Sugar transport by the marine chitinolytic bacterium Vibrio furnissii. Molecular cloning and analysis of the glucose and N-acetylglucosamine permeases.

Chitin catabolism by the marine bacterium Vibrio furnissii involves chemotaxis to and transport of N-acetyl-D-glucosamine (GlcNAc) and D-glucose. We report the properties of the respective permeases that complemented E. coli Glc− Man− mutants. Although the V. furnissii Glc-specific permease (55,941 Da) shares 38% identity with E. coli IIGlc (ptsG), it is 67% identical to MalX of the E. coli maltose operon (Reidl, J., and Boos, W. (1991) J. Bacteriol. 173, 4862-4876). An adjacent open reading frame encodes a protein with 52% identity to E. coli MalY. Glc phosphorylation requires only V. furnissii MalX and the accessory phosphoenolpyruvate:glycose phosphotransferase system proteins. The V. furnissii equivalent of IIGlc was not found in the 25,000 transformants screened. The GlcNAc/Glc-specific permease (52,894 Da) shares 47% identity with the N-terminal, hydrophobic domain of E. coli IINag, but is unique among IINag proteins in that it lacks the C-terminal domain and thus requires IIIGlc for sugar fermentation in vivo and phosphorylation in vitro. While there are similarities between the phosphoenolpyruvate:glycose phosphotransferase system of V. furnissii and enteric bacteria, the differences may be important for survival of V. furnissii in the marine environment.

Chitin, a ␤,134-linked polymer of N-acetylglucosamine (GlcNAc), is one of the most abundant organic compounds in nature. Huge quantities of this highly insoluble polysaccharide are turned over annually in the aquatic biosphere, and marine bacteria such as Vibrios are major biological components of this ecologically indispensable process.
In previous papers (1)(2)(3)(4), we reported that chitin degradation by one such organism, Vibrio furnissii, is extraordinarily complex, involving multiple signal transduction systems and many proteins. We proposed (4) that the means by which these cells locate chitin-containing organisms is by chemotaxis to components of the hemolymph and molt fluids including glucose, trehalose, GlcNAc and chitin oligosaccharides; all of these compounds are potent chemoattractants for V. furnissii.
Degradation of chitin by V. furnissii is initiated by chitinases that hydrolyze it to soluble oligosaccharides, which are further hydrolyzed in the periplasmic space to GlcNAc and (GlcNAc) 2 . Finally, each of these catabolites is taken up by specific transporters and converted intracellularly to Fru-6-P, NH 3 , and acetate. The cytoplasmic membrane permeases are, therefore, essential components of the chitin catabolic cascade. The disaccharide permease is described in an accompanying paper (5), while the present report is concerned with the GlcNAc and Glc chemoreceptors/permeases.
We have presented evidence that the uptake and phosphorylation of GlcNAc, Glc, and Man in V. furnissii is mediated by the bacterial phosphoenolpyruvate:glycose phosphotransferase system (PTS). 1 While the complete PTS is required for both chemotaxis and transport, the sugar chemoreceptors/translocators are the membrane-associated Enzyme II complexes (for reviews see Refs. 6 and 7). 2 The Enzyme II complexes often show overlapping substrate specificities. For example, in Escherichia coli and Salmonella typhimurium, GlcNAc is recognized and taken up by II Nag , and Glc by the protein pair II Glc /III Glc , but both substrates are also taken up by the less specific, more complex mannose system, II Man . For this reason, definitive characterization of the specificities of the V. furnissii Enzyme II complexes requires that they be separated from one another.
In this and accompanying papers (8,9), we describe the molecular cloning and characterization of genes and gene products from V. furnissii into E. coli that generate GlcNAc in the periplasmic space, and recognize and transport GlcNAc, Glc, and Man. The V. furnissii proteins are physiologically active in E. coli, and the deduced amino acid sequences of the proteins are similar to the corresponding proteins from the enteric bacteria (10), but they also show interesting and significant differences. 1 The abbreviations used are: PTS, phosphoenolpyruvate:glycose phosphotransferase system; MeGlc, methyl ␣-glucoside; cAMP-CAP, cyclic AMP-catabolite activator protein complex; kb, kilobase pair(s); bp, base pair(s); ORF, open reading frame; PCR, polymerase chain reaction; PAGE, polyacrylamide gel electrophoresis. 2 The nomenclature of the Enzymes II is complicated by the different variants that have been identified, and two systems of nomenclature have been proposed (6,56). In general, the sugar-specific, membraneassociated polypeptides are termed Enzyme II complexes. In the nomenclature used in this and the accompanying reports, the sugar receptors are designated "II" or Enzyme II if the integral membrane protein complex comprises a single polypeptide chain, such as II Glc , or II Nag . When the complex contains more than one integral membrane protein, the sugar receptor is called "IIB", such as II-B Man . The sugarspecific, soluble phosphotransfer proteins are called "III" or, if they are intrinsic membrane proteins, are designated "IIA". The terms IIA and III are therefore functionally equivalent.  (16). For sugar phosphorylation assays, E. coli strains were grown at 37°C in M9 medium (16) supplemented with 0.5% carbon source (lactate, Glc, GlcNAc, or Man) and 0.1% casamino acids (Difco). Mac-Conkey agar without lactose (Difco) was prepared with 1% sugar. Antibiotics were used at the following concentrations: ampicillin, 50 g/ml; kanamycin, 12.5 g/ml; chloramphenicol, 12.5 g/ml; and tetracycline, 15 g/ml. For growth experiments in LR2-175, Medium A (17) was supplemented with Met (20 g/ml), His (22 g/ml), Arg (22 g/ml), and the appropriate antibiotic.

Biochemicals and Molecular Biology Reagents-Restriction
Construction and Screening of V. furnissii DNA Libraries-Restriction enzyme digestions, ligations, transformations, and agarose gel electrophoresis were carried out as described by Maniatis (16). Genomic DNA was prepared from V. furnissii 1514 by lysozyme-SDS treatment (18). V. furnissii genomic DNA, digested to completion with BamHI and HindIII, was ligated into similarly digested, dephosphorylated pBR322. The ligation mixtures were transformed into E. coli HB101 and amplified by the method of Pulleyblank (19). The resulting recombinant plasmids were used to transform E. coli SR423 and SR425, screening for Glc ϩ Amp R colonies on MacConkey Glc/ampicillin agar. The plasmids pBR322 (vector) and pCB20 (carrying E. coli ptsG; Ref. 20) were used as negative and positive controls, respectively, for Glc fermentation in these studies. More than 12,000 transformants were screened from each library. Each Glc ϩ transformant was tested for the presence of a recombinant plasmid and for the ability of that plasmid to retransform SR423 or SR425 to Glc ϩ . Plasmids were isolated by the method of Pulleyblank (19).
DNA Sequencing and Analysis-Nucleotide sequencing was performed on double-stranded plasmid templates by the dideoxy chaintermination method (21) with the Sequenase kit (U. S. Biochemical Corp.) and [␣-35 S]dATP (1000 Ci/mmol) as described by the manufacturer. Sequencing primers (clockwise EcoRI 17-mer and counterclockwise HindIII 17-mer) were obtained from Life Technologies, Inc. and Pharmacia Biotech Inc. The remaining primers were designed and synthesized as needed while sequencing the cloned V. furnissii DNA. The primers were synthesized on an Applied Biosystems model 381AD oligonucleotide synthesizer, deprotected according to the manufacturer's instructions, and purified on C 18 Sep-Pak columns (Millipore, Milford, MA) as suggested by the manufacturer. The sequencing reaction products were separated on 8% polyacrylamide, 8.3 M urea gels in 3:1 TBE buffer (22). The analysis of DNA and amino acid sequences was carried out with the GCG sequence analysis package (Version 7, Genetics Computer Group, Madison, WI). The data bases searched for nucleotide and amino acid sequence similarities were: GenBank™ Release 79 and Swiss-Protein Release 26. Predicted amino acid sequences were aligned using the CLUSTAL W version 1.5; similarities were shown by shading using BOXSHADE/DOS 2.7, kindly performed by Michael Cleveland.
PCR Amplification of DNA-The regions corresponding to nucleotides 7-2284 and 2218 -3592 of p3B1S were amplified from V. furnissii genomic DNA by the polymerase chain reaction. The primers were synthesized and purified as described above. The primers consisted of the following nucleotides of the p3B1S sequence: for amplification of the malX/ptsG-like gene, nucleotides 7-19 with a 5Ј BamHI site and nucleotides 2284 -2266 with a 5Ј PstI site; for amplification of the malY-like gene, nucleotides 2188 -2210 with a 5Ј BamHI site and nucleotides 3592-3574 with a 5Ј PstI site. Amplifications were conducted in a Perkin-Elmer DNA thermal cycler in a reaction volume of 100 l con-taining 1x PCR buffer (U. S. Biochemical Corp.), 0.2 mM each deoxynucleotide triphosphate, 4 M each primer, 5 ng of V. furnissii genomic DNA, and 2.5 units of Taq DNA polymerase (Ampli-Taq, Perkin-Elmer), for 30 cycles of amplification (94°C, 30 s; 48°C, 1 min; 72°C, 2 min 30 s). The PCR products were ethanol precipitated, digested with BamHI and PstI, and were ligated into similarly digested pUC18. The resulting clones, the plasmids 3B1S-ORF1 and 3B1S-ORF2, were sequenced to confirm that no sequence errors were introduced during amplification.
Assays for PTS Enzymes II in Cell Membranes-E. coli transformants harboring plasmid clones of V. furnissii Glc permeases or control plasmids were grown to A 500 ϭ 0.8 as described above. The chilled cultures were harvested by centrifugation at 5400 ϫ g for 15 min. The cell pellets were washed twice and resuspended in 0.2 ml buffer (50 mM Tris-Cl, pH 7.5, at 25°C, 10 mM MgCl 2 )/g wet weight of cells. Cytoplasmic membranes were prepared and assayed for sugar phosphorylation as described by Waygood et al. (24). Incubation mixtures contained the following components in volumes of 50 l: 50 mM Tris-Cl pH 8 (25°C); 1 mM dithiothreitol; 10 mM phosphoenolpyruvate; 5 mM MgCl 2 ; 10 mM KF; 2 units of E. coli Enzyme I (25), 5 M HPr (26); 5 M E. coli III Glc (23); 1 mM [ 14 C]Glc (450 dpm/nmol) or [ 14 C]GlcNAc (264 dpm/nmol); and 0.5-5 g of membrane protein. Incubations were conducted for 30 min at 37°C, and the reactions were stopped at 100°C for 2 min. Sugar phosphate was separated on anion exchange columns and counted as described (24). Specific activities are expressed as nanomoles of sugar phosphate formed min Ϫ1 mg of membrane protein Ϫ1 . Protein was measured by the SDS-Lowry method (27).
In Vitro Transcription/Translation of Plasmid-Encoded Proteins-The plasmid-encoded gene products were analyzed with a Prokaryotic DNA-Directed Transcription/Translation System (Amersham) using [ 3 H]leucine (120 Ci/mmol) and 1-3 g of plasmid DNA. The reaction products were separated by SDS-PAGE (10% acrylamide) (8), and the labeled polypeptides were detected by fluorography (Enlightning, DuPont NEN).

RESULTS
Complementation of E. coli Glc Ϫ Man Ϫ Mutants-Two libraries of V. furnissii genomic DNA, in the BamHI and HindIII sites of pBR322, were screened for plasmids that could complement the Glc fermentation defect in E. coli strains SR423 or SR425. Five plasmids carrying V. furnissii DNA inserts of different molecular sizes were represented among the Glc ϩ Amp R transformants. DNA hybridizations (data not shown) demonstrated that unique V. furnissii fragments were carried by four of these plasmids. Since no hybridization was observed between the cloned DNA fragments and E. coli genomic DNA (data not shown), these experiments also conclusively demonstrated that the plasmids did not contain E. coli genomic DNA, which may have occurred by recombination with the host strain during preparation or screening of the V. furnissii libraries. To determine the minimum size necessary for complementation of Glc fermentation, restriction fragments of these four Glc ϩ plasmids were subcloned into pBR322, and screened for those that allowed Glc fermentation in SR423 or SR425. The restriction maps of two of the four clones (p3B1 and p5B16), and their subclones (p3B1S and p5B16H), are presented in Fig. 1.
Phenotypes of E. coli Transformants Carrying Glc ϩ Plasmids-Various E. coli pts mutants were transformed with p3B1 , and control plas-mids. These transformants were tested for fermentation of PTS and non-PTS sugars ( Table I). The plasmid p3B1 and its subclone, p3B1S, carried V. furnissii DNA fragments that allowed the three ptsG ptsM mutants to ferment Glc. Likewise, the plasmid p5B16 and its subclone, p5B16H, restored Glc fermentation to these strains. These four plasmids also complemented the Glc fermentation defect in the strains ZSC113 (ptsG ptsM glk; Ref. 28) and JM1100 (ptsG ptsM fruA; Ref. 29) (data not shown). In addition, the plasmids p5B16 and p5B16H restored GlcNAc fermentation to the nagE (Enzyme II Nag ) mutant LR2-175. Complementation of Gal, Man, and Fru fermentation was not observed by transformants harboring these plasmids when tested in the appropriate mutant strains (Table I).
The fermentation patterns of the transformants suggested that p3B1 and p3B1S might encode the V. furnissii Enzyme II Glc , and p5B16 and p5B16H the V. furnissii Enzyme II Nag . The clones were therefore tested in ptsI and ptsI-crr deletion mutants to confirm that they encoded membrane permeases of the PTS. PtsI and ptsI-crr strains should remain pleiotropically negative for PTS sugar fermentation when transformed by plasmids encoding PTS permeases. The results (Table I) showed that p3B1 and p3B1S require Enzyme I and III Glc for fermentation, as does E. coli II Glc . Interestingly, the putative GlcNAc permease encoded by p5B16 and p5B16H required both Enzyme I and III Glc for fermentation, in contrast to its E. coli counterpart (encoded by pKP1.1) which does not require III Glc . Negative fermentation results were obtained by transformation of S. typhimurium SB2950 (17), which carries a ptsHIcrr deletion (data not shown), confirming the requirement for these three soluble PTS proteins by the cloned permeases.
E. coli strain LR2-175 transformants carrying p3B1S and p5B16H grow on minimal media with Glc as the sole carbon source with generation times of 65 to 75 min (data not shown).
Enzyme II Activity in Membranes from E. coli Transformants-PTSdependent sugar phosphorylation was measured with membranes prepared from E. coli SR425 and LR2-175 transformants harboring p3B1S, p5B16H, and control plasmids. (p3B1 and p5B16 gave essentially the same results as p3B1S and p5B16H, respectively.) Table II summarizes these results. GlcNAc phosphorylation was not measured in SR425 transformants because of the wild-type nagE allele in this strain.
Sugar phosphorylation by control membranes (from pBR322 transformants) was very low. Membranes from p3B1S transformants phosphorylated Glc at 20 -60-fold rates greater than the controls, thereby explaining the Glc ϩ phenotype. In the one case tested, transformants of LR2-175, the membranes also showed an increased rate of phosphorylation of GlcNAc over controls, but this rate is comparable to that of LR2-175 transformants expressing cloned E. coli II Glc from the plasmid pCB20 (data not shown). Thus, the apparent in vitro nonspecificity of the II Glc system either does not function in vivo, or the rate is insufficient to permit growth on GlcNAc.
Membranes from p5B16H transformants phosphorylated  both Glc and GlcNAc, and in these cases, the increases ranged from 300-to 1000-fold over control sugar phosphorylation rates. The rate-limiting proteins in the assays were the membranes, and growth of the transformant on GlcNAc yielded an increased specific activity compared to membranes from Glcgrown cells. The results therefore agreed with those expected from the phenotypes of the transformants. However, there is a quantitative inconsistency in Table II. The ratio of GlcNAc to Glc phosphorylation varied with the sugar used for cell growth, although the ratio should be constant for a single Enzyme II. This quantitative effect may result from the difficulties associated with measuring membrane-bound II Nag activities. Also, bacterial membranes change in composition depending on the carbon source used for growth (30). Perhaps the V. furnissii II Nag protein inserts differently into E. coli membranes of dis-similar composition, giving rise to kinetic differences with respect to its sugar substrates.
Other radioactive sugars were tested as potential substrates in both systems. There was little to no detectable phosphorylation of the following labeled sugars (1 mM) by membranes from transformants of LR2-175: fructose, mannose, sucrose, 2-deoxyglucose, and MeGlc. The latter result was surprising, since to this point the results indicated that p3B1S encoded the V. furnissii equivalent of E. coli II Glc , and MeGlc is an excellent substrate of the latter (20,31).
coli equivalent of II Nag . The latter comprises an N-terminal membrane-bound domain and a cytoplasmic C-terminal domain, which is the functional equivalent of III Glc and accepts the phosphoryl group directly from phospho-HPr.
The requirement for III Glc , both in vitro and in vivo (Tables  I and II) were clarified by the experiments described below.
In Vitro Transcription/Translation of Plasmid-encoded Proteins-To further characterize the cloned genes from V. furnissii, the plasmid-encoded gene products were analyzed by in vitro transcription/translation. The plasmids p3B1S, p5B16H, and pBR322 were translated in a prokaryotic cell-free system and the labeled polypeptides were separated by SDS-PAGE as described under "Materials and Methods." The plasmid p3B1S produced labeled products with apparent molecular masses of 53 and 44.5 kDa as well as the vector-encoded ␤-lactamase (Fig. 2). A strongly labeled product with an apparent molecular mass of 49.5 kDa and ␤-lactamase were produced from the plasmid p5B16H (it should be noted that the background varies with this in vitro system). The isolates p3B1 and p5B16, carrying larger V. furnissii genomic DNA fragments, did not give clear results with this system, making it difficult to determine the number of gene products they express.
Nucleotide Sequencing and Deduced Amino Acid Sequence Analysis of p3B1S-The V. furnissii DNA fragment carried by p3B1S was sequenced (21) by the dideoxy chain-termination method (GenBank accession number U65013). The largest open reading frame in the 3997-bp SalI fragment (ORF 1, Fig.  3) extended from nucleotide 647 to 2218 and was preceded by a potential ribosomal binding site (32) centered 10 nucleotides upstream from the initiation codon. A second open reading frame was found between nucleotides 2273 and 3472 (ORF 2, Fig. 3). A potential promoter (positions 510 -534) was also identified by its similarity to the E. coli consensus sequence (33). However, no sequences with strong identity to the consensus CAP binding sequence (34) or a region with the structural characteristics of a -independent terminator (35) were found.
The deduced amino acid sequence of ORF 1 predicted a protein of 523 amino acids with a calculated molecular mass of 55,941 Da. This molecular mass is similar to the M r of the larger in vitro transcription/translation product of p3B1S (Fig.  2). Similarity searches (36, 37) of protein and nucleic acid data bases revealed homology between this V. furnissii protein and E. coli II Glc (38% identity), E. coli II Nag (34% identity), and Bacillus subtilis II Glc (37% identity). Surprisingly, the protein showed the greatest identity with the recently reported (10) E. coli MalX (67%). Its hydropathic profile also resembles that of E. coli MalX (data not shown), and its hydrophobicity is reflected in its amino acid composition (data not shown). An alignment of V. furnissii MalX with E. coli MalX is shown in Fig. 4. These sequences are well conserved, with 33% identical residues between the three proteins. Previous studies have suggested functional roles for several regions of high sequence identity among the PTS permeases (reviewed in Refs. 6 and 7). A putative site of phosphorylation (His-211 in E. coli II Glc ) was identified among the PTS membrane permeases by sequence comparison (6). Recent evidence obtained by mass spectrometry and biochemical analysis of phosphorylated peptides indicates that Cys-421 of E. coli II Glc is phosphorylated (38,39). These residues are conserved among E. coli II Glc , MalX, and the V. furnissii Glc permease. Additionally, four of the six residues of E. coli II Glc recently described to be important for sugar translocation (Met-17, Gly-149, Lys-150, Ser-157, His-339, and Asp-343; Ref. 40) are conserved in the V. furnissii homolog.
Translation of ORF 2 from p3B1S predicted a 399-amino acid protein with a molecular mass of 45,359 Da. This molecular mass compares favorably with that of the smaller polypeptide obtained by in vitro transcription/translation (Fig. 2). Since ORF 1 of p3B1S is similar to E. coli MalX, it is not surprising that the deduced amino acid sequence of ORF 2 shared 52% identity with E. coli MalY (10). The deduced sequence of the V. furnissii MalY-like gene product is aligned with the E. coli MalY sequence in Fig. 5. E. coli MalY has recently been identified as a ␤-cystathionase, an essential enzyme in methionine catabolism (41   providing even stronger evidence that the adjacent gene is indeed V. furnissii malX and not ptsG. It should be noted that V. furnissii grows very well on maltose and exhibits chemotaxis to this sugar (4).
Since two open reading frames were discovered in the V. furnissii SalI fragment in p3B1S, and both were expressed in vitro, we wished to determine whether both gene products were required for Glc transport and phosphorylation in vivo. Each reading frame was amplified by PCR with the addition of BamHI and PstI sites, and was ligated into pUC18. As shown in Table I, 3B1S-ORF1 alone was sufficient for Glc fermentation. A ptsG ptsM glk strain, ZSC113 (28), was also able to ferment Glc when transformed with p3B1S or 3B1S-ORF1, but not when transformed with 3B1S-ORF2 (data not shown). Therefore, glucokinase is not required for Glc fermentation in cells transformed with V. furnissii MalX. SR425 transformants carrying p3B1S or 3B1S-ORF1 grew well on Glc as the sole carbon source (doubling time, 65 min), whereas transformants of 3B1S-ORF2 did not (no detectable growth after 3 h).
Neither the deduced amino acid sequence nor any portion of the nucleotide sequence of p3B1S showed similarity to E. coli III Glc , a result consistent with the in vivo and in vitro requirements of p3B1S transformants for III Glc (Tables I and II).

Nucleotide Sequencing and Deduced Amino Acid Sequence
Analysis of p5B16H-The 2.2-kb HindIII fragment of p5B16H was also sequenced (GenBank accession no. U65014). The major open reading frame, indicated in Fig. 6, begins at nucleotide 140, terminates at nucleotide 1630, and is preceded by sequences similar to the E. coli ribosomal binding site (32) and Ϫ35 and Ϫ10 promoter elements (33). The predicted polypeptide is 496 amino acid residues in length with a molecular mass of 52,894 Da. This is similar to the apparent molecular mass of the major in vitro transcription/translation product (Fig. 2). Hydropathic analysis (42) showed the permease to be primarily hydrophobic, resembling the hydrophobic domain of E. coli II Nag . The predicted amino acid composition of the permease also showed a high proportion of hydrophobic amino acids (data not shown). FASTA and BLAST searches (36,37) of the Gen-Bank and Swiss Protein data bases revealed 48% identity (67% similarity) between E. coli II Nag (13) and the permease encoded by p5B16H. Other sequences showing similarity were E. coli II Glc (37%) and B. subtilis II Glc (41%). The plasmids p5B16 and p5B16H therefore carry the V. furnissii homolog of nagE. An alignment of the V. furnissii and E. coli GlcNAc permeases is presented in Fig. 7. Several extensive regions of the two enzymes are highly conserved (residues 256 -273, 295-330, and 424 -445, for example). However, the C-terminal, hydrophilic III Glc -like domain of E. coli II Nag is absent from the V. furnissii protein. The hydropathic profile of V. furnissii II Nag (data not shown), the transformant phenotypes (Table I), and the in vitro sugar phosphorylation data (Table II) are consistent with this. The residues surrounding the His phosphorylation site proposed by Meadow et al. (6) are identical in 9 out of 13 positions in the V. furnissii and E. coli GlcNAc permeases. A sequence, IDACITRL (residues 432-439), homologous to the phosphorylation site of E. coli II Glc identified by Meins et al. (43), is also present in the V. furnissii GlcNAc permease. No similarity to E. coli III Glc or the E. coli III Glc -like domain of E. coli II Nag was detected within the 2205-nucleotide V. furnissii DNA fragment in p5B16H.
The sequences upstream of the initiation codon and beyond the termination codon of V. furnissii nagE were analyzed for potential regulatory signals. Several potential promoter sequences were detected, and the one with the greatest identity to the consensus sequence for E. coli promoters (33) is highlighted in Fig. 6. The E. coli nag regulon is coordinately controlled by repressor (NagC) and cAMP-CAP binding to an operator located between nagE and nagBACD (15,44,45). The operator overlaps the Ϫ35 region of the nagE promoter (45). A site was identified upstream of V. furnissii nagE (Fig. 6), which is identical in eight of nine positions to the consensus sequence for NagC binding (TAXTTTXXXXTXCXAA; Ref. 45), and which overlaps the Ϫ35 region of the putative promoter sequence. No strong CAP binding sites were detected. A putative -independent terminator (35) with a GC-rich region of dyad symmetry was found within 100 bp of the nagE termination codon (Fig. 6). DISCUSSION The PTS is widely distributed in the eubacteria, including Gram-negative, facultative anaerobes such as E. coli and S. typhimurium. The families Enterobacteriaceae and Vibrionaceae are closely related, and Vibrio is one of the most widespread of the bacterial genera (46). The PTS was found in all Vibrio species tested (47)(48)(49). Kubota et al. (50,51) reported that Glc, trehalose, Fru, Man, and mannitol are PTS sugars in Vibrio parahaemolyticus, and separated four fractions from extracts (by gel filtration) that corresponded in their activities to Enzyme I, HPr, II Glc , and III Glc . Sucrose is a PTS sugar in Vibrio alginolyticus (52), and the Scr operon has been cloned and sequenced.
In V. furnissii, chemotaxis to and the transport of GlcNAc and Glc are part of a complex series of physiological events that we have designated the chitin catabolic cascade (1)(2)(3)(4). We have previously reported (4) that in this organism, the following are PTS sugars: GlcNAc, Glc, mannitol, mannose, fructose, trehalose, and sucrose. Each of these carbohydrates is a potent  (45) are indicated by asterisks (*) above the sequence. The putative ribosomal binding site is indicated with double underlines. Large arrows and bold type denote a region of dyad symmetry at the 3Ј end of the coding sequence, which is part of a potential -independent terminator-like structure (35). chemoattractant except fructose. The non-PTS sugars include glycerol, galactose, maltose, and N,NЈ-diacetylchitobiose. Galactose is a weak chemoattractant (unlike in E. coli), whereas both disaccharides are potent attractants.
The genes that encode the Glc and GlcNAc Enzyme II complexes and the proteins that serve both as the chemoreceptors and translocators of their respective sugar substrates are the subjects of this paper.
The soluble E. coli PTS proteins (Enzyme I, HPr, and III Glc ) cross-reacted with V. furnissii membranes in sugar phosphorylation assays in vitro (47). Therefore, it seemed likely that the V. furnissii Enzymes II would be functionally expressed in E. coli transformants, and based on this rationale, Glc Ϫ Man Ϫ E. coli strains SR423 and SR425 (11) were used for screening BamHI and HindIII libraries (in the plasmid vector pBR322) of V. furnissii genomic DNA. Four plasmids carrying nonhomologous DNA fragments each permitted the mutants to ferment Glc, and each isolate encoded a permease capable of transporting Glc. The accompanying article (8) reports a PTS-dependent Man complex; the second Glc transporter, which is PTS-independent, will be reported elsewhere. The present paper describes a plasmid that encodes a V. furnissii Glc-specific PTS permease, and one that encodes a V. furnissii GlcNAc/Glc permease.
We had expected that some of the Glc ϩ E. coli transformant colonies would express the V. furnissii homolog of II Glc , which is widespread in enterobacteria. It functions in concert with the cytoplasmic protein III Glc (also called IIA Glc ) 2 , and translocates/phosphorylates Glc and the analogue MeGlc. II Glc has also been assigned an important role in regulating the "glucose effect" (6,7).
Based upon complementation tests (Table I) and in vitro sugar phosphorylation assays, it appeared that the plasmid p3B1S encoded V. furnissii II Glc , although it showed little activity with MeGlc or with other Glc analogues that were tested. Searches of protein sequence data bases also suggested this interpretation, since one V. furnissii open reading frame in p3B1S is 38% identical to E. coli ptsG. But to our surprise, the same reading frame is 67% identical to E. coli malX.
The maltose regulon (53) Fig. 3) shares 47% identity with the N-terminal portion of E. coli MalI. Thus, the sequence data provide strong evidence that the gene cloned from V. furnissii is the analogue of malX, not ptsG.
The roles of MalX and MalY in regulation of the E. coli Mal system have been investigated (10). Our results are consistent with the previous study in that the II Glc -like (MalX) gene product does not require MalY for Glc fermentation (Table I).
The authors suggested (based on phenotypic behavior of mutants) that the malX gene product normally transports Glc (and maltose) by facilitated diffusion, but that under certain conditions, it could phosphorylate Glc; biochemical studies of MalX PTS activity were not reported. We show here that this protein functions like other Enzymes II of the PTS in Glc translocation (Tables I and II); it requires Enzyme I and III Glc (but not glucokinase) for Glc fermentation in vivo, and it phosphorylates Glc in vitro. While it is possible that another glucose permease may be the major PTS-driven Glc transporter, efforts to identify the V. furnissii homolog of E. coli ptsG were unsuccessful, although a total of 25,000 transformants were screened from two genomic libraries constructed with different restriction enzymes. Construction of an E. coli ptsG deletion will be required in order to use colony hybridization as an alternative method for screening the V. furnissii genomic banks. If this method also gives negative results, it is likely that V. furnissii does not contain the E. coli equivalent of ptsG, but that the gene products described in this and the accompanying papers (5,8)  selected Glc-fermenting transformants on rich medium, and (b) membranes from the transformants rapidly phosphorylated Glc, close to the same rate as GlcNAc (Table II). Previous reports (54) have shown that E. coli II Nag , when transformed into a ptsG ptsM crr nagE background, allows only slow growth on Glc and there are no reports that Glc is rapidly translocated by E. coli II Nag .
Thus, V. furnissii II Nag is apparently unique in its ability to translocate both Glc and GlcNAc. This permease also appears to be the sole transporter of GlcNAc; a V. furnissii mutant defective in this protein is unable to take up GlcNAc or to show taxis to this compound, although the cells behave normally on the disaccharide (GlcNAc) 2 (4).
The plasmid p5B16H contains the V. furnissii homolog of nagE, the gene that encodes the GlcNAc permease, II Nag . However, V. furnissii nagE is unique compared to all such genes that have been reported (6,7). (a) The size of the in vitro transcription/translation product from p5B16H (M r ϭ 49,500) compares favorably with the deduced amino acid sequence (52,594 Da) and is substantially smaller than the 69 kDa of E. coli II Nag . (b) The deduced amino acid sequence of the V. furnissii gene shows 47% identity (67% similarity) to the N-terminal domain of E. coli II Nag (Fig. 7). The C-terminal III Glc -like domain of E. coli II Nag , however, is absent from the V. furnissii permease, and no crr-like sequence was found in the p5B16H nucleotide sequence. (c) Unlike the other known proteins encoded by the nagE gene from other species, all of which interact directly with phospho-HPr, the V. furnissii GlcNAc permease requires a III Glc equivalent both in vivo (Table I) and in vitro (Table II).
We do not yet know whether V. furnissii nagE functions in the cell with a specific (previously undescribed) III Nag , or whether it functions with III Glc in V. furnissii as it does in the E. coli transformants and in vitro. If there is a gene that encodes a specific III Nag , it was not found in p5B16 and therefore is not linked to nagE on the V. furnissii chromosome.
The E. coli nag operon is inducible by growth on GlcNAc and is stimulated by cAMP-CAP (15,45,55), and the V. furnissii nag operon may be similarly regulated. E. coli LR2-175 transformants carrying p5B16H, grown on GlcNAc, gave 3 times greater Enzyme II Nag activity than Glc-grown cells (Table II). (The effect of cAMP-CAP on expression of the cloned genes was not tested.) Qualitatively similar results were obtained in sugar phosphorylation experiments with membranes (Table II) and toluene-permeabilized cells of wild-type V. furnissii (4). Because of the observed inducibility of GlcNAc phosphorylation, we believe that the V. furnissii nag promoter is present on p5B16 and p5B16H.
It is noteworthy (Table I) that the plasmid p5B16 complemented E. coli nagA mutants, defective in GlcNAc-6-P deacetylase, but not a nagB strain, defective in GlcNH 2 -6-P deaminase. This result indicates genetic linkage of V. furnissii nagE and nagA, analogous to the gene order of the E. coli nag operon (13); it seems likely that the remaining V. furnissii nag genes are in close proximity to the ones that we have cloned. Preliminary sequencing results show that p5B16 carries a nagC-like gene.
Certain functional domains have been defined in the Enzymes II, assignments that are based primarily upon sequence alignments and a few biochemical studies (6,7). These domains seem to be conserved among the sugar permeases described here. For example, of the six residues of E. coli II Glc recently reported to be important in sugar translocation but not phosphorylation (40), four are conserved in the V. furnissii ManX/ Glc permease and five are conserved in V. furnissii II Nag . The region surrounding the residue His-211 of E. coli II Glc , a puta-tive phosphorylation site (6), is very similar in the V. furnissii MalX/Glc, and Glc/GlcNAc permeases. Cysteine residues corresponding to the phosphorylated residue Cys421 of E. coli II Glc (38,39) are also present in the V. furnissii permeases, and the surrounding residues are highly conserved.
All II Nag proteins thus far reported (except the one described here) contain both the N-and C-terminal domains (7), and phosphorylate GlcNAc independently of the protein III Glc (also called IIA Glc ). V. furnissii II Nag therefore appears to be an anomaly. But is it? III Glc plays a critical role in regulating PTS and non-PTS transporters, and non-PTS enzymes such as glycerol kinase (6, 7) and adenylate cyclase. These regulatory roles in large part explain the diauxic effect when E. coli is presented with two sugars (57). Since GlcNAc appears to be an important, possibly a preferred nutrient for V. furnissii, perhaps it is essential that its metabolism involves III Glc . In this way, V. furnissii may continue to catabolize chitin, even in the presence of other potential nutrients.
In sum, two PTS-driven V. furnissii Glc transporters are described here. One is closely related to E. coli MalX, and the other to the E. coli GlcNAc permease II Nag (encoded by nagE), but in each case, the V. furnissii clones exhibited interesting and potentially important differences from their E. coli counterparts.