Specific tRNAs promote mRNA decay by recruiting the CCR4-NOT complex to translating ribosomes

The ribosomal P-site codon is a major determinant of CNOT3 recruitment to human ribosomes

We first examined whether human CNOT3 associates with translating ribosomes using sucrose density gradient ultracentrifugation of HEK293T cell lysates. Similar to the behavior of Not5, the CNOT3 ortholog in Saccharomyces cerevisiae (14), endogenous CNOT3, or stably-expressed Flag-tagged CNOT3, co-sedimented with higher order polysomes in human cells (Fig. 1A and fig. S1A). Polysome association of CNOT3 could be due to a direct interaction with translating ribosomes or could occur as a consequence of indirect recruitment to translated mRNAs via RBPs or miRNAs. To distinguish between these possibilities, lysates were treated with RNase prior to sucrose gradient fractionation and CNOT3 was immunoprecipitated from the resulting monosome fraction (Fig. 1, B and C and fig. S1, B and C). CNOT3 co-sedimented with monosomes and stably associated with large and small ribosomal subunit proteins, consistent with direct binding of CNOT3 to ribosomes.

We next performed selective ribosome profiling (27) to identify the features of mRNAs that were associated with CNOT3-bound ribosomes. Monosomes were prepared by RNase treatment and sucrose gradient fractionation, followed by immunoprecipitation (IP) of CNOT3 (Fig. 1D). Sequencing of the enriched ribosomal footprints confirmed the expected triplet periodicity within open reading frames (ORFs) (Fig. 1E and fig. S1, D and E), consistent with the stepwise translation of each codon and enabling assignment of the A-, P-, and E-site codons in each footprint.

In Saccharomyces cerevisiae, the Ccr4-Not complex preferentially associates with ribosomes with non-optimal codons in the A site (14). Codon optimality can be quantified using the tRNA adaptation index (tAI), a metric of cognate tRNA abundance (28). Re-analysis of published selective ribosome profiling data from yeast (14) confirmed that non-optimal codons with a low tAI (29) are highly enriched in the A-site of Ccr4-Not-bound ribosomes (fig. S1, F to H). In contrast, we observed no preferential enrichment of non-optimal codons in the A-site of CNOT3-bound human ribosomes and, accordingly, no correlation between tAI (20) and A-site codon enrichment (fig. S1, I to K). Similarly, codons with a low codon stability coefficient (CSC), a metric derived from global mRNA decay rate measurements which reflects whether a given codon has a positive or negative effect on overall mRNA stability (21), were not enriched in the A-site of CNOT3-bound ribosomes (fig. S1, L and M). In addition to tRNA availability, the rate of decoding is also impacted by additional parameters, such as amino acid concentration and tRNA charging (30). The relative decoding rate, which accounts for these additional variables, is approximated by the A-site dwell time of each codon, a parameter that can be estimated using ribosome profiling data (25). We observed a modest, but statistically significant, correlation between A-site dwell time and codon enrichment in the A-site of CNOT3-bound ribosomes (fig. S1N). Nevertheless, no specific A-site codons were strongly enriched in our dataset, with only a single codon exhibiting greater than two-fold enrichment (Fig. 1F). Overall, these data suggest that slow decoding of the A-site codon measurably promotes CNOT3 recruitment to translating ribosomes, but also raise the possibility that other determinants play a more dominant role in co-translational CNOT3 recruitment in mammalian cells.

To more broadly explore a possible connection between codon content and CNOT3 recruitment, we examined codon enrichment in the ribosomal P- and E-sites in our selective ribosome profiling data (Fig. 1F). This revealed that the most enriched codons in CNOT3-bound ribosomes were located in the P-site. In particular, select P-site arginine codons (CGG, CGA, and AGG) were the most enriched codons observed in any position within CNOT3-bound ribosomes, while other arginine codons (CGC, AGA, and CGU) exhibited weak enrichment. Analysis of the amino acids encoded by CNOT3-bound ribosomal footprints further documented a robust enrichment of arginine-centered tripeptides (Fig. 1G), with 18 of the top 20 enriched tripeptides containing arginine at the P-site (fig. S1O). Altogether, these data demonstrate that P-site codon identity is strongly associated with co-translational CNOT3 recruitment in human cells, with select arginine codons at the P-site providing the strongest detectable signal for CNOT3 binding.

CGG, CGA, and AGG arginine codons promote CNOT3-mediated mRNA decay

To examine the relationship between codon enrichment in CNOT3-bound ribosomes and mRNA decay, we used SLAM-seq (31) to globally measure mRNA half-lives in control and CNOT3-deficient HEK293T cells (fig. S2A). As reported (32), loss of CNOT3 resulted in reduced fitness (fig. S2B). Because P-site CGG, CGA, and AGG arginine codons were most strongly associated with recruitment of CNOT3 to translating ribosomes, we stratified mRNAs by calculating a weighted CGG/CGA/AGG score. Each of these codons was assigned a value equal to its enrichment in the P-site of CNOT3-bound ribosomes (Fig. 1F). The score for each mRNA was defined as the sum of the weighted values of these codons, normalized to the total number of codons, in the ORF. Transcripts with high weighted CGG/CGA/AGG scores were preferentially stabilized upon loss of CNOT3 (Fig. 2A). Transcripts rich in other arginine codons (CGC, AGA, and CGU) did not exhibit this behavior (Fig. 2A). Analysis of mRNA decay rates in CNOT3-depleted Jurkat cells (33) or steady-state mRNA levels in pro-B cells from Cnot3 knockout mice (34) further confirmed that transcripts rich in CGG, CGA, or AGG arginine codons, but not arginine encoded by CGC, AGA, or CGU, were preferentially stabilized in CNOT3-deficient cells (fig. S2, C and D).

To further assess whether select arginine codons are sufficient to promote CNOT3-mediated mRNA decay, we constructed a doxycycline-regulated reporter transcript encoding 42 tripeptides, each centered on a CGG, CGA, or AGG arginine codon (Fig. 2B). A control mRNA, with each arginine codon replaced with a codon that was not enriched in the P-site of CNOT3-bound ribosomes, was also generated. Consistent with our transcriptome-wide analyses, the arginine-encoding reporter decayed significantly faster than the control reporter and was selectively stabilized upon CNOT3 depletion. Together, these data provide strong evidence that CNOT3 recruitment to translating ribosomes by CGG, CGA, and AGG arginine codons results in accelerated mRNA degradation.

Mitochondrial ribosomal protein mRNAs are enriched in CGG/CGA/AGG codons and regulated by CNOT3

We next investigated the endogenous mRNAs that are most strongly regulated by CNOT3 due to the presence of destabilizing arginine codons. Gene set enrichment analysis (GSEA) (35) of mRNA decay rate data from HEK293T and Jurkat cells, as well as steady-state mRNA levels from Cnot3 knockout pro-B cells, revealed that genesets containing mitochondrial ribosomal proteins were highly upregulated upon CNOT3 depletion and were the only significantly upregulated genesets detected in all three datasets (Fig. 2C and fig. S2, E to G). To determine whether these effects were related to an enrichment of destabilizing arginine codons, we ranked genes according to their weighted CGG/CGA/AGG scores. Indeed, genesets containing mitochondrial ribosomal proteins were the most highly enriched genesets according to this metric (Fig. 2D and fig. S2H).

These findings indicate that mitochondrial ribosomal proteins are regulated by CNOT3. We confirmed that depletion of CNOT3 in HEK293T or Jurkat cells resulted in a strong increase in the steady-state abundance of these transcripts (Fig. 2, E and F) and a corresponding increase in mitochondrial mass (Fig. 2, G and H). Furthermore, fluorescent labeling of nascent mitochondrial peptides demonstrated an increase in mitochondrial translation (Fig. 2, I and J). Depletion of CNOT1, the scaffolding subunit of the CCR4-NOT complex, resulted in a similar increase in the steady-state abundance of mitochondrial ribosomal protein mRNAs (fig. S2, I and J). Thus, regulation of mitochondrial ribosomal proteins, which are rich in arginines encoded by CGG, CGA, and AGG codons, by CCR4-NOT impacts mitochondrial homeostasis in mammalian cells.

Structural analysis of co-translational CNOT3 recruitment by P-site arginine codons

To further corroborate our finding that select P-site arginine codons stimulate co-translational CNOT3 recruitment and to investigate the underlying mechanism, we established an efficient human in vitro translation system from HEK293T lysates. Endogenous mRNAs were removed by micrococcal nuclease, enabling assembly of polysomes on exogenously added mRNAs of any desired sequence (fig. S3A). To determine whether co-translational CNOT3 recruitment was recapitulated in this system, we constructed an mRNA encoding 41 repeats of leucine-arginine-aspartic acid (LRD), the most enriched tripeptide in CNOT3-bound ribosomes (fig. S1O), with arginine encoded by CGG (41×LR_CGGD). As a control, the arginine codon was replaced with a lysine codon (41×LK_AAGD). As predicted from our selective ribosome profiling data, in vitro translation of 41×LR_CGGD, but not 41×LK_AAGD, resulted in robust recruitment of Flag-tagged CNOT3 to polysomes (Fig. 3A). Other CCR4-NOT complex components were also selectively recruited to the 41×LR_CGGD mRNA (Fig. 3B). Moreover, the hierarchy of P-site arginine codon-mediated recruitment of CNOT3 was recapitulated in this system, as 41×LR_CGGD recruited the most CNOT3 to polysomes, followed by 41×LR_CGAD and 41×LR_CGUD, while no detectable CNOT3 association was observed upon translation of 41×LK_AAGD and 41×LK_AAAD mRNAs (Fig. 3C). Translation of an additional arginine-rich mRNA containing 73 CGG, CGA, or AGG codons also stimulated recruitment of CNOT3 to polysomes, while replacement of the arginine codons with codons that were not enriched in the P-site of CNOT3-bound ribosomes abolished recruitment (fig. S3B).

This system was then leveraged to determine the structural basis of CNOT3 recruitment to ribosomes with P-site arginine codons. Polysomes actively translating 41×LR_CGGD mRNA transcripts were enriched for CNOT3-bound ribosomes by Flag IP and analyzed by cryo-EM (fig. S3C). Micrographs of these samples showed linearly clustered ribosomes, indicating that polysomes remained intact during sample preparation (fig. S4A). We analyzed individual ribosomes (but not disomes or higher order polysomes) for structure determination. The single particle reconstruction of these ribosomes yielded a homogeneous structure with an overall resolution of 2Å (Fig. 3D, fig. S4, and table S1). Density interpretation and model building revealed an empty A-site, a tRNA in the P-site, and CNOT3 occupying the E-site (Fig 3, E and F). As observed in previous yeast and rabbit structures (14, 19), the N-terminal 3-helix bundle (aa 1–111) of CNOT3 bridges the two ribosomal subunits and contacts the D-loop, D-stem, and anticodon stem of the P-site tRNA (Fig. 3, F to J, and fig. S5, A to C). An additional helical domain of CNOT3, composed of amino acids 112–231, contacts uS7 and eS25, and extends out of the E-site. The C-terminal portion of CNOT3, which mediates interactions with other CNOT components and is connected to the N-terminal modules via a disordered linker, is not visible in our reconstruction. The codon-anticodon pairing (fig. S5D and fig. S6, A and B) and defined CCA-RLD density connecting the P-site tRNA and the nascent peptide chain (fig. S5E and fig. S6, A and C) unambiguously identified the arginine-encoding CGG codon in the P-site, as predicted from our selective ribosome profiling data. The CGG codon present in our transcript can be recognized by the two tRNA^Arg,CCG isodecoders or, due to wobble base-pairing, by one of the five tRNA^Arg,UCG isodecoders present in HEK293T cells (36). Northern blot analysis revealed that all of these tRNAs are present in the CNOT3-enriched sample (fig. S6, D and E). For model building, we chose the most abundant isodecoder, tRNA^Arg,CCG−1, including its post-transcriptional modifications, which are clearly identifiable in our reconstruction (fig. S6F). This structural model provided an opportunity to investigate the molecular interactions that promote CNOT3 recruitment when select arginine codons occupy the ribosomal P-site.

The P-site tRNA D-arm is a key determinant of co-translational CNOT3 recruitment

It has been reported that select codon pairs in the ribosomal P- and A-sites can distort the A-site mRNA configuration, resulting in impaired decoding (37). This could potentially lead to CNOT3 recruitment as a consequence of ribosomal stalling. Unlike previous reconstructions of Not5 and CNOT3-bound ribosomes, we were able to model the A-site mRNA. We therefore compared the mRNA geometry in our structure to the structure of mRNA in a translating ribosome with a tRNA probing the A-site (38) (fig. S7A). The conformation of the A-site mRNA, as well as the overall structure of CNOT3-bound ribosomes, were compatible with decoding, suggesting that recruitment of CNOT3 was not a consequence of ribosomal stalling due to mRNA distortion.

We next considered the possibility that the specific interactions between the N-terminal domain of CNOT3 and arginyl-tRNA visualized in our structure play a role in CNOT3 recruitment. The second and third N-terminal helices of CNOT3 contact the arginyl-tRNA D-loop and D-stem, while a short element named the tRNA clamp motif (tCM) (14) interacts with the tRNA anticodon stem (Fig. 3, G to J, and fig. S5, A to C). We hypothesized that sequence and structural differences among tRNAs could impact these interactions and thereby influence CNOT3 recruitment when distinct codons, and consequently distinct tRNAs, occupy the P-site.

The selective association of CGG, CGA, and AGG arginine codons with CNOT3 recruitment, in contrast to CGC, AGA, and CGU arginine codons, provides a framework for investigating tRNA sequence features that may impact CNOT3 binding. In human cells, the six arginine codons are decoded by five distinct isoacceptor tRNA families (Fig. 4A and fig. S7B). The D-loops of all arginine tRNAs are identical due to their essential role in recognition by arginyl tRNA synthetase (RARS) (39, 40). tRNA^Arg,CCG and tRNA^Arg,UCG decode the most enriched P-site codons in CNOT3-bound ribosomes (CGG and CGA) and, in addition to the conserved D-loop, also share identical D-stem and anticodon stem sequences (Fig. 4A and fig. S7B). The third most enriched arginine codon (AGG) is decoded by tRNA^Arg,CCU, which shares the D-arm sequence with tRNA^Arg,CCG/UCG, but possesses a distinct anticodon stem. The least enriched arginine codons (CGC, AGA, and CGU) are decoded by tRNA^Arg,ACG and tRNA^Arg,UCU, which have different D-stems compared to tRNA^{Arg,CCG/UCG/CCU}. Most notably, the D-stems of tRNA^{Arg,CCG/UCG/CCU}, tRNA^Arg,ACG, and tRNA^Arg,UCU engage in distinct triplet base interactions involving tRNA positions 13, 22 and 46 (Fig. 4B). These observations suggest that variations in the D- and anticodon stems of arginine tRNAs may alter their affinity for CNOT3 in the context of a translating ribosome. They also raised the possibility that nucleotide modifications, particularly the absence of m⁷G at nucleotide 46 in tRNA^{Arg,CCG/UCG/CCU}, which participates in the aforementioned D-stem triplet base interaction, could impact CNOT3 recruitment.

To examine these possibilities, we used in vitro transcription to generate a panel of arginine tRNA mutants. In vitro transcribed tRNAs, which lack nucleotide modifications, can efficiently participate in translation when introduced into cells or cell-free translation reactions (41–43), providing an efficient system for assessing the role of sequence features and modifications in CNOT3 recruitment. In vitro aminoacylation assays confirmed that none of the mutations impaired tRNA charging compared to each respective parental tRNA (fig. S8A). In vitro translation assays were then performed using the 41×LR_CGUD mRNA, which is decoded by tRNA^Arg,ACG and weakly recruits CNOT3 (Fig. 3C). As expected, addition of excess tRNA^Arg,ACG−1 to these reactions had no effect on CNOT3 recruitment (Fig. 4C and fig. S8B). In contrast, addition of a variant of tRNA^Arg,CCG−1, reprogrammed to decode the CGU codon by mutating the anticodon to ACG (tRNA^Arg,CCG-m1), promoted CNOT3 recruitment in a dose-dependent manner (Fig. 4C and fig. S8C). These data demonstrate that the sequence of the P-site tRNA plays a key role in co-translational CNOT3 recruitment and show that this effect is not dependent on specific tRNA nucleotide modifications.

We next tested whether the distinct D-stem sequences of tRNA^Arg,ACG and tRNA^Arg,CCG were responsible for the contrasting abilities of these tRNAs to recruit CNOT3. These D-stems differ by three nucleotides (Fig. 4C, C13 vs. U13; base-pair G12:C23 vs. C12:G23). Swapping the D-arm of tRNA^Arg,CCG-m1 (tRNA^Arg,CCG reprogrammed to decode the CGU arginine codon) with the D-arm of tRNA^Arg,ACG (generating tRNA^Arg,CCG-m2) abrogated CNOT3 recruitment to the 41×LR_CGUD mRNA (Fig. 4D and fig. S8D, compare m1 to m2). Moreover, mutation of U13 to C13 (tRNA^Arg,CCG-m3) was sufficient to impair CNOT3 recruitment, while reversing the C12:G23 to a G12:C23 base-pair had no effect (tRNA^Arg,CCG-m4). We further showed that addition of the D-stem sequence from tRNA^Arg,CCG to tRNA^Arg,ACG (tRNA^Arg,ACG-m5), or the single C13U mutation enabling formation of the U13:A22 base pair in tRNA^Arg,ACG (tRNA^Arg,ACG-m6) was sufficient to enable CNOT3 recruitment (Fig. 4E and fig. S8E, compare WT to m5 and m6). Again, flipping the G12:C23 base-pair (tRNA^Arg,ACG-m7) had no effect in this context. Altogether, these results pinpoint position U13 in the tRNA^{Arg,CCG/UCG/CCU} D-arm, which forms a triplet base interaction with A22 and A46 (Fig. 4B), as a critical feature associated with CNOT3 recruitment to ribosomes.

In tRNA^Arg,UCU, the triplet interaction is C13:G22:m⁷G46 (Fig. 4B), a frequent configuration of nucleotides at these positions among all tRNAs (fig. S9), whereas tRNA^Arg,ACG has an unusual C13:A22:A46 triplet. To further probe how this triplet base interaction influences CNOT3 recruitment, we mutated tRNA^Arg,ACG at positions 22 and 46 to introduce the common C13:G22:G46 triplet (tRNA^Arg,ACG-m8). These mutations did not increase CNOT3 recruitment (Fig. 4F and fig. S8F, compare WT to m8). Thus, the U13:A22:A46 triplet, present in tRNA^{Arg,CCG/UCG/CCU}, appears to be optimally configured to recruit CNOT3.

To determine why the U13:A22:A46 triplet favors CNOT3 recruitment, we further examined our structural model. Nucleotide A22 makes direct and water-mediated H-bonding interactions with CNOT3 residues E95 and K49, respectively (Fig. 4H). Supporting the importance of these interactions, an E95A substitution in CNOT3 greatly reduced co-translational recruitment to a CGG/CGA/AGG-rich transcript (Fig. 4K and fig. S8J). In the previously reported structures of Not5/CNOT3-bound ribosomes, yeast tRNA_i^Met or rabbit tRNA^Leu,UAA, which lack the U13:A22:A46 triplet, occupy the P-site (14, 19). We therefore examined the Not5/CNOT3-tRNA interaction interfaces in these structures to determine how alternative 13:22:46 triplet configurations affect Not5/CNOT3 binding. tRNA_i^Met has the most common triplet combination (C13:G22:m⁷G46), while tRNA^Leu,UAA has a trans-Hoogsteen G13:A22 pairing characteristic of type-II tRNAs (44). The distinct base configurations in tRNA_i^Met and tRNA^Leu,UAA increase the intermolecular distance to Not5 and CNOT3 in their respective complexes. This prevents the formation of the hydrogen-bonding network observed in our CNOT3/tRNA^Arg,CCG structure (Fig. 4, I and J). Because the C13:G22:m⁷G46 triplet of tRNA^Arg,UCU, visualized in a structure of this tRNA with a splicing endonuclease (45), is structurally similar to tRNA_i^Met (Fig. 4I, inset), we expect that interactions between tRNA^Arg,UCU and CNOT3 are similarly disrupted. These observations provide a structural basis for the enhanced recruitment of CNOT3 by arginine tRNAs containing the U13:A22:A46 triplet base interaction.

Impact of the P-site tRNA anticodon stem on co-translational CNOT3 recruitment

Our CNOT3-ribosome structure, as well as previously reported structures (14, 19), also revealed interactions between CNOT3 and the anticodon stem of the P-site tRNA. These interactions are mediated by the tCM, which forms several direct backbone interactions as well as water-bridged interactions with bases G42 and A43 of the P-site tRNA (fig. S8H). Supporting the importance of these interactions, a K105S substitution in the tCM abrogated recruitment of CNOT3 to a CGG/CGA/AGG-rich transcript (fig. S8, I and J). We hypothesized that these interactions might explain why tRNA^Arg,CCG and tRNA^Arg,UCG, which share an anticodon stem, recruit more CNOT3 compared to tRNA^Arg,CCU, which has a distinct anticodon stem but the same D-stem and D-loop structure (Fig. 4A and fig. S7B). The unique anticodon stem of tRNA^Arg,UCU may also contribute to the poor CNOT3 recruitment associated with this tRNA. To test whether the anticodon stem contributes to differential CNOT3 recruitment by these tRNAs, we replaced the anticodon stem of tRNA^Arg,ACG bearing the tRNA^Arg,CCG D-arm (Fig. 4F, tRNA^Arg,ACG-m5) with the anticodon stem from tRNA^Arg,UCU (Fig. 4F, tRNA^Arg,ACG-m9). Consistent with our hypothesis, this greatly reduced, but did not eliminate, CNOT3 binding (Fig. 4F and fig. S8F, compare m5 to m9). In addition, we reprogrammed tRNA^Arg,UCU−1 to recognize the CGU arginine codon (Fig. 4G, tRNA^Arg,UCU-m10), which showed that addition of the tRNA^Arg,CCG U13:A22:A46 triplet (Fig. 4G, tRNA^Arg,UCU-m11), but not the tRNA^Arg,CCG anticodon stem (Fig. 4G, tRNA^Arg,UCU-m12), was sufficient to increase CNOT3 recruitment (Fig. 4G and fig. S8G). Thus, the presence of the U13:A22:A46 triplet is the major determinant of CNOT3 recruitment by arginine tRNAs, while the anticodon stems play a lesser, but measurable, role. Although we lack the structural information to precisely determine how sequence variation in the anticodon stem impacts CNOT3 interactions, we suspect that different base-pairing configurations may alter the tCM-tRNA affinity. In keeping with this concept, the anticodon stems of tRNA^Arg,UCU and tRNA^Arg,CCU both contain G:U wobble base pairs, which can locally alter the twist of the RNA backbone (fig. S8K) (46), potentially affecting interactions with the CNOT3 tCM element.

An extra nucleotide upstream of the D-loop GG motif prevents CNOT3 recruitment

Our structural studies revealed that select arginine tRNAs possess distinct sequence and structural features that promote CNOT3 recruitment. To determine whether other tRNAs have features that inhibit recruitment of CNOT3, we examined the codons that were most depleted from the P-site of CNOT3-bound ribosomes (fig. S10A). We noticed that many of the most depleted codons, including codons for asparagine (N), lysine (K), isoleucine (I), tyrosine (Y), methionine (M), phenylalanine (F), and threonine (T), are decoded by tRNAs with an extra nucleotide in the α element of the D-loop, immediately preceding the universally conserved GG D-loop motif (Fig. 5, A and B, and fig. S9) (44). The GG motif forms a key interaction with the T-arm to stabilize the “elbow” structure of all tRNAs. α and β D-loop elements flank the GG motif and often contain modified dihydrouridine nucleotides (47). In our structure, CNOT3 contacts the α element of the D-loop, but not the β element (Fig. 5, B, D, and G), mirroring the interface between Not5/tRNA_i^Met in yeast and CNOT3/tRNA^Leu,UAA in rabbit reticulocyte lysate (Fig. 5, E and H, and data not shown) (14, 19). This interaction is critical for CNOT3 recruitment, as demonstrated by mutating the key CNOT3 residue (R59) that contacts this site in the tRNA (Fig. 5C and fig. S10B). A structural overlay of a tRNA with an extra nucleotide in the α element (tRNA^Lys,UUU) bound in the P-site (48) shows that the α expansion creates a steric clash with CNOT3 (Fig. 5, F and I). The same D-loop orientation is also observed in the crystal structure of isolated bovine tRNA^Lys,UUU (49), suggesting that it represents a structurally rigid bulge within the α-element (fig. S10C). These observations suggest that depletion of select P-site codons may result from a steric clash between CNOT3 and the cognate tRNAs for these amino acids that prevents effective CNOT3 binding in the E-site. Consistent with this hypothesis, we observed that mRNAs rich in N, K, I, Y, M, F, and T codons decoded by tRNAs with an extra nucleotide in the α element were less sensitive to CNOT3-mediated decay compared to other transcripts (Fig. 5J and fig. S10, D and E). Notably, cytosolic ribosomal proteins represent the class of transcripts most enriched for these codons (fig. S10, F and G), providing a possible explanation for our finding that depletion of CNOT3 selectively upregulates mitochondrial, but not cytosolic, ribosomal proteins.

To directly test the effect of the α element insertion on CNOT3 recruitment, we performed further tRNA mutagenesis experiments. We noted that, among tRNAs that decode methionine codons, the initiator tRNA_i^Met has a single residue in the α element, while all elongator tRNA^Met isodecoders have an additional nucleotide preceding the GG motif (Fig. 5A and fig. S9). A previous structural analysis of Not5-bound ribosomes in yeast identified tRNA_i^Met in the P-site (14). The structural similarity between the Not5/tRNA and CNOT3/tRNA interfaces suggests that CNOT3 recruitment may also be compatible with the presence of a similar tRNA in the P-site. Furthermore, mutations in the D-arm of elongator tRNA^Met that remove the extra α element nucleotide are expected to be compatible with tRNA charging and decoding of internal AUG codons (50, 51).

Confirming the inhibitory effect predicted from our structural studies, deletion of the extra α nucleotide upstream of the GG motif in elongator tRNA^Met,CAU−3 (tRNA^Met-m13) was sufficient to stimulate recruitment of CNOT3 to ribosomes translating a 41×LM_AUGD mRNA (Fig. 5, K and L, and fig. S10, H and I, compare WT to m13). Consistent with our structural analyses, changes to the D-loop β element did not impact CNOT3 recruitment, since introducing C20A (tRNA^Met-m14) or C20U (tRNA^Met-m15) mutations had no effect on CNOT3 binding (Fig. 5L and fig. S10, H and I, compare m13 to m14 or m15). Taken together, our results demonstrate that a D-stem U13:A20:A46 triplet and a short, single-nucleotide D-loop α element in the P-site tRNA are major determinants of CNOT3 co-translational recruitment, providing a structural basis to explain predominant patterns of P-site codon enrichment and depletion in CNOT3-bound ribosomes.

Slow decoding enables P-site tRNA-mediated CNOT3 recruitment to ribosomes

Notably, we observed that the A-site of human CNOT3-bound ribosomes was not occupied by tRNA (Fig. 3F). This configuration, indicative of slow decoding, was similar to the structure of Not5-bound ribosomes in yeast and CNOT3-bound rabbit ribosomes stalled on a highly non-optimal codon (14, 19). As discussed above, we detected a weak, but statistically significant, correlation between A-site dwell time and CNOT3 recruitment in our selective ribosome profiling data (fig. S1N), indicating that slow decoding measurably promotes CNOT3 recruitment in human cells. We found that the correlation between A-site dwell time and CNOT3 association was greatly enhanced when the P-site was occupied by a CGG, CGA, or AGG arginine codon (Fig. 6, A and B). In contrast, when any other codon was present in the P-site, the dwell time correlation was undetectable (Fig. 6C). These data suggest that, like in yeast and rabbit reticulocyte lysate, slow decoding increases the probability that the ribosomal A- and E-sites will be simultaneously vacant, providing an opportunity for CNOT3 to enter the E-site. Our results demonstrate that CNOT3 subsequently probes the D-arm of the P-site tRNA, which ultimately determines whether CNOT3 stably associates with the ribosome and initiates mRNA decay.

Discussion

The CCR4-NOT complex can associate with translating ribosomes to accelerate mRNA turnover. A major trigger for CCR4-NOT recruitment to ribosomes is slow decoding, which is sensed by entry of the N-terminal domain of the CNOT3/Not5 subunit into a vacated E-site when the A-site codon is non-optimal and therefore lacks a cognate tRNA. Here we present our finding that, in mammalian cells, the identity of the P-site tRNA plays a critical role in determining how efficiently CNOT3 is recruited to ribosomes. Based on our results, we propose the following model for co-translational CNOT3 recruitment in mammalian cells (Fig. 6D). As in yeast and rabbit reticulocyte lysate, slow decoding, resulting in a ribosomal conformation with empty A- and E-sites, appears to be a pre-requisite for CNOT3 entry into the E-site, enabling it to probe the P-site tRNA. At this stage, three outcomes are possible: (i) If the P-site tRNA has the U13:A22:A46 triplet and lacks the extended α-element (i.e., tRNAs that decode CGG/CGA/AGG arginine codons), CNOT3 binding will be stabilized by direct hydrogen bonding interactions with the P-site tRNA D-arm, resulting in accelerated mRNA decay. (ii) If the P-site tRNA is neutral, lacking both the U13:A22:A46 triplet and the extended D-loop α element, CNOT3 binding may be transient. However, if an extended ribosomal stall occurs due to scarcity of a charged tRNA that can enter the A-site, CCR4-NOT-mediated decay may ensue. (iii) If the P-site tRNA has the extended D-loop α element (i.e., tRNAs that decode N, K, I, Y, M, F, and T), CNOT3 accommodation will be sterically blocked, thereby favoring CNOT3 disengagement and resumption of translation. Because of the central role played by the P-site tRNA in dictating the outcome of CNOT3 recruitment, we propose the term P-site tRNA-mediated decay (PTMD) to refer to this mechanism of accelerated mRNA degradation.

These findings expand our understanding of how codon content influences post-transcriptional regulation. Prior to this work, the impact of specific codons on rates of co-translational mRNA turnover were largely attributed to their effect on decoding efficiency. The results reported here show that specific codons can also modulate the rate of mRNA decay due to the ability of their cognate tRNAs to promote or inhibit association of the CCR4-NOT complex. Thus, in addition to their canonical role in decoding, tRNAs directly participate in the recruitment of post-transcriptional regulators to translating ribosomes. This discovery raises the intriguing possibility that other regulatory complexes that engage ribosomes may be similarly impacted by tRNA identity. Evocative of this concept, it was recently shown that angiogenin, a ribonuclease with high specificity towards tRNA, binds to ribosomes with empty A sites, which activates its nuclease activity (52). Akin to CNOT3, interaction with the P-site tRNA appears to be important for stabilizing the association of angiogenin with ribosomes, although in this case, the identity of the P-site tRNA does not appear to be a determining factor in ribosome binding.

The presence of the U13:A22:A46 triplet and absence of the extra nucleotide in the α element is a conserved feature of metazoan tRNAs that decode arginine codons CGG, CGA, and AGG (fig. S11), suggesting that PTMD is operative across animal species. Nevertheless, it is less clear whether this mechanism is active in yeast, where codon optimality appears to be the dominant signal for Ccr4-Not recruitment. The high level of sequence and structural conservation between CNOT3 and Not5 (57% amino acid identity and 0.75 Å Cα RMSD), coupled with the presence of a U13:A22:A46 triplet and a short α element in arginine tRNAs that decode AGG and AGA (fig. S12A; tRNA^Arg,CCU/UCU), raises the possibility of a functional PTMD pathway in yeast. Indeed, codons with cognate tRNAs possessing an extra nucleotide in the D-loop α element were generally depleted from the P-site of Not4-bound ribosomes (fig. S12, B and C). Nevertheless, P-site AGG and AGA arginine codons were only marginally enriched. Instead, non-optimal codons were unexpectedly enriched, not only in the A-site as reported (14), but also to a similar extent in the P- and E-sites, whereas no such enrichment was observed in human cells (fig. S12, D and E). We speculate that these differences in codon enrichment patterns between yeast and humans are a consequence of distinct mechanisms of binding of CCR4-NOT to ribosomes between these species. Specifically, the E3 ubiquitin ligase Not4, which interacts with and ubiquitylates ribosomal protein eS7 (53, 54), is a constitutive subunit of Ccr4-Not in yeast (55, 56) and is required for recruitment of Not5 to ribosomes (14). In contrast, the metazoan Not4 homolog CNOT4 does not stably associate with CCR4-NOT (57, 58) and is dispensable for ribosomal recruitment of CNOT3 (19). These findings suggest that, when Not5 is recruited to the E-site of a slowly decoding ribosome in yeast, the associated Ccr4-Not complex may become tethered to the ribosome via Not4. If Not5 does not remain stably bound within the E-site, the ribosome may continue further cycles of elongation while remaining associated with Ccr4-Not, thereby moving the non-optimal codon to the P- or E-site. Importantly, this mechanism would render Ccr4-Not association with the ribosome less dependent upon the stable accommodation of Not5 within the E-site, and therefore less reliant upon specific Not5:P-site tRNA interactions. In agreement with this hypothesis, deletion of the N-terminal domain of Not5 has been shown to reduce but not eliminate its binding to ribosomes (14). Unlike the behavior of Not5 in yeast, we found that the R59S mutation of human CNOT3 that disrupted interaction with the P-site tRNA abolished its recruitment to ribosomes. Thus, mammalian ribosomes require CNOT3 accommodation within the E-site, which depends upon specific P-site tRNA interactions, to stably recruit CCR4-NOT, enabling the PTMD mechanism to dominate in this setting.

While we still have much to learn about the physiologic role of the PTMD pathway, we observed that mitochondrial ribosomal proteins, whose mRNAs are rich in CGG, CGA, and AGG arginine codons, are strongly regulated by this mechanism across diverse cell types. Cytosolic ribosomal proteins, in contrast, are rich in N, K, I, Y, M, F, and T codons, whose cognate tRNAs oppose CNOT3-mediated decay. As a consequence of this regulation, we observed that CNOT3 is a strong repressor of mitochondrial translation and mitochondrial mass. In keeping with these findings, mice with reduced CNOT3 expression exhibit increased respiration and are resistant to obesity, consistent with an increase in mitochondrial energy expenditure (59). Moreover, expression of CNOT3, but not other subunits of the CCR4-NOT complex, is upregulated in obese mice and downregulated after fasting (59), suggesting that regulation of mitochondrial activity by PTMD affects the physiological response to nutrient availability. Interestingly, in yeast, mitochondrial ribosomal protein mRNAs are rich in non-optimal codons, while cytosolic ribosomal proteins are mostly encoded by mRNAs with high levels of optimal codons (17, 20). Accordingly, loss of the Ccr4-Not complex also increases mitochondrial mass in yeast (60). Thus, regulation of mitochondrial homeostasis by CCR4-NOT appears to be highly conserved, although the underlying mechanism of recruitment of Not5/CNOT3 to target transcripts occurs through distinct mechanisms across species. Further investigation of how the PTMD pathway is deployed across metazoans, and the impact of this mechanism of gene regulation on physiology and disease, are important priorities for future investigation.

Materials and methods:

Supplementary Material

Data and materials availability:

All high-throughput sequencing data has been deposited in GEO (Accession GSE268325). The cryo-EM density maps and models have been deposited in EMDB and PDB databases under accession codes: EMD-45170/PDB-9C3H (human 80S/CNOT3/tRNA^Arg,CCG complex) and PDB-93CI for the rebuilt CNOT3/tRNA^Leu,UAA structure. All reagents generated in this study are available upon request from J.T.M with a completed Materials Transfer Agreement.