The rfaD Gene Codes for ADP-~-glycero-~-mannoheptose-6-epimerase AN ENZYME REQUIRED FOR LIPOPOLYSACCHARIDE CORE BIOSYNTHESIS*

The rfaD gene product, ADP-L-glycero-D-mannohep- tose-6-epimerase, is necessary for the conversion of ADP-D-glycero-D-mannoheptose to ADP-L-glycero-D- mannoheptose. The nucleotide ADP-D-glycero-D-man-noheptose accumulates in rfaD mutant strains. Two chimeric colicin E l plasmids carrying the coding se- quence of the rfaD+ locus have been selected and shown to complement the rfaD phenotype. These clones also result in an amplification of ADP-L-glycero-D-manno-heptose-6-epimerase activity. if content or is altered chemical treatment or by mutation D-glycero-D-mannoheptose was obtained from Dr. V. Ginsburg, National Institutes of Health. A filtered supernatant from an overnight mitomycin C (1.2 pg/ml)-induced colicinogenic strain (JC411/ColE1) was used as a source of crude colicin El.

The rfaD gene product, ADP-L-glycero-D-mannoheptose-6-epimerase, is necessary for the conversion of ADP-D-glycero-D-mannoheptose to ADP-L-glycero-Dmannoheptose. The nucleotide ADP-D-glycero-D-mannoheptose accumulates in rfaD mutant strains. Two chimeric colicin E l plasmids carrying the coding sequence of the rfaD+ locus have been selected and shown to complement the rfaD phenotype. These clones also result in an amplification of ADP-L-glycero-D-mannoheptose-6-epimerase activity.
Lipopolysaccharide is a major component of the outer membrane of Gram-negative bacteria. The outer membrane prevents the entry of molecules larger than 700 in molecular weight (1) such as oligopeptides (Z), hydrophobic antibiotics, and dyes (3). The barrier function can be breached if the lipopolysaccharide content or structure is altered by chemical treatment (4) or by mutation ( 5 ) , respectively.
The lipopolysaccharide of Escherichia coli K12 consists of lipid A and a core region. The core consists of an inner core and outer core. The inner core is composed of 2-keto-3-deoxyoctulosonic acid, L-glycero-D-mannoheptose, phosphate, rhamnose, and ethanolamine. L-Glycero-D-mannoheptose is found to be a typical component of the lipopolysaccharide core region of many Gram-negative bacteria (6). Little is known about the biosynthetic steps leading to the synthesis of L-glycero-D-mannoheptose. Eidels and Osborn ( 7 ) , using a transketolase mutant of Salmonella typhimurium, demonstrated a role for sedoheptulose-7-phosphate in lipopolysaccharide and aldoheptose biosynthesis. These results confirmed a role for sedoheptulose-7-phosphate in aldoheptose biosynthesis, as suggested earlier by Ginsburg et al. (8). Further, Eidels and Osborn (7) postulated a pathway for L-glycero-Dmannoheptose synthesis (Fig. 1) which predicted four enzymatic steps: an isomerase, a mutase, a nucleotide-diphosphate-heptose synthetase, and an epimerase activity. Later, Eidels and Osborn (9) demonstrated the conversion of sedoheptulose-7-phosphate to ~-glycero-~-mannoheptose"I-phosphate, thus demonstrating an isomerase activity equivalent to the first enzyme proposed in their biosynthetic scheme.
In a previous study (IO) we demonstrated that in E. coli K12, a single site mutation, designated rfaD, resulted in increased permeability to a large number of hydrophobic drugs and dyes. Further, it was shown that the mutation resulted in an altered lipopolysaccharide that contained primarily an atypical heptose and reduced sugar content. Based on available data, it was concluded that the new heptose was most * The costs of publication of this article were defrayed in pi ,t by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. likely D-glycero-D-mannoheptose. Thus, the rfaD gene mutation represented a tool for studying aldoheptose biosynthesis.
Starting with the rfuD mutation, the present paper provides evidence in the wild type organism for an epimerase activity necessary for the D to L conversion of the sugar nucleotide derivative of glyceromannoheptose. This report also demonstrates that in rfaD mutants (epimeraseless strains) ADP-Dglycero-D-mannoheptose accumulates. Thus, in the rfaD mutant one finds both an abnormal lipopolysaccharide (Dglycero-D-mannoheptose substituted for L-glycero-D-mannoheptose, and a markedly reduced sugar content) and accumulation of ADP-D-glycero-D-mannoheptose.
The above observation that an ADP-nucleotide accumulates in rfuD strains was recently and independently buttressed by the isolation of ADP-D-glycero-D-mannoheptose from R mutants of Shigella sonnei by Kontrohr and Kocsis (11). They also suggested a role for this nucleotide-diphosphate sugar in aldoheptose biosynthesis.
This paper further reports the identification of colicin E l hybrid plasmids carrying the coding sequence for ADP-Lglycero-~-mannoheptose-6-epimerase from the Clarke-Carbon bank (12).

EXPERIMENTAL PROCEDURES
Bacterial Strains and Growth Conditions-The E . coli K12 strains used in this study are listed in Table I. All cultures were grown aerobically in LB medium (13) or in the minimal media of Davis and Mingioli (14). Carbon sources (glucose or galactose) were added to a final concentration of 0.5%. Solid media were prepared by adding agar (1.5% final concentration) to the media described above. Cells were grown at 30 or 37 "C. Cultures were supplemented, where indicated, with 100 pg of required nutrients, 5 mM CaCIz, 100 pg of streptomycin sulfate, 14 pg of kanamycin sulfate, and 30 or 60 pg of novobiocin per ml. E . coli strains carrying hybrid plasmids were grown in the presence of sufficient amounts of colicin El to inhibit the growth of sensitive cells.
Genetic Analysis-Genetic transfers and miscellaneous techniques were based on procedures described by Miller (15). Transformations were done as described by Cohen et al. (16).
Plasmid DNA Isolation-Plasmid DNA was isolated following chloramphenicol amplication (17) using the alkaline extraction procedure described by Birnboim and Doly (18). Additional purification involved phenol extraction and recovery of supercoiled plasmid DNA from agarose gels as described by Tabak and Flavell (19).
Analytical and Quantitative Techniques-Agarose gel electrophoresis (0.7%) was carried out in a horizontal slab gel apparatus using a Tris-borate:EDTA buffer system (20). Sodium dodecyl sulfate-polyacrylamide slab gel electrophoresis was performed in the presence of urea (4 M) by the method of Laemmli (21). Lipopolysaccharide resolved in 15% sodium dodecyl sulfat,e-polyacrylamide slab gels was visualized by a silver staining method described by Tsai and Frasch (22). Note that the best resolution of the R mutant lipopolysaccharide samples is obtained using 15% sodium dodecyl sulfate gel and not'the 14% sodium dodecyl sulfate gel described by Tsai and Frasch (22). Thin layer chromatography for nucleotides was performed  the nucleotides were visualized under UV light as described by Verachtert et al. (23). Lipopolysaccharide preparation, hydrolysis, quantitation of lipopolysaccharide components, and alditol acetate derivatization of neutral sugars were described previously (IO). The degradation of the lipopolysaccharides and gel filtration of the released water-soluble oligosaccharides on Bio-Gel P2 was as described by Schmidt et al. (24). Carbohydrate analyses were performed by gasliquid chromatography of the alditol acetate derivatives on 3% SP2340 (see appropriate data legend for conditions).
In addition, qualitative and quantitative analyses of neutral sugars were accomplished using the automated sugar analyzer developed by Boykins and Liu (25). Paper chromatography of sugars was performed on Whatman No. 1 and No. 3 paper, and the chromatograms were visualized using the urea phosphate reagent of Greene (26). Acidlabile phosphorus and phosphate were assayed by the method of Ames (27). Determination of total sugar was by the phenol:HSOI method of Dubois et al. (28).
In Vivo Labeling-The in vivo labeling of nucleotides was accomplished by adding 1 mCi of [JH]adenosine and 0.22 mCi of ["C] sedoheptulose-7-phosphate to 100 ml of CL503 cultures. Tryptophan, phenylalanine, and tyrosine (100 pg/ml each), and 10 pg/ml of paminobenzoic acid andp-hydroxybenzoic acid were added to increase the specificity of labeling with [14C]sedoheptulose-7-phosphate. These cultures were grown to stationary phase, collected by centrifugation, and washed several times with minimal medium (0-4 "C), and pellets were stored at -20 "C.
Isolation of 'IC-labeled Sugar Nucleotide from rfaD Strains-Nucleotides were extracted from frozen pellets of rfaD mutant strains (from ['"Hladenosine-and/or ['4C]sedoheptulose-7-phosphate-labeled and unlabeled cultures) with boiling 80% ethanol as described by Ginsburg (29). The 80% ethanol extracts were mixed. The combined extracts were kept at -20 "C overnight, and the precipitated material was removed by centrifugation (8500 rpm) for 90 minutes a t -10 "C.
column AG I-X8 (2.6 cm X 40 cm, formate form; 200-400 mesh). The A sample of the mixture (100 ml) was applied to an anion exchange column was washed with deionized H 2 0 (27) and eluted with a linear gradient of ammonium formate, 0-2 M, as described by Ginsburg (29).
Fractions of 5 ml were collected, and the absorbance at 260 nm was determined. Samples of 0.1 ml were removed from every fifth fraction for radioactivity determination by scintillation counting. Radioactivity (I4C and "H) was found between fractions 230 and 259. The ultraviolet absorption spectra of the peak (e.g. 'H and I4C) fraction (244) were characteristic of an adenosine derivative. Fractions 230 to 259 were pooled and repeatedly lyophilized to remove ammonium formate. The lyophilized sample was resuspended in 2 ml of deionized H,O, and portions (0.5 ml) were applied onto Whatman No. 3 paper. Descending chromatography was developed in 95% ethanol, 1 M ammonium acetate (5:2 v/v). Four UV absorbing spots were detected, two of which were found to be radioactive. The radioactive spots were eluted with deionized H20, lyophilized, and resuspended in 70% ethanol for further purification (29).
Preparation of Crude Extracts-Bacterial cultures (100 or 200 ml of rfaD and wild type strains) were grown to midlog phase and harvested by centrifugation. The pellets were resuspended in 0.01 M Tris-HCI buffer, pH 8.0, containing I mM EDTA and 5 mM P-mercaptoethanol. These suspensions were passed through a French pressure cell at 12,000 p s i . and centrifuged a t 15,000 X g for 20 min. The centrifugations and all subsequent procedures were performed at 4 "C.
Preparation of Lipopolysaccharide Acceptor Solution-Several enzymes involved in lipopolysaccharide biosynthesis require a lipopolysaccharide acceptor and a phospholipid cofactor (30). The lipopolysaccharide acceptor was prepared as described by Creeger and Rothfield (31) by mixing lipopolysaccharide from the rfaD strain CL29 (10) and crude lipid extract (32) from E. coli K12 strain PL2.
This mixture [crude lipid (60 pmol of phosphate in 4 ml of methanol) plus lipopolysaccharide (7.0 pmol of 2-keto-3-deoxyoctulosonic acid)] was added to 20 pmol of EDTA, 4.65 ml of 1.0 M Tris-HC1 buffer (pH 8.5), and H20, in a total volume of 20 ml. The mixture was heated for 30 min a t 60 "C, cooled to room temperature, and sonicated with three 0.5-min bursts at 15 "C.
Epimerase Assay-The epimerase activity was assayed by the incorporation of ["C]glucose into lipopolysaccharide isolated from the rfaD strain CL29. In other words, enzyme activity was demonstrated by coupling the epimerase activity with subsequent incorporation of heptose and labeled glucose into lipopolysaccharide. The assumption is that the incomplete lipopolysaccharide chain(s) made by the rfaD mutant can serve as an acceptor for such an extension. The assay mixture contained 0.14 ml of the above lipopolysaccharide acceptor solution, 20 pmol of MgCI,, 34 nmol of NAD, 30 nmol of UDP-galactose, 30 nmol of UDP-['4C]glucose (3000 dpm/nmol), and a constant amount of an rfaD aqueous extract, in a total volume of 0.25 ml. The reaction was started by adding various amounts of rfaD' extracts. Note that preliminary experiments had indicated that the above reaction required a small molecule abundantly present in rfaD extracts. This requirement could be satisfied by adding small volumes of rfaD extracts, boiled rfaD extracts, or supernatants resulting from acetone-or chloroform-precipitated rfaD extracts. The reaction was terminated after 20 min a t 37 "C with 3 ml of cold 5% trichloroacetic acid containing 1 mg/ml of UDP-glucose. Following filtration on Millipore filters, the radioactivity of acid-insoluble materials was determined by liquid scintillation. [14C]Glucose was not incorporated from UDP-['4C]glucose in control experiments without lipopolysaccharide acceptor solution. Additional controls were performed which excluded rfaD extracts or rfaD' extracts, or both, from the reaction mixture, which allowed for corrections for any residual epimerase activity associated with assay components.
Identification of Plasmids Carrying the rfaD' Gene-The Clarke-Carbon colony bank (12) was screened for plasmids capable of abolishing the rfaD phenotype (e.g. hydrophobic antibiotic hypersensitiv-by guest on March 24, 2020 http://www.jbc.org/ Downloaded from ity). Transfer of colicin E l hybrid plasmids was performed by Ffactor-mediated transfer. Donor and recipient cells were mixed in a ratio of 1:3 at 2 X 10' cells/ml. After overnight incubation at 37 "C in Falcon plates, samples were transferred to LB plates containing 60 pg/ml of novobiocin and 100 pg/ml of streptomycin. Novobiocin-and streptomycin-resistant colonies resulted from 10 of the mating mixtures. These colonies were tested for colicin E l resistance. Survivors were checked for the genetic marker of the recipient and an isolatable plasmid capable of complementing a rfaD strain (CL515) in transformation experiments. Two of these isolated plasmids (pLC13-13 and pLC32-45) were selected for further study.

RESULTS
Previous gas-liquid chromatography and chemical studies of lipopolysaccharide from rfuD mutants demonstrated that the lipopolysaccharide contains primarily the stereoisomer, D-glycero-D-mannoheptose, rather than L-glyCerO-D-mannOheptose, and very little of the distal sugars (10). The identities of the sugar components were based on gas-liquid chromatographic behavior of alditol acetate derivatives of the lipopolysaccharide components and alditol acetate sugar standards on 5% ethylenesuccinate-cyanoethylsilicone copolymer (10). The resolution of the heptose components of rfuD mutant lipopolysaccharide by gas-liquid chromatography on a different stationary phase is shown in Fig. 2A (3% SP-2340). The mass spectra of the resolved heptose peaks are shown in Fig. 2B. A comparison of the mass spectrum of authentic L-glycero-Dmannoheptose (which is identical to the mass spectrum of the authentic D-epimer). (Fig. 2C) and the spectra of the resolved heptose peaks (Fig. 2B) confirms the presence of heptose in both peaks. Thus, the presence of both epimers of glyceromannoheptose in rfuD lipopolysaccharide was confirmed. The observation that the major heptose epimer was D-glyCer0-Dmannoheptose is consistent with the premise that D-glyCerO-D-mannoheptose is the precursor of L-glycero-D-mannoheptose via an epimerase activity (7). Thus, a mutation in the rfuD locus may result in the loss of an epimerase activity. An inversion of the configuration of a carbinol group, catalyzed by epimerases, occurs at the sugar nucleotide level (33). Therefore, a mutation in the necessary epimerase activity should result in the accumulation of the substrate sugar nucleotide. In order to test this assumption, cultures of an rfuD mutant, CL503, were supplemented with [14C]sedoheptulose-7-phosphate (see Fig. 1). These cultures were harvested, and the pellets were extracted with 80% ethanol (29'. From the ethanol extract, a I4C sugar nucleotide fraction was isolated (see "Experimental Procedures"). The behavior of the I4C sugar nucleotide on polyethyleneimine-cellulose layers was determined using a 0.5 and 1.6 M LiCl solvent system. The RF value of the unknown sample was larger than those determined for the co-chromatographed guanosine-D-glycero-Dmannoheptose in both solvent systems. The RF value of the I4C nucleotide was also larger than those determined for the co-chromatographed nucleotide diphosphate standards, except UDP, when the polyethyleneimine plates were developed  (25), indicated the presence of ribose and D-glycero-D-mannoheptose. In addition, when this sample was chromatographed on polyethyleneimine plates using 1.6 M LiC1, the UV spot remained at the origin. Similarly, authentic adenine also failed to migrate when the plates were developed using 1.6 M LiC1. These results suggested that the sugar nucleotide was ADP-D-glycero-D-mannoheptose. The identification of the nucleotide was further corroborated by a double-labeling experiment and ultraviolet spectroscopy of the sugar nucleotide. When CL503 (rfaD, Pur-) cultures were labeled with [14C]sedoheptulose-7-phosphate and ["Hladenosine, the isolated sugar nucleotide contained both 3H and I4C label. The UV absorption spectra (Fig. 3) of the isolated sugar nucleotide, at pH 1.0, 7.0, and 11.0, were identical to that of an ADP nucleotide derivative. Finally, chemical analysis of the sugar nucleotide, ADP-heptose, indicated the following stoichiometry per micromole of adenine: 0.82.0:0.90.7 pmol, respectively, for acid-labile phosphorus (27), total phosphorus (27), D-glycero-D-mannoheptose, and ribose.
Effect of the Hybrid Plasmids on the Lipopolysaccharide of rfaD Strains-Lipopolysaccharide was isolated from rfaD strain CL89 and the rfaD strain CL89 carrying the hybrid plasmid pLC13-13 or pLC32-45. The isolated lipopolysaccharide preparations were resolved by sodium dodecyl sulfatepolyacrylamide gel electrophoresis followed by visualization with silver staining (22). As shown in Fig. 4, the migration of a large fraction of the lipopolysaccharide from CL89 (lane A) was characteristic of a structure smaller than the lipopolysaccharide containing a complete core region, present in lane B, which was isolated from a rfaD' strain, JM15. These results suggested that the major lipopolysaccharide structure present in the mutant strain, CL89, contained few core components. In contrast to CL89, an rfaD mutant strain (CL89) carrying pLC32-45 or pLC13-13 plasmids was found to synthesize a complete lipopolysaccharide structure. As shown in the chromatogram (Fig. 5), mutant lipopolysaccharide (CL89) contained rhamnose, glucose, D-glycero-D-mannoheptose, and a small amount of L-glycero-D-mannoheptose, which was reported previously (10). L-Glycero-D-mannoheptose was the only heptose present in the lipopolysaccharide of rfaD strains carrying the plasmid pLC32-45 or pLC13-13. Galactose was also a component of the lipopolysaccharide isolated from a plasmid-complemented rfaD strain.

Content of neutral sugar components in the lipopolysaccharides obtained from the rfaD mutant strain and the rfaD plasmidcontaining strain
The quantitation of the sugar components was determined by chromatography of known amounts c.f authentic sugars on the sugar analyzer under the identical conditions used for experimental samples.  The lipopolysaccharide samples from rfaD and rfaD plasmid-complemented strains were subjected to quantitative analyses (Table 11). In addition, it was determined that the lipopolysaccharide of the rfaD strain, CL89, had a 2-keto-3deoxyoctulosonic acid:heptose (i.e. the D and L isomer) ratio of 2:1, respectively. Finally, a higher than expected glucose content (Table XI) was observed in lipopolysaccharide samples from rfaD and rfaD+ strains.
Fractionation of the Carbohydrate Moiety of rfaD Lipopolysaccharide by Gel Filtration-Following mild acid hydrolysis of rfaD lipopolysaccharide and removal of the waterinsoluble lipid A fraction, the released water-soluble oligosaccharides were characterized by gel permeation on Bio-Gel P2. Four peaks were identified (Fig. 6) as carbohydrate-containing. The peak tube of each carbohydrate fraction was tested for the presence of 2-keto-3-deoxyoctulosonic acid. Peaks 11, 111, and IV were found to contain 2-keto-3-deoxyoctulosonic acid. Each of the pooled fractions was hydrolyzed and then subjected to sugar analysis on the sugar analyzer. The only sugar detectable in the early eluting carbohydrate peak (I) was glucose.' Peak 11, the major carbohydrate fraction, consisted of glucose, D-glycero-D-mannoheptose, 2-keto-3-deoxyoctulosonic acid, and rhamnose. The third peak contained glucose, D-glycero-D-mannoheptose, and 2-keto-3-deoxyoctulosonic acid. The qualitative analysis of peak IV showed the presence of glucose and 2-keto-3-deoxyoctulosonic acid only.
A fraction containing L-glycero-D-mannoheptose was not observed, although qualitative and quantitative data (Figs. 1 and 2 and Table XI) presented above, strongly suggest its presence.
' The early elution and sugsr content of peak I suggest that the lipopolysaccharide samples were contaminated with the membranederived oligosaccharide described by Kennedy (34).

SCHEME 1
of an epimerase activity was demonstrated by coupling the epimerase activity with subsequent incorporation of heptose and labeled glucose into lipopolysaccharide. Very little epimerase activity was detected in the rfaD mutant strain, CL89. In contrast to the low activity in CL89, significant activity was present in a wild type strain, PL2. Dramatic increases in the epimerase activity were observed in rfaD strains carrying the hybrid plasmid, pLC32-45 or pLC13-13.
The substrate specificity of the epimerase activity was demonstrated by experiments in which ADP-D-glycero-Dmannoheptose or GDP-D-glycero-D-mannoheptose (8) was added as the substrate. As shown in Fig. 8, the addition of increasing amounts of ADP-D-glycero-D-mannoheptose resulted in a stimulation of the incorporation of ['4C]glucose into trichloroacetic acid-insoluble material. In contrast, no effect was observed with GDP-D-glycero-D-mannoheptose.

DISCUSSION
The biosynthesis of L-glycero-D-mannoheptose in Gramnegative bacteria has been postulated to occur by racemization of D-glycero-D-mannoheptose to the L isomer (7). However, the enzymatic activity responsible for this conversion remained unclear. A possible tool to investigate this process became available when a mutant designated rfaD was isolated which contained a defective lipopolysaccharide with an abnormal heptose (10). The present communication presents evidence for the existence of a specific epimerase activity (an rfuD gene product), which is involved in the conversion of Dglycero-D-mannoheptose to L-glycero-D-mannoheptose. It was demonstrated that a mutation in the epimerase gene, i.e. the rfaD mutation, resulted in the loss of the epimerase activity, predominantly an incomplete lipopolysaccharide structure which terminates with the heptosyl residue, D-glyCerO-D-mannoheptose, and the accumulation of the nucleotide, ADP-Dglycero-D-mannoheptose, in the extracts of these mutants. Recently and independently, Kontrohr and Kocsis (11) demonstrated the presence of ADP-D-glycero-D-mannoheptose in a S. sonnei R mutant that incorporates D-glycero-D-mannoheptose in its lipopolysaccharide. They also suggested an intermediate role for this sugar nucleotide in aldoheptose biosynthesis in Enterobacteria. The identification of ADP-D-glycero-D-mannoheptose and a substrate role for this nucleotide in E. coli lipopolysaccharide core biosynthesis support their contention.
Further, colicin E l hybrid plasmids were selected which complemented the rfuD phenotype, and amplified the ADP-~-glycero-~-mannoheptose-6-epimerase activity. In addition, only wild type lipopolysaccharide core structure was present in plasmid containing rfaD mutant strains.
The results presented here provide support for the proposed scheme of Eidel and Osborn (7) ( i e . steps 3 and 4 in Fig. 1) for L-glycero-D-mannoheptose biosynthesis (Scheme 1).