Regulation of membrane glycosyltransferases by the sfrB and rfaH genes of Escherichia coli and Salmonella typhimurium.

The role of sfrB and rfaH genes in the regulation of expression of membrane glycosyltransferases was studied in Escherichia coli and Salmonella typhimurium. The transferase enzymes form part of a multienzyme system involved in biosynthesis of the polysaccharide core of Gram-negative bacterial lipopolysaccharides. Several sfrB mutants of E. coli showed reductions of 90-98% in the activities of two of the glycosyltransferases (UDP-galactose:(glucosyl)lipopolysaccharide 1,6-galactosyltransferase and UDP-glucose: (glucosyl)lipopolysaccharide 1,3-glucosyltransferase). Introduction of a recombinant ColE1 plasmid restored the transferase levels to normal and simultaneously corrected the F-factor defects that also characterize sfrB mutants; recombinant plasmids containing other regions of the E. coli chromosome were ineffective. An amber mutation of the S. typhimurium rfaH gene (thought to be the homologue of the E. coli sfrB gene) resulted in 97% loss of activity of the Salmonella UDP-galactose:(glucosyl)lipopolysaccharide galactosyltransferase. Antibody precipitation studies showed that the loss of enzyme activity in the amber mutant was associated with a corresponding decrease in amount, but not in size, of the transferase protein, indicating that the gene is not the structural gene for the S. typhimurium galactosyltransferase. Taken together, the results indicate that the sfrB(rfaH) gene acts as a positive regulatory element in expression of multiple glycosyltransferases in E. coli and S. typhimurium.

Mutations of the sfrB gene of Escherichia coli are associated with abnormalities in several F-factor functions. These include surface exclusion properties, formation of F-pili, and ability to act as donors in conjugal genetic crosses (1). Beutin et al. (2) have shown that the SfrB-phenotype is associated with premature termination of transcription of the tray + Z operon of the F-factor, suggesting that the sfrB gene product may regulate expression of this operon by acting as an antiterminator of transcription. Consistent with this view is the observation that at least one gene product of the tray Z operon, the TraT protein of the outer membrane, is diminished in amount in sfrB cells (1).
In addition to these changes in F-factor-mediated properties, sfrB mutants also show other alterations in cell envelope + This work was supported by Grant AM-13407 from the United States Public Health Service. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "adoertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
f Present address, Becton Dickinson and Company, Research Triangle Park, NC 27709. structure. These include changes in bacteriophage sensitivity patterns, alterations in electrophoretic mobility of lipopolysaccharide, and defects in flagellar function. These observations led Beutin and Achtman (1) to suggest that the sfrB gene product may act to regulate the expression of a number of operons responsible for synthesis of cell envelope components.
In Salmonella typhimurium, mutations of the rfaH gene result in defects in F-factor function that are similar to those of sfrB mutants of E. coli (3). It is well established that rfaH mutants are also defective in lipopolysaccharide biosynthesis (4, 5). Their similarities in phenotype and map position suggest that sfrB and rfaH represent homologous genes in the two organisms, as suggested by Sanderson and Stocker (3).
In this paper we present evidence t h a t (i) sfrB mutations result in loss of activity of two E. coli membrane proteins, UDP-galactose:(glucosyl)lipopolysaccharide al,6-galactosyltransferase and UDP-glucose:(glucosyI)lipopolysaccharide CUI, 3-glucosyltransferase,' enzymes required for biosynthesis of the complete core region of the E. coli lipopolysaccharide; (ii) both transferase activities are restored by introduction of a recombinant plasmid that also corrects the conjugational abnormalities of the sfrB cells; and (iii) an amber mutation of the rfaH gene of Salmonella typhimuriurn, the homologue of the sfrB gene of E. coli, is associated with a decrease in amount, but not in apparent size, of a Salmonella membrane protein, UDP-galactose:(glucosyl)lipopolysaccharide galactosyltransferase. The results implicate the genes responsible for synthesis of the membrane glycosyltransferases as likely additional targets for regulation by the sfrB gene product.

EXPERIMENTAL PROCEDURES
Materials-Radioactive nucleotide sugars and 35S-labeled HzSO~ were purchased from New England Nuclear. Unlabeled nucleotide sugars were obtained from Calbiochem. Nucleotide sugars were checked for purity as previously described (6). IgGsorb was purchased from the Enzyme Center, Inc., and NCS tissue solubilizer from Amersham COT.
Crude colicin El was prepared from the supernatant of a culture of E. coli W3110 (7). This crude extract, sterilized by filtration, was mixed with the growth medium and was applied to selective plates in appropriate dilutions as previously described (8). Lipopolysaccharides were purified as previously described (9). Analytical Techniques-Ascending chromatography of sugars and nucleotide sugars was performed in 95% ethanol, 1 M ammonium acetate (7:3). Descending chromatography of sugars was performed in butanol/pyridine/water (6:4:3). SDS'-gel electrophoresis was carried out in 11% gels using the system of Lugtenberg et al. (10). Colicin E1 sensitivity and bacteriophage sensitivities were determined by spot tests as described previously (8).
Organisms-E. coli and S. typhirnuriurn strains are listed in Table  I. The galE strains M1174T1, M1162T2, and M1170T3 were constructed by PI-mediated co-transduction of galE and nad::TnlO from E. coli 112 (obtained from Dr. N. Kleckner, Harvard University). galE recombinants were identified by screening tetracycline-resistant transductants for galactose sensitivity. The GalE-phenotype of sfrB galE candidates was confirmed by analysis of radioactive sugars incorporated into assay reaction products (see below). Construction of SL3657-S98 and SL3657-S99 was accomplished by ES18hl-mediated transduction (11) of supDsul from strain TT2070 (supDsul zeb618::TnlO) to SL3657 (r/aH(Am) hisAm)). Tetracycline-resistant His+ recombinants were identified and were tested for sensitivity to bacteriophage P22 to look for suppression of the RfaH-(P22-resistant) phenotype. All His+ clones tested had simultaneously acquired sensitivity to P22, confirming the amber character of the rfaH mutation in SL3657. Two of these supD zeb618:TnlO clones (SL3657-S98 and S3657-S99) were used in the present study. Plasmids pLC10-7, pLC17-24, and pLC14-28 were originally constructed by Clarke and Carbon (12) by insertion of random fragments of E. coli DNA into ColEl and have been previously described (8,13).
Unlabeled cultures were grown in proteose peptone-beef extract medium (8). For labeling with 35S, a low sulfate minimal medium was used (M9(8) in which 0.2 g/liter of magnesium chloride and 3 mg/ liter of magnesium sulfate were substituted for the usual amount of magnesium sulfate). The low sulfate salts were supplemented with glucose and all non-sulfur-containing amino acids. Glycosyltransferase Assays: Preparation of Cell-free Extracts-All strains used for enzyme assays contained a galE mutation to prevent interconversion of UDP-galactose and UDP-glucose during the assay. Cells were grown, disrupted by sonication, and centrifuged at 39,000 X g as previously described (8). The supernatant fraction was used as a source of enzyme unless otherwise noted, S. typkimuriun Galactosyltransferase Assay-Galactosyltransferase activity was determined by measuring the ability of enzyme extracts to catalyze the transfer of [3H]galactose (4500 dpm/nmoI) from UDP-["HIgalactose into lipopolysaccharide acceptor from strain SL1060 as previously described (8). The lipopolysaccharide ofSL1060 can act as substrate for the a1,3-and al,6-galactosyltransferases of s. typhimurium (14) and no attempt was made to distinguish between the two activities in the present study. One unit of enzyme activity is defined as the incorporation of 1 nmol of 13H]galactose in 10 min at 37 "C.
E. coli Glucosyltransferase IIAssay-Glucosyltransferase 11 activity of E. coli extracts was determined by measuring the transfer of ["cl glucose from UDP-['4CJglucose into acceptor lipopolysaccharide. ACceptor lipopolysaccharide was prepared by preincubating galactosedeficient lipopolysaccharide acceptor from S. typhimurium G30 in the standard galactosyltransferase reaction (see above) using nonradioactive UDP-galactose and wild t p e E. coli enzyme from strain M1174T1. After 60 min at 37 "C to permit addition of the branch 1,6-galactosyl residue (see Fig. l), the wild type enzyme was inactivated by heating the reaction mixture at 100 "C for 2 min. Glucosyltransferase I1 activity was then determined by adding UDP-['4CI glucose (0.12 mM, 7,300 dpm/nmol) and the enzyme preparation to be assayed (5-60 gg of protein). Incorporation of ['4C]glucose into acid-precipitable material was determined after incubation for 10 min at 37 "C.
E. coli Galactosyltransferase Assay-Galactosyltransferase activity of E. coli extracts was determined by measuring transfer of I3H] galactose from UDP-[3H]galactose into acceptor lipopolysaccharide from S. typhimurium G30 as described above for the S. typhimurium galactosyltransferase; nonradioactive UDP-glucose was also included in the reaction mixture because of evidence3 that this resulted in a slight increase in the rate of incorporation of [3H]galactose. Experiments were also performed in the absence of nonradioactive UDPglucose; these gave similar results to those shown in Table 11.
Analysis of Reaction Products-Hydrolysis and chromatography of the acid-precipitated reaction products (5) confirmed that >95% of the incorporated radioactivity in the glucosyltransferase and galactosyltransferase assays co-chromatographed with authentic glucose and galactose, respectively.
Measurement of Efficiency of Conjugal Genetic Transfer-The donor strain E. coli KL181/KLF23 containing the F' trp' plasmid KLF23 was mixed with the trp-recipient strain E. coli D l at a ratio of 1:IO (donorlrecipient). After 90 min at 37 "C, serial dilutions of the mating mixture were plated on appropriate selective plates to determine the number of viable donor cells and the number of Dl/ KLF23 transconjugants. The number of transconjugants was expressed per 100 donor cells and was normalized to the value obtained for the s f r F control donor strain M1174/KLF23. Imnunopreclpitation-Antiserum to purified S. typhimuriun galactosyltransferase was obtained by immunization of female New Zealand white rabbits with purified UDP-galactose:(glucosyl)lipopolysaccharide galactosyltransferase. The transferase was purified by a modification of' the method of Romeo et al. (15). When chromatographed on Sephadex G-75, the purified enzyme showed a major peak of activity that eluted in a position corresponding to a molecular weight of 40,000 and a minor peak, accounting for ~5 % of the total activity, whose elution position indicated an apparent molecular weight of 80,000. SDS-gel analysis of the purified protein is described under "Results." The galactosylglucose linkage in the product of the reaction catalyzed by the purified transferase (originally thought to be the a1,3-galactosylglucose of the lipopolysaccharide core (5)) has not been definitely established. Recent studies4 suggest that the product may instead be the branch ul,6-galactosylglucose (Fig. 1).
Extracts to be assayed for immunoprecipitable material were prepared from exponentially growing cells exposed to ["S]S0, for several generations. Cultures of 20 ml containing 1.6 mCi of ["S]SO, were grown in parallel with 120 ml of nonradioactive cultures. The two cultures were mixed before harvesting. The resulting cell pellet contained approximately 5 X lo9 cpm/g of cells (wet weight). To obtain the supernatant and cell envelope fractions, the 3sS-labeled cells were resuspended in 0.2 M Tris-C1, pH 8.0,5 mM 2-mercaptoethanol (about 0.6 g of cells (wet weight)/ml) and treated with EDTA, sucrose, and lysozyme as described by Witholt et al. (16). After a 2-fold dilution with water, the suspension was incubated at 37 "C for 30 min. Spheroplast formation was monitored under the microscope. The spheroplast suspension was then chilled and sonicated for five 10-s bursts. Unbroken spheroplasts ( 4 % of total cells) were removed after pelleting by centrifugation at 500 X g for 2 min. The cell lysate was spun at 35,000 X g for 12 h at 4 "C and the supernatant fraction (soluble E. S. Creeger and L. I. Rothfield fraction) was reserved for immunoprecipitation. The membrane pellet was extracted twice by resuspending in 10 mM Tris-C1, pH 7.5, 2% Triton X-100, 10 mM EDTA, incubating at room temperature for 30 min, and spinning at 130,000 X g for 1 h at 4 "C. The combined supernatants of the two extractions were used as the cell envelope fraction for immunoprecipitation. For 35S-labeled whole cell lysates, cells were harvested, frozen, and sonicated as previously described (8). The whole cell lysate was extracted twice with 2% Triton X-100, 10 mM EDTA as described above. The combined supernatants were used as total cell extract for immunoprecipitation.
Immunoprecipitation was carried out by the following modification of the method of Shuman et al. (17). To 0.4-ml aliquots of the 35Slabeled extracts (about 10' cpm) was added 0.20 ml of buffer (150 mM Tris-CI, pH 7.5, 3 M NaCI, 0.1% aprotinin, 3% Triton X-100, 5 mM EDTA). After addition of 0.03 ml of a 1 5 0 dilution of normal rabbit serum, samples were vortexed and incubated at 37 "C for 30 min.
Samples were chilled, 0.03 ml of IgGsorb was added, and, after incubation at 0 "C for 20 min, the suspensions were spun for 3 min in a Microfuge (Beckman) to remove proteins that bind nonspecifically to rabbit serum or IgGsorb; the pellet was discarded. For quantitation, appropriate strips were cut from the dried gel. The strips were solubilized by incubation at 48 "C for 60 min in 0.1 ml of water and 0.7 ml of NCS tissue solubilizer and were counted in a toluene-based scintillation counting mixture. Counts/min in the strip from the control immunoprecipitate (normal rabbit serum) were subtracted from the counts/min in the corresponding strip from the anti-galactosyltransferase immunoprecipitate. The results were expressed as a fraction of the counts/min in the original "S-labeled cell extract.

RESULTS
F-factor Defects in sfrB Cells-As previously reported by Beutin and Achtman (I), the sfrB/F' strains were defective in their ability to act as donors in conjugal genetic transfers and were resistant to bacteriophages which utilize the F-pilus for adsorption to the cell surface (Table IV,  The E. coli galactosyltransferase catalyzes addition of the branch 1,6-galactosyl residue to the glucose I residue of the lipopolysaccharide of E, coli K12 (Fig. 1). The enzyme was assayed by measuring the ability of cell-free extracts to catalyze the transfer of galactose from UDP-galactose to an incomplete lipopolysaccharide acceptor prepared from a gulE strain of S. typhimurium (strain G30). This galactose-deficient lipopolysaccharide provides a suitable acceptor for assay of the E. coli c~1,6-galactosyltransferase because of the similarity of the inner core of the lipopolysaccharides of E. coli and S. typhimuriun (Fig. 1).
The E. coli glucosyltransferase I1 catalyzes addition of the  second glucosyl residue of the E. coli lipopolysaccharide core (Glcll in Fig. 1). Glucosyltransferase I1 activity was assayed by measuring the transfer of glucose from UDP-glucose into lipopolysaccharide acceptor prepared from S. typhimurium G30 which terminates with the glucose I residue of the core and therefore provides a suitable acceptor for assay of the E. coli glucosyltransferase I1 (Fig. 1).
The activity of both enzymes was markedly reduced in the two sfrB strains tested (M1162T2 and M1170T3) as compared with the sfrB+ parental strain (M1174T1) ( Table 11, lines 1-3). The defect in activity of the two glucosyltransferases was accompanied by the previously reported (1) alteration in bacteriophage sensitivity patterns (Table 111, lines 1-3).
Correction of Glycosyltransferase and F-factor Defects by Recombinant Plasmid pLCl4-28-Simultaneous correction of the glucosyltransferase and F-factor defects was achieved by introduction of a recombinant ColEl plasmid (pLC14-28) containing a cloned fragment of the E. coli chromosome that had previously been shown to correct the abnormal phenotype of rfaH mutants (13).
Introduction of pLC14-28 into E. coli sfrB strains M1162T2 and M1170T3 resulted in restoration of normal galactosyltransferase and glucosyltransferase I1 enzyme levels (Table  11, lines 4-6). The abnormal patterns of sensitivity to lipopolysaccharide-specific bacteriophages U3 and C21 was also cor-   rected by pLC14-28 (Table III), confirming that the in vitro changes in enzyme activity were paralleled by in uiuo changes in lipopolysaccharide structure. Introduction of the plasmid into the parental strain M1174Tl had no effect on the wild type levels of transferase activities. Two control ColEl plasmids pLC10-7 and pLC17-24, carrying cloned fragments from the main rfa cluster of the E. coli chromosome (8), failed to restore normal levels of glycosyltransferase activity in the sfrB mutants (Table 11, lines 7-12), confirming the specificity Correction of the glycosyltransferase defects was accompanied by correction of the defects in conjugal transfer efficiency and sensitivity to F-specific phages (Table IV, lines 4-7). The restoration of the normal F-factor functions was specific for pLC14-28 as shown by the inability of pLC10-7 to correct the defects.
S. typhimurium Galactosyltransferase Activity-Beutin et ul. (2) have suggested that the sfrB gene product acts as a positive regulator of synthesis of' affected cell envelope proteins. If the model is correct, sfrB(rfaH) mutants should contain a decreased amount of the affected glycosyltransferase proteins. The studies described above are consistent with this hypothesis but do not exclude the possibility that the sfrB gene may code for a common structural component of the affected cell envelope proteins. If the model of Beutin et al. is correct, sfrB (rfaH) amber mutations should result in a decrease in amount but not in size of the affected cell envelope proteins. The availability of an rfaH(Am) mutant provided an opportunity to test these predictions for the S. typhirnurium galactosyltransferase. An-of pLC14-28.  tibody directed against the purified protein was used to quantitate the amount of the protein and SDS-gel electrophoresis was used to estimate its size. As previously shown by Kuo and Stocker,' the rfuH(Am) strain SL3657 showed the characteristic resistance to C21 bacteriophage of other rfuH mutants and introduction of a supDsul suppressor resulted in restoration of the wild type bacteriophage sensitivity pattern (Table V, column 1). When the suppressed and unsuppressed strains were assayed for galactosyltransferase, transferase activity was markedly decreased in SL3657 (rfuH(Am)) to approximately 5% of the normal level (Table V) and introduction of the suppressor gene resulted in a 5-10-fold increase in activity (Table V, column 2). The extent of restoration of galactosyltransferase activity to approximately 40% of the wild type level is consistent with the partial suppression achieved by extragenic suppressors of amber mutations in other systems.
To identify the mutant gene product, anti-galactosyltransferase antibody was used to precipitate immunoreactive transferase protein from 35S-labeled total cell extracts; the immunoprecipitated material was then analyzed by SDS-gel electrophoresis. When analyzed in the same gel system, the purified wild type protein runs as a single major band with an apparent molecular weight of 39,000 (Fig. 2).
As expected, when the antibody was added to extracts of wild type (rfaH+) cells, the immunoprecipitate contained a major band with the same apparent molecular weight as the purified transferase protein (Fig. 3, lane d ) . Addition of antigalactosyltransferase antibody to labeled extracts of the rfaH(Am) strain resulted in immunoprecipitation of a radioactive protein whose electrophoretic mobility in SDS-gel electrophoretograms was indistinguishable from that of the wild type transferase, Although its electrophoretic mobility was unchanged, the amount of immunoprecipitable transferase in the amber mutant was markedIy reduced, to 2-3% of the wiId type level (Fig. 3u and Table V, lines 1 and 2 ) .
Introduction of the suppressor gene into the rfuH(Arn) strain resulted in an increase in amount but no detectable change in SDS gel electrophoretic mobility of the immunoreactive protein (Fig. 3 and Table V).
It has been shown previously that the glycosyltransferases T.-T. KUO Table V). The arrows indicate the positions and molecular weights (~1 0~) of standard proteins run on the same gel.
of lipopolysaccharide biosynthesis are present in both the soluble and cell envelope fractions obtained after mechanical disruption of cells (18). When immunoprecipitable galactosyltransferase was determined separately on the soluble and cell envelope fractions of rfaH(Am) and rfaH' cells (Fig. 4), the results were similar to those described above for the total cell extracts. Introduction of the suppressor gene resulted in a 6-fold increase in immunoprecipitable galactosyltransferase in both supernatant and cell envelope fractions (data not shown).

DISCUSSION
Effects of sfrB and rfaH on Glycosyltransferase Actiuities-The studies of Beutin et al. (2) indicate that the sfrB gene product acts as a positive regulatory element in transcription of the tray + Z operon. The present study indicates that the sfrB (rfaH) gene product also plays a role in expression of a t least three cell envelope glycosyltransferases that are involved in lipopolysaccharide biosynthesis. This explains the observation that sfrB mutants show abnormal patterns of sensitivity to bacteriophages that use lipopolysaccharide as their adsorption sites (1) and indicates that the altered bacteriophage sensitivity patterns are due to changes in the levels of lipopolysaccharide biosynthetic enzymes rather than reflecting the secondary effects of other cell envelope alterations.
In the case of the S. typhimurium galactosyltransferase, an absolute decrease in cellular content of the protein in RfaHcells was demonstrated by specific immunoprecipitation. The failure to see an increased cytoplasmic pool of the protein (Fig. 4) argues against the possibility that the rfaH (sfrI3) defect affects cell envelope proteins by interfering with their entry into the membrane. The fact that there was no detectable change in size of the immunoreactive transferase protein of an rfaH(Am) mutant supports the conclusion that rfaH (and sfrB) are regulatory rather than structural genes for components of the affected glycosyltransferases. It should be noted, however, that a small change in size due to an amber mutation close to the 3' end of a .structural gene would not have been detected by this analysis. All of the results are consistent with the suggestion of Beutin and Achtman (1) that the sfrB gene of E. coli (and presumably also the rfaH gene of S. typhirnurium) acts to regulate expression of several operons involved in synthesis of cell envelope proteins. It is not known whether the structural genes for the glycosyltransferases are part of a single operon that is positively regulated by the sfrB (rfaH) gene product or whether expression of each of the structural genes is individually regulated. The present studies also do not indicate whether the regulation occurs at a transcriptional or post-transcriptional level although the studies of Beutin et al. (2) suggest that an effect on transcription is most likely.
Effects of pLC10-7 on Glycosyltransferase Actiuities-It has been shown previously that pLC10-7 contains genes required for expression of several glycosyltransferase activities (8). Thus, pLC10-7 was capable of correcting the glucosyltransferase I defect of rfaG mutants of S. typhimurium and also induced the appearance of E. coli glucosyltransferase I1 (RfaM) activity when the plasmid was introduced into S. typhimurium host cells (8). In addition, pLC10-7 corrected the abnormal bacteriophage sensitivity patterns of rfaJ and rfaI mutants of S. typhimurium3 and therefore is likely to contain genetic information for expression of UDP-glucose:(galactosyl)lipopolysaccharide glucosyltransferase and UDP-galactose:(glucosyl)lipopolysaccharide al,3-galactosyltransferase, enzymes that catalyze addition of the GlcII and Gal11 residues of the S. typhirnurium lipopolysaccharide ( Fig.   1): An incidental finding in the present study was the observation that pLC10-7 induced a moderate increase in galactosyltransferase in both sfrB + and sfrB cells (Table 11). this is consistent with the idea that pLC10-7 may also encode the rfaB gene, responsible for synthesis of UDP-galactose:-(glucosy1)lipopolysaccharide al,6-galactosyltransferase (14), the enzyme responsible for addition of the Gal1 residue of the polysaccharide core (Fig. 1).
Taken together, these data indicate that pLC10-7 includes genetic information for expression of a t least five enzymes of lipopolysaccharide biosynthesis.