|Home | About | Journals | Submit | Contact Us | Français|
Small noncoding regulatory RNAs (sRNAs) play a key role in the posttranscriptional regulation of many bacterial genes. The genome of Caulobacter crescentus encodes at least 31 sRNAs, and 27 of these sRNAs are of unknown function. An overexpression screen for sRNA-induced growth inhibition along with sequence conservation in a related Caulobacter species led to the identification of a novel sRNA, CrfA, that is specifically induced upon carbon starvation. Twenty-seven genes were found to be strongly activated by CrfA accumulation. One-third of these target genes encode putative TonB-dependent receptors, suggesting CrfA plays a role in the surface modification of C. crescentus, facilitating the uptake of nutrients during periods of carbon starvation. The mechanism of CrfA-mediated gene activation was investigated for one of the genes predicted to encode a TonB-dependent receptor, CC3461. CrfA functions to stabilize the CC3461 transcript. Complementarity between a region of CrfA and the terminal region of the CC3461 5′-untranslated region (5′-UTR) and also the behavior of a deletion of this region and a site-specific base substitution and a 3-base deletion in the CrfA complementary sequence suggest that CrfA binds to a stem-loop structure upstream of the CC3461 Shine-Dalgarno sequence and stabilizes the transcript.
Caulobacter crescentus, an oligotroph that lives in nutrient-poor aquatic environments, is an alphaproteobacterium. Until recently, relatively little was known about the molecular mechanisms that enable C. crescentus to adapt to and survive in nutrient-poor environments. The sequence of the C. crescentus genome revealed that this bacterium possesses an unusually large number of TonB-dependent receptors, outer membrane proteins that bind to diverse substrates and facilitate their transport across the outer membrane and ultimately into the cytoplasm (28). It has been hypothesized that this diverse group of TonB-dependent receptors enables C. crescentus to diversify its nutrient uptake mechanisms during periods of nutrient deprivation, and two of these TonB-dependent receptors have been implicated in the uptake of the carbon sources maltodextrin (17) and N-acetylglucosamine (5). In addition to TonB-dependent receptors, C. crescentus also employs the intracellular second messenger ppGpp to adapt to carbon starvation (16). The product of the SpoT enzyme, ppGpp has been shown to play a key role in regulating the initiation of DNA replication during carbon starvation as well as enabling long-term survival in medium lacking a carbon source (16). Finally, microarray studies on carbon-deprived cells have identified a number genes encoding TonB-dependent receptors, carbon source catabolic enzymes, and transcriptional regulators whose expression is strongly upregulated during carbon starvation (7). The alternative sigma factor σ54 encoded by rpoN was shown to be necessary for the complete induction of many of these genes (7).
Small noncoding RNAs (sRNAs) have been identified as a key group of regulatory molecules that enable bacteria to adapt to environmental stress (19, 34). These molecules often regulate mRNA targets by directly hybridizing to nucleotide sequences in the 5′-untranslated regions (5′-UTRs) of target mRNAs and either activate or repress translation and/or mRNA stability (9). A global search for sRNAs in C. crescentus (14) identified and validated 31, 4 of which (tmRNA, 6S RNA, 4.5S, and RNase P) possess highly conserved functions in a wide spectrum of bacteria. Ten of the novel sRNAs were differentially expressed under various environmental or nutritional conditions, and four were cell cycle controlled (14).
In Escherichia coli, the regulatory targets and functions of some bacterial sRNAs have been determined through the use of multicopy plasmid library screens, where induction of sRNA overexpression led to a screenable phenotype (18, 20). To begin to assign functions to C. crescentus sRNAs, and to search for new sRNAs that may have been missed in the initial global survey, we designed a genetic screen to identify sRNAs that would cause a growth defect during overexpression. While the screen did not yield additional sRNAs, it did identify one of the sRNAs discovered in the earlier global survey, CrfA (CC3510_CC3511). Overexpression of CrfA in rich medium caused a striking growth defect. Analysis of sequence conservation of genes encoding sRNAs identified in C. crescentus and the closely related Caulobacter sp. K31 revealed that a surprisingly small number of C. crescentus sRNAs exhibit significant sequence conservation but that CrfA was one of those few sRNAs.
An investigation of the environmental stress signals that affect CrfA expression revealed that its expression is specifically induced by both carbon starvation and entry into stationary phase. Microarray studies identified downstream target genes that are either directly or indirectly regulated by CrfA, predominantly by functioning as an activator of gene expression. Most sRNAs are negative regulators of gene expression, and direct activation of a target mRNA by an sRNA is rare (9, 34). Analysis of the mechanism of CrfA activation of a specific target gene, encoding a putative TonB-dependent receptor, revealed that CrfA functions as an activator of its mRNA through direct hybridization of CrfA with a stem-loop structure at the terminus of its 5′-UTR, resulting in the stabilization of its mRNA. Unlike other previously described sRNA activators (15, 18, 26), CrfA does not appear to activate its target(s) by disrupting an inhibitory stem-loop structure in the 5′-UTR that occludes the Shine-Dalgarno sequence.
The Escherichia coli strains DH5α, DH10B, and TOP10 (Invitrogen) were used for cloning. E. coli strains were grown in Luria-Bertani (LB) medium at 37°C. All C. crescentus strains were derived from the synchronizable wild-type strain CB15N (NA1000) (8); a description of the strains used in this study is provided in Table Table1.1. C. crescentus was grown at 28°C in either peptone-yeast extract (PYE) or M2 minimal medium supplemented with 0.2% glucose (M2G) (6). Where indicated, 0.3% xylose or 0.5 mM sodium vanillate (Fluka) was added to either liquid or solid growth medium to activate transcription of genes cloned downstream of the xylose- or vanillate-inducible promoters. Plasmids were introduced into chemically competent E. coli cells by 42°C heat shock and into C. crescentus by electroporation. For E. coli, antibiotics were added at the following concentrations (liquid/solid medium concentrations, in μg/ml): gentamicin (15/20), chloramphenicol (20/30). For C. crescentus, medium was supplemented with gentamicin (0.5/2.5) and chloramphenicol (1/1).
Depletion for specific nutrients was achieved by first growing cells to mid-logarithmic phase (optical density at 660 nm [OD660] of 0.4) in M2G followed by washing three to four times in M2 minimal medium that lacked carbon, nitrogen, or phosphate. After the final washing, cells were resuspended in the same medium in which they had been washed to an OD660 of approximately 0.15 and incubated with aeration (250 rpm) at 28°C. Invariably, growth (measured by changes in the OD660 measurements) ceased in cultures subjected to sudden carbon, nitrogen, or phosphate depletion. As a negative control, cells were simultaneously washed and resuspended in M2G possessing all requisite nutrients. These cells were maintained in mid-logarithmic growth phase via periodic dilution with fresh M2G medium. Nutrient-depleted medium was obtained by manipulating the composition of M2G. Carbon was depleted by suspending the cells in M2 minimal medium lacking glucose as a source of carbon. Nitrogen was depleted by suspending the cells in M2G lacking ammonium chloride as a source of nitrogen. Phosphate was depleted by suspending the cells in M2G lacking potassium phosphate. Since potassium phosphate is both a source of phosphate and a buffering agent, the pH of M2G lacking phosphate was adjusted to pH 7.0 through the addition of 50 mM NaOH.
A deletion of crfA and a deletion of 14 bp that correspond to the first 14 residues of the 5′-UTR of CC3461 were generated by homologous recombination and allelic replacement using a two-step selection procedure with the nonreplicating vector pNPTS138 as previously described (10, 29, 31). To generate the ΔcrfA allelic replacement plasmid, 500-bp DNA fragments located upstream and downstream of the crfA open reading frame were PCR amplified and cloned into pNPTS138. The upstream DNA fragment was amplified using primers OSL100 (5′-ACT AGT GCC CCT TGC GTC TAT CGCA-3′) and OSL101 (5′-AAG CTT CCC GTC GCG ATT ATC CCCC-3′). This fragment was framed by SpeI and HindIII sites. The downstream DNA fragment was PCR amplified using primers OSL102 (5′-AAG CTT AAT TCG TGA CGG TGT CAA CGCA-3′) and OSL103 (5′-GCA TGC CCA ACG GGC TGA AGA TCC TCT-3′). This fragment was framed by HindIII and SphI sites. Triple ligation of the upstream and downstream DNA fragments into the SpeI and SphI sites of pNPTS138 generated plasmid pSL100. Confirmation of the desired plasmid constructions was obtained by DNA sequencing. To generate the CC3461 Δ5′-UTR allelic replacement plasmid, 500-bp DNA fragments located upstream and downstream of the 5′-UTR were PCR amplified and cloned into pNPTS138. The upstream DNA fragment was amplified using primers OJL156 (5′-TTT ACT AGT GCC CCG TCA TTG CGC ACA TAG TTG-3′) and OJL157 (5′-ATC GGC AGG GTT CAC CCT CGT CTT GA-3′). This fragment was framed by a SpeI site and a blunt end. The downstream DNA fragment was PCR amplified using primers OJL160 (5′-ACG CCA AAA ACA CCA GGG AGG AAA TCA TGA-3′) and OJL158 (5′-TTT GCA TGC TTG GGC TGG TTG GTG ATG TAG CGG AT-3′). This fragment was framed by a SphI and a blunt end. The upstream PCR product was cut with SpeI, while the downstream PCR fragment was cut with SphI. The blunt ends of both PCR products were phosphorylated with T4 polynucleotide kinase (Fermentas). Triple ligation of the upstream and downstream DNA fragments into pNPTS138 cut with SpeI and SphI generated plasmid pJL57. This vector harbored the Δ5′-UTR allele of CC3461 in which positions +1 to 14 in the 5′-UTR were deleted (corresponding to C. crescentus genome positions 3704201 to 3704188). Confirmation of the desired plasmid construction was obtained by DNA sequencing.
Primers Xyl_CrfA1 (CAT GTT AGC GCT ACC AAG TGC CGA GCA AGG ACG AAA CGA GCC CAC) and Xyl_CrfA2 (CTC GAG CTC TTC CGT CAA ATT TGT GAA ACA GGG) were used to PCR amplify the crfA gene starting at its random amplication of cDNA ends (RACE)-defined 5′ end at chromosomal position 3755518 and terminating 37 bp downstream of its Transterm-estimated 3′ end, at chromosomal position 3755354. The xylose-inducible xylX promoter (Pxyl) was amplified using primers Xyl1 (GGG AAT TCC ACC AGC CAC AGG CCC GTGC) and Xyl2 (TCG GCA CTT GGT AGC GCT AAC ATG). The crfA and Pxyl PCR products were designed so that 30 bp at the 3′ end of the Pxyl PCR product were complementary to a 30-bp extension at the 5′ end of the crfA PCR product. This design enabled us to fuse the +1 transcriptional start site of the Pxyl PCR product to the 5′ end of the crfA PCR product by using a previously described PCR overlap extension method (11). The fused Pxyl and crfA PCR products were digested and cloned into the EcoRI and XhoI sites of pMR31 to generate pXyl-CrfA-WT.
All Northern blot assays were done with DNA oligonucleotide probes. Probes were 5′ end labeled with [γ-32P]ATP and T4 polynucleotide kinase and twice purified using the Qiagen nucleotide removal kit. A 5′-end-labeled low-molecular-weight DNA ladder (New England Biolabs) was used to estimate sizes of RNA bands. Total RNA (5 to 10 μg) was separated on 6% PAGE-urea gels and transferred to Hybond N+ membranes (Amersham) in 0.5× Tris-borate-EDTA by using the TransBlot semidry transfer apparatus (Bio-Rad) according to the manufacturer's instructions. Blots were prehybridized in Ultrahyb-oligo buffer (Ambion) according to the manufacturer's instructions. After incubation with probe, blots were washed twice in 2× SSC (1× SSC is 0.15 M NaCl plus 0.015 M sodium citrate)-0.5% SDS and exposed. All hybridizations were done at 42°C.
5′-RACE was performed as described by Argaman et al. (2) using the same 5′ RNA adapter. Forty PCR cycles were performed with a 53°C melting temperature using 0.5 to 4 μl reverse transcription reaction mixture as template and Platinum Taq high-fidelity polymerase (Invitrogen). The following primers were used for CC3461 5′-RACE reactions: CC3461 RACE reverse transcription primer (GAC GAT GAT TTC CGC GAT), downstream PCR primer (ATG GCG TTC GGA TCA GCC TGCT), and upstream PCR primer (GAT ATG CGC GAA TTC CTG TAGA). 3′-RACE was performed as described by Argaman et al. (2) with the exception that a second PCR step using an internal 5′ primer was utilized to ensure isolation of the CrfA-specific product. The following oligonucleotides were utilized to perform 3′-RACE on CC3461: a 3′ RNA adapter (UUC AUC GUU CUU ACA AGC UUG AUG CUCiT), where iT is inverted deoxythymidine, a 3′-RACE reverse transcription primer (GAG CAT CAA GCT TGT AAG AA), a 3′ PCR primer (GAG CAT CAA GCT TGT AAG AACG), an external 5′ PCR primer (GAC GAA ACG AGC CCA CCA), and an internal 5′ PCR primer (AGA GCT CCC GTT CTT CGG).
Genomic DNA was isolated from C. crescentus using the DNeasy tissue kit (Qiagen). Genomic DNA was partially digested with the restriction enzyme HaeIII to produce DNA fragments in the size range from 200 to 700 bp. The digested chromosomal DNA was gel purified using a Qiaquick gel extraction kit (Qiagen), and purified genomic DNA fragments were ligated into plasmid pMT335 digested with SmaI. Ligation reactions were transformed into TOP10 to obtain approximately 100,000 individual colonies. The colonies were pooled, and plasmid DNA was isolated through the use of multiple plasmid miniprep columns (Qiagen). Purified plasmid DNA obtained by individual miniprep columns was then pooled to generate the library. The plasmid library was electroporated into C. crescentus to obtain approximately 30,000 transformants with approximately 150 transformants/plate. Transformants were subjected to triple velvet replica plating onto either PYE agar plates with gentamicin or PYE agar plates with gentamicin and vanillate (0.5 mM). Transformants that grew on PYE plates but were absent or displayed a marked growth defect on PYE plates supplemented with vanillate were identified as candidates. To eliminate false positives that arose due to unequal transfer of transformant colonies during replica plating, the growth defects of candidates were further evaluated by spotting serial 10-fold dilutions of an overnight culture onto both PYE gentamicin plates and PYE gentamicin plates supplemented with 0.5 mM vanillate. Plasmid DNA was isolated from candidates that displayed a consistent vanillate-dependent growth defect. The gene(s) responsible for the growth defect was determined by sequencing the insert cloned into pMT335 and using the M13 forward primer.
A 155-nucleotide antisense oligonucleotide of sequence tgcc ctcgag ACAC CAAA CCCG CCGC GGGG GATA ATCG CGAC GGGC TGGC TATT TTCT CGAT CATT CGAT CTAG GCGG GCGT TTCC TCCC CGAA ACGC CGAA GAAC GGGA GCTC TCTG GTGG GCTC GTTT CGTC CTTGC tcgg cact tggt agcg containing the RACE-determined CrfA sequence (in uppercase letters) preceded by the first 16 nucleotides of the xylose-inducible promoter Pxyl (in italics) (24) and followed by an XhoI site (underlined) was synthesized. For 121 nucleotides in the CrfA coding sequence (shown in bold) and excluding the 8-nucleotide T-rich sequence at the 3′ end of CrfA which follows the terminator hairpin, synthesis was from a phosphoramidite pool containing 99.9% of the wild-type residue mixed with 0.1% of an equimolar mixture of the three non-wild-type residues. The oligonucleotide was purified from a 6% PAGE-urea gel. Synthesis from these pools was expected to produce a family of molecules of which approximately 90% would contain the wild-type CrfA sequence, approximately 10% would contain a single point mutation, and less than 1% would contain more than one mutation. Sequencing of a sampling of molecules after cloning confirmed this approximate distribution.
To clone the CrfA mutant library, the forward strand of the Pxyl promoter region was asymmetrically amplified from 200 ng of the Pxyl PCR product (described above) using 10 cycles of PCR (94°C for 30 s, 53°C for 30 s, and 68°C for 45 s) in the presence of only primer pXyl1 (10 pmol). This reaction was mixed with 4 pmol of CrfA library oligonucleotide which had been gel purified, and a double-stranded Pxyl-CrfA fusion population was created by incubating at 94°C (2 min), 53°C (30 s), and 68°C (1 min). The product was amplified by 15 cycles of PCR (94°C for 15 s, 53°C for 30 s, and 68°C for 45 s) using primers pXyl1 and CrfA_rev (TGC CCT CGA GAC ACC AAA), purified with a Qiagen PCR purification kit, digested with EcoRI and XhoI, and ligated into the vector pMR31, which had been similarly digested. The ligation reaction mixture was transformed into E. coli strain DH-5α, yielding approximately 50,000 colonies, which were pooled, and plasmid DNA was isolated.
To screen the CrfA mutant library, 2 μg of library plasmid was electroporated into C. crescentus strain CB15N and plated on PYE/chloramphenicol in the presence and absence of 0.3% xylose. In the absence of xylose, approximately 30,000 large colonies were obtained, while in the presence of xylose, approximately 2,200 heterogeneously sized colonies were obtained. When the parent vector was transformed, no difference in transformation efficiencies in the presence or absence of xylose was observed, while approximately 98% fewer colonies were obtained in the presence of xylose when pXyl-CrfA-WT was transformed, and all colonies were extremely small. A total of 500 larger colonies from the CrfA library transformation were picked from PYE plates supplemented with 0.3% xylose and pooled, and plasmid DNA was isolated. To eliminate gross deletions within crfA, inserts from the pooled plasmid DNA were reamplified with primers Xyl_CrfA_for (ACA TGT TAG CGC TAC CAA GTG CCG A) and CrfA_revII (GGC CCC CCC TCG AGA CAC CAA A) by using 20 cycles of PCR (94°C for 15 s, 60°C for 15 s, and 68°C for 15 s). Amplified library sequences were cloned as described above for the initial library and subjected to sequencing.
In order to determine mRNA levels, 1.5 ml of C. crescentus was harvested from batch cultures growing in the presence (M2G) or absence (M2) of glucose as a carbon source. RNA was isolated by TRIzol extraction (Invitrogen) according to the manufacturer's instructions. Purified RNA samples were treated with DNase I (Invitrogen) to remove contaminating DNA, and RNA was repurified by phenol-chloroform-isoamyl alcohol extraction followed by ethanol precipitation. cDNA was synthesized and PCR amplified from 62.5 ng of total RNA using the Brilliant II SYBR green one-step master mix quantitative real-time PCR (qRT-PCR) kit (Stratagene). The CC3461 downstream primer (GAG TGG TCT GCA GGT TCT TG) and the CC0978 downstream primer (GTC ACT TCG GCG ATC AGTT) were used for cDNA synthesis, and PCR was performed using the following primers for CC3461 and CC0978: CC3461 upstream primer (CGG CTA CCA GAA CTA TGA CG)/CC3461 downstream primer and CC0978 upstream primer (CAA GCT GTC CAA GCT CGT C)/CC0978 downstream primer. Thermocycling conditions were as follows: 30 min at 50°C for reverse transcription, 10 min at 95°C, and 45 cycles (95°C for 30 s, 56°C for 1 min, and 72°C for 30 s). PCR primers were used at a concentration of 150 nM. Duplicate 20-μl reaction mixtures were performed in a Roche 480 thermocycler in a 384-well optical reaction plate. To control for genomic DNA contamination, identical reactions were performed in the absence of reverse transcriptase. The measured threshold cycle (CT) for each sample was determined using second derivative analysis carried out by instrument software. The measured CC3461 transcript CT values were normalized by subtracting the CT for the CC0978 transcript, which was also measured for each sample. The CC0978 transcript was chosen for normalization because the microarray experiments showed that the CC0978 transcript abundance was unaffected by carbon starvation. In addition, the CC0978 transcript abundance was unaffected by CrfA (no change in CC0978 transcript abundance was observed in either a ΔcrfA strain or the CrfA overexpression strain). The resulting ΔCT value derived from the wild-type strain grown in the presence of carbon was used as a reference to calculate relative level of transcripts for each condition (i.e., ΔΔCT). To determine the mRNA degradation rates in the presence and absence of carbon, the 0-time point CT was used as a reference to calculate the relative mRNA levels in samples removed at time points after rifampin (10 μg/ml) addition.
To identify CrfA-dependent genes during carbon starvation, a wild-type strain and a ΔcrfA strain were grown to mid-logarithmic phase (OD660, 0.4) in M2G, washed in M2, and resuspended in M2 for 20 min. Samples were pelleted at 16,000 × g for 1 min, and cell pellets were frozen at −80°C. To identify genes regulated by CrfA overexpression, a wild-type strain harboring either the empty vector pMR31 or the xylose-inducible CrfA overexpression vector pXyl-CrfA were grown to an OD660 of 0.4 and induced with 0.3% xylose for 15 min. Samples were pelleted at 16,000 × g for 1 min, and cell pellets were frozen at −80°C. Total RNA was extracted with TRIzol reagent (Invitrogen) according to the manufacturer's instructions, followed by additional purification with the Purelink Micro-to-Midi total RNA purification system (Invitrogen). Contaminating DNA was digested with DNase I, and RNA was further purified using the Purelink Micro-to-Midi total RNA purification system. Reverse transcription with random hexamers was conducted using SuperScript II (Invitrogen). cDNA was fragmented by DNase I treatment and hybridized to the C. crescentus CauloHi1 microarray as previously described (23). The RMA statistical algorithm (12) available in the Bioconductor software package of R was used for background noise removal, normalization, and summarization of the data. Each microarray experiment was performed twice, and genes were considered for further analysis if they showed an average 2-fold change relative to the control with less than a 50% standard deviation from the mean between replicates. The complete data set for all genes from the microarray experiments is available in Table S1 of the supplemental material.
To measure transcriptional activity of the CC3461 promoter, 500 bp upstream of the RACE-defined transcriptional start site for CC3461 was PCR amplified using primers 3461 β-Gal fusion transcription upstream (AAG AAT TCT TGC TGC AGG CCC GAC GAG CAGAA) and 3461 β-Gal transcription fusion downstream (AAC TGC AGC ATC GGC AGG GTT CAC CCT CGT). The PCR product was digested with EcoRI and PstI and cloned into the same sites on the promoterless lacZ vector pRKlac290 to generate pSL101. β-Galactosidase activity was measured as previously described (25).
A previous genome-wide screen for sRNAs in C. crescentus using a custom-tiled Affymetrix microarray, combined with an analysis scheme optimized for sRNA detection, identified 27 novel sRNAs (14). To identify sRNAs with functional roles regulating growth and development, we designed a genetic screen for short DNA sequences whose overexpression inhibited growth. Since many sRNAs regulate adaptation to environmental stress and nutrient deprivation, we predicted that overproduction of an sRNA during growth in rich medium, rather than under the stress condition in which its function was relevant, might disrupt cellular physiology and yield a growth defect. A plasmid library was constructed containing small chromosomal DNA fragments (size range, 200 to 700 bp) cloned downstream of the inducible vanillate promoter (30). The library was electroporated into wild-type C. crescentus strain NA1000, and colonies were triple velvet replica plated onto both plain PYE plates and PYE plates supplemented with 0.5 mM vanillate. Vanillate-sensitive colonies were identified after 2 to 3 days of incubation at 28°C. To eliminate false positives, overnight cultures of candidate colonies were serially diluted and spotted onto both plain PYE plates and PYE plates supplemented with 0.5 mM vanillate.
Of approximately 30,000 colonies screened by replica plating, 47 displayed reproducible vanillate sensitivity. Plasmids were isolated from these strains and plasmid inserts were sequenced. Two of the vanillate-sensitive strains had plasmids bearing intergenic regions with previously identified sRNAs (CC3510_CC3511 and CC3513_CC3514). The other 45 plasmids contained inserts with fragments of genes including the 5′ ends of genes, the 3′ ends of genes, and internal gene fragments oriented in both sense and antisense directions relative to the vanillate promoter. To determine if the sRNAs contained in these intergenic regions were responsible for the observed growth defect, each of the two sRNAs was subcloned downstream of a xylose-inducible promoter (construct for crfA shown in Fig. Fig.11 A), and growth was monitored in the presence and absence of sRNA induction. Only sRNA CC3510_CC3511 showed a significant growth defect during xylose-inducible overexpression. Cells from cultures containing the CC3510_CC3511 overexpression vector pXyl-CrfA stopped growing and dividing, while cells grown in the absence of the inducer exhibited growth kinetics indistinguishable from cells containing a control vector (Fig. (Fig.1B).1B). In the previous genome-wide screen for sRNAs in C. crescentus (14), the expression level of the sRNA CC3510_CC3511 was found to be higher during growth in minimal medium than it was during growth in nutrient-rich PYE medium. Due in part to this observation and others (see below), we named this intergenic sRNA CrfA, for caulobacter response to famine.
Many bacterial sRNAs have been identified by comparative sequence analysis of closely related species to identify conserved DNA sequence elements within intergenic regions (33). Surprisingly, of the 27 novel sRNAs identified in C. crescentus, only 8 showed significant sequence homology to intergenic regions within the closely related Caulobacter sp. K31 (14). CrfA was one of those eight with 93% sequence identity between C. crescentus and Caulobacter sp. K31 (Fig. (Fig.22 A). Overlaying the rare base changes in the Caulobacter sp. K31 sRNA onto the predicted secondary structure for CrfA from C. crescentus revealed that none of the sequence changes would have impacted the predicted secondary structure for the sRNA (Fig. (Fig.2B2B).
We investigated the effects of various environmental stresses on CrfA levels to determine conditions that induce crfA expression. Removal of glucose, the sole source of carbon in minimal medium (M2G), caused a dramatic induction of crfA (Fig. (Fig.3A).3A). The CrfA transcript increased more than 10-fold within 10 min after glucose was removed (Fig. (Fig.3A,3A, compare lanes 2 to 6 with lanes 7 to 11) This degree of induction is comparable to CrfA overexpression when induced from the xylose promoter. CrfA levels decreased within 10 min after the readdition of glucose, and they returned to previous levels within 30 min (Fig. (Fig.3A,3A, compare lane 10 with lanes 12 to 15). The level of 5S rRNA was used as a control, and it did not change over the course of the experiment.
To determine if the increase in CrfA accumulation was specific for carbon starvation, or the result of a general stress response to nutrient deprivation, CrfA levels were measured during depletion for nitrogen, phosphate, or carbon. Upon removal of a nitrogen source from the M2G growth medium by washing and resuspending mid-log-phase cells in M2G lacking ammonium chloride, CrfA levels decreased (Fig. (Fig.3B,3B, compare lanes 1 to 3 with lanes 8 and 9). Depletion for a phosphate source by washing and resuspending mid-log-phase cells in M2G lacking potassium phosphate did not affect CrfA levels (Fig. (Fig.3B,3B, compare lanes 1 to 3 with lanes 11 and 12). However, depletion for carbon by washing cells in M2G lacking glucose again stimulated a dramatic increase in CrfA levels that was reversible upon the introduction of glucose to the medium (Fig. (Fig.3B,3B, compare lanes 1 to 4 with 5 to 7).
In addition to the stress of nutrient deprivation, CrfA levels were also measured during the transition from exponential growth to stationary phase. As observed for carbon starvation, expression of crfA was strongly induced when cells entered stationary phase (Fig. 3C and D).
Most sRNAs function by base pairing with target mRNAs to either activate or repress their degradation (1). To identify genes that respond to CrfA, we used Affymetrix microarrays to examine global changes in RNA profiles under growth conditions in which CrfA was either absent or overexpressed. Accordingly, a ΔcrfA strain and a wild-type strain were grown in M2G, washed in M2 (M2G lacking glucose), suspended in M2, and incubated in the absence of glucose for 15 min. Total RNA from each culture was extracted, reverse transcribed, and hybridized to the C. crescentus microarray CauloHi1. Changes in RNA profiles in the ΔcrfA strain were determined relative to RNA profiles in the control wild-type strain. We reasoned that since CrfA is normally induced during carbon starvation, then the absence of CrfA during this growth condition might have a significant impact on the steady-state level of target mRNAs. We also examined the effects of CrfA overepression from a xylose-inducible promoter on a muticopy plasmid (pXyl-CrfA) during growth in rich PYE medium. Since CrfA overexpression in PYE inhibits growth (Fig. (Fig.1B),1B), we reasoned that target mRNA profiles should be altered by a 0.3% xylose-induced 15-min pulse of CrfA overexpression. In this case, alteration in RNA profiles was determined versus a control strain carrying an empty vector (pMR31). Each microarray experiment was performed in duplicate, and genes were considered significantly regulated if a ≥2-fold change was observed in their mRNA levels. Finally, genes that were significantly impacted in either or both of these two microarray experiments were compared relative to previously generated microarray data (L. Britos and L. Shapiro, unpublished data) from a study that examined alterations in gene expression in a wild-type strain subjected to carbon starvation for 30 min. We reasoned that since CrfA is strongly induced during carbon starvation, its target genes should also be affected by carbon starvation. A list of the magnitude of gene regulation observed in the three microarray experiments is shown in Table Table22.
Seven CrfA-regulated genes displayed decreased expression levels in the ΔcrfA strain during carbon starvation, increased expression levels in a CrfA overexpression strain, and increased expression levels in a wild-type strain subjected to carbon starvation (Fig. (Fig.44 and Table Table2).2). These observations are consistent with CrfA primarily functioning directly or indirectly as an activator of its target genes rather than as an inhibitor. A list of the seven genes significantly affected in all three microarray experiments is shown in Fig. Fig.44 and Table Table2.2. The putative functions of these seven genes are consistent with CrfA functioning as an activator of genes necessary for a bacterial cell's adaptation to carbon starvation. The genes are predicted to encode four TonB-dependent receptors, a putative membrane-bound proton-translocating pyrophosphatase, a putative 6-amino-hexanoate-dimer hydrolase, and a flavin-binding family monoxgenase.
In addition to the seven genes that were similarly regulated in all three microarray experiments, nine genes displayed decreased mRNA levels in the ΔcrfA carbon starvation microarray and increased mRNA levels in the wild-type carbon starvation microarray (Fig. (Fig.44 and Table Table2).2). The observation that these genes were not significantly affected in the CrfA overexpression microarray may have been due to the fact that the CrfA overexpression microarray was conducted with RNA extracted from cells growing in rich (PYE) medium, an environment in which CrfA levels are normally low and CrfA function may not be physiologically relevant. The rich growth medium may have generated competing regulatory signals that nullified some of the effects of CrfA overexpression.
Like the CrfA-regulated genes common to all three microarray experiments, the putative functions of all nine CrfA-regulated genes common to the ΔcrfA carbon starvation microarray and the wild-type strain carbon starvation microarray are consistent with CrfA functioning as an activator of genes necessary for adaptation to carbon starvation (Fig. (Fig.44 and Table Table2).2). Two of the nine genes encode a putative TonB-dependent receptor and a major facilitator family transport protein. Another five genes (CC0945, CC2494, CC0944, CC1634, and CC3338) encode putative enzymes whose upregulation could facilitate alterations in intermediary carbon metabolism necessary for adaptation to carbon starvation. An eighth gene (CC0081) encodes a putative transcriptional regulator, and the ninth gene (CC2031) encodes a putative TPR (tetratricopeptide repeats) domain protein whose relevance to carbon starvation adaptation is not known.
Finally, 11 genes displayed increased expression levels in the CrfA overexpression microarray and increased expression levels in the wild-type strain carbon starvation microarray (Fig. (Fig.44 and Table Table2).2). While these genes were not significantly affected in the ΔcrfA carbon starvation microarray, we still considered them potential targets of CrfA due to the possibility that redundant regulatory factors were compensating for the loss of CrfA. The proteins encoded by these 11 genes included 2 putative TonB-dependent receptors, a major facilitator family protein, and 8 proteins with potential roles in the regulation of intermediary metabolism (Table (Table22).
The majority of sRNAs in bacteria function as direct negative regulators of their target genes (9, 34). In most cases, these sRNAs hybridize to the 5′-UTR of the target mRNA, where they overlap the Shine-Dalgarno sequence, block ribosome access, and facilitate mRNA degradation by RNases such as RNase E (21, 27) or RNase III (4, 32). Only a few examples of sRNAs functioning as direct positive regulators have been described (9). The microarray studies to identify downstream targets of CrfA suggested that CrfA functions directly or indirectly as a positive regulator of at least 27 genes. To explore the mechanism whereby CrfA might directly activate its target mRNA, we dissected the mechanism by which it affects the expression of target gene CC3461, a putative TonB-dependent receptor that was one of the most strongly regulated targets of CrfA in both the CrfA overexpression microarray (4.8-fold increase in mRNA level) and the ΔcrfA carbon starvation microarray (10-fold decrease in mRNA level) (Table (Table2).2). Expression of CC3461 was also significantly induced in wild-type cells subjected to carbon starvation (3-fold induced) (Table (Table2).2). We considered three potential mechanisms to explain the effect of CrfA on CC3461 mRNA upon carbon starvation: CrfA activation of CC3461 transcription, CrfA inhibition of CC3461 mRNA degradation, and the simultaneous activation of transcription of CC3461 and inhibition of the degradation of CC3461 mRNA.
To determine if CrfA activates transcription of CC3461 during carbon starvation, we constructed a transcriptional reporter by mapping the +1 transciptional start site for CC3461 by using 5′-RACE (see Fig. Fig.6A)6A) and then fusing 500 bp of DNA sequence upstream of the +1 site to a lacZ reporter gene (pSL101). Output from the plasmid-borne transcriptional reporter was measured in both a wild-type and a ΔcrfA strain during growth in the presence and absence of carbon (Fig. (Fig.55 A). If CrfA activates CC3461 transcription during carbon starvation, then reporter output, measured as β-galactosidase activity, should increase in the wild-type strain during carbon starvation but remain unchanged in the ΔcrfA strain. The β-galactosidase assays revealed that transcription from the CC3461 promoter was unaffected by carbon starvation in both with wild-type and the ΔcrfA strains (Fig. (Fig.5A),5A), suggesting that the observed CrfA-dependent increase in the level of CC3461 mRNA during carbon starvation (Table (Table2)2) was not caused by a CrfA-dependent increase in CC3461 transcription.
To determine if CrfA inhibited degradation of CC3461 mRNA during carbon starvation, we measured the CC3461 mRNA half-life in both a wild-type and a ΔcrfA strain during growth in the presence and absence of carbon. mRNA half-life was determined by measuring CC3461 mRNA levels with qRT-PCR as a function of time following the inhibition of transcription with rifampin. If CrfA inhibits CC3461 degradation, then the half-life of CC3461 mRNA should increase dramatically during carbon starvation in the wild-type strain but remain unaffected by carbon starvation in the ΔcrfA strain. The mRNA degradation assays revealed that in a wild-type strain the CC3461 mRNA half-life increases 12-fold during carbon starvation, from 4.7 min in the presence of carbon to 55 min in the absence of carbon (Fig. (Fig.5B).5B). In contrast, in a ΔcrfA strain, the CC3461 mRNA half-life was unaffected by carbon starvation (Fig. (Fig.5C).5C). These results argue that the CrfA-dependent increase in CC3461 mRNA level during carbon starvation was caused by a CrfA-dependent inhibition of CC3461 mRNA degradation.
Many sRNAs affect target mRNA stability by interacting directly with the 5′-UTR. A secondary structure prediction analysis identified a prominent stem-loop structure at the 5′ end of the 5′-UTR for CC3461 mRNA and a lack of secondary structure around the Shine-Dalgarno site (Fig. (Fig.66 B). To identify a potential binding site for CrfA, the CrfA sequence was aligned with the 5′-UTR (Fig. 6B and C), and a region of complementarity was identified that overlapped extensively with the stem-loop structure at the extreme 5′ end of the 5′-UTR (Fig. 6B and C). Figure Figure2B2B shows the region of CrfA that is complementary to the 5′-UTR of CC3461 mRNA.
To determine if the stem-loop region of the 5′-UTR identified by sequence alignment to be complementary to CrfA is required for the cell's response to carbon starvation, a CC3461 mutant was generated in which DNA corresponding to a portion of the 5′-UTR predicted to interact with CrfA (from bases +1 to +14) was deleted (Fig. (Fig.6C).6C). The loss of this 5′-terminal end of the 5′-UTR would abolish the stem-loop structure in the remainder of the UTR. If CrfA directly hybridizes to the 5′-UTR as predicted by the sequence alignment, then CC3461 mRNA levels in the Δ5′-UTR strain should be unresponsive to carbon starvation as observed in a ΔcrfA strain. To test this hypothesis, CC3461 mRNA levels were measured by qRT-PCR in a wild-type strain, a ΔcrfA strain, and a Δ5′-UTR strain during growth in the presence and absence of carbon (Fig. (Fig.77 A). Consistent with the hypothesis that CrfA binds to the portion of the 5′-UTR of CC3461 mRNA identified by sequence alignment (Fig. 6B and C), the qRT-PCR assays revealed that the Δ5′-UTR strain behaved identically to the ΔcrfA strain (Fig. (Fig.7A).7A). Deleting a portion of the putative CrfA binding region in the 5′-UTR left CC3461 mRNA levels largely unresponsive to changes in carbon concentration (Fig. (Fig.7A7A).
To further characterize the regulation of CC3461 mRNA by CrfA, we sought to identify regions within CrfA that are crucial to its function. To identify these regions, we took advantage of the growth inhibition caused by CrfA overexpression to select for rare crfA mutations that rendered it unable to function and therefore unable to inhibit growth during overexpression. To generate these mutations, we utilized error-prone oligonucleotide synthesis to generate a library of crfA molecules containing an average of less than one mutation per molecule (3). This library was cloned downstream of the xylose-inducible (Pxyl) promoter on the multicopy plasmid pMR31 and transformed into wild-type C. crescentus, which was then plated onto PYE agar plates in the presence and absence of 0.3% xylose. Overexpression of wild-type CrfA under these growth conditions was sufficient to completely inhibit C. crescentus growth and prevent transformants from forming colonies (Fig. (Fig.7B).7B). However, some of the clones that harbored crfA mutations were able to form colonies. We picked and analyzed two of these xylose-resistant clones and confirmed that the phenotypes were plasmid mediated by isolating their plasmids and retransforming them into wild-type C. crescentus to reestablish the xylose resistance phenotype (data not shown). Plasmids were then sequenced to identify the DNA lesion(s) within crfA.
One library clone contained a single G52A base change within crfA, and another contained a 3-bp deletion, Δ52-54, in crfA, suggesting that the G52 position in CrfA is essential for mediating the overexpression growth arrest phenotype. To confirm this suggestion, we verified that both the G52A and Δ52-54 mutations conferred resistance to xylose-induced overexpression (Fig. (Fig.7B7B and Table Table3).3). In addition, we used Northern blotting to confirm that the G52A and Δ52-54 mutations did not affect CrfA steady-state levels during xylose-mediated induction (Fig. (Fig.7C).7C). Consistent with the hypothesis that the region of CrfA surrounding base G52 plays an important role in CrfA function, alignment of the CrfA sequence with the 5′-UTR of CC3461 mRNA revealed a significant region of complementarity encompassing CrfA residues 36 to 66 and indicated that G52 is predicted to hybridize directly with base C-17 in the 5′-UTR of CC3461 (Fig. 6B and C). Thus, mutations in either the 5′-UTR of CC3461 or in the region of CrfA predicted to interact with the 5′-UTR rendered CrfA unable to stabilize the CC3461 mRNA.
We have demonstrated here that CrfA, a novel Caulobacter sRNA, is required for the cellular stress response to carbon starvation. Expression of CrfA is rapidly induced in response to carbon starvation and rapidly downregulated when the carbon source is replenished. CrfA's induction by carbon starvation is specific in that other forms of nutrient depletion, such as nitrogen and phosphate starvation, do not induce CrfA accumulation. If CrfA is constitutively overproduced when C. crescentus is grown in rich medium, the culture exhibits a severe growth defect, but CrfA overexpression caused no growth defect in E. coli. Microarray analysis of a ΔcrfA strain and a CrfA overpexression strain identified 27 genes that showed a strong response to both CrfA expression and carbon depletion, with CrfA functioning as an activator of its direct or indirect target genes. Most of the genes strongly regulated by CrfA have potential roles in the cell's adaptive response to carbon starvation. One-third of the CrfA activated genes are predicted to encode membrane transport proteins, and the most common among them are the TonB-dependent receptors. These are outer membrane proteins that bind to extracellular substrates and facilitate their transport across the outer membrane and ultimately into the cytoplasm (13). The upregulation of transport proteins by CrfA during carbon starvation may enable the import of a greater variety of potential carbon sources. One of the CrfA-regulated genes (CC1363) encodes an enzyme with a putative function as a proton pump driven by pyrophosphate hydrolysis. Upregulation of this protein could help the cell maintain its proton electrochemical gradient and continue ATP synthesis during periods of carbon starvation. Other CrfA-activated genes encode enzymes capable of catabolizing alternative carbon sources, such as cyclic-6-aminohexanoate dimers (CC1323), as well as enzymes that modulate intermediary metabolic pathways.
A recent study of the C. crescentus response to carbon starvation identified 154 genes that are induced (7). A total of 112 of these genes showed at least partial rpoN dependence, suggesting a significant role for σ54 in Caulobacter's transcriptional response to carbon starvation (7). Many of the 42 rpoN-independent and carbon starvation-induced genes identified in that study overlap with the CrfA-regulated genes described here. All seven of the most significantly CrfA-regulated genes (Table (Table2,2, designated +++) overlap with the carbon starvation-induced and rpoN-independent genes identified in the previous study (7). This list of seven CrfA-dependent genes (Table (Table2)2) includes the three most significantly carbon starvation-induced genes identified in the previous study (7). Thus, σ54 and CrfA appear to be important transcriptional and posttranscriptional regulators, respectively, for Caulobacter's response to carbon starvation.
While CrfA is strongly induced by carbon starvation, and required for the accumulation of at least 27 mRNA targets during carbon starvation, a ΔcrfA strain did not exhibit a carbon starvation survival defect. This observation was not surprising, given that many bacterial sRNAs function as redundant components of adaptive stress responses (22), and there may be additional regulatory factors that substitute for the loss of CrfA function during carbon starvation. Indeed, the C. crescentus SpoT enzyme has recently been shown to activate the synthesis of the second messenger ppGpp during carbon starvation, and microarray studies have identified a significant number of regulatory factors whose steady-state mRNA levels increase substantially during carbon starvation (7, 16). Finally, because many of CrfA's mRNA targets encode putative outer membrane transport proteins, CrfA's primary mechanism for promoting survival during carbon starvation may be mediated through the diversification of transport proteins on the cell's surface which thereby enhance the cell's ability to import alternative carbon sources.
The function of CrfA was of particular interest because it was identified in a screen for sRNAs that inhibit growth of C. crescentus during overexpression in rich growth medium. From the list of 27 CrfA-regulated target genes, it is not obvious which target or group of targets is responsible for inhibiting C. crescentus growth. Cells overexpressing CrfA did not exhibit obvious morphological defects and appeared to have simply ceased growth in whatever stage of the cell cycle they occupied at the moment of CrfA overexpression. Since some of the CrfA target genes encode proteins with putative functions in modulating intermediary metabolism, it is possible that the inappropriate upregulation of these proteins during growth in nutrient rich medium, rather than during carbon starvation, disrupts the flux of carbon-containing molecules necessary for maintaining energy production and cellular growth.
We have explored the effect of CrfA on one of its target genes, CC3461. In the absence of carbon starvation, both CrfA and CC3461 mRNA accumulation levels were maintained at a low level. Once CrfA was allowed to accumulate in the absence of glucose, CC3461 mRNA was stabilized. Deletion of a region of secondary structure in the 5′-UTR of CC3461 mRNA, which is distal to the Shine-Dalgarno sequence, resulted in loss of CC3461 mRNA stabilization. This region of the 5′-UTR is complementary to a stretch of 31 bases in CrfA, shown in Fig. Fig.2B2B and 6B and C. A site-specific mutation in CrfA's complementary sequence, G52A, and a 3-base deletion, Δ52-54, abrogate CrfA function, possibly by destabilizing CrfA's ability to hybridize to the CC3461 5′-UTR. These observations suggest that CrfA binds directly to the 5′-UTR of CC3461 mRNA, thereby disrupting the secondary structure and inhibiting the degradation of the transcript. In fact, we observed that in the presence of CrfA, there was a 12-fold increase in the CC3461 mRNA half-life and an accompanying increase in the steady-state level of CC3461 mRNA.
It is not yet known how many of the other strongly responsive targets of CrfA regulation identified by the microarray studies are subject to direct regulation. Aligning CrfA and the CC3461 mRNA 5′-UTR using CLUSTALW identified a 31-nucleotide region of complementarity in which 21 bases out of 31 were paired. We used CLUSTALW to identify potential interactions between the CrfA region of complementarity and the other six CrfA-regulated genes identified by all three microarray experiments. For all six genes (CC1348, CC3161, CC2804, CC3336, CC1323, and CC1363), complementarity was detected between CrfA and the putative UTRs. For example, an alignment of the CC1363 mRNA 5′-UTR with CrfA identified complementarity for 19 out of 31 bases.
The most common form of posttranscriptional regulation by bacterial sRNAs involves the negative regulation of the target mRNA (34). Negative regulation occurs when the sRNA hybridizes to its target and obscures the Shine-Dalgarno sequence. Binding in this manner inhibits translation by preventing mRNA interaction with the ribosome. In addition, it creates a target for RNases, such as RNase E (21, 27) or RNase III (4, 32), leading to activated degradation of the mRNA target. A less common form of posttranscriptional regulation by sRNAs involves the positive regulation of the target mRNA (9). In these rare cases, the 5′-UTR of the mRNA forms an inhibitory stem-loop structure in the region corresponding to the Shine-Dalgarno sequence. This stem-loop prevents ribosome binding and reduces the translation efficiency of the transcript. The activating sRNA hybridizes to a region of the 5′-UTR that overlaps one arm of the stem-loop and results in the loss of secondary structure and the release of the Shine-Dalgarno sequence from the inhibitory stem-loop configuration. The exposed Shine-Dalgarno sequence is then free to bind ribosomes, and the translation efficiency and stability of the target mRNA increase. The direct stabilization and consequent activation of CC3461 mRNA by CrfA present a potentially novel mechanism for posttranscriptional activation by sRNAs. The Shine-Dalgarno sequence of the CC3461 mRNA is not obscured by an inhibitory stem-loop structure (Fig. (Fig.6B).6B). The only significant secondary structure within the CC3461 mRNA 5′-UTR is located at the extreme 5′ end of the transcript. Virtually all of the bases within the 5′-UTR predicted to hybridize with CrfA are located within this stem-loop, and virtually no hybridization is predicted at or around the Shine-Dalgarno sequence. The predicted CrfA hybridization site in the 5′-UTR was verified by deleting a portion of the UTR predicted to form the stem-loop. A CC3461 Δ 5′-UTR strain behaved identically to a ΔcrfA strain and was largely unable to regulate CC3461 mRNA levels in response to carbon starvation. Since the hybridization of CrfA to the extreme 5′ end of the 5′-UTR and away from the Shine-Dalgarno sequence is not predicted to alter ribosome binding, then activation of CC3461 mRNA by CrfA occurs by a unique mechanism by which CrfA binding at the 5′-UTR stabilizes the transcript against RNase degradation, ultimately leading to mRNA accumulation.
This work was supported by NIH grant GM32506 and by DOE grant DE-FG02ER64136 to L.S. S.G.L. and J.A.L. were supported by the Stanford NIH Genome Training Program.
Published ahead of print on 2 July 2010.
†Supplemental material for this article may be found at http://jb.asm.org/.