The Essential HupB and HupN Proteins of Pseudomonas putida Provide Redundant and Nonspecific DNA-bending Functions*

, A protein mixture containing two major components able to catalyze a b -recombination reaction requiring nonspecific DNA bending was obtained by fractionation of a Pseudomonas putida extract. N-terminal sequence analysis and genomic data base searches identified the major component as an analogue of HupB of Pseudomonas aeruginosa and Escherichia coli, encoding one HU protein variant. The minor component of the fraction, termed HupN, was divergent enough from HupB to pre-dict a separate DNA-bending competence. The determi-nants of the two proteins were cloned and hyperex-pressed, and the gene products were purified. Their activities were examined in vitro in b -recombination assays and in vivo by complementation of the Hbsu function of Bacillus subtilis. HupB and HupN were equally efficient in all tests, suggesting that they are independent and functionally redundant DNA bending proteins. This was reflected in the maintenance of in vivo activity of the s 54 Ps promoter of the toluene degradation plasmid, TOL, which requires facilitated DNA bending, in D hupB or D hupN strains. However, hupB/hupN double mutants were not viable. It is suggested that the requirement 0.3 M NaCl contained . 95% pure HupB. Fractions containing either pure HupN or HupB were pooled separately, dialyzed against buffer A, and added with glycerol to a concentration of 44% (v/v). The prepara- tions were stored at 2 20 °C and contained, respectively, 1.0 mg/ml (HupN) and 2.0 mg/ml (HupB). Cross-linking of HupB and HupN— The physical association between HupB and HupN monomers was examined through covalent binding of surface-exposed Lys residues on the polypeptides with suberic acid (disuccinimidyl suberate, Sigma). Stock solutions of suberic acid were prepared at a concentration of 10 mg/ml in dimethyl sulfoxide. A typical 100- m l cross-linking reaction contained 50 m M Tris-HCl, pH 7.5, 1 m M EDTA, 10 m M MgCl 2 , 25 m M NaCl, 4 m M HupB or HupN, 0.2 mg/ml suberic acid, in 100 m l reaction volume. Control reactions were set up by replacing the HU proteins by 30 m g/ml lysozyme, which exists only as a monomer. To examine the formation of multimers in the presence of DNA the reactions were added where indicated with 200 n M pCB8 plasmid. One way or the other, mixtures were incubated for 15 min at room temperature and the reactions stopped with an excess (4 m M ) of lysine. Samples were then dialyzed in Microcon-10 concentrators (Mil- lipore) and examined in denaturing 15% polyacrylamide-Tricine-SDS gels (33). Construction of a P. putida Ps-lacZ Reporter Strain— A specialized P. putida strain bearing a transcriptional lacZ fusion to the HU-depend- ent Ps promoter of the TOL plasmid (38, 39) was generated as follows.

Because of the topological changes that DNA must undergo during basic cellullar functions, factor-directed DNA bending is frequently involved in a variety of replication, recombination, and transcription mechanisms (1,2). The abundant polypeptides that assist the maintenance of the DNA structure and its conformational transitions in bacteria are generically termed nucleoid-associated proteins. This somewhat vague term includes, in the case of Escherichia coli, factors such as HU, H-NS, StpA, Lrp (leucine-responsive protein), Fis, and the integration host factor IHF (3,4). 1 With different degrees of specificity, these proteins bind distinct sites or attach to less specific DNA segments and compact, relax, or change the architecture of given chromosomal regions (5)(6)(7). Among these proteins, HU has been identified as the major component of the bacterial nucleoid, the function of which is to facilitate a whole spectrum of molecular processes that involve DNA bending. With at least 30,000 molecules (dimers)/cell, HU statistically can bind every 200 bp of the E. coli chromosome (3). The HU protein of E. coli consists of two genetically unlinked polypeptides of about 9 kDa, which are encoded by two highly similar genes, hupA and hupB (8 -11). In contrast to IHF, which binds strongly to a consensus DNA motif (12,13), HU does not generally bind in a sequence-specific manner. The HU proteins of E. coli are functional both as homodimers and heterodimers, although each form seems to be specialized in nonidentical roles (14 -16). HU Ϫ mutants, which bear deletions of both the hupA and hupB genes in E. coli, are viable but exhibit many growth defects (17,18). In addition, these mutants are sensitive to ␥ and UV irradiation because of the involvement of HU in DNA repair and homologous recombination (19,20). Although HU proteins are conserved in most microorganisms, only enteric bacteria seem to harbor two genes encoding polypeptides able to form a heterodimer (21). In other eubacteria such as Bacillus subtilis, an essential single gene (hbs) gives rise to a homodimeric HU protein (21,22), which is named Hbsu in this species.
The life cycle and the natural niches of E. coli are relatively simple compared with those of not-so-distant relatives, such as Pseudomonas putida, which thrive in soils polluted with toxic chemicals (23,24). In these niches, the transduction of environmental signals occurs through a whole collection of mechanisms that very frequently involve regulated DNA bending (1). This is particularly true for promoters of metabolic pathways that are dependent on factor sigma 54 ( 54 ), such as the Pu and Ps promoters of the toluene degradation plasmid (TOL) or the Po promoter of the pVI150 (catabolism of methyl phenols), the expression of which is finely regulated by multiple environmental inputs (25,26). Although such transduction pathways involving DNA bending can be reproduced to some extent in E. coli, it cannot be taken for granted that the same factors facilitating DNA bending are identical in bacteria living in more complex habitats. To address this issue, we fractionated a cell extract, P. putida, and sought proteins that were generically able to facilitate DNA bending. As shown below, the most active fraction contained two independent HU variants the function of which appeared to be to back up each other for essential DNA bending functions.

EXPERIMENTAL PROCEDURES
Strains, Plasmids, and Plasmid Construction-Plasmid pFBT8 contains a genomic 6-kb SacI fragment from the P. putida chromosome spanning the hupB sequence and cloned in vector pZERO-2 (Invitrogen). The hupB gene was amplified from pFTB8 with primers PLHU1N (5Ј-TTCATATG(NdeI)AACAA GTCG GAA CTGATTG-3Ј) and PRHU1B (5Ј-TGGGATCC(BamHI)GA CTT AGTTGA CGGC GTC-3Ј), which generated anϳ300-bp PCR product bound by a leading ATG overlapping a NdeI site and a stop codon, TAA, followed by a BamHI sequence. This fragment was digested with NdeI and BamHI and cloned in the corresponding sites of expression vector pT7-7 (gift of Stanley Tabor, Harvard Medical School), giving rise to plasmid pFBT17. Similarly, the hupN sequence was amplified with oligonucleotides PLNAK3 (5Ј-AC-CACCATGG(NcoI)CATTGACCAAAGACC-3Ј) and PRNAK4 (5Ј-TTTG-GATCC(BamHI)ATTACTTGTTGATGGCGTCG-3Ј), which left NcoI and BamHI sites at the start and end of the ORF. The resulting ϳ300-bp product was cloned as a NcoI-BamHI insert in pET-3d (Stratagene), originating plasmid pFBT12. The inserts of pFBT17 (hupB ϩ ) and pFBT12 (hupN ϩ ) were resequenced to ensure that no error had been entered in the ORFs during the cloning procedures.
For expression of hupB and hupN in B. subtilis, the corresponding inserts of pFBT17 and pFBT12 were transferred to shuttle vector pHP13 (27), which contains origins of replication functional in both E. coli and B. subtilis, as well as chloramphenicol (cat, Cm R ) and erythromycin resistance gene markers. To transfer hupB to such a shuttle plasmid, pFBT17 was digested with XbaI, a site for which is located upstream of the optimized Shine-Dalgarno sequences and the translation initiation regions present in the plasmids. The resulting XbaI end of the digested pFBT17 was then blunt-ended with T4 DNA polymerase and the plasmid subsequently digested with BamHI. This released the hupB sequence as a segment bounded by a blunt end and a BamHI site. Such purified fragment was then cloned in pHP13 digested with SmaI and BamHI, thereby giving rise to plasmids pHP13-hupB. The same procedure was used to generate an equivalent plasmid with hupN, except that the fragment inserted in the shuttle pHP13 was bounded by a blunt-ended extreme (formerly a XbaI site in pFBT12) and a HindIII site. Ligation of such a hupN ϩ fragment to pHP13, digested with SmaI and the HindIII originated plasmid pHP13-hupN. In these constructs, the hupB and hupN genes are located downstream of the cat gene of pHP13, so that their transcription originates from the cat promoter.
The plasmids employed for overproduction of hupB and hupN in P. putida (see below) were constructed by transferring the inserts of pFBT17 and pFBT12 into the broad host range and kanamycin resistance expression vector pVLT33 (28). To this purpose, both pFBT17 and pFBT12 were digested with XbaI. The resulting ends were filled in as described above, and digested with BamHI. This originated DNA segments each flanked by one blunt end an a BamHI extreme. The DNA fragments were ligated to vector pVLT33, which had been digested with EcoRI, filled in with the Klenow fragment of DNA polymerase ϩ dNTPs (29), and subsequently digested with BamHI. Because of this pedigree, the resulting plasmids, pFBT18 (hupB ϩ ) and pFBT16 (hupN ϩ ), bore each of the genes preceded by an optimal Shine-Dalgarno and translation initiation region sequences and located downstream of a strong Ptac promoter inducible with isopropyl-␤-D-galactopyranoside. A variant of pFBT18 named pFBT22 contained a xylE gene (encoding the enzyme catechol 2,3-dioxygenase) in addition to hupB. For the construction of pFBT22, xylE was entered in pFBT18 by digesting it with BamHI ϩ HindIII and ligating the result to the 960-bp fragment released from pXYLE1 (30) when cleaved with the same enzymes.
Fractionation of Cell Extracts of P. putida KT2442 and N-terminal Analysis of Predominant Proteins-The isolation of protein fractions enriched in the DNA bending activity was based on the protocol of Padas et al. (31). P. putida strain KT2442 (a rifampicin-resistant variant of P. putida KT2440 (32)) was grown at 30°C to stationary phase (A 600 ϭ 3) in 4 liters of LB medium (with 50 g/ml rifampicin). Cells were centrifuged at 10,800 ϫ g for 12 min at 4°C and stored at Ϫ70°C until further utilization. Subsequent manipulations were carried out at 0 -6°C. 12 g of the frozen cell paste were resuspended in 150 ml of buffer A (25 mM Tris-HCl, pH 7.4, 1 mM EDTA, 3 mM ␤-mercaptoethanol, 100 mM NaCl) added together with 3 mg of DNase I, and lysed with a French press. After centrifugation at 30,000 ϫ g for 30 min, the supernatant was added stepwise for 30 min with ammonium sulfate at 55 and 85%. After centrifugation of the 85% ammonium sulfate extract, the pellet was resuspended in 12 ml of buffer A and dialyzed overnight against 5 liters of buffer A using a MWCO/2000 tubes (Spectrum Medical Industries). The dialyzed protein solution was applied to a 2-ml heparin-Sepharose CL-6B (Amersham Pharmacia Biotech) column equilibrated with buffer A. Bound protein was eluted with a 30-ml gradient of 0.1-1.5 M NaCl in buffer A. The fractions enriched in proteins with a molecular size within the range of 7 to 12 kDa (as judged by denaturing polyacrylamide gel electrophoresis) were diluted in buffer A and subsequently loaded on a 4-ml SP-Sepharose High Performance (Amersham Pharmacia Biotech) column equilibrated with buffer A. A 40-ml gradient of 0.1-1.5 M NaCl in buffer A was applied. Fractions enriched in proteins within the same size range given above were selected for further analysis. For identification of major proteins, the fractions of interest were run in a denaturing Tricine-sodium dodecyl sulfate-gel electrophoresis system (33) and blotted on a polyvinylidene difluoride membrane (Millipore). The membrane was stained with Coomassie Brilliant Blue R250, washed in water, and air-dried. The protein bands in the range of 9 kDa were excised from the membrane, and their N-terminal amino acids were determined through and automated Edman degradation microsequencing protocol (34). Protein concentrations were determined with the Protein Assay ESL (Roche Molecular Biochemicals), which is based on complexing Cu 2ϩ ions by the protein to be measured in a reaction exclusively dependent on the number of peptide bonds.
In Vitro Monitoring of Nonspecific DNA Bending Activity-The assay (35) is based on the site-specific recombination between so-called six sites brought about by the ␤-recombinase encoded by plasmid pSM19035 of Streptococcus pyogenes as part of its replication mechanism (36). The 5.8-kb test plasmid pCB8 (36) contains two directly repeated target six sites separated by 2.2 kb. Reaction mixtures were set up in a volume of 25 l containing 10 nM pCB8, 25 mM Tris-HCl, pH7.5, 50 mM NaCl, 10 mM MgCl 2 , and 50 nM of purified ␤-recombinase. The materials tested for DNA bending included both crude extracts and protein fractions from P. putida as well as purified HupB and HupN proteins, which were added to the assays in amounts indicated in each case. Samples containing 50 nM of the chromatin-associated protein Hbsu from B. subtilis were also used as positive controls for the performance of the assay. After incubation for 30 min at 30°C, the reaction was stopped by heating (70°C, 10 min). The DNA was then digested in the same buffer with PstI and SalI, and the resulting fragments were analyzed by agarose-gel electrophoresis (37).
Separate Overproduction in P. putida of HupB and HupN and Purification of the Proteins-To avoid cross-contamination between the HU proteins of P. putida and E. coli, we systematically used a P. putida host for overexpression of the HupB and HupN products. To this end, the hupN ϩ plasmid, pFBT16, was electroporated into the wild type strain P. putida KT2442, whereas the hupB ϩ plasmid pFBT18 was placed into a ⌬hupN deletion mutant (see below). The purification procedure for HupB and HupN was carried out basically according to Ref. 31. Each of the expression strains were grown in 4 -5 liters of LB medium (with 50 g/ml kanamycin) at 30°C with vigorous shaking to an A 600 of 0.6 -0.8. Following the addition of 1 mM isopropyl-␤-D-galactopyranoside, the cultures were further incubated for 3-4 h at the same temperature. 13-18 g of wet cell paste were harvested by centrifugation for 10 min in the cold at 16,300 ϫ g. Cells resuspended in 3 ml of buffer A (see above)/gram of paste were lysed with a French press and cleared by centrifugation at 35,000 ϫ g for 30 min. The proteins that precipitated within the 55-85% saturation range of ammonium sulfate were resuspended in 2 ml of buffer A/gram of cells and dialyzed extensively against buffer A. The resulting protein solution (35-40 ml) was loaded at a flow of 1 ml/min in a 5-ml heparin-Sepharose CL-6B (Amersham Pharmacia Biotech) column equilibrated with buffer A. After washing the column extensively with buffer A, an 80-ml gradient of 0.1-1.7 M NaCl in buffer A was applied to the column at the same 1 ml/min flow rate. 2-ml fractions were collected and analyzed in Tricine-SDS-PAGE (33). HupN eluted from the column within 0.55-0.75 M NaCl with a purity in the best fractions Ն95%, and thus no further purification was pursued. On the contrary, HupB eluted at 0.51-0.76 M NaCl, but the fractions were significantly contaminated with a few additional proteins. The HupBcontaining fractions were therefore pooled and dialyzed overnight against buffer A, and 15 ml of the dialysate was loaded at a 1 ml/min rate on a 3-ml SP-Sepharose High Performance (Amersham Pharmacia Biotech) column equilibrated with buffer A. After washing the column with 3 ml of buffer A, a 60-ml gradient of 0.1-1.0 M NaCl was applied, also at 1 ml/min, to elute the protein. 1.5-ml fractions were collected and analyzed in Tricine-SDS-PAGE as before. The fractions corresponding to 0.3 M NaCl contained Ͼ95% pure HupB. Fractions containing either pure HupN or HupB were pooled separately, dialyzed against buffer A, and added with glycerol to a concentration of 44% (v/v). The preparations were stored at Ϫ20°C and contained, respectively, 1.0 mg/ml (HupN) and 2.0 mg/ml (HupB).
Cross-linking of HupB and HupN-The physical association between HupB and HupN monomers was examined through covalent binding of surface-exposed Lys residues on the polypeptides with suberic acid (disuccinimidyl suberate, Sigma). Stock solutions of suberic acid were prepared at a concentration of 10 mg/ml in dimethyl sulfoxide. A typical 100-l cross-linking reaction contained 50 mM Tris-HCl, pH 7.5, 1 mM EDTA, 10 mM MgCl 2 , 25 mM NaCl, 4 M HupB or HupN, 0.2 mg/ml suberic acid, in 100 l reaction volume. Control reactions were set up by replacing the HU proteins by 30 g/ml lysozyme, which exists only as a monomer. To examine the formation of multimers in the presence of DNA the reactions were added where indicated with 200 nM pCB8 plasmid. One way or the other, mixtures were incubated for 15 min at room temperature and the reactions stopped with an excess (4 mM) of lysine. Samples were then dialyzed in Microcon-10 concentrators (Millipore) and examined in denaturing 15% polyacrylamide-Tricine-SDS gels (33).
Construction of a P. putida Ps-lacZ Reporter Strain-A specialized P. putida strain bearing a transcriptional lacZ fusion to the HU-dependent Ps promoter of the TOL plasmid (38,39) was generated as follows. The 2.4-kb HpaI fragment of plasmid pTK19 (T. Köhler, University of Geneva) containing the complete xylR gene and the Ps promoter was cloned into the SmaI site of the lacZ vector pUJ8 (40), thereby generating a Ps-lacZ fusion. Such a fusion was excised from the resulting plasmid as a 6.5-kb NotI fragment, which was cloned into the NotI site of a tellurite resistance pUT/mini-Tn5 delivery plasmid (41). The hybrid xylR ϩ /Ps-lacZ mini-transposon generated thereby was integrated into the chromosome of P. putida KT2442 (described in detail in Ref. 42). One lac ϩ Tel R P. putida exconjugant sensitive to piperacillin (43) and displaying a good ␤-galactosidase induction upon cell exposure to benzylalcohol (44) was picked up as a reference strain to monitor the phenotypes of hupB and hupN mutants.
Construction of ⌬hupB and ⌬hupN Deletions in P. putida Ps-lacZ Strain-To generate P. putida variants lacking either hupB or hupN, we entered directed deletions of each gene into the chromosome by homologous recombination. In the case of hupB, plasmid pFBT8 was used as the template to generate two PCR products with the pairs of primers PLKB1 (5Ј-GGAATTC(EcoRI)ACGTGCCGATGTGGCCATGAC-CGGCG-3Ј)/PRKB2 (5Ј-TCAGGATCC(BamHI)TTAAGTATGTAATGA-AAAGCTTATGAAACGTGTGC-3Ј) and PLKB3(5Ј-TAAGGATC C(Bam-HI)TGACAAACCGGCAAGGCAATCAAGATCGAAGCC-3Ј)/PRKB4 (5Ј-ACGTAAGCTT(HindIII)GCATCAGTTGACGGCGCTGCATGTCGGC-3Ј). The first product (456 bp) was digested with EcoRI ϩ BamHI, whereas the second product (492 bp) was treated with BamHI ϩ Hin-dIII. The digested DNAs were purified and ligated together to vector pUC18Not (45) digested with EcoRI ϩ HindIII (46). This originated a plasmid that had a NotI insert spanning the genomic DNA sequence starting at Ϫ474 bp and going to Ϫ36 bp upstream of the leading ATG of hupB, followed by an artificial BamHI site and followed by the sequence ϩ189 downstream of the ATG to ϩ386 after the stop codon of the structural gene. Such a NotI fragment thus contains a genomic deletion of 224 bp that engages the hupB sequence. The presence of two in-frame stop codons around the BamHI site employed to create the deletion ensured the loss of the entire HupB function. The 947-bp NotI fragment was then passed on to suicide delivery vector pKNG101 (47) (resulting in plasmid pKNG101⌬hupB) and recombined in the chromosome of P. putida Ps-lacZ as described (47). The chromosomal deletion was confirmed by PCR with primers PLKOB5 (5Ј-GCCATGCCTCCCA-GCACACG-3Ј) and PRKOB7 (5Ј-GCGTCCTGGCTATGAGTGGC-3Ј). A similar approach was used to produce a ⌬hupN strain. In this case, genomic DNA was directly used to generate two PCR products with the pairs of primers PLKN1 (5Ј-GGAATTC(EcoRI)-TCGGCG CGCTGCAA-AAGTTGC-3Ј)/PRKN2 (5Ј-TCAGGATCC(BamHI)TTAATCAAA TTCG-ATGGTTATGCAGCGGACTACG-3Ј) and PLKN3 (5Ј-TAAGGATC C(BamHI)TGAGACCAACTGATTGCCGACATCGCCGAATCG3Ј)/ PRKN4 (5Ј-ACGTAAGCTT(HindIII)CCCGGGCAAGCCGGGCGCGGT-TTCG-3Ј). The PCR products digested, respectively, with EcoRI ϩ BamHI and BamHI ϩ HindIII were cloned simultaneously into pUC18Not treated with EcoRI ϩ HindIII as before. The resulting 690-bp insert was then placed into the NotI site of pKNG101, yielding pKNG101⌬hupN, and was subsequently recombined into the chromosome of the P. putida strain. This produced a genomic deletion of a small 31-bp segment that eliminated the region between coordinates Ϫ17 to ϩ14 in respect to the leading ATG codon of the gene. Such a deletion was confirmed with PCR using the primers PLKON5 (5ЈCGA-TAACGTCGTAGTCCGCTGC-3Ј) and PRKON6 (5Ј-GTCAGCAC-TTTGGCT-GGAACGAAC-3Ј).
Predictions of Evolutionary Distance among Homologous Proteins-Protein sequence alignment were assembled with programs CLUST-AL X (Version 1.63b (43)) and Protdist of PHYLIP (Phylogeny Inference Package, Version 3.5c (48) The sequence of the HupN protein from P. aeruginosa was deduced from the genomic data available at the www.pseudomonas.com site (49). The complete HupB and HupN sequences from P. putida were assembled with the assistance of the TIGR (The Institute for Genomic Research) genomic data bank. The GenBank TM accession numbers of the new HU proteins reported in this work are AF345628 (hupB of P. putida); AF345629 (hupN of P. putida), and AF345630 (hupN of P. aeruginosa).

Identification of Nonspecific DNA-bending Activities in Pro-
tein Extracts of P. putida-To identify the proteins of P. putida with an ability to promote DNA bending in a sequence-independent fashion, we resorted to the in vitro test summarized in Fig. 1. The ␤-recombinase encoded by plasmid pSM19035 (36) is able to catalyze DNA recombination between two directly oriented target sites (six sites) both in vivo and in vitro. However, this occurs only in the presence of native or heterologous factors that facilitate DNA bending in a nonspecific fashion. In Bacillus, such an activity is provided by the Hbsu protein (an HU homologue) but can be replaced by entirely heterologous factors such as mammalian or plant HMG B1-type proteins. This reaction can be followed easily with test plasmid pCB8, which contains two six sites in direct orientation and generates three DNA fragments (one of 4.8 kb and two of 0.5 kb each) when it is digested with PstI and SalI. Following ␤ recombination in vitro the digestion of the resulting catenane with the same enzymes produces two additional fragments of 2.7 and 2.1 kb (Fig. 1A). With this assay in hand, we tested protein fractions from P. putida extracts retained in and eluted from heparin columns (thus enriched in DNA-binding proteins) and then fractionated in a second cation exchanger column. The result of such a fractionation, as analyzed in a Tricine-SDS-PAGE system, is shown in Fig. 1B. The fraction that eluted at 0.3 M NaCl from the SP-Sepharose column (lane 1 in Fig. 1B) maximally stimulated the ␤-recombination reaction, to the point that a 100-fold dilution (ϳ8 g protein/ml) was as efficient as the positive control with 4.0 g/ml of purified Hbsu (Fig. 1C). The addition of nondiluted fractions to the reaction mixtures led to significant a degradation of the test DNA, perhaps reflecting some contamination by the DNase added during the preparation of the extracts.
Categorizing of Proteins Candidate to Assist ␤-Recombination-For identification of the protein(s) in the extract described above that accounted for the most predominant polypeptide(s) of fraction 1 (ϳ9 kDa, Fig. 1B), the gel was blotted and the band was subjected to N-terminal analyses as explained under "Experimental Procedures." The band turned out to be a non-equal mixture of two very similarly sized proteins. The most abundant polypeptide (two-thirds of the band) was led by the sequence MNKSELIDAIAASADIPKAVA-GRALD. Searches in the Pseudomonas genomic data bases of the University of Queensland, the PathoGenesis Corporation, and TIGR, as well as in individually deposited gene sequences in the EMBL, revealed that the product had a maximum match with a P. aeruginosa gene previously named hupB, which was proposed to be the HU protein of this species (50). Such a gene has its counterpart in the genome of P. putida, and therefore we named this component of the active fraction HupB (Fig. 2).
The second, less abundant protein (one-third of the band) yielded an N-terminal sequence ALTKDQLIADIXESIAAXXX-TAKN (i.e., lacking a first Met residue). Such a sequence could be matched to single ORFs present in both the P. aeruginosa and P. putida genomes, although no specific function had been assigned to them. These ORFs give rise to 77% identical products in both species, with the highest similarity in the Nterminal half of the predicted proteins. Apart from the mutual likeness, the ORFs had a 50% similarity to the hupA gene of V. proteolyticus (51). In addition, it matched an HU-like gene identified by Nakazawa 2 in P. putida mt-2 on the basis of complementation of an HU mutant of E. coli. We thus designed the second component of the ␤-recombination active fraction as HupN, to differentiate it from the previously described HU-like genes. It should be noted, however, that the HupN and HupB proteins still possess 45% identical amino acids (Fig. 2). Whether or not the relative ratio between HupB and HupN found in the extract bears any meaning or is just the result of the extraction and procedure remains unclear, although it is evident that both products are very abundant polypeptides. These results provided a basis for understanding the nonspecific DNA bending activity found in fraction 1, loaded in the gel of Fig. 1B. In fact, the HupB and HupN proteins became candidates, together or separately, for the HU homologue(s) of P. putida. Although the fractionation procedure employed did not rule out the possibility that other proteins with a similar activity might exist, it seems apparent that the bulk of proteins in the size range of typical nucleoid-associated factors included mostly these two products. To sort out these issues, we proceeded to the characterization in vitro and the genetic/ 2 T. Nakazawa, personal communication. phenotypic analysis in vivo of the roles of HupB and HupN in P. putida.
Purified HupB and HupN Are Competent in ␤-Recombination Assays-To assign unequivocally the DNA bending activity found in P. putida extracts to HupB, HupN, or a combination of the two, we proceeded to overexpress them in its native host and separately purify each of them with the method described under "Experimental Procedures" (Fig. 3A). To ensure the identity and quality of each of the purified polypeptides, the samples were separately subjected to an N-terminal analysis, which verified not only a purity Ն95% but also a complete absence of cross-contamination by other host nucleoid-associated proteins. Prior to running activity assays, the physical state of either protein was determined through in vitro crosslinking experiments with the lysine-specific agent suberic acid. As shown in Fig. 3B, both HupB and HupN predominantly formed dimers (lanes 4 and 7), although a small portion of each of the samples remained monomeric even after 15 min of treatment. Some tetrameric forms of HupN (but not of HupB) could be observed also, as well as two potential dimeric species, although their significance is unclear. The presence of DNA in the reaction did not enhance the formation of oligomeric forms of HupB (lane 5) or HupN (lane 8).
In a subsequent step, the activity of each protein was assayed in a ␤-recombination assay in vitro using as a template the pCB8 plasmid as described above. As shown in Fig. 3C, both HupB and HupN provided the DNA bending activity required for the reaction, albeit at somewhat different degrees and qualities. Although HupB abruptly stimulated ␤-recombination (100% efficiency) at 150 nM, HupN initiated the reaction at 75 nM, albeit at a lower efficiency. The reaction with HupN improved with increasing concentrations of the factor, to reach saturation (100% recombination) at 150 nM. These concentrations are in the range of other HU proteins tested in this type of assay (35). These results indicated that both HupB and HupN sufficed by themselves to bend DNA in a sequenceindependent fashion, and therefore, they both contribute to providing this capacity to the active extract (Fig. 1C) and, presumably, to the cells in vivo as well (see below).
Functional Complementation of B. subtilis Hbsu by HupB or HupN-The next step in the characterization of the HupB and HupN proteins was to examine their functionality in vivo in an entirely heterologous context. To this end, we took advantage of the fact that the whole requirement for nonspecific DNA bending functions in B. subtilis is met by the protein Hbsu protein (52), which is encoded by the hbs gene. This gene is in fact essential for cell viability (22), and knockout mutants do not exist. However, disabled Hbsu variants bearing amino acid substitutions F47W and R55A still allow cell growth, although the bacteria are deficient in DNA repair, homologous recombination, and site-specific recombination mediated by protein ␤ (53). Such an allele thus provides a good assay system in vivo to examine the functional replacement of Hbsu by any other factors that supply the same activities. Although HupB and HupN of P. putida exhibit a respectable sequence similarity to Hbsu (53 and 43% respectively, not shown), the divergence is great enough to make the test more significant that a simple replacement of one HU variant by another. On this basis, both hupB and hupN were cloned separately in shuttle vector pHP13, and the resulting constructs were transformed into Bacillus strain BG405, which bears the mutated hbs gene (hbs-4755). To monitor functional complementation between Hbsu and the P. putida proteins, the transformants and their controls were then exposed to the DNA-damaging agent methyl methane sulfonate (10 mM), and the surviving cells were monitored as described previously (53). As shown in Fig. 4, the Hbsu Ϫ strain harboring vector pHP13 without any insert exhibited a high sensitivity to methyl methane sulfonate, whereas the wild type Bacillus strain was only marginally affected. But B. subtilis BG405 (hbs) transformed with either pHP13-hupB or pHP13-hupN was as resistant to methyl methane sulfonate exposure as the wild type and as the B. subtilis BG405 (pCB188) strain, the plasmid of which encodes the wild type hbs gene (see "Experimental Procedures"). These results indicated that both HupB and HupB can complement in full the loss of Hbsu in B. subtilis and strengthened the notion that each protein is capable by itself of meeting in vivo any physiological demand of generic DNA bending. In this respect, it is worth noticing that the HU-like hlpA gene of the plastid genome of the cryptomonad alga Chryptomonas ⌽ was not as efficient as the P. putida hup genes in the same complementation assay (54).
Phenotypes Endowed by ⌬hupB and ⌬hupN Deletions in P. putida-To determine the functions of HupB and HupN, together and separately, in their native context in vivo, we produced deletion mutants of each gene through homologous recombination of DNA segments into the chromosome of P. putida Ps-lacZ. The replacement of DNA fragments between the delivery plasmid and the chromosome occurred at high frequencies (Ն10%). The changes in the corresponding genomic sites caused by each of the deletions are represented in Fig. 5. Single ⌬hupB and ⌬hupN mutants did not display any evident phenotype. They grew at the very same rate and yield that the wild type cells in a variety of rich and minimal media tested and were as resistant to UV radiation as the wild type (not shown). As a more specific test, we examined the effect of a lack of HupB or HupN on the activity of the 54 promoter Ps of the TOL plasmid pWW0 of P. putida mt-2 (55). This promoter and its cognate activator, XylR, form part of an intricate regulatory system that controls the expression of a plasmid-encoded pathway for the biodegradation of toluene, m-xylene, and p-xylene (26). XylR belongs to the family of the prokaryotic enhancerbinding proteins that act in concert with the 54 -containing form of RNAP. The activation mechanism of Ps requires the looping out of the distant XylR⅐UAS complex to contact the 54 -RNAP bound downstream (56). This process involves the exacerbation of an intrinsically curved DNA sequence located between the bound XylR⅐UAS complex and the polymerase. When Ps and XylR are placed in E. coli, the promoter activity becomes dependent on the host HU protein (39); this probably reflects the need of a helper DNA bending factor to produce an optimal promoter geometry (39). To examine whether such a requirement is fulfilled in P. putida by either HupB or HupN, we monitored the activity of the chromosomal Ps-lacZ fusion born by the strain host to the deletions when induced with the XylR effector benzylalcohol. Fig. 6 shows the results of triplicate ␤-galactosidase assays carried out with the hupB and hupN single-deletion mutants compared with the parental strain with no deletions. Although we could detect some minor changes in the mutants (more evident in the case of the ⌬hupB strain), such variations did not exceed ϳ30% and could thus hardy be considered significant. Taken together, the data presented above suggested either that HupB and HupN are in fact redundant proteins, having the same functions in a variety of systems, or that yet another factor(s) may exist in P. putida that compensate the loss of each of them. This was ascertained with the assays in vivo shown below.
HupB and HupN Provide Essential functions for P. putida-Because the function(s) of HupB and HupN could only be hinted at ultimately by examining a double hupB/hupN mutant, we set out to produce a ⌬hupB/⌬hupN strain. This was attempted through a sequential recombination of each of the deleted DNA segments into the chromosome of the target strain. Unfortunately, after screening more than 1000 clones, we failed to produce such a double mutant, regardless of the order in which each delivery vector was used (see "Experimental Procedures"). That recombination was not the problem was evidenced by the observation that double chromosomal deletions ⌬hupB/⌬hupN could easily be generated if the delivery plasmid for the ⌬hupN DNA segment was recombined in the ⌬hupB strain that had been transformed with the hupB ϩ plasmid pFBT22. This gave us a clue that perhaps cells lacking both HupB and HupN were not viable. To test this notion, we set up the experiment shown in Fig.  7. In this assay, we employed the double-deleted ⌬hupB/⌬hupN strain P. putida transformed with pFBT22 (hupB ϩ ). This plasmid bears as well a xylE ϩ insert (encoding catechol 2,3dioxygenase), which allows an easy identification of plasmidcontaining clones by spraying the plates with catechol. This turns the cells yellow because of the production of hydroxymuconic semialdehyde (57). As shown in Fig. 7, the growth of this strain for Ͼ100 generations in LB medium without any antibiotic (i.e. no external selection for pFBT22 maintenance) always produced 100% yellow colonies when spread with catechol. On the contrary, when the same plasmid was placed in the single deletion ⌬hupB mutant, there was a progressive loss of pFBT22 when cells were grown in nonselective medium. One round of plating without kanamycin sufficed to caused the spontaneous loss of Ն80% of the plasmid in viable cells. As little as eight generations later, only Յ10% of cells bearing the single ⌬hupB deletion maintained the hupB ϩ construct. The addition of 50 mg/ml kanamycin to the cultures ensured full plasmid retention under every con-dition tested (Fig. 7). These results suggested that plasmid pFBT22 could not be cured spontaneously from the hupB/hupN double-deletion mutant, because at least one of the two genes must be kept functional for viability. DISCUSSION All of the results reported in this article support the idea that HupB and HupN are two distinct proteins, albeit structurally related, which are necessary and sufficient by themselves to provide essential factor-mediated but nonspecific DNA bending functions in P. putida. That no other factors can take over such a function is suggested by the nonviability of hupB/hupN doubledeletion (Fig. 7) and, to a minor extent, by the virtual absence of any additional DNA bending activity in cell extracts (Fig. 1). The phylogenetic tree in Fig. 8 shows a prediction of the relationship between the HupB and HupN proteins from both pseudomonads and other eubacterial HU-like proteins. The most salient feature of such an analysis is that HupN clearly branches away from the HupA/HupB-type proteins of ␥-proteobacteria. In fact, it has been generally believed that the existence of two genes encoding HU proteins was exclusive to enteric bacteria (21). This view appeared to be true for P. aeruginosa, for which only one gene (hupB) and its corresponding gene product had been assigned to the HU function (50). The N-terminal sequence of the HU protein from P. aeruginosa was determined and found to be a HupB homodimer (58,59). However, a hupN analogue very close to the P. putida gene does exist in the P. aeruginosa chromosome (49) but was detected neither biochemically nor genetically in this genus; whether this reflects a genuine lack of expression or a limitation of the techniques employed is uncertain. On the contrary, our results demonstrate unequivocally that both HupB and HupN are expressed and are perfectly active in P. putida.
The presence of two genes for the HU function in some  7. Comparison of the loss of the hupB ؉ /xylE ؉ plasmid pFBT22 from P. putida ⌬hupB and ⌬hupB⌬hupN. Cultures of strains P. putida ⌬hupB (pFBT22) and P. putida ⌬hupB⌬hupN (pFTB22) grown in the presence of kanamycin to ensure plasmid retention were used to inoculate fresh media with or without the antibiotic. Cultures were then plated on either selective or nonselective medium immediately or after eight generations. After overnight growth, the plates were sprayed with 1% catechol to reveal colonies carrying xylE gene and thus the pFBT22. The bars represent the percentage of plasmid-containing clones detected in each case. Solid bars, P. putida ⌬hupB (pFBT22); hatched bars, P. putida ⌬hupB ⌬hupN (pFTB22). bacterial genera and only one in others is intriguing. Unlike the IHF protein, which necessarily requires the assembly of a heterodimer to function (60,61), each of the two HU proteins (hupA and hupB) of E. coli seem to be able to function independently, although they do form functional heterodimers as well (15,16). It has been claimed (15) that the subunit composition of HU from E. coli changes during growth and that this may reflect a certain specialization of the functions of each protein species. We cannot distinguish, however, any gross phenotypic differences between P. putida cells expressing either one of the Hup proteins or both of them. Although we did not examine the formation of heterodimers between the two proteins, it seems clear as well that each polypeptide fulfills by itself (i.e. homodimers, Fig. 3B) every function assigned to HU or HU-like proteins (Fig. 8), suggesting that if the formation of heterodimers between them does occur, it is irrelevant for cell physiology.
A second aspect of the redundancy of HU activities is that double hupA hupB mutants of E. coli (i.e. lacking any HUrelated function) do exist that are viable. In our studies and those of others (17,18), however, such mutants survive very poorly, develop multiple morphologies, and are quite unpredictable, surely because of the accumulation of compensatory mutations. Although it could be argued that such mutations may originate in a decrease of DNA-binding specificity of IHF that takes over some HU functions, the fact is that HU Ϫ /IHF Ϫ exist as well (62), an issue that deserves further studies. Under the conditions of the experiment shown in Fig. 7, which is short enough so as not to allow the overgrowth of bacteria bearing compensatory mutations, it is clear that the functionality of at least one HU protein is essential for gross viability in P. putida.
Although we could detect none but subtle differences in vivo and in vitro between HupB and HupN in the laboratory tests described in this article, we cannot rule out that small changes can make a difference when P. putida thrives in its natural habitats. Ecological fitness under the tougher environmental conditions that predominate in the polluted sites that P. putida colonizes depend to a large extent on an ability to integrate different environmental signals in the outcome of catabolic promoters (63). Such signals frequently involve DNA bending as a signal transmission mechanism (1,2), and thus small differences in expression, like those shown in Fig. 6, may turn out to be significant. With the data at hand, however, the simplest explanation for the presence of two proteins fully competent in nonspecific DNA bending in some bacterial species (including P. putida) is that one simply acts as a backup for the other. The redundancy of HU factors in the same cell would thus reveal the evolutionary importance of such a function, not a specialization in a particular role in the physiology of the bacteria.