Codon-specific ribosome stalling reshapes translational dynamics during branched-chain amino acid starvation

Abstract

Background: Cells regulate protein synthesis in response to fluctuating nutrient availability through mechanisms that affect both translation initiation and elongation. Branched-chain amino acids, leucine, isoleucine, and valine, are essential nutrients. However, how their depletion affects translation remains largely unclear. Here, we investigate the immediate effects of single, double, and triple branched-chain amino acid deprivation on translational dynamics in NIH3T3 cells using RNA-seq and ribosome profiling.

Results: All starvation conditions increased ribosome dwell times, with pronounced stalling at all valine codons during valine and triple starvation, whereas leucine and isoleucine starvation produced milder, codon-specific effects. Notably, stalling under isoleucine deprivation largely decreased under triple starvation. Positional enrichment of valine codons near the 5' end and downstream isoleucine codons potentially contributes to these patterns, suggesting a possible elongation bottleneck that influences translational responses under branched-chain amino acid starvation. The presence of multiple valine stalling sites was associated with decreased protein levels. Finally, codon-specific dwell time changes correlated strongly with patterns of tRNA isoacceptor charging.

Conclusions: Together, these findings suggest that differential ribosome stalling under branched-chain amino acid starvation reflects a balance between amino acid supply, tRNA charging dynamics, codon position, and stress-response signaling.

Keywords: Amino acid starvation; Codon-specific stalling; Elongation bottleneck; Nutrient stress response; Ribosome profiling; Translation regulation; tRNA charging.

Fig. 1

Translation elongation rates are modulated in a codon-specific, non-additive manner upon limitation for BCAAs. A–E Codon-specific ribosomal dwell times (DTs) in Ctrl vs. starvation (Double: -Leu & -Ile, Triple: -Leu & -Ile & -Val). Shown is the mean DT of triplicates over a ± 3 codon region centered on the P-site. Significance was evaluated with ANOVA followed by Tukey’s HSD test. Annotated are codons with significantly upregulated DT relative to Ctrl (p-value < 0.01). F Changes in Ribo-seq read density around annotated codons relative to the Ctrl. Mean reads per million (RPM) values were normalized to Ctrl condition, and the mean log₂FC computed for a ± 100 nt window around codons. G Heatmap of log₂FC in intracellular amino acid levels under different starvation conditions relative to Ctrl. Values below detection limit (LOD) are indicated. H Scatter plot of DT under Ctrl condition (A–E, log₂) vs. codon frequency (%). Codons for Val (orange), Ile (blue), and Leu (green) are highlighted. A linear trendline (red line) shows the relationship between DT and codon frequency

Fig. 2

Stress response pathways are differentially activated upon Leu, Ile, and Val starvation. A Expression levels of Atf4 and Chop in RNA-seq and Ribo-seq under different deprivation conditions. Values represent log₂-transformed reads per million (RPM). Each point represents an individual replicate. Statistical significance was determined by unpaired t-tests comparing each starvation to Ctrl, with all comparisons resulting in p < 0.01. B qPCR analysis of mRNA levels for Asns, p62, and Atg3 under different starvations at 3 h and 6 h. Expression levels were normalized using the ΔΔCt method (normalized to Gapdh expression) and are shown as fold change relative to Ctrl. Bars represent mean fold change, with error bars indicating standard error of mean (SEM). C Heatmap and quantification of log₂FC for the top 150 ATF4 target genes from ChIP-Atlas [55] at the mRNA level. Genes are clustered based on their expression profiles using k-means clustering (k = 12). In the boxplot, red points indicate the mean expression change. Significance was assessed using unpaired t-tests. D Representative Western blot showing phosphorylation levels of 4E-BP1 (Phos_Ser65) and RPS6 (Phos_Ser235/236) under different amino acid starvations. Total protein levels and Tubulin level serve as loading controls. 1 µM Torin1 treatment and full medium conditions are included as controls. E Quantification of phosphorylation levels of 4E-BP1 and RPS6, normalized to total protein levels and relative to the control (Ctrl). Quantification was performed using ImageJ. Data represent mean ± SEM from three independent experiments. Statistical significance was determined using unpaired t-tests. F Heatmap and quantification of log₂FC for the known TOP motif containing transcripts [22] in our Ribo-seq. Genes are clustered based on their expression profiles using k-means clustering (k = 12). In the boxplot, red points indicating the mean expression change. Significance was assessed using unpaired t-tests

Fig. 3

BCAA starvations differentially modulate translation initiation and elongation. A Schematic representation of detection of newly synthesized proteins by OPP incorporation assay. Newly synthesized proteins incorporate OPP, which causes translation termination and allows subsequent labeling with TAMRA-azide. This enables their detection and quantification via fluorescence microscopy, distinguishing them from pre-existing proteins that lack OPP. B Representative fluorescence microscopy images of TAMRA and Hoechst staining. Wild-type cells and cells treated with 0.3 mM CHX to inhibit translation are shown. TAMRA (orange) represents OPP staining, marking newly synthesized proteins. Hoechst stains nuclei (blue), and DPC (digital phase contrast) provides cellular morphology. Scale bar: 100 µm. C,D Quantification of OPP incorporation under C Ctrl and D starvation conditions. Boxplot representations display the relative mean fluorescence intensity (MFI) of the TAMRA signal; each dot represents an individual cell, and red points indicate the mean MFI. Data represent three biological replicates. Statistical significance between conditions was determined using unpaired t-tests with Bonferroni correction. E Representative polysome profiles of Ctrl (black line) and starvations (orange line). F–H Quantification of polysome-to-monosome (P/M) ratio (F), monosome (G), and polysome (H) fractions. I Number of dysregulated proteins (p < 0.01 and |log₂FC|> 0.26) across starvations. J Correlation between measured log₂FC proteomic changes upon Triple starvation and published protein half-lifes in wildtype NIH3T3 cells [59]. Lines indicate theoretical log₂FC assuming standard degradation kinetics under total synthesis block (orange) and 50% synthesis block (sienna). K Distribution of protein half-lives (see J) for down- and upregulated proteins under any BCAA starvation. L–P Scatter plots showing the relationship between log₂FC measured in proteome and Ribo-seq under the given conditions. Red dots represent transcripts corresponding to upregulated proteins (p < 0.01 and log₂FC > 0.26) with significant changes in Ribo-seq (p < 0.01 and |log₂FC|> 0.26), while blue dots highlight transcripts corresponding to downregulated proteins (p < 0.01 and log₂FC < − 0.26) with significant changes in Ribo-seq (p < 0.01 and |log₂FC|> 0.26), indicating translational repression

Fig. 4

BCAA starvation leads to ribosome density shifts toward the 5′ of the CDS. A Metagene profiles showing the mean normalized RPF count of the Ribo-seq along the CDS for each starvation (colored) compared to Ctrl (black). Solid lines represent the mean signal at each normalized CDS position, and shaded ribbons indicate the standard error of the mean (SEM). B Bar plot quantifying the regional “ramp” in ribosome densities. Log₂-transformed ratios of summed RPF counts in the 5′ region (1–20% of CDS positions), and 3′ regions (80–100% of CDS positions) are shown for starvation relative to Ctrl. Bars indicate the mean value with SEM as error bars. Statistical significance between Ctrl and starvation was determined using unpaired t-tests with Bonferroni correction. C Bar plot indicating the mean transcript-specific polarity scores in the indicated conditions. D Bar plot indicating the number of transcripts displaying a significant reduction of the polarity score (Δpi < − 0.15; p < 0.05) relative to the control condition upon indicated starvations. E UpSet plot summarizing the overlap of transcripts with significant 5′ polarity score shifts (Δpi < − 0.15; p < 0.05) across BCAA starvations. Each bar in the upper panel represents the number of transcripts showing polarity shifts in the conditions indicated by the connected black dots below. Single dots indicate condition-specific sets, while connected dots indicate shared transcripts between multiple starvations. F Dot plot showing the log₂ fold change in mean codon frequencies for transcripts exhibiting a significant 5' polarity shift. For each condition, the codon frequencies within the first 20% of the CDS of polarized transcripts were compared to the frequencies in the same region of all other expressed transcripts. Dot size reflects the − log₁₀(p-value) from unpaired t-tests (BH-adjusted), and color indicates the magnitude and direction of the log₂ fold change. G,H Representative ribosome profiling tracks for the genes Cbx1 (G) and Hint1 (H), with Val and Ile codons annotated

Fig. 5

Non-uniformly distributed codons create elongation bottlenecks. A Bar plots showing the number of extracted stalling sites (positions) and the corresponding transcripts harboring stalling sites across different starvations. Peaks were defined based on the following criteria: log₂FC(RPM) > 2 and log₂FC(Norm_RPM) > 2 and FC_relative_to_gene_mean > 1. Peaks that met these thresholds and showed statistically significant differences (p < 0.05) between conditions were considered robust. B Number of significantly downregulated proteins (log₂FC < − 0.26, adjusted p < 0.01) that also exhibit at least the indicated number of stalling sites. The red line shows the statistical enrichment of overlap between stalling and downregulation, calculated using a hypergeometric test and plotted as − log₁₀(p-value). The dashed horizontal line marks the significance threshold (p = 0.05). The first bar (0) represents the total number of downregulated proteins per condition, irrespective of stalling. C Count graphs depicting the valine and isoleucine codon counts at indicated positions around the extracted stalling sites in the annotated starvation condition. The region surrounding each stalling site was extracted and the amount of Val or Ile codons was summed at each position; accompanying pie charts indicate, in percentage, the proportion of extracted stalling sequences that contain an Ile or Val codon versus those that do not. D,E Venn diagrams displaying the overlap of identified D stalling positions and E transcripts that harbor stalling sites in Ile, Val, and Triple starvation. F Metagene distribution of the extracted stalling sites across the CDS in all starvation conditions. G Spatial analysis of stalling sites on the 1683 transcripts harboring stalling sites in Ile, Triple, and Val starvation conditions. Per transcript, the distance between the first stalling site caused by -Ile starvation and the first stalling site caused by -Val starvation was calculated, revealing that under Val starvation the stalling occurs further upstream on the transcript

Fig. 6

Differential tRNA charging patterns reveal a mechanistic basis for codon-specific stalling under BCAA starvation. A Schematic representation of the different tRNA isoacceptors for Leu, Ile, and Val. B–F Differential charging of tRNA isoacceptors under the indicated starvation conditions. Yellow indicates starved cells without treatment. Green represents cells treated with 100 µg/mL cycloheximide (CHX) in the last 30 min of starvation to inhibit translation. Blue represents cells treated with 10 µM MG132 and 160 nM Bafilomycin A1 to inhibit proteasomal and autophagic degradation (amino acid recycling inhibition). Certain tRNA isodecoders could not be measured separately and are therefore represented with IUPAC ambiguity codes: W (A or U), R (A or G), and M (A or C). Large outlined dots indicate significant downregulation (p < 0.05). Sample sizes: untreated (n = 5), translation and AA recycling inhibition (n = 2). Displayed is the mean of the replicates with the SEM indicated as error bars