Small RNA-mediated control of the Agrobacterium tumefaciens GABA binding protein

Summary

Wounded plants activate a complex defence programme in response to Agrobacterium tumefaciens. They synthesize the non-proteinogenic amino acid γ-aminobutyric acid (GABA), which stimulates degradation of the quorum sensing signal N-(3-oxo-octanoyl) homoserine lactone. GABA is transported into A. tumefaciens via an ABC transporter dependent on the periplasmic binding protein Atu2422. We demonstrate that expression of atu2422 and two other ABC transporter genes is downregulated by the conserved small RNA (sRNA) AbcR1 (for ABC regulator). AbcR1 is encoded in tandem with another sRNA, which is similar in sequence and structure. Both sRNAs accumulate during stationary phase but only the absence of AbcR1 resulted in significant accumulation of Atu2422 and increased GABA import. AbcR1 inhibits initiation of atu2422 translation by masking its Shine–Dalgarno sequence and thereby reduces stability of the atu2422 transcript. It is the first described bacterial sRNA that controls uptake of a plant-generated signalling molecule. Given that similar sRNAs and ABC transporter genes are present in various Rhizobiaceae and in Brucella, it is likely that such sRNA-mediated control impacts a number of host–microbe interactions.

Introduction

Agrobacterium tumefaciens induces tumours, so-called crown galls, on plants upon transfer of a segment of its DNA (T-DNA) from its tumour-inducing (Ti) plasmid to the plant nuclear genome (Wood et al., 2001). Gene products encoded on the T-DNA reprogramme the infected cells to produce growth factors leading to the development of plant tumours, and opines, which are used by the pathogen as nitrogen and carbon source (Zupan et al., 2000). Some opines stimulate the synthesis of the quorum sensing signal 3-oxo-octanoyl homoserine lactone (OC8HSL) in A. tumefaciens resulting in increased aggressiveness of the bacterium (White and Winans, 2007). As part of a complex defence mechanism, plants synthesize γ-amino butyric acid (GABA). GABA is a remarkably versatile signalling molecule, functioning in animals as inhibitory neurotransmitter, as well as mediating communication between plants or plants and microbes (Shelp et al., 2006). Plant-produced GABA is imported into A. tumefaciens, where it induces the lactonase AttM (BlcC), which degrades the quorum sensing signal and attenuates bacterial virulence (Chevrot et al., 2006). Accumulation of proline antagonizes the GABA effect. Both GABA and proline are taken up by a specific ABC transporter in concert with the periplasmic binding protein Atu2422 (Haudecoeur et al., 2009a), but many aspects of the regulatory mechanisms that connect the uptake of GABA with its known effects are not yet understood.

In this study, we demonstrate that atu2422 expression and GABA uptake are tightly controlled by a small RNA (sRNA). Over the last decade, sRNAs have been recognized as widespread regulators of gene expression in bacteria (Narberhaus and Vogel, 2009). They range from 50 to 500 nucleotides (nt) in length, usually remain untranslated and carry out a variety of regulatory functions. Most sRNAs act through base pairing with target mRNAs, frequently modulating their translation and/or stability (Majdalani et al., 2005; Waters and Storz, 2009). They often display short, imperfect complementarity to their targets. A ‘seed sequence’ of only seven nucleotides can be sufficient to confer specificity (Kawamoto et al., 2006; Bouvier et al., 2008; Balbontín et al., 2010). This makes the identification of target mRNAs by computational means notoriously difficult (for review see Backofen and Hess, 2010). Furthermore, many sRNAs regulate not only a single target but modulate expression of multiple mRNAs (Altuvia et al., 1997; Lease et al., 1998; Majdalani et al., 1998; Sharma et al., 2007).

Bacterial sRNAs have been reported to regulate a number of critical cellular processes, e.g. chromosome replication and cell division (DicF: Bouche and Bouche, 1989), transcription (6S RNA: Wassarman and Storz, 2000), virulence and quorum sensing (Qrr: Lenz et al., 2004), RNA processing (RNase P: Kazantsev and Pace, 2006), protein stability (tmRNA: Muto et al., 1998), stress adaptation (OxyS: Altuvia et al., 1998), transition from exponential to stationary growth phase (DsrA and RprA: Majdalani et al., 1998; 2002), plasmid replication (RepE: Chai and Winans, 2005), carbon storage (CsrBC: Weilbacher et al., 2003) and oligopeptide transport (GcvB: Urbanowski et al., 2000).

During standard genome analysis, sRNA genes are not annotated. However, various computational and experimental strategies (reviewed in Backofen and Hess, 2010, and Sharma and Vogel, 2009, respectively) have revealed numerous sRNAs in diverse prokaryotes, not only in well-studied γ-proteobacteria but also in α-proteobacteria (Berghoff et al., 2009; Vercruysse et al., 2010), in the ε-proteobacterium Helicobacter pylori (Sharma et al., 2010), in Gram-positives (Toledo-Arana et al., 2009) and in archaea (Jäger et al., 2009).

The only sRNA of A. tumefaciens experimentally analysed so far is RepE, which is encoded on the Ti plasmid and controls its replication (Chai and Winans, 2005). However, in the closely related plant symbiont Sinorhizobium meliloti, eight sRNAs were experimentally confirmed on the basis of computational predictions by del Val et al. (2007) and 14 each by Ulve et al. (2007) and Valverde et al. (2008). 173 sRNAs were identified in a comprehensive approach using deep sequencing and microarray hybridizations (Schlüter et al., 2010). Some sRNAs were consistently identified in several of these studies and homologues in other α-proteobacteria including A. tumefaciens were predicted. Despite the substantial conservation of some of these sRNAs, their biological functions have remained enigmatic.

In the present study, we used a comparative bioinformatic approach to predict sRNAs in A. tumefaciens. We validated the presence of four sRNAs previously reported in S. meliloti and other rhizobia (Ulve et al., 2007; del Val et al., 2007; Valverde et al., 2008; Schlüter et al., 2010). We focussed our attention on two related sRNAs, which we suggest to name AbcR1 and AbcR2 for ABC transport regulator. One of the targets codes for a periplasmic binding protein that is required for uptake of the plant-derived defence signal GABA. This regulatory mechanism likely is part also of other host/pathogen or host/symbiont interactions because homologous sRNAs exist in several related bacteria.

Results

Identification of sRNAs on the circular chromosome of A. tumefaciens

The genome of A. tumefaciens is composed of a circular and a linear chromosome and the two megaplasmids pAT and the pTi (Wood et al., 2001). Here, we screened the intergenic regions of the circular A. tumefaciens chromosome and the genomes of S. meliloti and Rhizobium etli for conserved sRNA candidates using the comparative approach established previously by Voss et al. (2009a). Briefly, intergenic regions longer than 50 nucleotides were extracted from the three genomes and aligned using BLASTN. Sequences that produced a significant blast hit (E-value < 10⁻⁵) were collected in clusters. For each cluster a multiple sequence alignment was computed, based on which the ncRNA scoring was performed. This yielded 231 sRNA candidates, among them several that have previously been reported in S. meliloti and R. etli (Ulve et al., 2007; del Val et al., 2007; Valverde et al., 2008; Voss et al., 2009b; Schlüter et al., 2010; Vercruysse et al., 2010). The presence of four sRNAs already described in Rhizobiaceae and of nine arbitrarily chosen candidates was examined by Northern hybridization in total RNA from A. tumefaciens cells grown to different growth phases. The four previously described sRNA species were also detected in A. tumefaciens (Table S1). Despite their widespread occurrence, the physiological function of none of these sRNAs is known.

In this study, we focussed on the structure and function of the related sRNAs C2A and C2B, which will henceforth be called AbcR1 and AbcR2 respectively. These RNAs are transcribed from two closely spaced loci within the 600 nt long atu2186-atu2187 intergenic region. The flanking genes code for putative transcriptional LysR-type (atu2186) and ArsR-type (atu2187) regulators (Fig. 1A). Homologues of AbcR1 and AbcR2 (Fig. 1B) have been identified experimentally in S. meliloti and R. etli (Ulve et al., 2007; del Val et al., 2007; Valverde et al., 2008; Voss et al., 2009b; Schlüter et al., 2010; Vercruysse et al., 2010). Reflecting their sequence identity of 65% (see Fig. 3C), both A. tumefaciens sRNAs were detected by the same probe (Fig. 1C). AbcR1 and AbcR2 were present during exponential growth. Their abundance peaked towards the end of the exponential growth phase reaching a maximum at an OD600 of 2.0.

Figure 1. The two sRNAs AbcR1 and AbcR2. A. Schematic diagram of AbcR1 and AbcR2 and their flanking genes. B. Sequence alignment of AbcR1 and AbcR2 in A. tumefaciens (A.t.), R. etli (R.e.) and S. meliloti (S.m.). C. Growth phase-dependent expression of AbcR1 and AbcR2. Cells were grown in YEB medium and harvested at different growth phases (indicated above). 8 µg of total RNA was separated on 10% polyacrylamide gels containing 7 M urea and detected by Northern blot analysis using a DNA probe (generated by primers DNAprobe_C2AB_fw/rv). Sizes of the sRNAs are given on the right. Ethidiumbromide-stained 5S RNA and tRNAs were used as loading control.

Figure 3. Secondary structures and sequence alignment of AbcR1 and AbcR2. A. Secondary structure prediction of the AbcR1 sRNA and summary of the probing results. Filled arrows indicate cleavage sites of RNase V1. T1 cleavage sites are represented by open arrows. B. Secondary structure prediction of the AbcR2 sRNA. C. Sequence alignment of AbcR1 and AbcR2. Asterisks indicate identical nucleotides. Arrows indicate the three-stem-loop structure. D. Enzymatic cleavage of 5′ end-labelled AbcR1 RNA was carried out at 37°C. The RNA fragments were separated on an 8% polyacrylamide gel. RNase T1 (0.1 U) and RNase V1 (0.01 U) were used. The first lane (H₂O Contr.) indicates the incubation control with water instead of RNase. The second lane (T1 Contr.) shows the T1 ladder of hydrolysed, denatured AbcR1 RNA. T1 cuts, also indicated in (A), are marked by asterisks. Lane OH: alkaline ladder.

AbcR1 and AbcR2 are similar in sequence and structure

Bacterial sRNAs generally adopt particular secondary structures, which are functionally relevant. Therefore, it was important to know the exact ends of AbcR1 and AbcR2. Their 5′ ends were determined by primer extension analysis (Fig. 2A and B). Almost identical promoter-like sequences in the −10 and −35 regions (Fig. 2C) suggest that both sRNA are transcribed individually from separate promoters. Another experiment supported the presence of two independent transcription units. Cloning AbcR1 or AbcR2 preceded by 60 nucleotides of their upstream regions on a promoterless plasmid resulted in high expression of the RNAs in A. tumefaciens (data not shown).

Figure 2. Determination of the 5′ ends of AbcR1 and AbcR2. Primer extension analysis for AbcR1 (A) and AbcR2 (B) was carried out on RNA extracted from late stationary phase cultures. Reverse transcription was performed using ³²P-labelled oligonucleotide PE_C2A for AbcR1 and PE_C2B for AbcR2 5′ end mapping. T, G, C and A indicate the sequencing reaction. Asterisks mark the 5′ ends of the sRNAs. C. Alignment of putative promoter regions of AbcR1 and AbcR2. −10 and −35 regions are boxed.

Typical Rho factor-independent termination sequences (Fig. 3A and B) define the 3′ ends of both RNAs suggesting a total length of 117 and 99 nucleotides for AbcR1 and AbcR2 respectively. These sizes correspond well to the Northern blot signals (Fig. 1C) and indicate a gap of 85 nucleotides between the two sRNA genes. AbcR1 and AbcR2 have similar sequences (Fig. 3C) and both are predicted to fold into similar structures comprised of three hairpins (Fig. 3A and B). The most obvious difference is an extended hairpin I of AbcR1, which is the functionally most relevant part of the sRNA (see below).

As AbcR1 was found to play a more central role in gene regulation (see below), its structure was mapped by enzymatic probing using RNases T1 (cuts 3′ of single-stranded guanines) and V1 (specific for double-stranded and stacked regions). The overall cleavage pattern (Fig. 3D) was in good agreement with the predicted secondary structure (Fig. 3A). As expected, the terminal loops and single-stranded regions connecting the three stems were not cut by RNase V1 whereas many nucleotides in the predicted double-stranded stem regions were cleaved by this ribonuclease. RNase T1 cut at G48 positioned in the single-stranded region between hairpins I and II. This also agrees well with the calculated structure. The accessibility of G31 in hairpin I to RNase T1 suggested a much wider terminal loop than predicted.

AbcR1 targets two periplasmic binding proteins of the ABC transporter family

To address the biological function and identify potential target genes of AbcR1 and AbcR2, marker-less single mutants (ΔAbcR1 and ΔAbcR2) and a double mutant (ΔAbcR1/2) were generated. Cultures of the wild type (WT) and the deletion strains were grown to different growth phases and total protein extracts were separated on a 12% SDS gel. The protein composition of the double mutant showed no significant difference to the WT in exponential phase (Fig. 4A). In stationary phase, however, two proteins were clearly overrepresented. The corresponding bands were extracted from the gel, digested with trypsin and subjected to mass spectrometry. The proteins were identified as Atu2422 and Atu1879, both predicted periplasmic binding proteins of the ABC transporter family. Atu2422 has been reported as GABA and proline transporter (Haudecoeur et al., 2009a). Substances bound by Atu1879 are presently unknown.

Figure 4. Altered protein expression in sRNA deletion mutants and target validation. A. Total protein samples from different growth phases (indicated above) of the A. tumefaciens wild-type and the ΔAbcR1/2 (ΔR1/2) deletion mutant were loaded on a 12% SDS-PAGE. The positions of marker proteins are given on the left in kDa. Two prominent protein bands (arrows) accumulating in the mutant in late stationary phase were analysed via MALDI-TOF and identified as Atu2422 and Atu1879. B. Total protein samples from the wild-type, the single mutants ΔAbcR1 (ΔR1) and ΔAbcR2 (ΔR2) and the double mutant ΔAbcR1/2 (OD600 = 2.0) were loaded on a 12% SDS gel. Protein bands representing Atu2422 and Atu1879 are marked by arrows.

Protein fractions from the single mutants at an OD600 of 2.0 revealed that AbcR1 was much more effective in repressing the production of Atu2422 and Atu1879 (Fig. 4B). If at all, overexpression of the ABC transporter proteins in the ΔAbcR2 strain was only marginal.

AbcR1 accelerates turnover of the target mRNAs atu2422 and atu1879

AbcR1 and AbcR2 may modulate translation of atu2422 and atu1879 mRNAs as well as affect their stability. To determine the mRNA levels in the WT and sRNA mutants, we performed Northern blot analyses with total RNA from cells harvested at different growth phases. Probes against atu2422 and atu1879 detected transcripts of approximately 1119 and 774 nucleotides, respectively, suggesting the formation of monocistronic transcripts (Fig. 5A). Both mRNAs were present in WT cells at early exponential phase but disappeared towards stationary phase (when the two sRNAs are present, see Fig. 1C). The absence of AbcR1 in the single or double mutant resulted in significantly elevated transcript levels of atu2422 and atu1879 indicating that the sRNA reduces their expression in the WT. In the AbcR2 mutant, the ABC transporter mRNA levels were only slightly elevated in comparison with the WT suggesting that this sRNA does affect expression of these genes only marginally, if at all.

Figure 5. AbcR1 (and AbcR2) stimulate turnover of atu2422 and atu1879 mRNAs. A. Northern blot analysis of atu2422 and atu1879 transcript levels in different growth phases. The wild-type and the ΔAbcR1 (ΔR1), ΔAbcR2 (ΔR2) and ΔAbcR1/2 (ΔR1/2) strains were grown to optical densities indicated above in YEB medium. 5 µg of total RNA was separated on 1.2% denaturing agarose gels. Probes used for the blots are given on the right (primers: probe_atu2422_fw/rv and probe_atu1879_fw/rv) and apparent sizes of the corresponding transcripts are given in brackets. Ethidiumbromide-stained 16S rRNA fractions were used as loading control. B. Target–mRNA stability. Wild-type and ΔAbcR1 (ΔR1), ΔAbcR2 (ΔR2) and ΔAbcR1/2 (ΔR1/2) mutant cells were grown to OD600 of 1.0 in YEB medium. Samples were taken before and 1 and 6 min after addition of rifampicin (250 µg ml⁻¹). C. Quantification of transcript stabilities based on the above Northern blots. Calculated half-lives of the atu2422 and atu1879 mRNAs in the different strains are given on the right.

To investigate transcript stability of the mRNA targets, WT and sRNA mutants were grown to mid stationary phase (OD600: 1) before transcription was stopped by adding rifampicin to the cultures. Both atu2422 and atu1879 transcripts were significantly stabilized in the ΔAbcR1 and the ΔAbcR1/2 mutant strains (Fig. 5C) clearly demonstrating that AbcR1 accelerates turnover of the target mRNAs.

AbcR1 controls expression of a third ABC transporter

Computational target predictions (targetRNA: Tjaden et al., 2006, sRNAtarget: Cao et al., 2009) suggested that AbcR1 controls at least one other ABC transporter that escaped detection by our SDS-PAGE approach. The top hit for AbcR1 retrieved by the targetRNA programme was frcC encoding a component of a putative sugar ABC transporter (Fig. 6A). Northern blot experiments showed that expression of the frcCAK operon is controlled by AbcR1 in a growth phase-dependent manner (Fig. 6B). In summary, we conclude that AbcR1 regulates the expression of at least three ABC transporters in A. tumefaciens.

Figure 6. The additional AbcR1 target frcC. A. Predicted RNA duplex between AbcR1 and frcC. SD and AUG start codon sequences are boxed. Numbering of bases is given relative to the AUG start codon. B. Northern blot analysis of the frc operon in two different growth phases. The wild-type and the ΔAbcR1 (ΔR1), ΔAbcR2 (ΔR2) and ΔAbcR1/2 (ΔR1/2) strains were grown to optical densities as indicated in YEB medium. 5 µg of total RNA was separated on 1.2% denaturing agarose gels. Ethidiumbromide-stained 16S rRNA fractions were used as loading control. A schematic picture of the frc operon and the position of the probe (generated by primers probe_frcC_fw/rv) are shown below.

Interaction between AbcR1 and atu2422 in vitro

Like in the frcC mRNA (Fig. 6A), the programme RNA hybrid predicted that AbcR1 targets the SD sequences of atu2422 and atu1879 (Fig. 7A). Gel retardation experiments with in vitro synthesized RNAs were performed to examine this interaction. Target RNAs used for these experiments consisted of 50 nucleotides upstream and downstream from the AUG start codon. First, ³²P-labelled AbcR1 RNA was incubated with increasing concentrations of unlabelled target mRNA fragments before separation on a non-denaturing PAGE gel. In support of the observed in vivo effects, addition of atu2422 RNA resulted in complex formation with AbcR1 (Fig. 7B). The reciprocal experiment, in which the sRNA was added to ³²P-labelled atu2422 RNA also resulted in a clear band shift. The predicted interaction between AbcR1 and atu1879 could not be demonstrated in vitro probably because this interaction is weaker (Fig. 7A), suggesting that Hfq or other factors, facilitate complex formation in vivo.

Figure 7. AbcR1 binds to the atu2422 RNA in vitro. A. Predicted RNA duplexes between AbcR1 and the two target mRNAs atu2422 and atu1879. SD and AUG start codon sequences are boxed. Numbering of bases is given relative to the AUG start codon. B. Band shift experiments of AbcR1 to atu2422 and atu1879 mRNA fragments (+/− 50 nt from the AUG start codon) were performed as follows. Left: ³²P-labelled AbcR1 sRNA (< 0.05 pmol) was incubated with increasing concentrations of unlabelled atu2422 or atu1879 RNA at 30°C for 20 min. Right: ³²P-labelled atu2422 RNA (> 0.05 pmol) was incubated with unlabelled AbcR1 sRNA. Final concentrations of unlabelled added RNA were 50 (lanes 2, 6 and 12), 100 (lanes 3, 7 and 13), 200 (lanes 4, 8 and 14) and 500 (lanes 5, 9 and 15) fold excess. Lanes 1 and 11 show water controls. Lane 10 shows a control, in which sRNA and target RNA fragments were incubated at 75°C for 1 min following incubation for 20 min at room temperature. Samples were run on a 6% native gel at 250 V. C. Band shift experiment of AbcR1 to an atu2422 fragment with WT-Shine–Dalgarno sequence (left) and with a mutated SD sequence (right) in which GGGAG was altered to CCCUC. atu2422 RNA fragments (+/− 30 nt from the AUG start codon) were added with increasing concentrations (0, 50, 100, 200, 500, 1000 fold excess). Samples shown in lanes 7 and 14 were incubated at 75°C and cooled down to room temperature.

To support the assumption that AbcR1 blocks translation of atu2422 by interacting with its SD sequence, we used synthetic 60-nucleotides long atu2422 fragments with either the WT SD sequence or a mutated SD sequence (Fig. 7C). As expected, only the fragment with the WT SD sequence was shifted.

AbcR1 interferes with translation initiation at the atu2422 RNA

Toeprinting analysis was used to precisely map the interaction site between AbcR1 and atu2422. The strategy is outlined in Fig. 8A. In a control experiment demonstrating binding of the 30S ribosome to the SD sequence, a 100 bp atu2422 mRNA fragment (50 nt +/− from the AUG start codon) was annealed to an end-labelled primer complementary to the atu2422 coding region. Incubation with 30S subunits and tRNA^fMet followed by cDNA synthesis resulted in a typical toeprint at the characteristic + 15 position relative to the AUG start codon (‘ribosome toe’ in Fig. 8B). This signal was significantly decreased when AbcR1 RNA was added before incubation with 30S subunit and tRNA^fMet. Instead, prominent signals close to the SD sequence appeared (‘sRNA toe’ in Fig. 8B and arrows in Fig. 8C). This experiment shows that binding of AbcR1 blocks 30S binding. Strikingly, the sRNA region complementary to the SD sequence of atu2422 is located in the accessible loop of hairpin I.

Figure 8. AbcR1 blocks ribosome binding to the atu2422 RNA. A. Principle of the toeprinting experiment. Top: without 30S ribosome added, reverse transcription (RT) starting from a primer complementary to the target-mRNA, produces a full length cDNA. Middle: In the presence of 30S subunit, reverse transcription terminates at the bound ribosome, resulting in a truncated cDNA fragment (ribosome toe). Bottom: Pairing of the sRNA with the Shine–Dalgarno (SD) sequence prevents ribosome binding. The reverse transcriptase stalls at the sRNA binding site (sRNA toe). B. Ribosome toeprinting of the atu2422 RNA fragment (+/− 50 nt from the AUG start codon; 0.5 pmol) was performed as described in Experimental procedures. +/− indicate the presence or absence of 30S ribosomal subunit (1 pmol µl⁻¹) and AbcR1 sRNA (5 pmol µl⁻¹). On the left the atu2422 AUG start codon is marked. Full-length product, sRNA toe and ribosome toe are indicated on the right. C. Predicted RNA duplex formed by AbcR1 with atu2422. SD and AUG start codon sequences are boxed. Numbering of bases is given relative to the AUG start codon. Arrows indicate the sRNA toeprint on the atu2422 mRNA. The binding site on the AbcR1 secondary structure is shown below.

AbcR1 regulates GABA uptake

The atu2422 mRNA, which is efficiently silenced by AbcR1, encodes a binding protein responsible for GABA uptake in A. tumefaciens. GABA is synthesized by wounded plants as part of a complex defence mechanism, whereas proline, which also is released by plant tumours, acts as a GABA competitor (Haudecoeur et al., 2009a; Haudecoeur and Faure, 2010).

To test whether GABA uptake is influenced by AbcR1 and/or AbcR2, uptake of ¹⁴C-labelled GABA into A. tumefaciens was monitored. WT and sRNA mutants were grown to stationary phase in minimal medium (OD600: 0.9) and uptake was measured after adding 2 µM labelled GABA. The WT took up about 2% of the available GABA (Fig. 9). In contrast, cells lacking AbcR1 (ΔAbcR1 single mutant and the double mutant) imported significantly larger quantities of GABA demonstrating an important role for AbcR1 in this process. When proline or unlabelled GABA was added in 100-fold excess, GABA uptake was strongly reduced in all four strains showing the specificity of the uptake process.

Figure 9. The absence of AbcR1 increases GABA uptake. Uptake of radioactively labelled (¹⁴C) GABA (% of total available) in the WT compared with the sRNA deletion mutants (ΔR1, ΔR2 and ΔR1/2). Black bars show GABA uptake without any competitor. The effect of the competitors (100-fold excess) proline (grey bars) and non-labelled GABA (white bars) are demonstrated. The experiment was performed in triplicate with comparable results.

Discussion

Comparative genomics-based prediction of sRNAs has become one of the standard approaches for sRNA identification (Axmann et al., 2005; Mandin et al., 2007; Ulve et al., 2007; del Val et al., 2007; Padalon-Brauch et al., 2008; Voss et al., 2009a). In this study, we validated four out of 13 predicted sRNAs via Northern blot hybridizations in the ubiquitous plant–pathogen A. tumefaciens (Table S1). Although numerous sRNAs have been identified in various bacteria, the elucidation of their target RNAs (or proteins) by experimental or biocomputational approaches has remained a major challenge (Vogel and Wagner, 2007; Backofen and Hess, 2010). In this study, two promising targets of AbcR1 were apparent from an inspection of the protein profile of deletion mutants, another one derived from computational target prediction.

AbcR1, a conserved sRNA, controls expression of three ABC transporters

Two protein bands were visibly overrepresented in one-dimensional SDS-PAGE gels of protein extracts from an A. tumefaciens mutant lacking both AbcR1 and AbcR2, or AbcR1 alone. Atu2422 and Atu1879, the discovered targets of AbcR regulation, belong to the ABC transporter family, a major class of amino acid uptake systems that commonly use periplasmic solute-binding proteins to take up substrates upon transit of the outer membrane (Hosie and Poole, 2001; Davidson and Chen, 2004). Expression of ABC transporters must be tightly controlled in response to the environmental conditions. In enterobacteria, several sRNAs are involved in this process (Antal et al., 2005; Sharma et al., 2007). Our report indicates that sRNA-mediated control of ABC transporter genes might be a general phenomenon.

The molecule(s) bound by Atu1879 are unknown whereas the role of Atu2422 in GABA and proline uptake has previously been documented in A. tumefaciens and in Rhizobium leguminosarum (Hosie et al., 2002; Haudecoeur et al., 2009a). We demonstrated that the atu2422 mRNA is a direct target of AbcR1. Base pairing occurred without additional factors in vitro and binding to the SD sequence prevented ribosome binding. As a consequence, the sRNA-mRNA duplex without ribosome protection was prone to ribonucleolytic degradation in vivo.

Homologues of AbcR1 and AbcR2 have been found in S. meliloti and R. etli. Here, they were called SmrC15/16, sm3/3′, C15/16, sra41, SmelC411/412 in S. meliloti and ReC58/59 in R. etli (Ulve et al., 2007; del Val et al., 2007; Valverde et al., 2008; Schlüter et al., 2010; Vercruysse et al., 2010). In each case, they are encoded in tandem and are similar in sequence and predicted secondary structure suggesting a common function in these plant-associated bacteria.

Multiple sRNA copies are not unusual and can have different roles (Waters and Storz, 2009). They may act redundantly, serving as back-up in critical pathways. OmrA and OmrB are two such highly similar sRNAs encoded by adjacent genes on the E. coli chromosome, both regulating the same targets (Guillier and Gottesman, 2008). Alternatively, repeated sRNAs can function additively, as in the case of the four quorum regulatory RNAs 1–4 (Qrr1-4) in Vibrio harveyi (Tu and Bassler, 2007). A third possibility is that duplicated sRNAs act hierarchically upon each other, as in case of GlmY and GlmZ from E. coli (Reichenbach et al., 2008; Urban and Vogel, 2008).

Although AbcR1 and AbcR2 are induced simultaneously during stationary phase in A. tumefaciens, AbcR2 seems to play only a minor role in controlling atu2422 and atu1879. Slightly elevated mRNA levels and stabilities in the AbcR2 single mutant as compared with the WT (Fig. 5) and consistent differences between the AbcR1 mutant and the double mutant in the level of target mRNAs at high cell densities (Fig. 5A) and in GABA uptake (Fig. 9) might suggest some contribution of AbcR2 to regulation of atu2422 and atu1879. AbcR1 and AbcR2 do not seem to act hierarchically upon each other, because the absence of one sRNA did not influence the amount of the other (data not shown). AbcR2 might serve as back-up for AbcR1 under certain conditions. In addition, AbcR2 might have other not yet discovered targets in the cell. Computational target predictions (targetRNA: Tjaden et al., 2006, sRNAtarget: Cao et al., 2009) resulted in a long list of putative AbcR2 targets, many of them different from those of AbcR1 (data not shown) suggesting that both sRNAs have a distinct set of target genes.

sRNA–target interactions and the possible need for Hfq

Competition with ribosome binding explains the inhibitory action of many sRNAs that bind within or in the vicinity of ribosome binding sites (Argaman and Altuvia, 2000; Chen et al., 2004; Huntzinger et al., 2005; Udekwu et al., 2005; Sharma et al., 2007). Diverse sRNAs interact with their targets via residues in extended single-stranded regions (Huntzinger et al., 2005; Boisset et al., 2007; Johansen et al., 2008). AbcR1 binds the atu2422 mRNA at the SD sequence and it hybridizes to the mRNA via the terminal loop of hairpin I (nucleotides 18 to 35 on the sRNA). According to our structure probing results, this wide loop is perfectly suited for target interaction. The absence of such an anti-SD region in hairpin I of AbcR2 might explain the poor performance of this sRNA in in vivo experiments and the absence of atu2422 and atu1879 binding in vitro (data not shown). The failure of both AbcR RNAs to bind the atu1879 transcript in vitro despite clear-cut upregulation of this ABC transporter in the sRNA mutants suggests that some RNA–target interactions of AbcR1 (and probably AbcR2) in vivo are aided by additional factors. The most likely candidate is the RNA chaperone Hfq, which facilitates RNA–RNA interactions in particular when complementarity is limited (reviewed in Aiba, 2007; Brennan and Link, 2007; Waters and Storz, 2009). The three-stem-loop structure of AbcR1 and AbcR2 with an extended single-stranded AU-rich region between the second and the third hairpin is typical for Hfq-binding sRNAs (Zhang et al., 2002). As the S. meliloti homologues of AbcR1 and AbcR2 are highly stabilized by Hfq (Voss et al., 2009b; Torres-Quesada et al., 2010), it is likely that the A. tumefaciens sRNAs rely on Hfq protein for efficient target binding. This assumption is supported by a recent microarray analysis in S. meliloti. Compared with the WT, transcripts of 140 transporter-related genes accumulated in an hfq mutant (Gao et al., 2010). Torres-Quesada et al. (2010) obtained similar results in microarrays and proteome analyses. One of the upregulated genes in both studies was the atu2422 homologue livK indicating a crucial role of Hfq in control of this ABC transporter.

Small RNA-controlled GABA uptake

Successful tumour formation by A. tumefaciens on plants is preceded by a sophisticated cross-kingdom signalling process involving numerous plant and bacterium-derived molecules. Among the plant signals are salicylic acid, indole-3-acetic acid and GABA (Chevrot et al., 2006; Liu and Nester, 2006; Yuan et al., 2007; Anand et al., 2008). The bacterial response to these compounds is complex and partially overlapping. 95 A. tumefaciens genes were found to be specifically regulated by GABA; 19 were differentially expressed in response to GABA plus one or two of the other signalling molecules (Yuan et al., 2008).

Perception and uptake of GABA depends on the periplasmic binding protein Atu2422 containing a typical Venus flytrap domain (Morera et al., 2008). GABA attenuates virulence functions by inducing the expression of the lactonase AttM in A. tumefaciens (Chevrot et al., 2006). Preferentially in young emerging tumours, the lactonase cleaves the bacterial quorum sensing signal 3-oxo-octanoyl homoserine lactone (OC8HSL) thereby modulating quorum sensing-dependent functions (Carlier et al., 2004; Haudecoeur et al., 2009b; Khan and Farrand, 2009). Proline is an alternative high-affinity substrate of the Atu2422/Bra system (Planamente et al., 2010). It antagonizes the GABA-induced response (Haudecoeur et al., 2009a; Haudecoeur and Faure, 2010). In line with these observations, enrichment of GABA in the AbcR1 mutants was efficiently inhibited in the presence of proline (Fig. 9).

Our finding that a stationary phase-induced sRNA silences expression of the GABA uptake system provides an interesting potential link between bacterial quorum sensing, virulence and plant defence. From the bacterial perspective, repression of GABA uptake by AbcR1 at high cell densities in the WT is important to prevent interference of the plant signal with the bacterial quorum sensing system. Shutting off synthesis of the GABA uptake system might be a successful strategy to subvert the plant defence mechanism. In support of this hypothesis, recent studies have shown that plants infected by an atu2422 deletion mutant produce more tumours than plants infected by the WT strain. Wild-type like virulence was restored in the complemented mutant (Haudecoeur et al., 2009a).

Planamente et al. (2010) found that the Atu2422 amino acids Tyr275 and Phe77 are important for GABA specificity in A. tumefaciens. Interestingly, most bacteria containing Atu2422-related periplasmatic binding proteins with these two key residues interact with animal or plant hosts. The function of GABA in animal–microbe interactions still remains to be explored. The Atu2422 orthologue RL3745 in the plant symbiont R. leguminosarum is known to be involved in GABA transport (Hosie et al., 2002). Although not strictly required for nitrogen fixation activity of bacteroids, GABA provided by the plant partner plays an important role in energy generation of the bacterial symbiont (Prell et al., 2009). Interestingly, homologues of AbcR1 as well as atu2422 exist or have been predicted not only in Rhizobium, Sinorhizobium and Agrobacterium but also in Mesorhizobium, Brucella and Ochrobactrum (Fig. S2). Hence, it is an attractive hypothesis that sRNA-mediated control of ABC transporters relevant for host–microbe interactions is a common theme.

Experimental procedures

Biocomputational prediction of sRNAs

A comparative prediction of sRNAs within intergenic regions on the circular chromosome of A. tumefaciens was performed as described (Voss et al., 2009a).

Bacterial growth conditions

Bacterial strains and antibiotics used in this study are listed in Table S2. A. tumefaciens strains were cultivated in YEB medium or AB minimal medium at 30°C. E. coli was grown in LB medium at 37°C.

Strain and vector constructions

ΔAbcR1, ΔAbcR2 and ΔAbcR1/2 mutant strains were constructed as described in Paulick et al. (2009) with minor modifications. Upstream and downstream fragments (400 nt) of the desired sRNA gene regions were amplified by PCR using the corresponding primer pairs (Table S3). To construct the ΔAbcR1/2 double mutant, the up-fragment of AbcR1 and the down-fragment of AbcR2 were used. The fragments were ligated into the suicide vector pK19mobsacB. The resulting plasmid was introduced into A. tumefaciens by electroporation. Single cross-over integration mutants were selected on LB plates containing kanamycin. Single colonies were grown over night in liquid LB without antibiotics and plated on LB containing 10% (w/v) sucrose to select for plasmid excision by double cross-over events. Kanamycin sensitive colonies were then checked for the targeted deletion by colony PCR and Northern analyses.

RNA preparation

Cells (10 ml) were harvested by centrifugation. After washing in ice-cold AE-buffer (20 mM Na acetate, pH 5.5), pellets were immediately frozen in liquid nitrogen. Total RNA of cultured bacteria was isolated using the hot acid phenol method (Aiba et al., 1981). To measure mRNA stability, rifampicin was added to the cell cultures in a final concentration of 250 µg ml⁻¹ and samples for RNA isolation were collected before (0 min) and 1 and 6 min after addition of the transcriptional inhibitor.

Northern analysis

For the detection of small RNAs (50–200 nt), 8 µg total RNA was separated on 10% polyacrylamide gels containing 7 M urea and subsequently transferred onto nylon membranes by semi-dry electroblotting. Hybridizations were carried out at 37°C over night using digoxygenine-labelled DNA or RNA probes, which were produced according to the instruction manual (Roche, Mannheim, Germany). After hybridization, membranes were first washed for 5 min at room temperature with solution 1 (5 × SSC, 0.1% SDS) followed by incubation with solution 2 (1 × SSC, 0.1% SDS) and 3 (0.5 × SSC, 0.1% SDS). Detection was carried out by exposing the blot to a luminescence detector, using chemiluminescence substrate (CDP-Star; Roche Molecular Biochemicals, Mannheim, Germany). For target mRNA detection (500–3000 nt), Northern blot analysis was performed as described previously (Waldminghaus et al., 2005).

Primer extension analysis

Primer extensions were carried out as described before (Babst et al., 1996). Primer PE_C2A was used for reverse transcription to map the AbcR1 transcriptional start site. Primer PE_C2B was used for AbcR2 respectively (Table S3). The DNA sequence reactions were performed using the Thermo Sequenase cycle sequencing kit (USB, Cleveland, Ohio, USA) using plasmid pUC_IGR_C2AB as template for the sequencing reactions (Table S2).

RNA structure probing

AbcR1 RNA was synthesized in vitro by run-off transcription with T7 RNA polymerase from the linearized plasmid run-off_C2A. 5′ end labelling was performed as described (Brantl and Wagner, 1994). Partial digestions of 5′ end labelled RNA with ribonucleases T1 (Ambion, Austin, USA) and V1 (Ambion, Austin, USA) were carried out as follows. RNA corresponding to 30 000 c.p.m. was mixed with 1 µl of 5 × TMN buffer (100 mM tris acetate, pH 7.5; 10 mM MgCl₂; 500 mM NaCl). 0.4 µg tRNA and distilled water were added to a volume of 4 µl. Samples were pre-incubated for 5 min at 37°C before 1 µl of T1 (0.1 U) or 1 µl of V1 (0.01 U) nuclease was added. After 5 min of cleavage, 5 µl formamide loading dye was added, samples were denatured for 5 min at 95°C and aliquots were separated on a denaturing 8% polyacrylamide gel. The alkaline ladder was generated as described (Brantl and Wagner, 1994). The RNase T1 ladder was obtained by incubating labelled RNA (30 000 c.p.m.) in 1 × sequencing buffer (Ambion, Austin, USA) for 1 min at 95°C.

Protein identification

Proteins were separated by SDS-PAGE as published previously (Schägger and von Jagow, 1987). Proteins from excised gel bands were subjected to in-gel digestion as described (Bandow, 2010). 0.5 µl of peptide extracts containing α-cyano-4-hydroxycynamic acid as matrix were spotted on a stainless steel MALDI target (Bruker Daltonik, Bremen). The tryptic peptide mixture was then analysed using the ultraflex II MALDI TOF instrument (Bruker Daltonik, Bremen) using a modification of a method described elsewhere (Grzendowski et al., 2010).

Gel shift experiment

5′ end labelling of AbcR1 or atu2422 with ³²P was carried out as described (Brantl and Wagner, 1994). RNA band shift experiments were performed in 1× structure buffer (Ambion, Austin, USA) in a total reaction volume of 10 µl as follows. 5′ end labelled AbcR1 or atu2422 RNA (RNA corresponding to 5000 c.p.m.) and 1 µg of tRNA were incubated in the presence of unlabelled atu2422 or atu1879 mRNA fragments (50 or 30 nt +/− from the AUG start codon) or AbcR1 sRNA at 30°C for 20 min. The final concentrations of added RNA fragments are given in the figure legends. Before gel loading, the binding reactions were mixed with 3 µl of native loading dye (50% glycerol, 0.5× TBE, 0.1% bromophenol blue and 0.1% xylenxyanol) and run on native 6% polyacrylamide gels in 0.5× TBE buffer at 250 V for 3 h.

Toeprint analysis

Toeprint assays were carried out according to Hartz et al. (1988) with minor modifications. Annealing mixtures contained 0.5 pmol unlabelled atu2422 mRNA fragment (50 nt +/− from the AUG start codon) and 1 pmol of 5′ end labelled primer run-off_atu2422_rv in VD buffer without magnesium. Annealing mixtures were heated for 3 min at 80°C and snap-frozen in liquid N₂. After incubation on ice for 30 min 5 pmol sRNA AbcR1 was added and incubated at 37°C for 20 min. All following incubations were performed at 37°C. 6 pmol 30S ribosome (from E. coli) or water (negative control) was added, following incubation for 5 min. After addition of 16 pmol tRNA^fMET (Sigma-Aldrich, St Louis, Missouri, USA) incubation for another 15 min followed before 2 µl of MMLV-Mix [VD + Mg²⁺, BSA, dNTPs and MMLV reverse transcriptase (USB, Cleveland, Ohio)] was added. cDNA synthesis was performed at 37°C. Reactions were stopped after 10 min by adding formamide loading dye and aliquots were separated on a denaturing 8% polyacrylamide gel. Toeprint signals were identified by comparison with sequences generated with the same 5′ end labelled primer.

GABA uptake assay

GABA uptake assays were performed on cultures of A. tumefaciens grown in AB minimal medium (OD600: 0.9). Reactions were initiated by adding [4-¹⁴C]-4-Aminobutyric acid (¹⁴C-GABA; Hartmann Analytic GmbH, Braunschweig, Germany) to cell suspensions at a final concentration of 2 µM. Samples were incubated for 5 min at 30°C under constant shaking. The reaction was terminated by rapid filtration of samples (1 ml) on ultrafiltration membranes (0.45 µm Millipore HA, Millipore). Filters were washed four times with 1 ml AB medium. Filter-bound radioactivity was quantified in a scintillation counter. 200 µM proline or unlabelled GABA was added to some samples (indicated in the figure legend) and incubated for 10 min at 30°C previous to ¹⁴C-GABA addition.

Bioinformatic tools

Sequence alignments were generated by the ClustalW software obtained from http://www.ebi.ac.uk/Tools/clustalw. Secondary structures were predicted by the programme mfold (Version 3.2) from the website http://frontend.bioinfo.rpi.edu/applications/mfold/cgi-bin/rna-form1.cgi (Zuker, 2003). Target RNA predictions were performed with TargetRNA (http://snowwhite.wellesley.edu/targetRNA/contact.html), described in (Tjaden et al., 2006) and sRNATarget, obtained from http://ccb.bmi.ac.cn/srnatarget/index.php (Cao et al., 2009). sRNA-mRNA duplexes were predicted with RNA-hybrid (Rehmsmeier et al., 2004), obtained from http://bibiserv.techfak.uni-bielefeld.de/rnahybrid. 5′ regions of A. tumefaciens genes (+/− 50 nt from the AUG start codon) were used as target sequence input.

Acknowledgements

We are grateful to Jose Jimenz-Zurdo for providing plasmid pBBSyn. We thank Rosemarie Gurski for excellent technical assistance, Birgit Klinkert for 30S ribosome isolation and Aaron Overlöper for construction of the AbcR1/2 deletion mutant. Parts of this study were conducted in the Central Unit for Ionbeams and Radionuclides (RUBION). This work was supported by a grant from the German Research Foundation (DFG priority programme SPP 1258: Sensory and regulatory RNAs in Prokaryotes) to FN and to WRH and a fellowship from the RUB Research School to IW.