Polysome profile under isoleucine starvation
The translatome profile was determined for L. lactis cells in conditions of isoleucine starvation (total exhaustion of isoleucine from the growth medium). In this stress condition, L. lactis grew constantly but very slowly: 0.05 h-1 to be compared to 0.88 h-1 in optimal growth conditions. Translation elongation was arrested and cells lyzed, and mRNA-ribosome complexes were separated according to the polysome size. Typical polysome profile and peak assignment under stress conditions are shown in Figure 1. In stress conditions, the polysomal profile was less clearly resolved than optimal growth conditions[18] and showed a broad initial peak of absorbance (during the first three minutes) corresponding to low molecular weight proteins confined in the upper layer of the sucrose density gradient. Elution fractions were pooled into five fractions named S1, S2, S3, S4 and S5 (see Methods). RNA concentrations in the first and last eluted fractions were lower in cells under stress than optimal growth conditions, leading to a fractionation into a total of five fractions instead of seven in optimal growth conditions[18].
Fraction S1 contained free mRNAs or mRNAs loaded with an incomplete ribosome; S2 contained mRNAs bound to one ribosome (monosome fraction). The number of ribosomes per transcript in the polysomal fractions S3, S4 and S5 has been estimated as previously[18] and the averages are in the ranges 1.4-2.9 (mean value 2.1), 2.9-5.1 (mean value 3.9), 5.1-15.1 (mean value 8.4), respectively. In stress conditions, the maximum number of ribosomes per transcript was 15 whereas it is 18 in optimal growth conditions[18]. The total RNA concentration in cells in conditions of isoleucine starvation was almost half that in optimal growth conditions (5.43 ± 0.50 and 9.01 ± 0.16 g.100 g-1 cells, respectively). Ribosomal RNAs are the major components of the total RNA population; therefore this suggests that there is a significant reduction of the total number of ribosomes and of the polysome size during stress. We checked that mRNA as a proportion of total RNA (estimated by summation of all microarray spot intensities) did not differ between optimal growth conditions and isoleucine starvation. The percentage of ribosomes engaged in translation, estimated by area integration of the absorbance of the polysomal profile as described previously[18], was also not significantly affected by isoleucine starvation: the mean values were 59 ± 3% in conditions of isoleucine starvation and 61 ± 2% in optimal growth conditions[18].
Influence of isoleucine starvation on ribosome density and ribosome occupancy
In L. lactis cells starved of isoleucine, the peak fraction corresponding to the most frequent number of ribosomes bound on each mRNA was identified for each of 1317 genes and the fractions in which it was found were identified with a 95% bootstrap confidence interval. Only 29 genes had a peak fraction overlapping two or three fractions. We therefore focused our analysis on the 1288 genes with a peak fraction confined to a single fraction. Most of these genes were heavily ribosome-loaded with the peak fraction in fraction S5, corresponding to 5.1-15.1 bound ribosomes. Only one gene (yohD, unknown function) had a peak fraction in the monosome fraction (S2). For each of the 1288 genes, ribosome density was calculated as the ratio of the ribosome number in its peak fraction to the coding sequence length. However, this calculation attributed ribosome densities above the maximal theoretical ribosome density (3.33 ribosomes/100 nucleotides) to 47 genes of the 1288 gene set. These 47 genes with aberrant ribosome densities were omitted from the subsequent analyses. Among 1241 remaining genes, ribosome density was obtained in both optimal growth and stress conditions for 816: the ribosome density ranged from 0.04 to 3.31 ribosomes per 100 nucleotides (Figure 2A). For cells in isoleucine starvation, the median ribosome density was 0.89 ribosomes per 100 nucleotides whereas it was 1.41 ribosomes per 100 nucleotides in optimal growth conditions[18]; indeed, the ribosome density was significantly lower in stress than in optimal growth conditions for 95% of the genes (Figure 3B, lower line of genes). The interpretation of ribosome density can vary according to the limiting step of translation. For genes exhibiting initiation-limited translation, high ribosome density is correlated with high translation levels, but for genes with elongation limitation, high density can result from ribosome congestion leading to low translation efficiency. The ribosome density far lower than the maximal theoretical density suggests that the translation of most mRNAs in stress conditions may be initiation–limited, as previously suggested in optimal growth conditions[18]. Further evidence for this was recently provided by kinetic modeling of bacterial translation from experimentally determined ribosome densities for L. lactis mRNAs in both conditions: translation was mainly initiation-limited for 89% and 69% of the genes in optimal growth and stress conditions, respectively[20]. Most L. lactis mRNAs are expected to be subject to initiation-limited translation, a decrease of the ribosome density in stress conditions presumably reflects less translation. Only 39 genes showed increased ribosome density in conditions of isoleucine starvation (Figure 3B, upper line of genes): the corresponding peak fractions moved from the monosome fraction under optimal growth conditions to the polysomal fractions under stress conditions. This set of 39 genes included several prophage genes (pi116, pi218, pi228, pi231, pi245, pi331, ps207 and ps214).
Ribosome occupancy was calculated for each transcript species as a measure of efficiency of translation initiation; values for each mRNA were compared between stress and optimal growth conditions. In stress conditions, the distribution of ribosome occupancy was shifted toward higher values (Figure 2B). For the set of 816 selected genes (see above), the median ribosome occupancy value for their mRNAs was 74% during isoleucine starvation and 68% in optimal growth conditions[18]; there was a statistically significant difference of ribosome occupancy between the two conditions for 381 genes (46%). Ribosome occupancy in stress conditions was higher for 375 mRNAs and lower for only six (accB and fabG1 related to fatty acid biosynthesis, ptnD and ypcG involved in sugar transport and ychG and yvcA with unknown functions) (Figure 3A).
Impact of isoleucine starvation on gene translation regulation
To examine the consequences of isoleucine starvation on the regulation of gene translation, we studied the effects of stress on both ribosome occupancy and ribosome density: the differences in both ribosome occupancy and ribosome density between the two conditions were analyzed for each mRNA from the set of 816 genes. We plotted the variation in ribosome density under stress and optimal growth conditions (Figure 3B). The upper line indicated genes moving from the monosome fraction in optimal growth conditions to an average 8.4 ribosomes per transcript in stress conditions. One gene yohD remained in the monosome fraction in both conditions. The lower line displayed genes with a higher average number of ribosomes per transcript in optimal growth conditions than in stress conditions, 14 against 8.4. Black, red and green symbols indicate respectively genes with similar, higher and lower ribosome occupancy in stress than optimal growth conditions. The resolution of the translatome experiment under stress conditions was lower than in optimal growth conditions (5 elution fractions instead of 7). Therefore, part of the variability of the translational response in stress conditions could be hidden in our ribosome density comparison and thus not displayed in Figure 3B. Nevertheless, the observation of linear profiles suggests that except for one gene, most genes can be clustered into two classes of “density regulation” reflecting variation in translation elongation under isoleucine starvation may be related to isoleucine codon bias or other sequence specificities.
We identified only one gene yohD for which neither measure was significantly modified. yohD was thus considered to not be translationally regulated by stress conditions. There were 39 genes for which translation was up-regulated under stress conditions: for 19 genes, both translational measures were higher (Figure 3B, red genes in the upper gene line) and for 20 genes, the ribosome density was higher and the ribosome occupancy was not significantly affected (Figure 3B, black genes in the upper gene line). The translation of the mRNAs for 420 genes was down-regulated under stress conditions: the six genes described above for which both translational measures were lower (Figure 3B, green genes in the lower gene line); and 414 genes for which the mRNA showed lower ribosome density and similar ribosome occupancy (Figure 3B, black genes in the lower gene line). For a surprisingly large number of genes (356 genes), the mRNA showed a lower ribosome density but a higher ribosome occupancy in stress than optimal growth conditions (Figures 3B, red genes in the lower gene line). Thus, the translation of the mRNAs of these genes appears to be subject to two antagonistic regulations.
Functional analysis based on category enrichment tests showed a relationship between translation regulation and gene function. Prophage genes (including genes pi116, pi218, pi228, pi231, pi245, pi331, ps207 and ps214) were over-represented in the translationally up-regulated group indicating that the mRNAs of these genes were more translated in stress conditions. Genes involved in growth-related functions were enriched in the translationally down-regulated group: translation [ribosomal genes encoding proteins involved in protein synthesis (fmt, gatB, ksgA, miaA, prmA, rplI, rpsC, truA, truB, ycjD and yhdC), several translational factors (efp, Frr, fusA, lepA and prfB) and amino acyl tRNA synthetases (alaS, aspS, hisS, leuS, lysS, pheS, pheT, proS, thrS and valS)], transcription [genes related to RNA polymerase (rpoB, rpoC and rpoE), to pseudouridine synthase (rluA, rluB, rluC and rluD) and to RNA degradation (pnpA and vacB1)], base metabolism [genes of the purine (guaC, purB, purC, purD, purE, purF, purH, purK, purl, purM and purN) and pyrimidine (carA, carB, pydB, pyrB, pyrC, pyrF and pyrZ) biosynthetic pathways] and fatty acid and phospholipid metabolism (accB, accC, accD, fabF, fabG1, fabZ2, fadD, lplL, plsX and yscE). These growth-related functions were thus down-regulated at the translational level during isoleucine starvation, and are presumably responsible for the decreasing growth rate during the stress. The complete leu-ilv operon involved in the biosynthesis of isoleucine was subject to antagonistic translational regulation: ribosome occupancy was higher and ribosome density lower in stress than optimal growth conditions for all the seven transcripts detected (leuB, leuC, leuD, ilvB, ilvC, ilvD and ilvN; targets of ilvA and leuA were not available on the array).
Different regulatory patterns of gene expression for adaptation to isoleucine starvation
Translation is one of the downstream processes at which gene expression can be controlled. Indeed, protein production is dependent on the regulation of translation (related to ribosome density and ribosome occupancy) and also on the level of mRNA concentration. The concentration of any particular mRNA depends on the balance between transcription rate, degradation rate and rate of dilution by growth. We investigated the relative importance of translation regulation in this multilevel network during isoleucine starvation. We calculated ratios of the transcription, mRNA degradation and dilution rate constants between stress and optimal growth conditions and compared these ratios with those for ribosome occupancy and ribosome density (Additional file1: Table S1).
In the stress condition studied, rate constants of transcription, mRNA degradation and dilution by growth were all substantially lower than in optimal growth conditions (at least 85% of the rate constant ratios were lower than 0.9). Ribosome densities were significantly lower in stress conditions (96% of the ratios ≤ 0.6) whereas ribosome occupancy behavior was diverse (higher (ratios above 1.1) for 42% of the mRNAs and lower for 10%). Isoleucine starvation thus resulted in a general trend of reduced mRNA production by transcription, but slower mRNA loss, both degradation and dilution by growth being slower.
These various ratios were used for hierarchical clustering of the genes (Figure 4). Six gene clusters were thereby defined with different patterns of regulation of gene expression. In clusters 4 to 6, the general regulatory pattern described above was respected, and the differences between clusters mostly involved the magnitude of the difference of the transcription rate constant. Patterns in clusters 1, 2 and 3 differed from the general regulatory pattern. Clusters 1 and 3 consisted of genes for which the mRNA degradation rate constant was higher in stress than optimal growth conditions. The regulation of the expression of genes in cluster 2 involved a much higher ribosome density under stress than optimal growth conditions.
At the functional level, only the prophage category was over-represented in the groups with atypical patterns: 20 prophage genes (Additional file1: Table S1) were found in clusters 1 or 3. This is consistent with the destabilization of their mRNA exhibiting half-lives from more than 47 minutes under optimal growth conditions to an average of 17 minutes under isoleucine starvation[19]. Average mRNA half-lives were 6.2 minutes in optimal growth conditions and 12.1 minutes in conditions of isoleucine starvation, so these prophage mRNAs can still be considered to be stable transcripts. Prophage genes pi116, pi228 and pi245 were classified into cluster 2, in agreement with their presence in the group of genes with enhanced ribosome density under isoleucine starvation.
The isoleucine starvation response has been studied previously by transcriptome analysis: the concentration of mRNAs associated with growth-related functions was demonstrated to decrease whereas that of mRNAs of the isoleucine biosynthetic pathway specifically increased[2]. Our results (Additional file1: Table S1) indicated that the regulation of ilvB, ilvC, leuC and leuD gene expression (classified in clusters 4 or 5) followed the general trend: the higher mRNA levels during isoleucine starvation were not related to an increase of the transcription rate but to a decrease of the transcription rate counteracted by mRNA stabilization and reduced dilution by growth. The higher mRNA levels for ilvD, ilvN and leuB (genes classified in clusters 1 and 3), were associated with increased transcription rates, and for ilvN and leuB with a concomitant antagonistic mRNA destabilization. Genes encoding proteins with the growth-related functions, translation, transcription, fatty acid and phospholipid metabolism and base metabolism were almost exclusively classified in clusters 4, 5 and 6, and displayed the general trend of gene expression regulation: their lower mRNA levels under isoleucine starvation corresponded to lower transcription rate with only small contributions from the opposite influences of mRNA stabilization and reduced dilution by (slow) growth.
Influence of the regulatory pattern of gene expression on protein profiles under isoleucine starvation
We tested whether the patterns of regulation of gene expression, corresponding to the six clusters, were associated with different protein expression profiles. Ratios of protein concentration between stress and optimal growth conditions were calculated for the products of 107 genes, from available proteomics analyses[2]. These ratios were compared to the clustering of the regulation of the expression of the corresponding gene. No particular protein ratio could be associated to any of the six clusters, and the protein ratios did not allow the different clusters to be discriminated (Additional file2: Figure S1). This indicates the absence of any direct relationship between a type of regulatory pattern of gene expression and the resulting regulation of protein levels under isoleucine starvation.
We analyzed, in more detail, proteins related to translation, and in particular ribosomal proteins (Fmt, GatB, RplI, RplV, RpsC and RpsS), translational factors (Frr, FusA, PrfA and PrfC) and amino acyl tRNA synthetases (AspS, LeuS, LysS, MetS, PheT, ProS, ThrS and TrpS). The genes for RpsS, PepP and PrfC were in clusters 1 or 3, and the genes of all the other proteins were in clusters 4 to 6 displaying the general trend of gene expression regulation. Half of these proteins were more abundant in isoleucine starvation than optimal growth conditions, whereas the other half were less abundant. Similarly, other growth-related proteins (five proteins of fatty acid and phospholipid metabolism, 13 of base metabolism and three of transcription), also encoded by genes in clusters 4 to 6, were either more or less abundant under stress; indeed, in this group of growth-related proteins, some of the expression ratios were extreme (0.44 for the reductase FabG1 of fatty acid biosynthesis and above 20 for the synthetase PurM of purine biosynthesis). Two dehydratases of the isoleucine biosynthesis pathway, IlvD and LeuC, encoded by genes in cluster 3 and cluster 5, were significantly more abundant in isoleucine starvation than optimal growth conditions (ratios of 4.3 and 2.8, respectively).