The Proteomic Response to Mutants of the Escherichia coli RNA Degradosome

Abstract

The Escherichia coli RNA degradosome recognizes and degrades RNA through the coordination of four main protein components, the endonuclease RNase E, the exonuclease PNPase, the RhlB helicase and the metabolic enzyme enolase. To help our understanding of the functions of the RNA degradosome, we quantified expression changes of >2,300 proteins by mass spectrometry based shotgun proteomics in E. coli strains deficient in rhlB, eno, pnp (which displays temperature sensitive growth), or rne(1-602) which encodes a C-terminal truncation mutant of RNaseE and is deficient in degradosome assembly. Global protein expression changes are most similar between the pnp and rhlB mutants, confirming the functional relationship between the genes. We observe down-regulation of protein chaperones including GroEL and DnaK (which associate with the degradosome), a decrease in translation related proteins in Δpnp, ΔrhlB and rne(1-602) cells, and a significant increase in the abundance of aminoacyl-tRNA synthetases. Analysis of the observed proteomic changes point to a shared motif, CGCTGG, that may be associated with RNA degradosome targets. Further, our data provide information on the expression modulation of known degradosome-associated proteins, such as DeaD and RNase G, as well as other RNA helicases and RNases – suggesting or confirming functional complementarity in some cases. Taken together, our results emphasize the role of the RNA degradosome in the modulation of the bacterial proteome and provide the first large-scale proteomic description of the response to perturbation of this major pathway of RNA degradation.

Introduction

The RNA degradosome is a vital component of mRNA degradation in Escherichia coli¹ and a main contributor to post-transcriptional gene regulation. The catalytic core enzyme of the RNA degradosome, the endonuclease RNase E (RNase E), is conserved amongst proteobacteria, but homologs have also been identified in archaea and plants ¹. It consists of an N-terminal catalytic region and a C-terminal non-catalytic region which provides a scaffold for binding of other proteins, RNA or other cellular components (Figure 1). The canonical RNA degradosome in E. coli is a 500-700KDa complex formed by the assembly of polynucleotide phosphorylase (PNPase, the product of the pnp gene) which functions as an exonuclease, the DEAD-box helicase RhlB, and the glycolytic enzyme enolase (Eno) on the C-terminal end of the RNase E ¹. Cells lacking RNase E are not viable, most likely due to defects in tRNA transcript processing ². However, truncation mutants lacking the C-terminal scaffolding domain and thus incapable of assembling the degradosome complex are viable although they exhibit slow growth and temperature sensitivity². RNase III (Rnc) and RNase G (Rng) can complement rne mutants whereas RNase R (Rnr) and RNase II (Rnb) can substitute for PNPase¹.

Figure 1. The Escherichia coli RNA degradosome complex.

The figure shows the four main RNA degradosome components, RNase E, Enolase, RhlB, and PNPase. RNase E comprises a catalytic N-terminal region and a non-catalytic C-terminal region to which the other three proteins, RhlB, Eno and Pnp, bind. In addition to interacting with RNase E, RhlB and Pnp also physically interact with each other ⁷. The RNase E truncation mutant contains only the catalytic N-terminal domain; the other three mutants are gene deletions.

Several other proteins have been suggested to associate with the degradosome or play a role in RNA degradation. For example, the poly(A) polymerase PcnB interacts with the DEAD-box helicases SrmB, CsdA and RhlE ³. RNase E co-precipitates with Ppk, DnaK and GroEL in sub-stoichiometric amounts relative to the major degradosome components ¹. Pull-down experiments have further identified a number of proteins bound to degradosome components (e.g. see Figure 3), but the role of these interactions is unclear ⁴.

Figure 3. Expression changes of RNA degradosome proteins and related helicases and RNases.

The figure shows the significance (Z-score) and direction of the protein expression change of the four main mutants as determined by mass spectrometry experiments for the degradosome components and other known RNases and RNA helicases that can potentially buffer for the loss of gene function. Color codes follow the designation in Figure 1. A Z-score of |Z|>1.96 is significant within a P-value of 0.05. The figure shows a range of Z between {-2, 2} for clarity reasons; actual Z-scores range from -15 to 8 (Supplement). Corresponding fold-changes of protein expression can be found in Supplement, Figure S4. The value for the gene in the respective mutant is left out (see Supplement, Figure S1). Triangles indicate the significance of the combined Z-score at the 5% level (see Methods).

While the physiological significance of the E. coli degradosome is well-established, there are still many unresolved questions regarding its function¹. For example, the exact role of the RNA degradosome in modulating mRNA decay and thus protein abundance in vivo is unclear⁵. The significance of the enolase binding to the C-terminal half of RNase E has not been defined apart from its role in modulating the mRNA stability of the glucose transporter ptsG during metabolic stress ⁶. Enolase does not bind RNA, and only 5-10% of the total enolase in cell extracts is complexed with the degradosome ⁵. RhlB and PNPase can also form a complex that degrades double-stranded RNA independently of RNase E ⁷. Furthermore, it is not known how the individual component enzymes cooperate in degrading RNA or how changes in the degradosome composition affect the decay of different transcripts ⁸.

To answer some of these questions at a systems level, several studies have analyzed changes in transcript abundance and mRNA half-lives in E. coli strains deficient in degradosome components, namely in cells carrying null mutations in pnp, eno or rhlB or an RNase E C-terminal truncation that abrogates degradosome assembly (rne(1-602)) ^9-12. Interference with degradosome function may lead to accumulation of un-degraded RNAs in the cell and an increase in their RNA half-lives. However, transcript studies observed a wide range in responses, with a substantial number of transcripts that decreased in half-lives or did not change in concentration ¹¹ suggesting that the regulation of RNA degradation is complex.

The ultimate effect of mutations that influence mRNA half-life is a change in the steady state proteome. Here we present a large-scale quantitative analysis of changes in the proteome of strains lacking the canonical degradosome components and compare our data with the published information on the mRNA half-lives in these mutants. Deep protein abundance analysis was carried out using high-resolution mass spectrometry and a label-free quantitation method developed in our lab ^{13, 14} to estimate the concentration changes of >2,300 proteins. This data provide novel information on how the function of the degradosome shapes the E.coli proteome.

Methods

Data collection

E. coli strains carrying pnp, rhlB, eno mutations or rne(1-602) were kindly provided by Dr. S. N. Cohen (Stanford University) ¹⁰ (Table 1). Cells were grown at 30 °C in M9 media supplemented with 0.2% tryptone, 0.2% glycerol, 1mM MgSO₄ and 0.0001% (w/v) thiamine. For strains 40mM succinate was added to strains K10 and DF261 as described in ref. ¹⁰. Cell pellets were harvested when the culture absorbance reached A₆₀₀=0.6, the cells were resuspended in lysis buffer (25mM Tris HCl, pH7.5, 1mM EDTA , 2.5mM DTT, and 1× Roche (Nutley, NJ) Protease inhibitor) and then lysed by passing three times though a French Press at 20,000psi. Soluble fractions were collected by centrifugation (10,000×g, 10min, at 4°C) and diluted to 4mg/ml. Protein concentration was determined by Nanodrop (Thermo Scientific). Protein samples were heat denatured at 95 °C for 15min. After cooling down to room temperature, trypsin (Sigma) was added at 1:50 (w/w). Digestions were carried out at 37 °C for 24 hours. The sample was again lyophilized to 10μl, resuspended in 120μl buffer (95% H₂O, 5% acetonitrile, 0.1% formic acid) and filtered through a Microcon-10 filter at 12,000g. The samples were stored at -80°C until LC-MS/MS analysis.

Table 1. Strains carrying mutations in the major degradosome components.

All experiments were conducted as described in ref. ¹⁰. The protein expression levels were quantified in mutant and wild-type strains separately, and the Z-score calculation from ref. ¹⁴ used to estimate significance and direction of the expression changes. See Supplement for assessment of data quality.

Name of experiment	Mutant strain	Genotype	No. of proteins quantified in expression change (5% FDR)	Wild-type	Genotype
Δeno	DF261	K10 except for eno-2	1,576	K10	garB10, fhuA22, ompF627(T2R), fadL701(T2R), relA1, pit-10, spoT1, rrnB-2, mcrB1, creC510
Δpnp	YHC012	N3433 except for Tn5∷pnp	1,651	N3433	lacZ, relA, spot1, thi1
ΔrhlB	SU02	N3433 except for ΔrhlB	1,540	N3433	lacZ, relA, spot1, thi1
truncated rne	BZ453	SH3208 except for RNase E truncation	1,545	SH3208	his ΔtrpE5(λ)

Samples (10μl) were injected into an LTQ-Orbitrap Classic (Thermo Electron) mass spectrometer and peptides were resolved using a 5 to 90% acetonitrile gradient over five hours via reverse phase chromatography on a Zorbax BioBasic-18 column 150 mm × 0.10 mm ID (Thermo Fisher). Each of the runs was analyzed independently with Bioworks (Thermo Fisher), searching a database of E. coli protein sequences imported from the NCBI genomic database. The results were combined for analysis by PeptideProphet ¹⁵, ProteinProphet ¹⁶ and post-processed in the APEX pipeline ^{13, 14} to estimate absolute and differential protein expression based on adjusted spectral counts. We accepted proteins as confidently identified if the ProteinProphet probability was above a cutoff corresponding to <5% global false discovery rate. Relative protein expression changes were calculated with respect to measurements in the respective wild-type strain (Table 1), using a Z-score calculation described by Lu et al. ¹⁴. The Z-score adjusts the error model to the absolute concentration (spectral counts). A Z-score of |Z|>1.96 corresponds to a P-value<0.05. Note that Z-scores can be calculated even if one of the samples (mutant or wild-type) has zero counts measured, but a fold-change in expression cannot. We collected at least five technical replicates per sample, i.e. repeat mass spectrometry measurements which were pooled for the quantitative analysis. These replicates also included repeat sample preparations. Replicate concentration measurements correlated well (R²=0.74 to 0.94, not shown) and were pooled for further analysis. The raw data is available at Tranche (ProteomeCommons), data hash: zO+qZFSpP8Z5XzXrSSTm3KYpF2fE2J8F3pDUZpSL5WtS0dIdQ+lF/eGXJHUoXMsRtcdkHzfFmStufDON4S9BaOSjQq0AAAAAAAADIA==. More information on parameters of the mass spectometry experiment and a discussion of data quality is provided in the Suppl. Section 1.

Computational analysis

The mRNA microarray expression data and estimated half-lives were obtained from Bernstein et al. ¹⁰ and consist of relative mRNA expression values for 1,825 genes with missing values estimated using a K nearest neighbor method. Tiling-array data for rne(1-602) cells were from Stead et al. ¹¹ and consist of 2,720 genes for which significant up- or down-regulation is reported.

All data were analyzed with a combination of Perl scripts, R, and the software Perseus ¹⁷. The combined Z-score for a protein's expression change (Figures 3, 4) was calculated as the sum of all Z-scores across all four mutants divided by the square root of the number of Z-scores available for the particular protein: Z_comb=Σ[Z of all contributing mutants]/√[number of contributing mutants]. The combined Z-score is a measure for the general trend in the data: we assume that a protein (mRNA) which is consistently up-regulated across the four RNA degradosome mutants is more likely a direct target of the degradosome than proteins (mRNAs) with low combined Z-scores.

Figure 4. Expression changes of proteins associated with the RNA degradosome.

The figure shows the significance (Z-score) and direction of the protein expression change of several genes with known association with the RNA degradosome or RNA degradation. A Z-score of |Z|>1.96 is significant within a P-value of 0.05. The figure shows a range of Z between {−2, 2}; actual Z-scores range from -15 to 8 (Supplement). Triangles indicate the significance of the combined Z-score at the 5% level (see Methods).

For clustering of the data in Figure 2, up to two missing values were accepted for the proteomics data, and all missing values in the tiling array data were substituted by o. Data were re-normalized to a mean of 0 and standard deviation 1. Clustering was based on the Euclidean distance measure using an average linkage algorithm (Perseus ¹⁷), with a 1.8 Euclidian distance cutoff to determine genes clusters with similar expression profiles. Gene function analysis was performed for the largest clusters using FuncAssociate ¹⁸. Sequence motifs were analyzed using MEME ¹⁹ and fuzznuc ²⁰.

Figure 2. Protein and RNA expression response of degradosome mutants.

The heatmap shows the Z-scores of the protein expression data (first four columns) of the four degradosome mutants mapped to tiling-array based mRNA expression data of the RNase E truncation mutant ¹¹ (rightmost column). Significant function enrichments (adjusted P-value<0.01, compared to entire E. coli genome) are indicated for the four largest clusters (yellow boxes), with cluster sizes in brackets. Only genes without missing data are shown (N=904). Colors saturate at values of |2| to illustrate statistically significant expression changes (P-value<0.05) – actual Z-scores range from -15 to 8. Fold-changes of expression range from -6.0 to 5.7 on a log scale (base 2). t-rne denotes the truncation mutant rne(1-602).

Results

Effects of Δpnp, Δeno, ΔrhlB or rne(1-602) on the E.coli proteome – data quality and general results

We grew E.coli strains carrying Δeno, ΔrhlB, Δpnp, or rne(1-602) alleles to mid-exponential phase and used mass spectrometry-based proteomics of cell lysates to quantify soluble protein expression changes relative to the respective parental (wild-type) strains, in a manner similar to the published mRNA analysis ¹². We provide statistical scores of expression changes for a total of 2,390 proteins, quantifying >1,500 proteins per mutant (Table 1). Figure 2 shows the direction and statistical significance (Z-scores) of the protein expression changes between the mutant strains and their respective wild type cells for the subset of 904 proteins that are completely characterized (no missing values, capped at |Z|=2 corresponding to a P-value=0.05).

Label-free quantitative proteomics reports protein expression changes with a precision of about 2- to 3-fold ¹⁴. The proteomics experiment conducted here correctly detects expression changes of the four mutant proteins with highly significant Z-scores: as expected, the proteins encoded by eno, pnp, and rhlB are absent in the respective null strains whereas in rne(1-602) cells the truncated RNase E(1-602) protein is over-expressed, consistent with previous observations ¹² (Figure S1). For further quality control, we conducted western blot experiments on two soluble proteins, GlmS and CspE (Figure S2). For both proteins, the proteomics and western blot data are in strong agreement for the PNPase mutant, and less so for the RhlB mutant. GlmS concentration drastically increases in the PNPase mutant compared to wild-type both in the proteomics and western blot data. This increase contrasts the concentration decrease at the mRNA level which can be explained by strong regulation of GlmS translation and mRNA degradation by two small non-coding RNAs (GlmY and GlmZ) which in turn interact with the RNA degradosome ²¹. This result is also consistent with our recent study in which we searched for mutants which can up-regulate rraB transcription. In our study, we isolated a glmS mutant in which a Tn5 insertion disrupts the coding region of the GlmS protein ²². RraB disrupts PNPase associated in RNA degradosome both in vivo ²³ and in vitro (L. Zhou, G. Georgiou, unpublished). The proteomics data is consistent with the notion that there is negative feedback between PNPase and GlmS.

In general, the fold-change in expression values between mutant and wild-type for a protein appears to be of similar if not larger magnitude to that of the corresponding mRNA (Figures S4, S6). Examples of fold-changes of protein expression for some proteins are provided in the Suppl. Figure S5. About 25% of the proteins detected increase in relative abundance in all four mutant strains (Δeno, ΔrhlB, Δpnp, rne(1-602)) (Figure 2). When grouping protein expression profiles according to Euclidean distance, the ΔrhlB and Δpnp cells proved to be more similar to each other than any other mutant pairs (Figure 2, Suppl. Figure S4), supporting the known physical and functional interaction between the two enzymes ⁷.

Function analysis of the four largest clusters in Figure 2 indicates a global down-regulation of translation-related proteins (chaperones, ribosomal subunits), primarily in the Δpnp and ΔrhlB mutants (cluster B). Such down-regulation is consistent with regulatory coupling between mRNA availability and translation. However, the proteomics data also demonstrates an up-regulation of 14 of the 20 E. coli aminoacyl-tRNA synthetases in all four mutants (cluster A, Figure 2, S3, P-value<0.001). To the best of our knowledge, a possible link between the RNA degradosome and aminoacylation of tRNAs has not yet been described. We hypothesize that this up-regulation may occur as the cell's effort to counteract the imbalanced concentrations of mRNAs and mature tRNAs due to impaired degradosome function, or possibly as part of a general stress response that produces alarmones ²⁴.

Next, we examined likely degradosome targets for enrichment of sequence motifs which may be recognized by components of the RNA degradation machinery. We derived these likely targets by combining the Z-scores describing the expression change in the four individual mutants into one combined Z-score. Across the protein and RNA expression data, as well as mRNA half-life data ²⁵, we consistently observe enrichment in the CGCTGG motif in which the center CTG triplet is most conserved part (Supplementary notes and datafile 3). This motif occurs more frequently in mRNAs of short half-life than mRNAs with long half-life, and it is enriched in sequences that increase in either mRNA or protein expression levels when RNA degradosome function is impaired. Thus, the CGCTGG can be loosely associated with direct targets of the RNA degradosome. The Supplementary notes discuss this observation and analysis of other sequence features.

The expression response in degradosome components and protein family members

Figure 3 indicates the protein expression changes for the four main degradosome components as well as for other E. coli RNases and RNA helicases detected by the proteomics analysis. Note that not all RNases and RNA helicases are associated with the degradosome, and that the figure caps expression changes at a Z-score of >2 which corresponds to significant expression changes at a P-value of <0.05.

Rnase R (Rnr) is a 3′ to 5′ exoribonuclease closely related to RNase II (Rnb) which is important for decay of mRNA with extensive secondary structures ²⁶. Since the coordination between RNA degradosome components organized by the C-terminal region of RNase E is required for decay of structured RNA substrates, it is tempting to assume that the observed increase in the cellular concentration of RNase R in cells expressing the truncated RNase E(1-602) (Figure 3) is due a defect in structured RNA observed in this strain. Similarly, RNase BN (Rbn) is implicated in the processing of tRNA precursors ²⁷. RNase BN abundance decreases in both the rhlB and pnp mutants (Figure 3), but is unchanged in rne(1-602) (or in the eno mutant). The RNase BN's (Rbn) inverse expression pattern with RNase II (Rnb) is consistent with their overlapping functions in vivo ²⁸: down-regulation of one enzyme may be compensated for by up-regulation of the other enzyme and vice versa.

E. coli has five known DeaD-box RNA helicases (ribonucleases): RhlB, RhlE, SmrB, DeaD (also called CspA), and DbpA ¹. Null mutants of these helicases result in viable cells, but can affect growth rate ²⁹. DeaD, SrmB and RhlE have been reported to associate with RNase E in vitro ^{5, 30, 31}. DeaD and RhlE have evidence for their role in RNA degradation ¹; they can complement the RNA helicase function in the absence of RhlB ³², whereas SrmB cannot ³³. The binding sites for DeaD, SrmB and RhlE within the C-terminal end of RNase E are different from those of RhlB, and overlap with the Enolase binding site (residues 833-850, Figure 1). We observe an overall increase in the protein expression levels of DeaD, SrmB, and RhlE across the four mutants which is largely consistent with the functional complementarity between the enzymes and degradosome components (Figure 3).

HrpA is also an RNA helicase (although not from the DEAD-box family) that functions in mRNA processing ³⁴. Its expression level is drastically down-regulated in both the rhlB and pnp mutants, similar to the effect observed for another RNA helicase, YejH (Figure 3). The expression patterns of these two helicases contrast those of the DeaD-box helicases.

RNase III (Rnc) is involved in ribosomal RNA processing ³⁵, an activity which overlaps with the essential role of RNase E in E. coli. Similar to RNase R, we observe significant up-regulation of RNase III in the rne(1-602) mutant consistent with their complementary roles in rRNA processing and in ribosome biogenesis (Figure 3). Although RNase G (Rng) is an RNase E parolog that cannot fully complement RNase E function ³⁶, RNase G concentration also increases in three of the four mutants (Figure 3). Both the endonuclease RNase G and the DeaD helicase are up-regulated in the eno, pnp and rne mutants, a finding consistent with a recent genetic analysis by Tamura et al. showing that mutation in DeaD can suppress RNase E in a temperature-sensitive rne mutant¹².

RNase A (rna) is involved in tRNA and rRNA processing ³⁷, and our data shows slight down-regulation in the RNase E mutant (Figure 3). The oligonucleotide-ribonuclease encoded by orn acts downstream of the RNA degradosome where it regenerates single nucleotides from oligonucleotides ³⁸. Its expression levels changes are similar to RNase E and RNase G except for the rne mutant (Figure 3).

The expression response of other degradosome associated proteins

We also assembled a number of genes which have some known association with the RNA degradosome based on pull-down interaction data and other evidence ¹, and for which our proteomics data provided information (Figure 4). Since these proteins are neither RNases nor RNA helicases which are discussed above, their role in RNA degradation is presumably indirect.

In particular, three chaperones (GroEL, GroES, and DnaK) and two ribosomal proteins (RplB, RpsC) are strongly down-regulated in the four mutants (somewhat less in the RNase E truncation)(Figure 4) – corresponding to cluster B in Figure 2. The RNA polymerase RpoA is significantly up-regulated in the rhlB and pnp mutants, but down-regulated in the eno and rne(1-602) mutant strains (P-value<0.05)(Figure 4), suggesting feedback between transcription and degradation regulation. The poly(A) polymerase PcnB is up-regulated in all four mutants in our data as might be expected since these strains display varying degrees of defects in RNA processing and PcnB may counteract this defect. The addition of poly(A) tails (by PcnB) generally destabilizes RNA in bacteria ³⁹, and it facilitates degradation of REP-stabilizers, i.e. structural elements preventing degradation, with the help of RhlB ¹.

The protein expression response to perturbation of the RNA degradosome differs from the RNA response

We compared the protein expression changes to published steady state transcriptome and tiling array data ^{10, 11} to investigate the relationship between transcript and protein expression. Since partially degraded RNA can still hybridize to arrays, and translation efficiency of partially degraded RNAs is likely affected in the mutants as well, we did not expect much agreement between the protein and mRNA data. Indeed, the protein expression data do not correlate with either mRNA expression changes or mRNA half-life changes measured by microarrays (Suppl. Figure S4)¹⁰.

mRNAs that are not processed in the rne(1-602) mutant (which is unable to form the degradosome complex) may still be translated into proteins. Inspection of the published tiling array data ¹¹ in Figure 2 (rightmost column) indicates that about one third of the transcripts for which protein abundance data could be obtained are possibly be processed in a degradosome-dependent manner (Figure 2, clusters A, D, (227+98)/904=36%): their expression levels increase in tiling array in the rne(1-602) mutant. Of these, a large number of transcripts also show positive protein expression changes in the rne(1-602) mutant (Figure 2, second column from the right), but the agreement between tiling array data and protein expression data for the RNase E mutant is generally low. Only one quarter (227/904, cluster A) of the proteins in the dataset are largely up-regulated in their protein expression patterns across all four mutants and in the tiling array data. The remaining set of proteins may accumulate at the mRNA level, but their functional half-life does not seem to be affected, and protein expression levels are not up-regulated (cluster D). Similarly, a number of genes are either unaffected in their RNA expression levels by the RNase E truncation or down-regulated (clusters B, C), and these genes may be degraded by pathways other than the canonical degradosome. Again, the protein data matches the mRNA data only partially (cluster B).

Conclusions

We present a large-scale proteomics analysis of mutants of the E. coli RNA degradosome, complementing existing data on mRNA expression and half-life changes ¹⁰. Since measurements of mRNA half-lives do not inform on translation activity of the transcripts and the final effect of RNA degradation on protein abundance, proteomics analysis provides an important additional angle to our understanding of RNA degradation. The proteome-wide data provided here indicates several trends and novel associations that can be followed up by future experiments. Expression data of >2,300 proteins reveal substantial differences between the effects of perturbed RNA degradation on transcript and on protein concentrations: fewer than half of the genes in our dataset appear to follow the canonical model of RNA degradation (Figure 2). A trivial explanation of this discrepancy lies in technical measurement noise. However, the estimated error in protein expression measurements is smaller than the observed variation between experiment and control here ^{14, 40}, supporting the existence of true biological trends.

The complexity of E. coli RNA degradation can be partially explained by the up-regulation of several RNA helicases and RNases who may buffer for loss of enzyme function in the degradosome mutants. We are able to provide proteomic characterization for RNA helicases and RNases known to have such buffering roles (DeaD, Rnr, Rng)¹, but also suggest possible complementary functionality for additional enzyme (HrpA, YejH, Rna, Orn) – but their precise interaction with the degradosome needs to be tested in future experiments. We provide two novel findings that have not yet been discussed in the context of the RNA-degradosome. First, we observe a large number of aminoacyl-tRNA synthetases to be up-regulated in the pnp, rhlB, and rne mutants, suggesting a link to tRNA metabolism. Second, we observe enrichment of the CGCTGG motif in sequences that are likely direct targets of the RNA degradosome – the role of which is entirely unknown.

Our results provide insights into the cellular role and structure of the RNA degradosome itself. Since only a small fraction of the PNPase and enolase molecules participate in the degradosome complex ^{5, 41}, it is unsurprising that their mutant expression profiles are highly discordant with those of the other mutants, and highlight the roles of these proteins outside the degradosome. However, the profiles of the rhlB and pnp mutants correlate better with each other than the profiles of the other mutants (R=0.72), consistent with their physical and functional interaction independently of the RNase E scaffold ⁷. Interestingly, this correlation between the rhlB and pnp mutant is observed for protein expression (this study) and mRNA half-lives ¹⁰, but not for published mRNA expression data (Figure S4), suggesting a highly dynamic and complex system of RNA degradation.

The lack of correlation amongst the datasets also suggests extensive regulation at the level of translation and protein degradation as well as RNA degradosome functions outside the removal of mRNAs. Indeed, even in unperturbed wild-type bacterial systems, a low correlation between mRNA and protein concentrations has indicated that translation and protein degradation regulation account for substantial amounts of protein expression variation ^{14, 42}. The same is true for eukaryotes; transcription and RNA degradation account for only half of the variation in steady-state protein expression levels ^{40, 43}. Further, mRNA half-lives and transcript concentrations are inversely correlated in E. coli ⁹, suggesting additional layers of regulatory complexity.

Finally, the discordance between protein and RNA expression levels observed is also consistent with a model that distinguishes between the chemical and functional half-life of mRNAs ¹. The mRNA expression data indicates changes in the chemical half-life, i.e. the physical presence of the transcript. The functional half-life may be very different from the chemical half-life, affecting the translation activity associated with the mRNA and influencing the final protein expression levels. An mRNA may be physically present in the cell, but translationally inactive and thus non-functional. Our proteomics data delivers supporting evidence for the distinction between chemical and functional half-lives of mRNAs, but detailed biochemical studies will be required in order to confirm and further characterize this mode of regulation.