcripts by MeRIP. We also assessed the methylation levels of Rluc mRNAs of the two populations expressing different copy numbers of Lenti-Rluc (Figure 3G) and found them to be very similar (Figure S6D), suggesting that the differences observed in MeRIP are unlikely to stem from unequal mRNA expression levels. Furthermore, we observed differences in the post-MeRIP recovery of Rluc mRNAs transcribed from either TATA-containing or TATA-less promoters (Figure S6E), suggesting that promoters with intact TATA elements yield mRNAs with relatively lower m⁶A content. Next, we in vitro transcribed Rluc in m⁶A-compatible nuclear HeLa extract (Shimba et al., 1995), using conditions that reduce rate of transcription (e.g., elevated MgCl₂ [Wildeman et al., 1984] or a template with a mutated TATA box), and tested the different transcripts for m⁶A content. Indeed, both changes resulted in the relatively lower levels of Rluc mRNA, which were associated with higher deposition of m⁶A (Figure 5B). Importantly, both the UTRs and polyA tails of the different transcripts were similar (Figures S3E and S3F). Furthermore, treatment with CPT reduced mRNA levels in parallel to the relative enhancement of their m⁶A content (Figure 5C), suggesting a negative correlation between transcription elongation and m⁶A deposition. Additionally, we performed MeRIP analysis of genes that exhibited reduction of both mRNA levels and TE after CPT treatment (Figure 4E). These genes indeed displayed lower levels of mRNA coupled with enhanced m⁶A content (Figure 5D). To further address the link between the rate of transcription and m⁶A deposition in an additional independent system, we interrogated α-amanitin resistant RNAPII mutants exhibiting either normal (N792D) or slow (C4/R749H) rates of elongation (Fong et al., 2014). Similarly to the previous observations, we found that cells expressing the slow RNAPII mutant contained relatively lower levels of mRNA, associated with elevated methylation content (Figure 5E). Thus, various experimental approaches indicate that a suboptimal rate of transcription results in higher m⁶A deposition on mRNAs.

This conclusion raised the possibility that m⁶A modification of mRNAs occurs co-transcriptionally, and therefore we anticipated a physical interaction between RNAPII and the methyltransferase complex (MTC). To test this, we immunoprecipitated RNAPII from either normally growing or CPT-treated cells and probed the eluate for METTL3, a catalytic core of MTC (Wang et al., 2016). Remarkably, whereas in untreated conditions we observed no visible interaction, upon CPT treatment, METTL3 was efficiently co-precipitated (Figure 5F), suggesting co-transcriptional m⁶A modification and enhancement of the interaction between RNAPII and MTC upon impediment of RNAPII elongation dynamics.

Next, we tested whether elevated levels of m⁶A could negatively affect translation. We first examined the translation capacity of the in vitro transcribed Rluc mRNAs in the conditions of reduced transcription and enhanced m⁶A (as presented in Figure 5B). In line with our hypothesis, conditions associated with attenuated transcription resulted in lower efficiency of protein production compared with the normal transcription conditions (Figure 6A). Next, we examined the global effects of CPT treatment and expression of the slow RNAPII mutant on polysomes. Figures 6B and 6C show that these two treatments considerably reduced the RNA content of the polysomal fractions, suggesting a general reduction of translation and providing further evidence for the positive feedback between transcription and translation.

To directly examine the role of m⁶A in translation, we knocked down two genes in the m⁶A pathway (METTL14 and YTHDH1, Figure S6F) and examined the effect on the expression of Rluc mRNA. Interestingly, knocking down either factor resulted in a significant boost in the expression of the non-induced Rluc, whereas the effect on the induced gene was moderate (Figure 6D). Importantly, while global polysome segregation analysis indicated only a slight effect of each knockdown on general translation (Figure S6G), possibly due to compensatory mechanisms, BPP assay verified a considerable increase in the TE of the non-induced Rluc mRNA in these conditions (Figure 6E). Given that methylation of coding regions (CDRs) has been recently suggested to attenuate translation (Choi et al., 2016), we re-examined the experiment presented in Figure 5A, in order to compare the relative changes in methylation of the different Rluc transcript regions. Remarkably, the highest gain of methylation upon transcriptional attenuation was observed throughout the CDR of Rluc (up to ∼11% near the end of the non-induced mRNA CDS), whereas both UTRs (i.e., 5′ and 3′) displayed the lowest levels (<1%, Figure 6F). This result therefore suggests a plausible explanation for the strong inhibitory effect of transcriptional repression on TE that we observed (Figure 3D). Lastly, we sought to directly examine the effect of m⁶A levels on protein production and produced Rluc mRNAs in vitro with different proportions of methylated adenosines. Strikingly, increased incorporation of m⁶A led to a progressive attenuation of the TE of Rluc (Figure 6H), suggesting an overall inhibitory effect of m⁶A modification on translation. Altogether, these results demonstrate that covalent m⁶A modification of mRNAs depends on the dynamics of the transcribing RNAPII, and tends to negatively affect TE. Therefore, we conclude that m⁶A links transcription and translation.

Key Resources Table

Contact for Reagent and Resource Sharing

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Reuven Agami (r.agami@nki.nl).

Experimental Model and Subject Details

Human breast adenocarcinoma MCF7 cells (female) were purchased from ATCC and authenticated by detecting the deficiency of caspase-3 using western blotting. These cells were grown in DMEM media (GIBCO) supplemented with 10% bovine serum and penicillin/streptomycin (GIBCO) in 5%CO₂-buffered incubators at 37°C. Cells were split twice per week and kept in culture for up to 8 weeks. To create cells compatible with FLP recombination system (ThermoFisher scientific), MCF7 cells were transfected with the pFRT/lacZeo and grown in the presence of Zeocin (400μg/ml) until stable colonies appeared. These cells were named MCF7/FRT and used for integration of Pro-Lib. To integrate Pro-Lib, these cells were transfected with Pro-Lib vectors along with pOG44 vector (ThermoFisher scientific) in quantities of 1.6μg+3μg respectively, per 35-mm well. After 48h, the cells were collected, re-plated onto 10-cm dishes, and selected for three weeks in the presence of Hygromycin (0.1 mg/ml f.c.). After appearance of the stable colonies, the cells were collected, frozen, and used for further experiments. To create cells compatible with T-REx inducible system (Invitrogen), MCF7/FRT cells were transfected with pcDNA6/TR vector encoding for the tetracycline-sensitive repressor and separate clones were selected for their resistance to Blasticidin S (5mg/ml). The induction efficiency was examined in separate clones by transient transfection with pcDNA4/TO-Rluc construct, incubation with Doxycycline (Sigma, 1 μg/ml final conc.) for 17 hr and measurement of Renilla protein activity. One suitable clone was chosen for further experiments and the cells were named MCF7/FRT/TR. These cells were stably transfected with pcDNA5/FRT/TO plasmid encoding for barcoded Renilla ORF (TRex-Rluc) in the presence of pOG44 vector (ThermoFisher scientific) and grown for three weeks in the presence of Hygromycin (0.1 mg/ml f.c.). Individual colonies were isolated and tested by treatment with Doxycycline (Sigma, 1 μg/ml final conc.). 293 cells bearing integrated alpha-amanitin resistant mutants of Rpb1(“wild-type” N792D, and “slow” C4/R749H), were a kind gift of D. Bentley. They were maintained in DMEM supplemented with 10% FBS, 1% penicillin/streptomycin, 200 μg/mL hygromycin, and 6.5 μg/mL blasticidin. To stimulate the expression of Rpb1 mutants, cells were induced with 2.0 μg/mL doxycycline for 16 hr and treated with 2.5 μg/mL α-amanitin for additional 42 hr.

Method Details

Data and Software Availability

The deep sequencing datasets generated in this study have been deposited in the GEO database under accession number GEO: GSE96643.