Region 2.1 of the Escherichia coli heat-shock sigma factor RpoH ({sigma}32) is necessary but not sufficient for degradation by the FtsH protease

All organisms react to heat stress with the expression of heat-shock genes in order to cope with accumulating unfolded proteins. In Escherichia coli, the heat-shock response is initiated at the transcriptional level by the alternative sigma factor σ³² (RpoH). During heat stress, RpoH associates with RNA polymerase (RNAP) and initiates transcription of the heat-shock regulon, coding mainly for molecular chaperones and proteases. The transient induction of the heat-shock response is tightly controlled by a regulatory feedback loop coupling the amount of heat-shock proteins to the cellular protein folding state (reviewed by El-Samad et al., 2005; Gross, 1996; Yura et al., 2000).

The activity and stability of RpoH is modulated by the chaperone systems DnaKJ and GroESL (Gamer et al., 1996; Guisbert et al., 2004; Straus et al., 1990). At physiological temperature, e.g. 30 °C, DnaK and DnaJ associate with RpoH, rendering it inaccessible for RNAP. The sequestered sigma factor is made available for degradation by the protease FtsH by a still unknown mechanism. Under heat-shock conditions, however, DnaK and DnaJ are titrated away by a high concentration of unfolded proteins. As a consequence, RpoH becomes available to associate with RNAP and to initiate transcription of heat-shock genes.

The major protease degrading RpoH is FtsH, a membrane-bound, hexameric, zinc- and ATP-dependent metalloprotease, which itself is a heat-shock protein (Herman et al., 1995; Tomoyasu et al., 1995). Apart from general protein quality-control functions in degrading misfolded inner-membrane proteins, the FtsH protease has important regulatory functions (reviewed by Ito & Akiyama, 2005). Its active site is oriented towards the cytoplasm and it degrades specifically a number of soluble proteins such as LpxC, the key enzyme in LPS biosynthesis (Führer et al., 2006; Ogura et al., 1999), phage λ proteins cII, cIII and Xis (Herman et al., 1997; Kihara et al., 1997; Leffers & Gottesman, 1998; Shotland et al., 1997), and the sigma factor RpoH. It is still largely unknown how this repertoire of entirely unrelated substrates is recognized and processed by the FtsH protease.

Multiple regions of RpoH have been implicated in its activity and stability (Arsène et al., 1999; Bertani et al., 2001; Blaszczak et al., 1999; McCarty et al., 1996; Nagai et al., 1994; Narberhaus & Balsiger, 2003). Recently, an important turnover element was narrowed down to the highly conserved region 2.1 (Horikoshi et al., 2004; Obrist & Narberhaus, 2005). Several point mutations in this region markedly stabilized RpoH against FtsH-mediated proteolysis. Horikoshi et al. (2004) provided evidence that the enhanced activity and stability of RpoH proteins mutated in region 2.1 is not the result of an increased interaction with RNAP core enzyme. Since the critical residues line up on the same face of a putative α-helix, we speculated that these amino acids might be exposed for interaction(s) with components of the proteolytic machinery, either chaperones or the FtsH protease itself (Obrist & Narberhaus, 2005).

Proteolysis of alternative sigma factors has emerged as a common principle in bacterial gene regulation (Gottesman, 1999; Hengge & Bukau, 2003; Jenal & Hengge-Aronis, 2003). Interestingly, even closely related sigma factors, like RpoH and RpoS (σ³⁸), are targeted to different proteases. Under optimal growth conditions, the starvation and general stress sigma factor RpoS is degraded by the ClpXP protease. Region 2.5 and the N-terminal region are critical for degradation and are recognized by the adaptor protein RssB and hexameric ClpX rings, respectively (Becker et al., 1999; Stüdemann et al., 2003). Region 2.1 of RpoS has not been implicated in ClpXP-mediated turnover of the sigma factor.

Although region 2.1 is a rather conserved region in all σ⁷⁰-type sigma factors (Lonetto et al., 1992), it must be sufficiently diverse to allow for its critical role in protease discrimination. We asked whether transfer of RpoH region 2.1 to FtsH-resistant sigma factors would render them susceptible to degradation by FtsH. The corresponding regions were exchanged between E. coli RpoH and Bradyrhizobium japonicum RpoH₁, which is known to be active and stable in E. coli (Urech et al., 2000). Introduction of RpoH region 2.1 into RpoS and into a bacterial one-hybrid system, developed to screen for protease recognition sites (Dautin et al., 2000), was used to examine whether region 2.1 would confer FtsH sensitivity to these hybrid proteins. Finally, the interaction between RpoH and its partner proteins was assessed. Our results suggest that region 2.1 is involved in direct interaction with FtsH. In addition we conclude that regions outside region 2.1 are required for efficient turnover of the sigma factor.

Bacterial strains and growth conditions.
Bacterial strains used in this study are listed in Table 1. All strains were grown in Luria–Bertani medium (Sambrook & Russell, 2001). Antibiotics and chemicals were added as follows: ampicillin (Ap) (200 µg ml^–1), chloramphenicol (Cm) (200 µg ml^–1), kanamycin (Km) (30 µg ml^–1), nalidixic acid (Nal) (15 µg ml^–1) and IPTG (1 mM final concentration), unless indicated otherwise. E. coli DH5α, GB2 and GB28 were grown at 37 °C, E. coli DHM1 at 30 °C, E. coli ΔftsH at 23 °C and E. coli ΔrpoH at 25 °C.

Table 1. Bacterial strains and plasmids used in this study

Plasmids and recombinant DNA techniques.
Plasmids used in this study are presented in Table 1 and primer sequences are shown in Table 2. DNA manipulations were performed according to standard protocols (Sambrook & Russell, 2001). To introduce single amino acid exchanges and to swap nucleotides coding for region 2.1 between E. coli rpoH and B. japonicum rpoH₁ primer-based mutagenesis was used (Quik-Change Site-Directed Mutagenesis kit, Stratagene). The plasmids pEC5364, pEC5377 and pEC5388 resulted from site-directed mutagenesis with template plasmid pEC5217 using primer pairs MO11/12, MO19/20 and MO21/22, respectively (exchanged nucleotides in Table 2 are underlined twice). To construct pRJ5389, plasmid pRJ5000 served as template and primers MO23 and MO24 were used. By transfer of the NcoI–PstI fragment of pEC5385 into pEC5217, replacing the wild-type fragment, plasmid pEC5387 was obtained. Plasmids coding for C-terminally histidine-tagged RpoH proteins were based on a PCR fragment amplified from pEC5217 using primers MO3 and MO4. The 870 bp fragment was cloned via NdeI/XhoI into the expression vector pET24b(+) (Novagen), resulting in pEC5261. Point mutations were transferred from rpoH into rpoH-His by digesting pEC5217 derivatives with MluI/PstI. A 526 bp fragment was used to replace the equivalent wild-type fragment in pEC5261. To generate pRJ5399, plasmid pRJ5389 carrying B. japonicum rpoH₁ with the 2.1 region of E. coli rpoH was digested with SacI/HindIII and the 296 bp fragment was cloned into pRJ5086, replacing the SacI–HindIII wild-type fragment. To construct plasmids pBO711, pBO712, pBO713, pBO714 and pBO715, PCR was performed with primers MO39 and MO40 using pEC5217, pEC5357, pEC5356, pEC5360 and pEC5387 as templates. PCR fragments were digested with NheI/KpnI and introduced into pEC5352, replacing the rpoH fragment between the subdomains of the Bordetella pertussis adenylate cyclase gene.

Table 2. Oligonucleotides used in this study

The construction of plasmid pRpoS18 was described previously (Becker et al., 1999). Codons for amino acids 47–55 of helix 12a of E. coli σ³² were introduced into pRpoS18 by a four-primer/two-step PCR mutagenesis protocol as described previously (Germer et al., 2001). The internal primer with the coding sequence for helix 12a of σ³² and the external primers with an EcoRI restriction site at the start of the rpoS coding sequence and an AatII restriction site 672 bp downstream of the EcoRI site in the coding sequence of rpoS are listed in Table 2. For PCR reaction, primer Int1 was used with Ext1, and Int2 with Ext2. These two restriction sites (EcoRI, AatII) were used for cloning the PCR fragment into pRpoS18, resulting in pRpoS18(2.1_H).

In vivo degradation assay and immunoblot analysis.
In vivo degradation and sample preparation of E. coli ΔrpoH cells carrying pEC5217 derivatives was performed as previously described (Obrist & Narberhaus, 2005). Equivalent amounts of protein (10 µg) were separated on 12 % SDS-polyacrylamide gels and transferred to nitrocellulose membranes (Hybond-C; Amersham). RpoH proteins were detected using an anti-RpoH antibody (polyclonal rabbit anti-RpoH antibody; 1 : 4000 dilution) and a secondary antibody [goat anti-rabbit immunoglobulin G(H+L)-horseradish peroxidase conjugate; Bio-Rad; 1 : 3000 dilution], followed by chemiluminescence detection (SuperSignal, Pierce). RpoH bands on X-ray films (ECL Hyperfilm, Amersham) were scanned and quantified using the AIDA program (Advanced Image Data Analyser, version 4.13, raytest).

Pulse labelling of cells with L-[³⁵S]methionine and immunoprecipitation of σ^S was described previously (Lange & Hengge-Aronis, 1994). E. coli GB2 bearing plasmid pRpoS18(2.1_H), grown in M9/0.4 % (v/v) glycerol supplemented with 100 µg ampicillin ml^–1 was harvested at an optical density (578 nm) of 0.6 and pulse-labelled for 1 min. Chase times were between 1 and 10 min. For immunoprecipitation, a polyclonal antiserum against RpoS was used. RpoS bands were quantified on a FLAG2000G phosphoimager (Fuji Photo Film Co.). The intensity of the RpoS band was calculated relative to the intensity of bands representing stable proteins that cross-reacted with the RpoS antiserum.

In vivo degradation of RpoS was also assayed by using spectinomycin to stop protein biosynthesis. After addition of spectinomycin (1.5 mg ml^–1) at an OD₅₇₈ of 0.6, samples were taken at the indicated time points, precipitated with 10 % cold trichloracetic acid, prepared for SDS-PAGE and transferred to a PVDF-membrane (Carl Roth) as previously described (Lange & Hengge-Aronis, 1994); 30 µg of total cellular protein was applied per lane. A polyclonal serum against σ^S, a goat anti-rabbit IgG-alkaline phosphatase conjugate (Sigma) and a chromogenic substrate (BCIP/NBT, Carl Roth) were used for visualization of the σ^S bands.

Protein expression and purification.
Plasmids coding for C-terminally histidine-tagged RpoH proteins were freshly transformed into E. coli BL21(DE3) cells. Cultures (500 ml) were grown at 37 °C to an OD₆₀₀ of 0.6, then production of recombinant proteins was induced by the addition of 0.5 mM IPTG. After incubation for 2 h at 30 °C the cells were harvested. Pellets were resuspended in 20 ml binding buffer (0.5 M KCl, 20 mM Tris/HCl, 5 mM imidazole, pH 7.9), 1 mM PMSF and 5 µg DNase I ml^–1 were added and cells were disrupted by sonication (6x15 s at 20 % output level; Branson Sonifier 250). Cell extracts were centrifuged at 15 000 g for 30 min and the supernatant was loaded onto a 0.5 ml Ni-NTA-agarose column (Qiagen). Proteins were eluted with increasing imidazole concentration and stored in 20 % (v/v) glycerol at –80 °C. GST-FtsH was expressed, purified and stored as described previously (Shotland et al., 2000b).

Cell growth, harvesting and preparation of protein extract for co-purification experiments were essentially performed as described above except that cells from a culture of 100 ml were resuspended in 4 ml binding buffer. Extracts were loaded on a column of 0.5 ml and bound proteins were washed with 1.5 ml washing buffer and eluted with 1 ml elution buffer, consisting of binding buffer supplemented with increasing imidazole concentrations: W1 (5 mM), W2 (25 mM), W3 and W4 (50 mM), E1 (100 mM), E2 (125 mM), E3 (150 mM) and E4 (1 M). From each fraction, a sample of 15 µl was analysed by 12 % SDS-PAGE and proteins were visualized by Coomassie Brillant Blue staining or immunodetection with specific antisera. Anti-DnaK, -DnaJ and -GroEL sera (Stressgen Bioreagents) were diluted 1 : 3000, 1 : 2000 and 1 : 5000, respectively, and detected by secondary anti-rabbit sera as described above.

In vitro degradation assay.
Degradation tests with purified GST-FtsH and RpoH-His derivatives were carried out essentially as described previously (Tomoyasu et al., 1995; Urech et al., 2000). The total volume was adjusted to 85 µl and contained 17 µg GST-FtsH and 4.25 µg RpoH-His. The reaction was started by addition of 5 mM ATP and incubated at 42 °C. Samples of 10 µl were taken at different time points and analysed on 10 % SDS-polyacrylamide gels followed by staining with Coomassie Brillant Blue. The relative amount of RpoH proteins on scanned gels was quantified with the AIDA program (Advanced Image Data Analyser, version 4.13, raytest).

Other methods.
SDS-PAGE and β-galactosidase assays were performed as described by Sambrook & Russell (2001) and Miller (1972), respectively.

Additional point mutations in region 2.1 that render E. coli RpoH more active in vivo
Point mutations in RpoH region 2.1 at positions 47, 50, 51, 54 or 55 were shown to turn the sigma factor into a stabilized protein, resulting in enhanced transcription of an RpoH-dependent promoter in vivo (Horikoshi et al., 2004; Obrist & Narberhaus, 2005). The predicted three-dimensional structure suggested that amino acids at positions 47, 50 and 54 are positioned on the same side of the putative α-helix 12a in region 2.1 of RpoH (Obrist & Narberhaus, 2005). The sequence of region 2.1 from 21 bacterial species shows that several RpoH proteins contain a proline residue at position 47 (Fig. 1). It is known that proline can act as a helix breaker introducing a kink into the α-helix. To assess whether a proline residue would affect the activity and stability of the E. coli protein, we introduced the corresponding codon at this position. In another mutant, leucine 47 was changed to a cysteine residue. An RpoH protein with an isoleucine to threonine exchange at position 54 that derived from our previous mutagenesis screen but had not been analysed yet was included in this study. The different rpoH alleles coding for RpoH-L47P, RpoH-L47C and RpoH-I54T were transformed into an E. coli ΔrpoH strain carrying a chromosomally integrated RpoH-dependent groE-lacZ fusion. Functionality of the mutated sigma factors was assessed by β-galactosidase assays. Mean values of three independent experiments showed that all RpoH derivatives tested were functionally intact. They induced higher β-galactosidase activities than the wild-type (WT) sigma factor (set to 100 %) (Fig. 2). Sigma factor activities differed substantially among RpoH derivatives mutated at the same position. Reproducibly, RpoH-L47C (274 %) and RpoH-L47P (246 %) were more active than RpoH-L47Q (142 %), which originates from the original genetic screen (Obrist & Narberhaus, 2005). Variants at position 54 showed 213 % (RpoH-I54T) and 135 % (RpoH-I54F) activity in comparison to RpoH-WT.

(82K): Fig. 1. Sequence alignment, generated with CLUSTAL W (Thompson et al., 1994), of RpoH region 2.1 from different proteobacteria. Subdivisions of the listed proteobacteria are indicated on the left. The predicted secondary structure derives from our previous study (Obrist & Narberhaus, 2005). Residues 47, 50 and 54, which are critical for turnover of the protein, are indicated. The RpoH sequences belong to: E.co, Escherichia coli; B.ja_1, Bradyrhizobium japonicum RpoH1; B.ja_2, Bradyrhizobium japonicum RpoH2; S.me, Sinorhizobium meliloti; A.tu, Agrobacterium tumefaciens; C.cr, Caulobacter crescentus; R.sp, Rhodobacter sphaeroides; R.ca, Rhodobacter capsulatus; Z.mo, Zymomonas mobilis; H.th, Hydrogenophilus thermoluteolus; R.me, Ralstonia metallidurans; C.fr, Citrobacter freundii; C.ma, Colwellia maris; E.cl, Enterobacter cloacae; P.ae, Pseudomonas aeruginosa; P.pu, Pseudomonas putida; P.mi, Proteus mirabilis; S.ma, Serratia marescens; V.ch, Vibrio cholerae; X.ca, Xanthomonas campestis; M.xa, Myxococcus xanthus; S.au, Stigmatella aurantiaca. For comparison, the E. coli RpoS sequence is given at the bottom of the alignment. Amino acids identical to conserved RpoH residues are underlined.

(34K): Fig. 2. Sigma factor activity of RpoH variants derived from E. coli RpoH and B. japonicum RpoH1. β-Galactosidase activity was measured in E. coli ΔrpoH cells carrying a chromosomally encoded transcriptional PgroE-lacZ fusion. An activity of 100 % corresponds to 1000 Miller units. Error bars indicate standard deviations from at least three independent assays. RpoH-WT is wild-type E. coli RpoH and RpoH1-WT is the B. japonicum RpoH1 protein. RpoH-L47Q, -L47C, -I54F and -I54T contain the indicated amino acid exchanges in E. coli RpoH. In RpoH-2.1_H1, amino acids 47–55 (helix 12a) of E. coli RpoH are exchanged for the equivalent B. japonicum RpoH1 region. Conversely, in RpoH1-2.1_H B. japonicum helix12a was replaced by the E. coli RpoH region.

Exchange of the turnover element in region 2.1 between E. coli RpoH and B. japonicum RpoH₁ affects sigma factor activity and stability
The amino acid sequence of the comparatively stable RpoH₁ protein of the α-proteobacterium B. japonicum deviates from its enterobacterial homologue at several positions in region 2.1 (Fig. 1). Only two alanine residues (A49 and A50) and L53 are identical in the predicted α-helix 12a, which is flanked by a conserved aspartic acid residue (D46) and an SHLR-motif. Nevertheless, the corresponding region of RpoH₁ is also predicted to form an α-helix (data not shown).

To test whether E. coli RpoH could be stabilized by the corresponding RpoH₁ helix and vice versa, the nine amino acids located between the conserved residues D46 and S56 were swapped between the two proteins. Both hybrid proteins were active sigma factors (Fig. 2). E. coli RpoH containing RpoH₁ region 2.1 (designated RpoH-2.1_H₁) exhibited strongly enhanced activity (256 %) as compared to RpoH-WT (100 %), whereas RpoH₁-2.1_H was almost as active (356 %) as RpoH₁-WT (395 %).

The half-life of the plasmid-encoded RpoH derivatives was determined in the E. coli ΔrpoH strain in order to avoid interference with chromosomally encoded sigma factor. The RpoH proteins mutated in region 2.1 were stabilized in comparison to the wild-type protein (Fig. 3a). RpoH-L47P was the most stable protein (about 10 times as stable as RpoH-WT), and RpoH-L47C and RpoH-I54T were stabilized approximately fivefold (Fig. 3b). Both RpoH-RpoH₁ hybrids showed intermediate stability (Fig. 3c). E. coli RpoH was stabilized by the RpoH₁ helix 12a (RpoH-2.1_H1) but did not reach the stability of RpoH₁. B. japonicum RpoH₁ was slightly destabilized by the corresponding E. coli segment (RpoH1-2.1_H). However, it remained much more stable than E. coli RpoH. The finding that both helix replacements caused intermediate protein stability suggests that amino acids outside region 2.1 contribute to efficient degradation of RpoH and high stability of RpoH₁.

(56K): Fig. 3. In vivo stability of E. coli RpoH and B. japonicum RpoH1 proteins. (a) RpoH decay was determined by immunoblot analysis after stopping translation by chloramphenicol (Cm) addition. (b) Quantification of the stability of RpoH variants carrying single amino acid exchanges. (c) Quantification of the stability of RpoH/RpoH1 hybrids. All experiments were repeated three times with independent extracts and the corresponding mean values are shown.

Region 2.1 alone is not sufficient for FtsH-dependent proteolysis
All stabilized RpoH derivatives were still degraded by FtsH with slow kinetics, suggesting that other regions in RpoH also contribute to degradation. To test whether region 2.1 alone was sufficient for degradation by FtsH, residues 37–70 of RpoH, which include the C-terminal six amino acids of region 1.2 and the entire region 2.1, were introduced between two subdomains of the adenlyate cyclase (AC) of the bacterial one-hybrid system used previously (Obrist & Narberhaus, 2005). The system is based on two domains (T25 and T18) of the Bordetella pertussis AC, which produce cAMP if they are connected by a stable linker (Dautin et al., 2000) (Fig. 4a). cAMP production can be scored by LacZ activity in the AC-negative E. coli reporter strain DHM1. If the linker between T25 and T18 is cleaved by a cellular protease, cAMP production is lost.

Table 1.

We introduced RpoH region 2.1 together with the final six residues of region 1.2 to ensure that the putative α-helix 12a was formed properly. The constructs were transformed into the AC-deficient reporter strain E. coli DHM1 and cAMP production was assessed by β-galactosidase assays. AC-RpoH-WT (containing full-length wild-type RpoH) and AC-RpoH-NT (containing the N-terminal half of RpoH) were degraded as described previously (Dartigalongue et al., 2001; Obrist & Narberhaus, 2005). All region 2.1 fusions induced significantly higher β-galactosidase activities than the RpoH-WT fusion protein (set to 100 %) (Fig. 4b). Most constructs produced β-galactosidase activities as high as the positive control AC-p5, which is known to be stable in E. coli (Dautin et al., 2000). The high activities suggest that introduction of region 2.1 between the AC subdomains is not sufficient to produce an FtsH substrate.

To further substantiate this finding, amino acids 47–55 of RpoH region 2.1 were used to replace the equivalent region in RpoS. The helix 12a region deviates substantially between RpoH and RpoS (Fig. 1). The product (RpoS-2.1_H) retained susceptibility towards the ClpXP protease. Turnover was slightly delayed (Fig. 5a) but remained strictly ClpXP-dependent. Neither RpoS nor RpoS-2.1_H was degraded in a clpX mutant over a period of 80 min (Fig. 5b). Hence, introduction of region 2.1 from RpoH was not sufficient to commit the starvation sigma factor to degradation by FtsH. Despite being present at similar levels, RpoS-2.1_H was inactive in initiating transcription of the RpoS-dependent osmY promoter (Fig. 5c), suggesting that the contribution of region 2.1 to RNAP core binding is reduced in this mutated version of RpoS.

(18K): Fig. 5. In vivo degradation and sigma factor activity of E. coli σS with helix 12a from σ32, RpoS(2.1_H). (a) E. coli GB2 harbouring the plasmids pRpoS18 or pRpoS18(2.1_H) was grown to an OD578 of 0.6 and pulse-labelled for 1 min with L-[35S]methionine. Chase times (in min) are indicated. A polyclonal serum against σS was used for immunoprecipitation. Quantification: •, RpoS-WT; , RpoS(2.1_H). (b) GB28 (GB2 clpX : : cat) harbouring plasmid pRpoS18 or pRpoS18(2.1_H) was grown to an OD578 of 0.6. After addition of spectinomycin (1.5 mg ml–1) samples were taken at the indicated time points and prepared for SDS-PAGE. The gel was blotted and developed with a polyclonal serum against σS and an anti-rabbit alkaline phosphatase conjugate. Quantification: •, RpoS-WT; , RpoS(2.1_H). In vivo degradation experiments were performed at least three times with similar results. (c) Cells were grown in M9 minimal medium with 0.4 % glycerol as a carbon source at 37 °C. At the indicated time points arabinose was added to a final concentration of 0.02 % to induce pRpoS18 and pRpoS18(2.1_H) in E. coli strain GB2. •, , Growth curves of GB2 bearing pRpoS18 or pRpoS18(2.1_H), respectively; , , β-galactosidase activity derived from the σS-dependent osmY : : lacZ fusion of strain GB2 bearing pRpoS18 () or pRpoS18(2.1_H) ().

Chaperone binding to RpoH
The rapid turnover of RpoH in E. coli depends on molecular chaperones. DnaK, DnaJ, GrpE and GroESL have been shown to interact with RpoH (Gamer et al., 1996; Guisbert et al., 2004; Liberek et al., 1992). One possible explanation for the stabilization of RpoH mutated in region 2.1 might be an impaired interaction with the chaperones. To test this assumption, different rpoH alleles were transferred into a pET24b expression vector. Upon moderate expression, the His-tagged RpoH proteins were purified on Ni-affinity columns and the elution fractions were analysed (Fig. 6). In each case, DnaK and DnaJ co-eluted with RpoH regardless of whether wild-type or mutated RpoH was used (shown as immunodetected DnaK and DnaJ in purified fractions of RpoH-WT and the hyperstable variants RpoH-L47Q and RpoH-I54F in Fig. 6). The amounts of DnaK were also visible in Coomassie-stained SDS-PAGE gels (data not shown). Quantification of these bands from three independent purification experiments for each RpoH protein demonstrated that the relative amount of DnaK/RpoH in the elution fractions ranged from 2 to 6 %, with 4 % for RpoH-WT (data not shown). All other RpoH derivatives, RpoH₁ as well as the hybrid proteins between E. coli RpoH and B. japonicum RpoH₁, showed similar results (data not shown). In control experiments with extracts from cells without histidine-tagged RpoH proteins, the chaperones did not bind to the column resin (data not shown). The interaction between RpoH and DnaK was also examined by glutaraldehyde cross-linking using Strep-tagged DnaK and His-tagged RpoH. RpoH-L47Q, L47C, L47P, A50V and I54F were able to interact with DnaK like RpoH-WT (data not shown).

(90K): Fig. 6. Co-elution of chaperones with RpoH derivatives. RpoH proteins were purified by Ni-NTA chromatography. In each top panel, RpoH (H) is indicated in Coomassie-stained SDS-PAGE gels. The corresponding lower panels show immunodetected DnaK (anti-K) and DnaJ (anti-J) bands. Imidazole concentrations were as follows: W3 and W4, 50 mM; E1, 75 mM; E2, 100 mM; E3, 125 mM; E4, 150 mM; E5, 1M.

Since Guisbert et al. (2004) recently reported an interaction between RpoH and GroESL, we subjected our RpoH elution fractions to immunodetection with anti-GroEL sera. GroEL indeed co-eluted with RpoH but did not discriminate between wild-type RpoH and the mutated RpoH proteins (data not shown). Thus, it appears that region 2.1 of RpoH is not critical for an interaction with molecular chaperones.

In vitro degradation of RpoH by FtsH
To analyse whether stabilization of the RpoH variants was due to an altered interaction with the protease FtsH, we performed in vitro degradation experiments using purified histidine-tagged RpoH proteins and GST-FtsH. Degradation of all proteins was ATP-dependent (data not shown). β-Casein, which is known to be a loosely folded protein, was readily degraded (Fig. 7) (Holt & Sawyer, 1988; Shotland et al., 2000a). As observed previously, the kinetics of RpoH degradation in vitro was much slower than in vivo (Urech et al., 2000). RpoH-WT was degraded with a half-life of approximately 20 min, whereas RpoH-L47Q, -A50V and -I54T were clearly stabilized (Fig. 7a, b). Consistent with previous in vivo results (Obrist & Narberhaus, 2005), RpoH-I54F was the most stable RpoH protein.

(48K): Fig. 7. FtsH-dependent in vitro degradation of RpoH proteins carrying single point mutations. (a) Coomassie-stained RpoH bands after SDS-PAGE. (b) Quantification of RpoH degradation. Mean values of three independent experiments are shown.

Degradation of the RpoH/RpoH₁ hybrids in the purified system was also in line with their stability in vivo (Fig. 8). The poor FtsH substrate RpoH₁ was destabilized by the presence of helix 12a from E. coli RpoH. On the other hand, RpoH was moderately stabilized by the equivalent RpoH₁ region. The sum of our in vivo and in vitro results suggests that proteolysis of RpoH might involve a direct interaction between RpoH and FtsH in region 2.1.

(37K): Fig. 8. FtsH-dependent in vitro degradation of RpoH/RpoH1 hybrids with exchanged α-helix 12a of region 2.1. (a) Coomassie-stained RpoH bands after SDS-PAGE. (b) Quantification of RpoH/RpoH1 degradation. Means of three independent experiments are shown.

Initiation of transcription in bacteria is a multistep process involving binding of the sigma factor to other subunits of RNAP core enzyme and to the –10 and –35 regions of the promoter. Accordingly, sigma factors of the σ⁷⁰ family (to which RpoH and RpoS belong) are characterized by several conserved regions. These functional domains are separated by flexible linkers (Gruber & Gross, 2003). The remarkable structural flexibility of free RpoH was demonstrated by amide hydrogen exchange studies (Rist et al., 2003). Only in contact with interaction partners do sigma factors adopt a well-defined conformation amenable to structure determination (Campbell et al., 2003; Murakami et al., 2002; Vassylyev et al., 2002).

Apart from binding to components of the transcription machinery, sigma factors interact with many other cellular proteins, which mediate post-translational control by influencing the activity and stability of the bacterial transcription initiation factors. Altogether at least seven cellular macromolecules (promoter DNA, the β and β' subunits of the RNAP core enzyme, DnaK, DnaJ, GroEL and FtsH) compete for binding to the heat-shock sigma factor RpoH. Some proteins might share overlapping binding sites because interaction of some partners is mutually exclusive. Incubation of RpoH with DnaK or GroEL recruits the sigma factor away from the RNAP and inhibits transcription initiation in vitro (Gamer et al., 1996; Guisbert et al., 2004; Liberek et al., 1992). Conversely, in complex with RNAP the sigma factor is protected from attack by the FtsH protease in vitro (Blaszczak et al., 1999; Urech et al., 2000).

Region 2.1 of E. coli RpoH carries a critical turnover element promoting rapid degradation by FtsH in vivo (Horikoshi et al., 2004; Obrist & Narberhaus, 2005). In this study, we provide further evidence for the importance of this region. Rationally designed point mutations stabilized the sigma factor and resulted in elevated RpoH-dependent gene expression. Swapping of the turnover element between RpoH and the intrinsically stable RpoH₁ protein from B. japonicum produced proteins with intermediate stability. Differences in region 2.1, which contains highly conserved and some variable residues (Fig. 1), and yet undefined residues outside of region 2.1 seem to account for differences in stability of RpoH proteins from different bacteria (Narberhaus et al., 1997; Urech et al., 2000).

It is important to note that the mutated RpoH proteins were not entirely protected from proteolysis. Some residual degradation occurred both in vivo and in vitro, indicating that RpoH regions outside region 2.1 also contribute to efficient turnover. The fact that transfer of RpoH region 2.1 into the AC of Bordetella pertussis or the E. coli RpoS sigma factor did not suffice to convert these proteins into FtsH substrates supports this assumption.

As the key residues for protein turnover in RpoH region 2.1 are predicted to be exposed on one face of a α-helix, they might provide an interaction surface for proteins of the degradation machinery (Obrist & Narberhaus, 2005). There is a striking analogy in the turnover element of RpoS that is located in region 2.5 and interacts with RssB. Only amino acid exchanges on one side of an α-helix, in particular at position K173, have an effect on RpoS proteolysis (Becker et al., 1999).

Little is known about the precise contact sites between DnaK, DnaJ, GroEL or FtsH with RpoH. Multiple DnaK interactions sites have been proposed by a peptide-scanning approach using cellulose-bound peptides (McCarty et al., 1996). Region 2.1 of RpoH was among those potential DnaK sites, which led us to examine chaperone binding to our stabilized RpoH proteins. Binding of neither DnaK, DnaJ or GroEL was abolished by point mutations in amino acids 47, 50 or 54, suggesting that region 2.1 is not a major specific chaperone-binding site. Apparently, several of the membrane-displayed short peptides (13-mers) that served as good DnaK-binding sites in the peptide library do not bind the chaperone in the context of the native RpoH protein. This was also shown for region C (Arsène et al., 1999). A remaining promising candidate as DnaK-binding site in RpoH lies around amino acid 200 in region 3.2 (McCarty et al., 1996).

Since helix 12a did not seem to be a chaperone-binding site, the FtsH protease itself was left as potential interaction partner. The direct effect of mutations in region 2.1 on proteolysis was examined by using purified sigma factors and FtsH. Despite its low unfoldase activity, FtsH is able to degrade RpoH without the aid of chaperones (Herman et al., 2003). As described previously, degradation in vitro was slower than in vivo (Urech et al., 2000). Interestingly, the stability determined in vitro largely reflected the in vivo half-lives, i.e. proteins that were stabilized by mutations in region 2.1 were also stabilized in vitro. The RpoH₁ protein carrying the destabilizing E. coli RpoH helix 12a was rapidly degraded in vitro. These results strongly suggest that chaperones do not play a role in stability control by region 2.1. Based on our results we propose that residues 47, 50 and 54 of region 2.1 are important for establishing a contact to the FtsH protease.

Again, it should be noted that some residual degradation was observed with all constructs. Hence, it appears that region 2.1 is not the only RpoH region required for in vitro degradation by the FtsH protease. It will be interesting to track down additional amino acids involved in this process. It is remarkable that neither genetic screen came up with good candidates outside region 2.1 (Horikoshi et al., 2004; Obrist & Narberhaus, 2005). Since the N-terminal half of RpoH mediates efficient turnover in the one-hybrid screen (Dartigalongue et al., 2001; Obrist & Narberhaus, 2005), critical residues might be located in the first 156 residues of RpoH.

F. N. and M. O. thank Hauke Hennecke for long-standing support and Blanka Kutscher for technical assistance. Daniel Ladant generously provided strains and plasmids for the one-hybrid system. This study was supported by grants from the Swiss National Foundation for Scientific Research (grant 3100A0-100713/1), the Deutsche Forschungsgemeinschaft (DFG priority program SPP 1132, grants NA240/3-2 and He1556/10-3) and the Fonds der Chemischen Industrie.

Edited by: M. Paget