AnkB, a Periplasmic Ankyrin-Like Protein inPseudomonas aeruginosa, Is Required for Optimal Catalase B (KatB) Activity and Resistance to Hydrogen Peroxide

ABSTRACT In this study, we have cloned the ankB gene, encoding an ankyrin-like protein in Pseudomonas aeruginosa. TheankB gene is composed of 549 bp encoding a protein of 183 amino acids that possesses four 33-amino-acid ankyrin repeats that are a hallmark of erythrocyte and brain ankyrins. The location ofankB is 57 bp downstream of katB, encoding a hydrogen peroxide-inducible catalase, KatB. Monomeric AnkB is a 19.4-kDa protein with a pI of 5.5 that possesses 22 primarily hydrophobic amino acids at residues 3 to 25, predicting an inner-membrane-spanning motif with the N terminus in the cytoplasm and the C terminus in the periplasm. Such an orientation in the cytoplasmic membrane and, ultimately, periplasmic space was confirmed using AnkB-BlaM and AnkB-PhoA protein fusions. Circular dichroism analysis of recombinant AnkB minus its signal peptide revealed a secondary structure that is ∼65% α-helical. RNase protection and KatB- and AnkB-LacZ translational fusion analyses indicated that katBand ankB are part of a small operon whose transcription is induced dramatically by H2O2, and controlled by the global transactivator OxyR. Interestingly, unlike the spherical nature of ankyrin-deficient erythrocytes, the cellular morphology of anankB mutant was identical to that of wild-type bacteria, yet the mutant produced more membrane vesicles. The mutant also exhibited a fourfold reduction in KatB activity and increased sensitivity to H2O2, phenotypes that could be complemented in trans by a plasmid constitutively expressing ankB. Our results suggest that AnkB may form an antioxidant scaffolding with KatB in the periplasm at the cytoplasmic membrane, thus providing a protective lattice work for optimal H2O2 detoxification.

Ankyrins are structural proteins in human erythrocytes and brain that bridge the spectrin exoskeleton to the cytoplasmic surface of the plasma membrane (5). They are composed of three domains: (i) an N-terminal membrane-binding domain, (ii) a spectrin-binding domain, and (iii) a C-terminal domain with an apparent regulatory function (5). Integral membrane proteins that associate with ankyrin both in vivo and in vitro include the band III anion exchanger (15), Na ϩ /K ϩ -ATPase (15), and multiple sodium channels (51,52). The protein-binding N-terminal domain harbors a series of 33-amino-acid tandem repeats (herein termed ank repeats) that extend over 740 residues. The tandem repeat motif is present in 22 contiguous copies with 30 to 35% identity between the repeats (reference 5 and references therein). Closely related repeats (30 to 35% identity to brain ankyrin) were found in seemingly dissimilar proteins of lower and higher eukaryotes that regulate the cell cycle in yeast (e.g., products of cdc10 and SWI6) and are involved in intercellular signaling during development and cell differentiation of Caenorhabditis elegans (products of lin-12, glp-1, and fem-1), Drosophila (Notch), or Xenopus (Xotch) (references 2,9,19, and 50 and references therein). Subsequently identified ankyrin-like proteins (ALPs) include transcription factors (e.g., GABP-␤ and NF-B), toxins (e.g., black widow spider venom), enzymes (e.g., rat liver-specific glutaminase), and a viral host range factor (Vaccinia hr gene product) (listed in references 5 and 50); a protein-tyrosine kinase in Hydra vulgaris (14); and the Chlorella virus long terminal repeat gene product (GenBank accession no. D14469). Two ALPs were also identified in the higher plant Arabidopsis thaliana (GenBank accession no. M82883), one of which was implicated in membrane transport (GenBank accession no. X62907). So far, more than 150 genes possessing ank repeats have been reported in eukaryotic systems (GenBank search, May 2000). Due to the success in whole genome sequencing, however, genes encoding ankyrin homologs found most recently reside in bacteria.
The first bacterial ALP-encoding gene (phlB), from Serratia liquefaciens, was not recognized as such (21) until Bennett (5) identified an ank repeat consensus sequence (-G-TA/PLM/H-AA--GH---V/A--LL--GAD-N/D--D-). According to various databases, bacterial ALPs have been identified in several actinobacteria (Streptomyces verticillus [13], Streptomyces argillaceus [ [32] [AF133262 and AF133263]). Interestingly, unlike eukaryotic ankyrin or ALPs, bacterial ALPs seem to belong to divergent operons: bleomycin and mithramycin antibiotic resistance in S. verticillus (13) and S. argillaceus (U43537), respectively; periplasmic flavocytochrome c and cytoplasmic tetraheme cytochrome c in C. vinosum (17); and a catalase with proposed periplasmic and cytoplasmic locations in P. syringae (32) and P. fluorescens (U83328). The ankyrin gene in V. cholerae (http:// www.tigr.org) is also downstream of a gene encoding a type I bacterial catalase. A putative open reading frame (ORF) upstream of a gene encoding a histidinol phosphate aminotransferase, an enzyme required for ethanol tolerance, was found in Acetobacter pasteurianus (DDBJ accession no. D14440) (54). The ALP of S. liquefaciens, whose gene is located downstream of the gene encoding periplasmic phospholipase A1, has a putative regulatory function regarding phospholipase activity (21). Taken together, the bacterial ALP genes are located in close proximity to genes encoding proteins involved in either (i) nutrient acquisition and uptake or (ii) tolerance or resistance to antibiotics, starvation, or oxidative stress.
In this study, we demonstrate the first functional characterization of a bacterial ALP, AnkB, in P. aeruginosa. AnkB was found to be a cytoplasmic membrane-periplasmic protein whose expression is increased upon exposure of bacteria to H 2 O 2 . AnkB was also found to be essential for optimal resistance to H 2 O 2 , which we believe is in part due to its ability to bind to and stabilize KatB, a type I bacterial catalase.

MATERIALS AND METHODS
Bacterial strains, plasmids, and media. All P. aeruginosa and Escherichia coli strains used in this study are listed in Table 1 and were maintained on Luria (L) agar (10 g of tryptone, 5 g of yeast extract, and 10 g of NaCl) or M9 minimal medium (6 g of Na 2 HPO 4 , 3 g of KH 2 PO 4 , 1 g of NH 4 Cl, 0.5 g of NaCl, 1 mM MgSO 4 ⅐ 7H 2 O, and 0.2% glucose [per liter]) plates, with each medium solidified with 15 g of Bacto agar per liter. All strains were stored indefinitely at Ϫ80°C in a 1:1 suspension of overnight-grown culture and either 25% glycerol or 10% skim milk.
Growth conditions. All bacteria were grown from single-colony isolates or overnight cultures in L broth or M9 minimal medium. Liquid cultures were grown at 37°C with shaking at 300 rpm or on a roller wheel at 70 rpm unless otherwise indicated. Culture volumes were 1/10 of the total Erlenmeyer flask volume to ensure proper aeration.
Cloning and sequence analysis of ankB. Steps involved in the cloning of the P. aeruginosa PAO1 ankB and radA genes are described in Results. DNA sequencing was performed on both strands using the PRISM Dye Deoxy Terminator Cycle Sequencing Kit and analyzed on an ABI model 373A DNA sequencer. Oligonucleotides for DNA sequencing reactions and PCR analysis were synthesized in the DNA Core Facilities in the Department of Molecular Genetics, Biochemistry and Microbiology at the University of Cincinnati College of Medicine or in the Department of Microbiology at the University of Colorado Health Sciences Center. Sequence analyses were performed using either MacVector 6.5.1 (Eastman Chemical Co., New Haven, Conn.), Sequencher 3.1 (Gene Codes Corp., Ann Arbor, Mich.), or Gene Runner (Hastings Software, Inc.) software. Amino acid alignments were performed using either the BLASTP program provided by the National Center for Biotechnology Information (1) or the Align Plus 3 Global Alignment Program (Sci-Ed Software, Durham, N.C.). Potential membrane-spanning domains (MSDs) of AnkB and other bacterial ankyrin-like proteins were determined using TOP-PRED II (57) or the SOSUI program (http://www.tuat.ac.jp/ϳmitaku/adv_sosui/submit.html).
Manipulation of recombinant DNA and RNA. Plasmid DNA was transformed into E. coli DH5␣-MCR (Gibco-BRL, Gaithersburg, Md.) or SM10 (48). 5-Bromo-4-chloro-3-indolyl-␤-D-galactopyranoside (X-Gal) (40 g/ml) was routinely added to agar medium to detect the presence of insert DNA. Restriction endonuclease, alkaline phosphatase (AP), Klenow fragment, T4 DNA polymerase, and T4 DNA ligase were used as specified by the vendor (Gibco-BRL). Plasmid DNA was isolated using miniprep kits from Qiagen. DNA fragments used for cloning or in the construction of radiolabeled probes were recovered from agarose gels using SeaPlaque low-melting-point agarose (FMC BioProducts, Rockland, Maine) or with the GeneClean II kit (BIO 101, Inc., La Jolla, Calif.). RNA was isolated by the hot-phenol method and analyzed by RNase protection assays as described in detail elsewhere (3). Radiolabeled riboprobes were generated from cloned DNA fragments (Table 1) using an in vitro runoff transcription system (Promega), and excess probe was hybridized to 20 g of total RNA.
Overexpression of recombinant AnkB in E. coli. Overexpression of recombinant AnkB as a His 6 -tagged protein in E. coli was performed using the T7 promoter-T7 RNA polymerase system (53). A 0.53-kb fragment containing the ankB gene minus the first 19 codons comprising its signal sequence was amplified using primers (XhoI)-ctCGAGGTGCATGGGGTCGAGGT and (BamHI)-gga TCCAGACTAGCCCAGCAGGC (bases in the XhoI and BamHI restriction sites are underlined, and nonmatching bases near the 5Ј end are in lowercase type). This PCR product was cloned into pCRII (Invitrogen), sequenced, and directionally cloned as an XhoI-BamHI fragment into pET14b (Novagen). The resulting plasmid, pET14b-ankB, allowed T7-inducible expression of AnkB containing an in-frame His 6 tag at its N terminus spaced by a thrombin cleavage site encoded on pET14b. Similarly, a 0.485-kb fragment containing the ankB gene minus the first 17 codons and lacking the last two codons was amplified using the primers (NheI)-GCTaGCGAGGTGCATGGGGTCGA and (NotI)-GCgGCCG CAGTGCCGTTCAGTTC and subsequently cloned as an NheI-NotI fragment into pET23a linearized with NheI and NotI. In the resulting plasmid, pET23-ankB, the AnkB protein was fused in frame to a carboxy-terminal His 6 tag encoded by pET23a. Recombinant plasmids were first selected in E. coli DH5␣-MCR and then transformed into E. coli BL21(DE3), which harbors a single genomic copy of the T7 RNA polymerase gene under control of the lacUV5 promoter. These bacteria were grown in 1 liter of L broth containing ampicillin at 100 mg/ml to an optical density at 600 nm (OD 600 ) of 0.3. At this point, the synthesis of T7 polymerase was induced by the addition of 0.5 mM isopropyl-␤-D-thiogalactopyranoside (IPTG), and the cells were allowed to grow for an additional 3 h at 37°C. Recombinant AnkB proteins were then purified using a nickel-nitrilotriacetic acid column as specified by the manufacturer (Qiagen).
Construction of ankB::blaM, ankB::phoA, and ankB::lacZ fusion plasmids. For construction of a translational ankB::blaM fusion plasmid, the N-terminal region of ankB was PCR amplified to create a 551-bp fragment harboring the entire ankB gene. This fragment was cut with NaeI, and the 162-bp 5Ј fragment containing the predicted MSD and a portion of the predicted periplasmic region of AnkB was ligated into the SmaI site of pKM1 (10). For construction of a translational ankB::phoA fusion plasmid, a 174-bp 5Ј fragment of the ankB gene containing an AflIII restriction site was first cloned into a T/A version of pKS(Ϫ) (35). The ankB fragment was removed from this plasmid and cloned into the NcoI-SmaI sites of pEX30 (H. P. Schweizer). pPHO7, a broad-host-range phoAcontaining plasmid encoding AP (24), was first cut with BamHI, blunted, and then excised with PstI. The resulting phoA gene was then ligated to pEX-ankB that had been previously digested with EcoRI (blunted) and PstI, allowing for directional cloning of ankB in frame with phoA. The resulting construct fuses the first 58 amino acids of AnkB with PhoA. Translational fusions of ankB to the lacZ reporter gene were constructed as follows. PCR products were generated with primer (BamHI)-ggatCCTCGCATCGTCGTCATCTC containing a BamHI site after the third codon of the ankB gene and either primer TACAA GGCTGACAGCGACT (0.16 kb upstream), primer GCAGCGAGGTCAATA CCGTC (0.47 kb upstream), or primer CTTGGAACTGCGCCATGCAG (1.9 kb upstream). The PCR products were cloned into pCRII and directionally ligated as EcoRI-BamHI fragments into pPZ30 linearized with EcoRI and BamHI, yielding pPZ-ankB-160, pPZ-ankB-470, and pPZ-ankB-1900, respectively.
Construction of isogenic P. aeruginosa ankB and katB ankB mutants. To generate a clean ⌬ankB deletion mutant, the DNA sequences flanking ankB were PCR amplified using primers GCAGCGAGGTCAATACCGTC and (KpnI)-ggtaCCGGTTGCGATCAATCCTGG (0.408-kb ankB upstream fragment) and primers (XbaI)-tctaGATGCGTTGGGCAACAGCGT and (HindIII)-aaGCTTCAACCCCTGCAGCGCCA (0.334-kb ankB downstream fragment). The two PCR products were cloned into pCRII, sequenced, and cloned as an EcoRI-KpnI ankB upstream fragment and an XbaI-HindIII ankB downstream fragment on each side of and in proper orientation to a 1.7-kb aac1 (Gm r ) cartridge in pUC-Gm, yielding pUC⌬ankB::Gm. A 2.8-kb PvuII fragment of pUC⌬ankB::Gm containing the ⌬ankB::Gm construct was cloned into the SmaI site of the suicide vector pEX100T (46), and the resulting pEX100T-⌬ankB::Gm was transformed into E. coli SM10, which then served as the donor strain in a biparental plate mating (16 h, 37°C) with P. aeruginosa PAO1 or with PAO1 ⌬katA::Tc (35). The mating mixture was plated on brain heart infusion agar containing gentamicin (75 g ml Ϫ1 ) and irgasan (Ciba-Geigy) (50 g ml Ϫ1 ) as a counterselective agent. Several colonies were grown to late logarithmic phase in L broth, and serial dilutions were spread onto L agar containing gentamicin (75 g ml Ϫ1 ) and sucrose (5%). Chromosomal DNA from individual colonies was evaluated for deletion of the ankB gene by PCR and Southern blot analysis (data not shown). A PAO1 ⌬katB ankB deletion mutant was obtained as follows. A 2.7-kb fragment containing the katB-ankB region was PCR amplified using the primers CTTGGAACTGCGCCATGCAG and GCTTCAACCCCTGCAGCG CCA and cloned into pCRII. A 1.75-kb SstII fragment containing most of the katB and ankB genes was excised and replaced by a 1.2-kb Gm r cartridge by blunt-end ligation after filling in the ends with Klenow enzyme. A 2.6-kb PvuII fragment of the resulting plasmid, pCRII-⌬katB ankB::Gm, was cloned into the SmaI site of pEX100T, yielding the donor plasmid pEX100T-⌬katB ankB::Gm for a mating as described above.
Cell fractionation: periplasm, cytoplasm, and cytoplasmic membrane. Bacteria were grown aerobically in 1 liter of L broth at 37°C until the OD 600 reached 0.6. At this point, organisms were treated with 350 M paraquat (Sigma) for 1 h to stimulate katB-ankB transcription prior to harvesting the bacteria by centrif-ugation at 10,000 ϫ g for 10 min. The pellet was washed twice in ice-cold 10 mM Tris-HCl-30 mM MgCl 2 (pH 7.3) (Tris-Mg) and resuspended in 1/25 the volume of the same buffer. Chloroform (15 l/ml of buffer) was then added, and the cells were incubated on ice for an additional 15 min followed by dilution with an additional 1 ml of buffer. The bacteria were pelleted by centrifugation at 10,000 ϫ g for 10 min at 4°C, and the supernatant was further subjected to centrifugation at 150,000 ϫ g for 2 h at 4°C to remove potential contaminating membranes. Finally, the periplasmic preparation was stored on ice. The bacteria were washed again, resuspended in 5 ml of Tris-Mg, and disrupted by sonication with a Heat Systems-Ultrasonics (Farmingdale, N.Y.) model W-225 sonicator equipped with microtip at output setting 5 at 4°C. Cell debris and membrane fractions were clarified by centrifugation at 35,000 ϫ g for 1 h at 4°C. The supernatant was designated the cytoplasmic fraction.
For preparation of cytoplasmic membrane proteins, the above-described growth conditions were employed. The pellet of paraquat-stimulated bacteria was resuspended in 10 ml of 10 mM Tris-HCl (pH 7.5) (T buffer), containing 20% sucrose, treated with 0.5 mg of DNase I (Gibco) and RNase (Sigma) per ml, and incubated for 15 min at 22°C with periodic agitation. The suspension was a Abbreviations used for genetic markers were as described by Holloway et al. (29). mob, mobilization site (ColE1); Tra ϩ , conjugative phenotype; oriT, origin of transfer (RK2); Ap r , ampicillin resistance; Cm r , chloramphenicol resistance; Gm r , gentamicin resistance; Km r , kanamycin resistance. placed on ice for 20 min, followed by two passages through a French pressure cell at 1,200 lb/in 2 at 4°C. The cell debris was removed by centrifugation at 5,500 ϫ g for 10 min at 4°C. The supernatant containing the membranes was subjected to a two-stage sucrose gradient centrifugation. The first stage involved layering 10 ml of membranes on 14 ml of 50% sucrose and 14 ml of 70% sucrose in T buffer. The membranes were then separated by centrifugation in an SW28 swingingbucket rotor at 130,000 ϫ g for 17 h at 4°C. Cytoplasmic membranes (the top red band) and outer membranes (the bottom white band) were collected by dropwise collection and diluted to 7 ml in cold T buffer-20% sucrose. The second centrifugation stage involved layering the membrane fractions on 9 ml of 52% sucrose, 9 ml of 58% sucrose, 9 ml of 64% sucrose, and 3 ml of 70% sucrose, each prepared in T buffer. The membranes were again separated by centrifugation in an SW28 swinging bucket rotor at 130,000 ϫ g for 17 h at 4°C. The purified cytoplasmic and outer membrane fractions were concentrated by dropwise collection and dialyzed exhaustively against T buffer for 17 h at 4°C. The purity of each cellular fraction was gauged by measuring enzymatic activities specific for the cytoplasm, periplasm, and inner membrane. Stripping of AP activity from cytoplasmic membrane preparations was accomplished by incubating the membranes in either T buffer (control) or 0.1 M Na 2 CO 3 for 30 min on ice. Thirty milliliters of the membrane suspension was carefully layered on 4 ml of 0.3 M sucrose in T buffer and centrifuged at 150,000 ϫ g for 2 h at 4°C. The pellet was solubilized in 2% Triton X-100 in T buffer and assayed for AP activity as described below. Glucose-6-phosphate dehydrogenase, a cytoplasmic marker, was assayed by monitoring the reduction of NADP ϩ at 340 nm as previously described (33). ␤-Lactamase, one periplasmic marker, was assayed by the decomposition of 50 M nitrocefin (CalBiochem, La Jolla, Calif.) at 486 nm (34). AP, another periplasmic marker, was assayed by monitoring the production of p-nitrophenol through hydrolysis of 1 mM p-nitrophenyl phosphate (Sigma) (20). AP activity in bacterial colonies on L-agar plates containing 10 mM potassium phosphate (pH 7.0) to inhibit endogenous AP activity was monitored using 5-bromo-4-chloro-3-indolyl phosphate (BCIP) (40 g/ml) in the medium. Succinate dehydrogenase, a cytoplasmic membrane marker, was assayed by monitoring the reduction of dichlorophenol-indophenol as previously described (12).
TEM. Cells were harvested from exponential-growth-phase cultures, treated with 1 mM H 2 O 2 to induce the katB-ankB operon, and negatively stained for transmission electron microscopy (TEM). For staining, the cells were put on a carbon-and Formvar-coated 200-mesh TEM grid, stained for 15 s with 2% (wt/vol) uranyl acetate, and blotted dry. The negatively stained bacteria were imaged using a Philips EM300 instrument operating at 60 kV under standard conditions with the liquid nitrogen anticontaminator in place.
Circular dichroism spectropolarimetry. Circular dichroism spectra of purified recombinant AnkB proteins (AnkB-23a and AnkB-14b) were obtained using a JASCO J-710 spectropolarimeter calibrated with d 10 -camphorsulfonic acid. All samples were dissolved in 10 mM sodium phosphate (pH 7.0) and equilibrated at 22°C for 10 min prior to data collection. The spectra represent the means of four independent scans obtained at a scanning rate of 20 nm min Ϫ1 . Samples were added to a water-jacketed, cylindrical quartz cuvette with a 0.05-cm path length. Protein content was calculated based upon the equation [protein] g/ml ϭ 144(A 215 Ϫ A 225 ), using a 1-cm light path. Percent ␣-helix was estimated from [] 222 as described by Greenfield and Fasman (23).
Cell extract preparation and biochemical assays. Routine cell extracts were prepared from bacteria harvested by centrifugation at 10,000 ϫ g for 10 min at 4°C. Bacteria were washed twice in ice-cold 50 mM potassium phosphate buffer (pH 7.0) and sonicated in an ice-water bath for 10 s as described above. The sonicate was then clarified by centrifugation at 13,000 ϫ g for 10 min at 4°C. Cell extract preparation for native gel electrophoresis was performed as described above except that 50 mM T buffer was used as the diluent. Catalase activity was assayed spectrophotometrically at 240 nm by monitoring the decomposition of 18 mM H 2 O 2 using a Spectronic Genesys 5 spectrophotometer (Spectronic Unicam, Rochester, N.Y.) (4). The concentration of H 2 O 2 was determined using an extinction coefficient of 43.6 mM. One unit of activity is that which degrades 1 mol of H 2 O 2 per min at 23°C. Nondenaturing polyacrylamide gels (5%) were stained for catalase activity according to the method of Wayne and Diaz (58). ␤-Galactosidase assays were performed using o-nitrophenyl-␤-D-galactopyranoside, and the results were expressed as international units with an extinction coefficient for o-nitrophenyl-␤-D-galactopyranoside of 3.1 mM (37). Protein concentrations in cell extracts were estimated by the method of Bradford (8), using bovine serum albumin as a standard.
Nucleotide sequence accession number. The P. aeruginosa katB, ankB, and radA sequences have been assigned GenBank accession number U89384.

RESULTS
Sequence analysis downstream of the P. aeruginosa katB gene: identification of ankB, encoding an ALP, and radA, a DNA repair protein. In a previous study, we cloned and characterized the katB gene, encoding a 228-kDa tetrameric catalase (11). The katB gene was recently found to be under the control of the global transactivator OxyR (39), and its transcription is markedly induced upon exposure to H 2 O 2 or the redox-cycling agent paraquat (11). DNA sequence analysis downstream of the katB locus revealed a small, 549-bp ORF (Fig. 1). This ORF, ankB, is predicted to encode a protein of 183 amino acids with a monomeric molecular mass of 19,360 Da and a pI of 5.55. The deduced amino acid sequence demonstrated the highest similarity with genes harboring ank repeats in other bacteria (13,17,21,32) and various ankyrin or ALP genes in eukaryotes (5) (see below). When we sequenced further downstream of the ankB locus, we discovered a large inverted repeat spanning 33 bp (Fig. 1). We then identified a gene downstream of the inverted repeat that was 70% identical to the radA gene of E. coli (49). The radA gene in E. coli encodes a protein that repairs DNA damaged (alkylated) by gamma irradiation. However, because virtually nothing is known of the function of ALPs in bacteria, we chose to focus our efforts on the functional characterization of P. aeruginosa AnkB.
Amino acid comparison of AnkB with other bacterial ALPs. Ankyrins are proteins that are characterized by 33-amino-acid ank repeats that are thought to represent an ancient motif that has evolved to allow for functional diversity without compromising specificity (5). Each of the five bacterial ALPs in Fig. 2 possess the 33-amino-acid tandem, nonidentical ank repeats. These repeats are based upon the consensus ank repeat motif put forth by Michaely and Bennett (36). AnkB (PaAnkB), not surprisingly, is similar to AnkF (PsAnkF, in the related plant pathogenic species P. syringae, strain 61 [32]; PsAnkB [in P. syringae B301D; accession no. AF133262]; and PfAnkB [in P. fluorescens; U83328]) in that it possesses four tandem repeats and an 18-residue signal sequence (Fig. 2). However, the Ank signal peptides from P. syringae differ from those of PaAnkB and PfAnkB in that they are followed by a proteolytic prolinethreonine-box (hinge [32]). Interestingly, PaAnkB and PfAnkB, but not PsAnkF and PsAnkB, possess an RGD motif, which, in eukaryotes, is responsible for physical interaction with integrins, structural proteins that play a role in the homing and action of immune cells. Other bacterial ALPs (some of which are not listed in Fig. 2 verticillus; L26954). The 33-amino-acid tandem repeats (underlined) were revealed by using the Ank motif of conserved residues (boldface) as identified by Bennett using the erythrocyte ank repeat consensus sequence (5). Proposed signal sequences are indicated by a double underline. The conserved ank repeat sequences for erythrocyte ankyrin are given below the selected bacterial ALP sequences. RBC, erythrocyte concensus ank repeat sequence. N-terminal MSD of 12 residues. The ALPs of S. verticillus (13) (SvAnkA and SvAnkB) possessed only two 33-amino-acid ank repeats. All other identified bacterial ALPs contain a putative cleavage site following an MSD either in the middle or close to the C terminus, or they do not contain a hydrophobic region long enough to incorporate into the membrane.
AnkB is a periplasmic protein: AnkB-␤-lactamase and AnkB-alkaline phosphatase protein fusion analysis. Using the membrane topology program TopPred 3.0, it was predicted that the cytoplasmic N terminus of P. aeruginosa AnkB (3 amino acids) is followed by a 20-amino-acid stretch that is predicted to be the hydrophobic inner-MSD (for a von Heijne schematic, see Fig. 3A) (TMRGWILAGLL-LAALAAQAEVH), with the remaining portion of the protein (being highly hydrophilic) predicted to reside in the periplasm. The ALPs of other selected gram-negative bacteria, including C. vinosum, S. liquefaciens, and other Pseudomonas species, are also predicted to span the cytoplasmic membrane (Fig. 3A). To test whether the predicted cytoplasmic membrane location was correct, we constructed AnkB-BlaM and AnkB-PhoA protein fusions linking the C terminus of AnkB with both reporters (Fig. 3B). E. coli and P. aeruginosa harboring an ankB::blaM fusion plasmid were resistant to ampicillin and carbenicillin, respectively. In addition, E. coli and P. aeruginosa harboring pEX30-ankB:: phoA were blue on indicator plates containing BCIP, suggesting that the C terminus of AnkB resides in the periplasm and confirming the AnkB-␤-lactamase fusion plasmid results. To determine the precise location of AnkB within the cell, membrane fractionation techniques (see Materials and Methods) were employed. Cytoplasm, periplasm, cytoplasmic membrane, and outer membrane fractions were assayed for AP activity (Fig. 3C). The vast majority of AP activity was detected in the periplasmic space and the cytoplasmic membrane. The contaminating AP activity in the cytoplasm is likely due to overexpression and problems associated with transport of a larger-than-normal AnkB-PhoA protein. To show that AnkB traverses the cytoplasmic membrane and is not simply nonspecifically bound to it, cytoplasmic membranes were stripped by treatment with Na 2 CO 3 at high pH (56). Our results showed that AnkB is tightly bound to the cytoplasmic membrane (data not shown). These results suggest that AnkB first exists as a cytoplasmic membrane protein, followed by proteolytic cleavage via a putative LepB signal peptidase cleavage site (GEVHG) to finally reside in the periplasmic space (42).
Overexpression of AnkB in E. coli: AnkB secondary structure is predominantly ␣-helical. To obtain some preliminary structural analysis of AnkB, we overexpressed and purified two recombinant AnkB proteins with N-terminal (pET23a-AnkB) and C-terminal (pET14b) His 6 -tagged fusions in E. coli BL21(DE3) without their predicted MSDs. Figure 4A demonstrates purified pET23a-AnkB. Gorina and Pavletich revealed that the secondary structure of an ank repeat in protein 53BP2, which binds to the p53 tumor suppressor, consists of an L-shaped structure with a ␤-turn and 2 ␣-helices (22). Circular dichroism spectropolarimetric analysis of recombinant AnkB-23a suggested that AnkB is ϳ60 to 70% ␣-helical (Fig. 4B). This structure is consistent with the 66% ␣-helical nature of the ank repeats of the 53BP2 protein (22). To determine if katB and ankB are part of a small operon, RNase protection assays were performed. Figure 5 demonstrates that transcription of both katB and ankB is stimulated by paraquat in a concentration-dependent fashion. The transcriptional start site was found to be a G 227 bp upstream of the katB start codon. Furthermore, transcription of katB-ankB is dependent upon OxyR, since no katB and very little ankB transcript could be detected in an oxyR mutant. These results were also confirmed using ankB::lacZ reporter fusion studies (data not shown).
Phenotypes of a P. aeruginosa ankB mutant. (i) Normal cell size and shape. Humans with hereditary spherocytosis (HS) suffer from an ankyrin deficiency. Erythrocytes from individuals with HS lack deformability and stability (40) and are unable to pass through capillaries, resulting in hemolytic anemia and hypersensitivity to osmotic lysis. This disorder has been reproduced in nb/nb (normoblastosis, ankyrin-deficient) mice (7), which have a severe hemolytic anemia throughout life (41). In these settings, it is predicted that a loss of ankyrin from the lipid bilayer causes a reduction in the critical surface area/ volume ratio, leading to a shift in the morphology of erythrocytes from discoidal to spherical. Thus, there is a definitive structural role for ankyrins in erythrocytes. In contrast to the case for HS erythrocytes, the ultrastructure of wild-type and ankB mutant bacteria was observed by TEM and no significant differences in overall cell shape were found (Fig. 6). However, the ankB mutant produced more membrane vesicles than wildtype bacteria (Fig. 6B). This implies that a fundamental difference exists between the surfaces of the wild type and the ankB mutant and that their ability to package periplasmic constituents in natural membrane vesicles has changed (i.e., the ankB mutant has more packaging potential). The differ-ence in quantities in membrane vesicles has been confirmed by thin sections (6).
(ii) Enhanced sensitivity to H 2 O 2 . Because we found that ankB is part of a small operon with katB, we postulated that its gene product may play a role in resistance to H 2 O 2 . To test this hypothesis, the wild type and ankB, katB, and katB ankB mutants were screened for H 2 O 2 sensitivity. As shown in Fig. 7, an ankB mutant was only slightly more susceptible to H 2 O 2 than wild-type organisms (bars 2). However, when the mutant was pretreated with a sublethal dose of H 2 O 2 , which activates the katB-ankB operon, sensitivity was increased dramatically (bars 2 versus bars 1 [shaded bars]). Bars 6 demonstrate that provision of a plasmid that allows for constitutive expression of ankB restored wild-type resistance regardless of H 2 O 2 pretreatment. The katB (bars 3) and katB ankB (bars 4) mutants were equally susceptible to H 2 O 2 , and more so than the ankB mutant. Interestingly, provision of ankB alone to the katB ankB mutant dramatically helped these organisms resist H 2 O 2 (bars 8 relative to bars 7).
(iii) Absence of AnkB decreases KatB activity. Because of the close proximity of katB and ankB, their organization in a small operon, and the enhanced H 2 O 2 sensitivity of the ankB mutant, we postulated that AnkB could play a role in KatB function. To test this hypothesis, the catalase isozyme profiles of several mutant organisms were examined. RNase protection assays indicate that katB and ankB comprise an operon and are regulated by OxyR. Riboprobes specific for the katB promoter (katB rp) and for the katB-ankB overlapping region (katB-ankB rp) were used to detect the corresponding transcripts in P. aeruginosa PAO1 or oxyR mutant total RNA isolated during the exponential growth phase in aerobic M9 minimal medium. Paraquat (PQ) was added to final concentrations of 10 and 100 M 1 h prior to harvest as indicated. Also shown are the digested probes in the absence of any P. aeruginosa RNA as a control. A DNA sequencing reaction was run in parallel and served as a size marker. Numbers on the left are base pairs. ously undiscovered catalase in P. aeruginosa, although analysis of the recently completed P. aeruginosa genome suggested otherwise (data not shown). We now know that this paraquat-inducible catalase band is one of the alkyl hydroperoxide reductases, AhpA, that possesses weak catalase activity (39). The KatB activities of the ankB mutant (Fig. 8, right panel, lane 2) and a katA ankB mutant (right panel, lane 5) were still reduced relative to that of wild-type bacteria (right panel, lane 1).
(iv) Quantitative effect of AnkB on KatB activity. To quantify the effect of AnkB on KatB activity, we examined catalase activity in katA and katA ankB mutants that had been treated with paraquat in stationary-phase culture, where the only catalase activity that can be detected and quantified spectrophotometrically is KatB. As shown in Fig. 9, the KatB activity of a paraquat-treated katA mutant is ϳ21.5 U/mg (bar 1). Provision of ankB in trans to the katA mutant had no effect on KatB activity (bar 2). Interestingly, KatB activity in a katA ankB mutant was reduced fourfold (bar 3) relative to that of the katA mutant and was fully complemented by providing ankB in trans (bar 4). There was no observation of the AhpA activity band under these conditions.

DISCUSSION
The major catalase gene of P. aeruginosa, katA, encoding a constitutive 170-kDa heteromultimer, is positively regulated by iron (35) and maximally expressed in stationary phase, in part through a process of cell-to-cell communication known as quorum sensing (27). Thus, it is not surprising that KatA contributes significant protection against H 2 O 2 in both planktonic and biofilm cultures (26,27,35).
In contrast to katA, which is minimally responsive to H 2 O 2 , we found in this study that the katB-ankB operon is transcribed dramatically in its presence and requires the global transactivator OxyR (25,39). When we discovered ankB downstream of katB, we immediately classified its gene product as an ALP because it possessed the characteristic 33-amino-acid ank re-FIG. 6. Ultrastructural analysis of wild-type PAO (A) and PAO ankB mutant (B) bacteria. Bacteria were grown aerobically in L broth to mid-exponential growth phase and treated with 1 mM H 2 O 2 for 1 h at 37°C to stimulate transcription of katB-ankB. Organisms were then prepared for TEM examination as described in Materials and Methods. The arrows in panel B point to the larger number of membrane vesicles being produced by the ankB mutant. The width of the cells is ϳ800 nm.
peat motifs and showed high similarity to ALPs in the related organisms P. fluorescens and P. syringae (32). Bacterial ALPs differ dramatically from their eukaryotic counterparts in that they contain only the ank repeats, which we believe to be involved in protein-protein interactions (5). Bacterial ALPs should, therefore, lack the structural role in the cell, similar to the function of prototypical erythrocyte spectrin-binding ankyrin. Examination of transmission electron micrographs of the ankB mutant confirmed this assumption (Fig. 6). Due to the polycistronic nature of katB-ankB and the conservation of this operon in the pseudomonads and other proteobacteria such as V. cholerae, we postulated that AnkB might belong to a group of evolutionarily related proteins with a novel, unrecognized function(s), one of which could contribute toward protection against H 2 O 2 . Indeed, AnkB appears to play a role in the response of P. aeruginosa to H 2 O 2 , because an ankB mutant was more sensitive to it than wild-type organisms (Fig.  7). Furthermore, the enhanced H 2 O 2 sensitivity of an isogenic katB ankB mutant did not change when only katB was provided in trans (39). Although unproven, the nearly fourfold reduction in KatB activity in the ankB mutant suggests that there could be a physical interaction between the two proteins. We found KatB activity in the cytoplasm, periplasm, and cytoplasmic membrane (data not shown). Because AnkB is a cytoplasmic membrane protein whose bitopic integration into the inner membrane ultimately causes its ank repeat domain to reside in the periplasm, we postulated that one function of AnkB may be to bind KatB near inner membrane targets that are sensitive to H 2 O 2 (e.g., F 1 F o -ATPase [55]). H 2 O 2 must first enter a protein channel leading to the heme catalytic site of the catalase molecule (47). Without entering this channel, the H 2 O 2 is free to damage cellular components, especially sensitive respiratory chain components and DNA (16). Thus, AnkB may position or anchor KatB so that its H 2 O 2 channel is in the optimal orientation for H 2 O 2 entry. Alternatively, AnkB may serve to stabilize KatB, allowing it to persist longer and function better upon exposure of bacteria to H 2 O 2 .
An alternative hypothesis is that AnkB may reinforce the cytoplasmic membrane and prevent crippling of the proton motive force. Microscopic oxygen bubbles could be produced upon H 2 O 2 degradation, thereby increasing cellular turgor pressure. Although unexplained, such cell swelling has been shown in E. coli (38) and in mitochondria treated with H 2 O 2 or agents that generate it (30). Upon H 2 O 2 degradation by cata- FIG. 7. Effect of ankB, katB, and katB ankB on sensitivity to H 2 O 2 . All bacteria were grown aerobically overnight in M9F medium at 37°C. Fresh prewarmed medium (1 volume of culture in 10-volume flasks) was inoculated with 1/50 of the final culture volume and allowed to reach an OD 600 of 0.6. Some bacteria were pretreated with a sublethal (1 mM) dose of H 2 O 2 for 1 h (shaded bars) relative to control bacteria (open bars). The suspensions were diluted 100-fold in 7 ml of M9F 0.6% top agarose kept at 37°C and poured onto M9F plates. Filter paper disks (7 mm) impregnated with 8.8 M H 2 O 2 were placed on the top agar surface. Zones of growth inhibition were measured after a 24-h aerobic incubation at 37°C. Bars: 1, PAO1; 2, ankB; 3, katB; 4, katB ankB; 5, ankB plus pUCP22; 6, ankB plus pankB; 7, katB ankB plus pUCP22; 8, katB ankB plus pankB. Error bars indicate standard errors of the means.
FIG. 8. Absence of AnkB causes a decrease in KatB activity. Bacteria were grown aerobically overnight in L-broth medium at 37°C. Fresh prewarmed medium (1 volume of culture in 10-volume flasks) was inoculated with 1/100 the final culture volume, and the organisms were grown to an OD 600 of 0.6. Some bacteria (right panel) were then treated with a sublethal (0.35 mM) (11) dose of paraquat for 1 h, and the others (left panel) served as controls. Cell extracts were prepared, and 10 g was subjected to nondenaturing polyacrylamide gel electrophoresis (5% polyacrylamide). The gels were then stained for catalase activity (58). Lanes: 1, PAO1; 2, ankB; 3, katA; 4, katB; 5, katA ankB. Catalase activity and activity gel staining were monitored in cell extracts. Bars: 1, katA plus pUCP22; 2, katA plus pankB; 3, katA ankB plus pUCP22; 4, katA ankB plus pankB. Error bars indicate standard errors of the means. The inset photograph is KatB activity staining of cell extracts in a representative native polyacrylamide gel of each strain. It should be noted that no AhpA activity band (see Fig. 8) could be detected under these conditions. lase, oxygen gas nuclei could be stabilized and even grow in the bacteria at hydrophobic sites. With the production of gas at a rate that saturates the cytoplasm, gas bubbles could readily appear, be stabilized by lipid and/or protein adsorption, and take up considerable volume inside a cell, thereby creating a turgor pressure (31). Thus, AnkB could serve to stabilize the inner membrane against swelling due to the mounting intracellular pressure built by H 2 O 2 degradation. Both hypotheses are being tested experimentally.