The Protein Interaction of RNA Helicase B (RhlB) and Polynucleotide Phosphorylase (PNPase) Contributes to the Homeostatic Control of Cysteine in Escherichia coli*

PNPase, one of the major enzymes with 3′ to 5′ single-stranded RNA degradation and processing activities, can interact with the RNA helicase RhlB independently of RNA degradosome formation in Escherichia coli. Here, we report that loss of interaction between RhlB and PNPase impacts cysteine homeostasis in E. coli. By random mutagenesis, we identified a mutant RhlBP238L that loses 75% of its ability to interact with PNPase but retains normal interaction with RNase E and RNA, in addition to exhibiting normal helicase activity. Applying microarray analyses to an E. coli strain with impaired RNA degradosome formation, we investigated the biological consequences of a weakened interaction between RhlB and PNPase. We found significant increases in 11 of 14 genes involved in cysteine biosynthesis. Subsequent Northern blot analyses showed that the up-regulated transcripts were the result of stabilization of the cysB transcript encoding a transcriptional activator for the cys operons. Furthermore, Northern blots of PNPase or RhlB mutants showed that RhlB-PNPase plays both a catalytic and structural role in regulating cysB degradation. Cells expressing the RhlBP238L mutant exhibited an increase in intracellular cysteine and an enhanced anti-oxidative response. Collectively, this study suggests a mechanism by which bacteria use the PNPase-RhlB exosome-like complex to combat oxidative stress by modulating cysB mRNA degradation.

mRNA is unstable, and its stability is post-transcriptionally controlled to quickly change gene expression to adapt growth to new environments. In Escherichia coli, the RNA degradosome comprises many enzymes and minor components (1)(2)(3)(4) that can help to respond to environmental changes by modulating mRNA stability. The major components include the endoribonuclease RNase E, the 3Ј to 5Ј single-stranded RNA degrading enzyme polynucleotide phosphorylase (PNPase), 3 the DEAD box RNA helicase RhlB, and the glycolytic enzyme enolase (5)(6)(7). Although PNPase and RhlB can interact independently of RNase E (8,9), the importance of this interaction has not been studied in depth.
The bacterial PNPase complex is a trimeric complex, and each PNPase has two RNase PH domains followed by S1 and KH domains, all of which are important for RNA binding (10). The six RNase PH domains of the trimeric complex are arranged around a central pore with a diameter of ϳ14 Å-just large enough to accommodate a single single-stranded RNA molecule (11). This structural feature ensures that the bacterial PNPase complex exhibits a preference for the degradation of single-stranded RNA, and functionality ceases when the machinery arrives at a long double-stranded stem in the 3Ј end of mRNA (12). Thus, in vivo, the degradation of RNA by PNPase probably requires cofactors such as RNA helicase to efficiently unfold RNA secondary structures.
In E. coli, the helicase RhlB interacts with PNPase independently of RNA degradosome formation (8,9). In vitro analysis has shown that RhlB helps PNPase to degrade double-stranded RNAs (8). Moreover, the reconstitution of a minimal RNA degradosome demonstrated that RhlB enables PNPase to mediate the degradation of a repetitive extragenic palindrome sequence-containing transcript, i.e. malEF (13). Interestingly, eukaryotic and archaea cells have 3Ј to 5Ј exoribonuclease complexes with a core-exosome that is structurally similar to PNPase (14,15). It has also been shown that the eukaryotic exosome associates with a variety of accessory factors in a cell compartment-and species-dependent manner to mediate RNA degradation and processing (16 -23). It is not yet understood how a ribonuclease-protein complex selects its specific mRNA substrate and thus specifically controls degradation.
In this study, we examined the importance of the protein interaction between RhlB and PNPase for mRNA stability in the absence of the degradosome. We isolated an RhlB mutant, RhlB P238L , with an impaired PNPase, but not RNase E, interaction. Microarray analysis of cells bearing this mutant protein revealed altered expression profiles of cysteine regulon genes responsible for control of cysteine biosynthesis. In E. coli, cysteine biosynthesis is mediated by the cysteine regulon that includes 7 operons: sbp, cysPUWA, cysM, cysJIH, cysDNC, cysK, and cysE, containing 14 genes. With the exceptions of sbp and cysM, deficiency of any other gene of the cysteine regulon renders the cell a cysteine auxotroph (24,25), underscoring the importance of controlled expression. Cells with reduced RhlB-PNPase interactions also have a prolonged half-life of cysB mRNA, a dual transcription factor (26) that activates the expression of all cysteine regulon genes except cysE. Further, cells expressing the RhlB P238L mutant exhibit increased intracellular cysteine and increased anti-oxidative ability. These data provide a possible mechanism by which bacteria may modulate cysteine biosynthesis through the exosome-like complex to combat oxidative stress.
Screening of a Mutant RhlB That Reduces Interactions between RhlB and PNPase Using a Bacterial Two-hybrid System-Random mutagenesis was performed using an error prone (1-3 mutations per rhlB DNA fragment) PCR kit (GeneMorph II random mutagenesis kit; Stratagene), and mutants with weakened protein interactions were identified as per the method described by Karimova et al. (31). In brief, rhlB DNA fragment PCR products resulting from the error prone PCR were digested with PstI and BamHI, followed by cloning into a pT25 plasmid that expresses a T25 fragment corresponding to amino acids 1-224 of CyaA (adenylate cyclase) as an N-terminal tag. The resulting plasmid was named pT25RhlB. Wild-type pnp with a T18 plasmid expressing the T18 fragment corresponding to amino acids 225-399 of CyaA as a C-terminal tag was also prepared (pPNPT18). Only tagged interacting protein partners can induce CyaA activity by bringing the N-and C-terminal regions of CyaA together. Mutated pT25RhlB pool and wildtype pPNPT18 (8) were cotransformed into a DHP1 strain (an adenylate cyclase-deficient derivative of DH1) to screen for protein-protein interactions as described (31). ␤-Galactosidase activity assays were performed as described previously (8) to measure the strength of interactions between mutant RhlB and PNPase in vivo. To exclude false positive results caused by different protein expression levels, we performed Western blotting using ␣-RhlB antibody to confirm the expression level of mutant T25RhlB proteins. White transformants with a similar expression protein level to wild-type T25RhlB were selected and isolated for sequencing confirmation.
BIAcore Surface Plasmon Resonance Analysis-Real time protein-protein interaction strength was examined by means of a Biacore instrument (Biacore X) as described (8). pFlagRhlB (8), pFlagRhlB P238L , and pFlagPNP (8) were used for protein purification as described previously (6). The identified mutation (P238L) was introduced into pFlagRhlB by QuikChange II XL site-directed mutagenesis kits (Stratagene). Purified FlagPNPase was immobilized on a CM5 sensor chip using an amine-coupling kit (Amersham Biosciences). Different concentrations, ranging from 0.75 to 24 M of purified FlagRhlB wt or FlagRhlB P238L , were injected with a constant (10 l/min) flow rate at 25°C. The kinetic analysis was performed using BIAcore evaluation software.
Helicase Activity Assay-Helicase activity was measured as described previously (8), with minor modifications. The 5Ј-labeled longer strand and the unlabeled shorter strand RNA were hybridized to form duplex RNA in a 1:1 ratio. The duplex RNA was incubated with purified FlagRhlB wt or FlagRhlB P238L at 30°C in final reaction volumes of 20 l containing 20 mM Tris-HCl, pH 7.5, 5 mM MgCl 2 , 0.1 mM DTT, 100 mM NaCl, 5% glycerol, and 0.1% Triton X-100. Samples were separated on a 16% polyacrylamide (19:1 bis-acrylamide) native gel, visualized by autoradiography, and quantified by LAS-1000 plus (Fuji Film).
Microarray Measurements of RNA Abundance and Data Processing-BL21(DE3) rne131 ⌬rhlB carrying FLAG-tagged wild-type or mutant (P238L) RhlB were grown at 30°C in LB medium to an OD 600 of 0.55-0.6. To distinguish the effect of the RhlB-PNPase interaction from that of the RNA degradosome, we used an E. coli strain with a truncated rne gene (rne131) encoding the N-terminal domain of RNase E (residues 1-584) (32,33). RNA was isolated according to the RNeasy mini kit (Qiagen) manufacturer's protocol. Technical support for microarray experiments was provided by the Institute of Molecular Biology (Academia Sinica) Microarray Core Facility. The relative mRNA levels were determined by parallel twocolor hybridization of DNA microarrays as described previously (34,35). RNA samples taken from BL21(DE3) rne131 ⌬rhlB expressing FlagRhlB wt or FlagRhlB P238L were synthesized into cDNA and then labeled with Alexa Fluor 647 (Molecular Probes, Invitrogen). Relative mRNA abundance was measured using BL21(DE3) rne131 ⌬rhlB cells expressing FLAG tag only as reference, and the RNA sample was then synthesized into cDNA and labeled with Alexa Fluor 555 (Molecular Probes, Invitrogen). Synthesis of cDNA, hybridization, and analysis of spots were performed as described (35). The microarray data have been deposited at GEO database (GSE: 57784). Assistance with data analysis was provided by the Institute of Molecular Biology Bioinformatics Core Facility. The microarray data were first subject to intensity-dependent LOWESS normalization using the "per spot and per chip" setting in the GeneSpring software (Agilent Technologies). To find the significantly expressed genes within each of the sample triplicates, we subjected gene lists to significance analysis for the microarray package, implemented in the TIGR MultiExperiment viewer (The Institute for Genomic Research, Rockville, MD). The missing values were imputed before testing using the K nearest neighbor method, where K ϭ 6. The false discovery rates within and among sample groups were estimated by a bootstrap resampling method, and false discovery rate thresholds of 5% or less were established to obtain significantly expressed genes.
RNA Stability Assay-Bacteria were grown in LB medium at 30°C to an OD 600 of 0.5-0.6, and total RNA was extracted as described previously (36) at different time points after the addition of rifampicin (0.5 mg/ml). Specific RNA was detected by RNA probe labeling with digoxigenin. Briefly, 5 g of total RNA of each strain was separated on 1.2% formaldehyde agarose gel and transferred to a Hybond Nϩ membrane (Amersham Biosciences) through capillary action with a stack of towels 10 centimeters high. For hybridization, RNA probes were internally labeled with digoxigenin-11-UTP and hybridized to the membrane with a digoxigenin Northern starter kit (Roche) (37). Northern blot signals were visualized using a UVP digital image system and quantified via ImageJ software 1.50b.
Growth Inhibition Assay-Bacteria were grown at 37°C to an OD 460 ϭ 0.4 in Vogel-Bonner medium E (0. 8 (38), supplemented with 0.5 mM glutathione as the organic sulfur source and 0.2% glucose as the carbon source. The culture was further divided into two flasks supplied with either 0.5 mM cysteine or an equal volume of autoclaved double distilled H 2 O. The OD 460 was measured every 20 min for an hour, and growth inhibition (%) was calculated using the following formula, where ⌬OD 460 is the difference in A after 1 h of incubation.
Anti-oxidative Stress Assay-The bacterial strain BL21(DE3) rne131 ⌬rhlB ⌬cysE ⌬tnaA was used to examine whether impaired RhlB-PNPase interactions resulted in impaired antioxidative resistance. To measure the effect of cysteine biosynthesis, we removed chromosomal cysE and induced expression of a cysteine-insensitive mutant (CysE M256I ) under the control of its own promoter (39 -41). A PCR-generated EcoNI-NdeI fragment encoding the full transcription unit and the promoter of cysE was cloned into pACYCDeut-1 (EMD Millipore), and the M256I mutation was introduced into pACYCDeut-CysE by QuikChange II XL site-directed mutagenesis kits (Stratagene). To analyze the effects of weakened RhlB-PNPase interactions on cysteine synthesis, chromosomal tnaA was removed and replaced by a kanamycin cassette as described under "Bacterial Strains and Plasmids" above. The strains containing pFlagRhlB wt or pFlagRhLB P238L were grown in LB medium at 37°C overnight. The overnight cultures were further diluted to an OD 600 value of 0.1, and 2 l of each diluted culture was spotted on LB plates with 0 -1 mM H 2 O 2 (Merck) as indicated in Fig. 8. For the oxidative stress induced by paraquat (PQ; Sigma), 2 l of each culture was serially diluted and spotted on the plate with or without 0.4 mM PQ. The plates were incubated at 37°C overnight, and the stress resistance was measured by observing colony formation ability.
Measurement of Cysteine Content from Bacterial Metabolite Extract by LC-MS/MS-Bacteria were grown at 37°C to an OD 460 ϭ 0.4 in M9 medium (2.2 mM KH 2 PO 4 , 1.87 mM NH 4 Cl, 4.23 mM Na 2 HPO 4 , 0.86 mM NaCl) supplemented with 0.2% glucose, 1 mM MgSO 4 ⅐7H 2 O, 0.1 mM CaCl 2 , 1 g/ml vitamin B 1 , and 100 mg/liter each of L-isoleucine, L-leucine, L-methionine, and glycine amino acids. The amino acids were added to avoid the growth inhibition caused by increased L-cysteine (42,43). The metabolites were extracted as described previously (44), with 21 g/ml L-[1-C13] cysteine (Icon Services Laboratory) as an internal standard. The final extracted metabolites were lyophilized using a miVac Duo concentrator (GeneVac) and then resuspended in 700 l of 50:50 methanol/water. LC-MS/MS analysis was supported by the Metabolomics Core Facility of the Scientific Instrument Center (Academia Sinica). LC-MS/MS was performed on a LTQ-Orbi Elite system (Thermo Scientific) equipped with an ACQUITY ultra performance LC (Waters). For ultra performance LC-MS analysis, the mobile phases for positive electrospray ionization consisted of acetonitrile/water (50:50)/10 mM NH 4 OAc, pH5.0 (buffer A) and acetonitrile/water (99:1)/10 mM NH 4 OAc, pH5.0 (buffer B). The samples were separated on an ACQUITY ultra performance LC BEH Amide column (Waters, 2.1 mm ϫ 100 mm, 1.8 m) by gradient elution (50 -99% buffer A in 1-4 min with a flow rate 400 l/min), with a column temperature of 25°C. Mass spectrometric conditions were set to spray voltage of 3.2 kV, sheath gas flow rate of 50 liters/min, auxiliary gas flow rate of 15 liters/min, and capillary temperature of 360°C. Full scan MS spectra were acquired in the Orbitrap (m/z 50 -250). The most intense ions were selected, and a 38eV higher energy col- DECEMBER 11, 2015 • VOLUME 290 • NUMBER 50

JOURNAL OF BIOLOGICAL CHEMISTRY 29955
lision dissociation with a 2 m/z isolation width was utilized. The signal to noise ratio was set to 3.0.
We determined the limit of detection using L-12C-cysteine (Sigma) standard solutions prepared in extraction buffer (50:50 methanol/water), and the limit of detection was 3.5 g/ml. The standard curve was generated using L-12C-cysteine at six different concentrations (3.5-35 g/ml), and 21 g/ml of L-[1-13C]-cysteine was added into each standard solution as an internal control. The standard curve was generated using the peak areas ratio of L-C12-cysteine and L-[1-13C]-cysteine. The cysteine contents extracted from different strains were calculated using the standard linear equation, where y is the peak area ratio, and x is the concentration of L-12C-cysteine (45).

P238L Mutation on RhlB Affects Its Interaction with PNPase but
Not RNase E-It has previously been shown that PNPase and RhlB can interact independently of RNase E degradosome formation by bacterial two-hybrid and immunoprecipitation experiments (8,9), but the importance of this interaction has remained unclear. To address this question, we isolated a mutant RhlB with impaired ability to interact with PNPase through a bacterial two-hybrid system. Random mutations were introduced into the rhlB gene by error prone PCR, and the resulting constructs were cloned into the pT25 plasmid (31). This pT25RhlB plasmid with randomly introduced mutations was cotransformed with wild-type pPNPT18 into the DHP1 strain for blue and white colony selection. T25RhlB with normal PNPT18-interacting ability induces CyaA activity, and its interaction with the promoter of the reporter gene LacZ results in blue colonies on isopropyl ␤-D-thiogalactopyranoside/X-gal plates (31). We isolated plasmids from 314 white colonies of a total of 2000 on isopropyl ␤-D-thiogalactopyranoside/X-gal plates, indicative of cells potentially harboring a mutant RhlB with impaired PNPase interactions. After sequencing, we identified a mutation, P238L, located in the C-terminal region of RhlB. Because residues 194 -421 of RhlB have previously been shown to interact with PNPase and with residues 684 -784 of RNase E, RE12 (8), we further examined this RhlB P238L mutant to see whether it interacts with RE12 like wild-type RhlB. The RhlB P238L mutant interacted normally with RE12 (Fig. 1A,  panel a versus panel b) but exhibited reduced interaction with PNPase compared with that of wild-type RhlB (Fig. 1A, panel d  versus panel e). Consistently, the ␤-galactosidase activity of E. coli cells carrying the pT25RhlB P238L mutant and pPNPT18 was less than that of pT25RhlB wt and pPNPT18 (Fig. 1B) but similar between the strains carrying pT25RhlB wt and pRE12T18 or pT25RhlB P238L and pRE12T18. Western blots confirmed that the protein abundance of T25RhlB P238L was comparable to that of T25RhlB wt (Fig. 1C), confirming that the reduced color in the E. coli two-hybrid assay and the reduced ␤-galactosidase activity are due to weakened interaction between the RhlB P238L mutant and PNPase. These data suggested that Pro-238 of RhlB is important for interactions with PNPase.
Mutation P238L Reduces the Association Rate of RhlB and PNPase-We next examined whether the association or dissociation rate constants of RhlB and PNPase were affected by the P238L mutation. The binding profiles of purified proteins FlagPNPase, FlagRhlB wt , and FlagRhlB P238L were analyzed using BIAcore surface plasmon resonance (Fig. 2, A and B). We found that the association rate constant of RhlB P238L -PNPase (3.11 ϫ 10 2 M Ϫ1 s Ϫ1 ) is only 25% that of RhlB wt -PNPase (1.26 ϫ 10 3 M Ϫ1 s Ϫ1 ) (Fig. 2C). In contrast to the association rate constant that was affected by the P238L mutation, the dissociation rate constant remained similar for interactions between both the wild-type and P238L-mutated RhlB and PNPase (Fig. 2C). These results indicate that the P238L mutation decreases the binding affinity of RhlB with PNPase by reducing the rate of association.
RhlB P238L Mutant Retains Similar ATP-dependent Helicase Activity as RhlB Wild Type-To clarify whether the P238L mutation on RhlB affects the unwinding activity of RhlB, we performed an RNA helicase activity assay. As shown in Fig. 3, FlagRhlB wt and FlagRhlB P238L were able to unwind duplex  (8). The colonies carrying pT25 with pRE12T18 (panel c) or pPNPT18 (panel f) were negative controls. B, ␤-galactosidase assays were used to study the in vivo strength of protein-protein interactions. The interactions of PNPT18 (black bar) or RE12T18 (gray bar) with different T25RhlB constructs are shown. The reaction activities are given relative to the activity of T25RhlB wt , which was set as 100%. C, Western blot; total cell lysate of cells carrying the plasmids encoding different test proteins as indicated were resolved and detected using ␣-RhlB antibody.
RNA (lanes 2-5 and 7-10, respectively), and these reactions were ATP-dependent (compare lanes 2-5 to lane 6 and lanes 7-10 to lane 11). This result indicated that the P238L mutation on RhlB does not affect its ATP-dependent RNA unwinding activity.
RhlB P238L Mutation Alters mRNA Abundance of Cysteine Regulon Genes by Regulating the Degradation of cysB-Our previous in vitro study showed that RhlB helps PNPase degrade duplex RNA (8). To investigate the effects of RhlB and PNPase interactions independently of RNase E degradosome formation on mRNA abundance in vivo, we designed a microarray assay to examine changes in mRNA expression in cells harboring the RhlB C-terminal mutation P238L. Wild-type and mutant RhlB were introduced into an E. coli strain lacking the rhlB gene and carrying a truncated rne gene (BL21(DE3) rne131 ⌬rhlB), with the rne131 mutation encoding a truncated RNase E (residues 1-584) that prevents it from forming a degradosome complex with RhlB and PNPase (29); both wild-type and mutant RhlB were expressed at comparable levels (data not shown). The microarray data were analyzed as described under "Experimental Procedures." Among the 220 significantly expressed genes, expression of FlagRhlB P238L increased the abundance of the messages encoded by 69 genes (31.4%) more than 1.5-fold (Fig. 4A). Interestingly, 11 of the 69 genes belong to the cysteine regulon, strongly suggesting that the interaction of RhlB and PNPase is involved in regulating the expression of this set of genes, which is responsible for cysteine biosynthesis.
We next explored whether the altered expression of the cysteine regulon occurs at the transcriptional or post-transcriptional level. Northern blots revealed no significant differences in the stability of cysJIH and cysK transcripts in the strains expressing FlagRhlB wt and FlagRhlB P238L (Fig. 4, B and C). However, consistent with microarray data, the abundance of both cysJIH and cysK transcripts were increased in cells with a weakened RhlB-PNPase interaction (Fig. 4, B and C), indicating altered transcriptional control over these transcripts.
Taken together with the changes in multiple cysteine regulon genes and the fact that this control appeared to be occurring at the transcriptional level, we focused attention on the regulatory protein CysB. CysB is a dual transcription factor that is responsible for activating expression of the cysteine regulon with the exception of cysE. The attenuated RhlB-PNPase interaction resulted in stabilization of the cysB transcript ϳ1.8-fold (Fig.  4D). These data indicate that RhlB-PNPase post-transcriptionally regulates the stability of cysB and controls the expression of the cysteine regulon.
RhlB-PNPase Regulates the Stability of cysB Independently of the RNA Degradosome-To clarify the effect of RNA degradosome formation on the RhlB-PNPase target cysB, we compared the stability of cysB in the strain with full-length or C-terminally truncated RNase E and found that RNA degradosome formation does not affect the stability of cysB (Fig. 5A). We found that disruption of the interaction between RhlB and PNPase increases the stability of cysB. Thus, we wondered whether RhlB is required to regulate the degradation of cysB. The results showed that deletion of rhlB stabilized cysB more than 1.5-fold in the background of both full-length and C-terminal truncated RNase E (Fig. 5, A versus B). These results demonstrated that   DECEMBER 11, 2015 • VOLUME 290 • NUMBER 50

JOURNAL OF BIOLOGICAL CHEMISTRY 29957
RhlB is required for regulation of the stability of cysB independently of RNA degradosome formation.
The Degradation of cysB Requires the Activity of RhlB, PNPase, and PcnB-Our results showed that interaction between RhlB and PNPase is required to regulate degradation of cysB, so we then set out to ascertain whether individual RhlB or PNPase activity is also required. The cysB stability was determined with DEAD box motif mutated RhlB E166K (29), a PNPase deletion, or low phosphorolysis activity PNPase N435D (30) (Fig.  6, A-C). The results showed that the stability of cysB was comparable in rhlB deletion and FlagRhlB E166K -expressing strains (Fig. 6F). Furthermore, the stability of cysB was comparable in the pnp deletion strain complemented with wild-type PNPase and in the parental BL21(DE3) rne131 strain (Fig. 6F). Those results suggested that in addition to the protein-protein interaction, the activities of both RhlB and PNPase are required for cysB degradation. Combining pnp and rhlB deletions had no additional effect (Fig. 6D), thus demonstrating that PNPase and RhlB were regulating cysB degradation in the same pathway. Additionally, it has previously been shown that the Rho-independent terminator serves as a signal for polyadenylation by PcnB (46). The polyadenylated transcript has also been suggested to facilitate RNA degradation by PNPase (47). To examine the effect of polyadenylation on the degradation of cysB, we determined stability of cysB in the pcnB deletion background and found that the half-life of cysB is similar to that of the FlagRhlB P238L -expressing strain (Fig. 6E versus right panel of  Fig. 4B). Consistent with our result that the RhlB-PNPase com- . RhlB P238L mutation alters mRNA abundance of cysteine regulon genes by regulating the degradation of cysB. A, relative mRNA abundance was measured as described under "Experimental Procedures" and arranged according to the n-fold change (Ϫ2 Ͻ n Ͻ 2) between FlagRhlB P238L and FlagRhlB wt . The microarray data show the top 30 genes (shown in three replicates), whose expression levels were altered by the expression of the P238L mutant RhlB. Gene names are shown to the right of the panel. A reference color bar on the left indicates the correlation between the observed color patterns and quantitative changes. Closed circles indicate genes belonging to the cysteine regulon. B and C, the steady-state level and the stability of cysJIH (B) and cysK (C) in BL21(DE3) rne131 ⌬rhlB containing FlagRhlB wt or FlagRhlB P238L were analyzed by Northern blot. D, the stability of the transcript cysB was analyzed and showed 1.8-fold stabilization upon expression of FlagRhlB P238L . The culture samples were collected at different time points after rifampicin treatment (Time af Rif.) as indicated. The half-life (T1 ⁄2 ) of each transcript was determined using the intensity of signals normalized to 23S rRNA, and the values are shown in a semi-logarithmic plot at the right of each panel.
plex regulates the degradation of cysB in a 3Ј to 5Ј direction, PcnB also facilitates degradation of cysB at its 3Ј-end. Based on these results, under "Discussion," we present a molecular mechanism for regulation of cysB mRNA degradation in a 3Ј to 5Ј direction that is influenced by RhlB, PNPase, and PcnB, independently of degradosome formation.
The Effect of Disrupting RhlB-PNPase Complex Formation in Controlling the Homeostatic Level of Cysteine-We have shown that interacting RhlB-PNPase regulates the stability of cysB and, further, that it modulates the expression of the cysteine regulon. Because cysteine is essential for many normal cellular functions, we examined the broader biological consequences of impaired interactions between RhlB and PNPase. Cysteine plays a major role in maintaining an intracellular reducing environment, thus protecting against oxidative stress (48). We treated E. coli carrying the wild-type or P238L mutant RhlB with the oxidative stress inducers H 2 O 2 and PQ (an agent that diverts electrons from NADH or NADPH to molecular oxygen to generate a flux of superoxide (49)) and examined growth ability. We observed improved survival in the P238L mutant, suggesting improved oxidative stress resistance (Fig. 7,  A and B).
Cysteine is also known to inhibit cell growth (50), so we also studied growth inhibition in wild-type and P238L mutants by adding 0.5 mM cysteine to the growth media, which we anticipated would amplify the growth inhibition induced by intracellular cysteine levels. We expected to observe reduced growth in our mutant with up-regulated cysteine regulon expression. As expected, cells expressing RhlB P238L did exhibit 2-fold increased growth inhibition compared with the wild-type (Fig.  7C). To directly measure the cysteine content in vivo, we performed LC-MS/MS analysis with extracted metabolites from E. coli as described under "Experimental Procedures." We determined the cysteine content by using the standard curve of  L-12C-cysteine, which is generated using peak area ratios of L-C12-cysteine and L-[1-13C]-cysteine (Fig. 7D). Although the intracellular content of cysteine in cells expressing FlagRhlB wt was 13.5 g/liter/OD 460 , cells bearing the FlagRhlB P238L mutant had an increased cysteine content of 25.3 g/liter/ OD 460 (Fig. 7E), indicating that the P238L mutation resulted not only in increased cysteine regulon expression but, ultimately, increased intracellular cysteine content.

Discussion
RhlB has been shown in vitro to facilitate PNPase degradation of double-stranded RNA (8). Here, we report that the P238L mutation attenuates the interaction of RhlB and PNPase and is important in regulating the stability of cysB, which further modulates the homeostatic levels of cysteine. This is the first report linking RhlB-PNPase to cysteine biosynthesis and showing that the protein-protein interaction is necessary for this regulation. In addition, the regulation of cysB by RhlB-PNPase is independent of the RNA degradosome (Fig. 5). In contrast, rpsT and the repetitive extragenic palindrome-con-taining transcript malEF has been shown to be regulated by RhlB and PNPase, but this regulation requires RNA degradosome formation (13). Previous reports have shown that the RhlB-PNPase complex can form independently of the RNase E-based degradosome (9), but we show that this degradosomeindependent interaction plays a distinct role in controlling gene expression in cells.
We show that the interaction of RhlB-PNPase regulates the stability of the cysB transcript and the abundance of cys transcripts. Weakening the protein interaction of RhlB-PNPase can thus stabilize cysB, increase CysB protein levels, and ultimately lead to the increased abundance of cys transcripts that we observed. However, CysB itself auto-regulates the expression of cysB and, as a result, the weakly interacting RhlB-PNPase increased the steady-state level of cysB only 1.3-fold, although this small increase of cysB mRNA had a significant impact on controlling the cysteine regulon. This regulation is independent of its association with the RNA degradosome but dependent on PNPase (Fig. 5-8). In addition, PcnB also facilitates the degradation of cysB by polyadenylation (Fig. 6E). However, FlagRhlB P238L -expressing cells with normal PcnB activity can sufficiently stabilize cysB mRNA, suggesting that PNPase requires RhlB association to unwind the secondary structure and that polyadenylation may enhance the binding of PNPase. We suggest a molecular mechanism of cysB degradation in a 3Ј to 5Ј direction independently of degradosome formation (Fig. 8). As shown in Fig. 8a, the PcnB adds polyadenosine at the 3Ј-end of cysB, which facilitates the binding of the degradosome-independent RhlB-PNPase complex. RhlB further unwinds the strong secondary structure downstream of the stop codon (which cannot be degraded by PNPase alone), thereby allowing the unwound transcript to be degraded by PNPase. Because CysB plays an important role in controlling the homeostasis of cysteine, the expression levels of its transcript must be tightly controlled. In the absence of PNPase (Fig. 8b) or RhlB (Fig. 8c), the degradation of cysB may be compensated by other helicases and exoribonucleases. This possibility is supported by previous studies that show pnp deletion increases the expression level and activity of ribonuclease II (51) and that the expression level of SrmB was increased in an rhlB deletion strain (52). The degradation of cysB may also begin with endonucleolytic cleavage by RNase E and may be faster than 3Ј exonucleolytic attack when either PNPase or RhlB is absent (Fig. 8d). Because our experiments were done with the rne131 strain that lacks degradosome formation, we cannot exclude the possibility that additional pathway(s) involved in cysB degradation can occur in a strain having RNase E-degradosome formation.
Recently, the exosome of the bacterium Deinococcus radiodurans was reported as an Rsr-Y RNA-PNPase RNP complex, wherein Rsr and PNPase form a complex with Y RNA to facilitate the degradation of double-stranded RNA (53). However, neither Rsr protein nor Y RNA homologs exist in E. coli. In E. coli, purified RhlB and PNPase interact directly without assistance (9). Here, we found that disruption of the interaction of RhlB and PNPase increased the intracellular cysteine content and enhanced anti-oxidative resistance (Fig. 7). Hence, our study suggests a role for the RhlB-PNPase complex in modulating oxidative stress responses by controlling cysteine synthesis through regulating the stability of cysB. Previous studies have reported that pnp deletion strains exhibit heightened resistance to the oxidative stress agents H 2 O 2 and PQ (54,55). Under similar oxidative stress conditions, other studies have also shown that the activity and protein level of hPNPase are reduced in HeLa cells (56) and that PNPase protein abundance is reduced in bacteria (57). Moreover, it has been shown in E. coli that binding of metabolites such as citrate inhibits both the degradation and polymerization activities of PNPase (58). Whether the binding of metabolites can also regulate the interaction of RhlB and PNPase remains to be studied.