Mechanism of Assembly of the Outer Membrane of

SUMMARY The site of synthesis of lipopolysaccharide has been investigated in cytoplasmic and outer membrane fractions isolated by isopycnic sucrose gradient centrifugation of the total membrane fraction from lysozyme-EDTA spheroplasts. synthesis of O-antigen occurs exclusively in cytoplasmic membrane was obtained by pulse-chase experiments in viva and by assay of biosynthetic enzymes in isolated membrane fractions. O-Antigen chains pulse-labeled in viuo with appeared initially in cytoplasmic membrane, but were rapidly transferred to outer membrane during a subsequent chase with nonradioactive mannose. specific activities of enzymes of O-antigen in isolated cytoplasmic 15 3%fold greater in


SUMMARY
The site of synthesis of lipopolysaccharide has been investigated in cytoplasmic and outer membrane fractions isolated by isopycnic sucrose gradient centrifugation of the total membrane fraction from lysozyme-EDTA spheroplasts. Evidence that synthesis of O-antigen occurs exclusively in cytoplasmic membrane was obtained by pulse-chase experiments in viva and by assay of biosynthetic enzymes in isolated membrane fractions. O-Antigen chains pulse-labeled in viuo with [14C]mannose appeared initially in cytoplasmic membrane, but were rapidly transferred to outer membrane during a subsequent chase with nonradioactive mannose. In accord, the specific activities of enzymes of O-antigen synthesis in isolated cytoplasmic membrane fractions were 15 to 3%fold greater than in outer membrane.
The cytoplasmic membrane fractions were also enriched for glycosyltransferase activities involved in biosynthesis of the core region of lipopolysaccharide.
However, unequivocal localization of core transferase enzymes in the isolated membranes was not possible since these activities were also found with soluble fraction, and secondary binding of soluble enzyme to both cytoplasmic and outer membrane was shown to occur during the isolation procedure. Evidence that the cytoplasmic membrane is the site of synthesis of the core region was derived from pulse-chase experiments in a mutant in which [14C]galactose is incorporated exclusively into the core portion of the polysaccharide.
Over 90% of the [14C]galactose incorporated into lipopolysaccharide during a 1 -min pulse was recovered in the cytoplasmic membrane fraction. The mechanism of translocation of lipopolysaccharide to the outer membrane is unknown. Evidence suggesting that the process is not readily reversible was obtained in experi- The external layer of the cell envelope of Salmonelln nrld IF lated gram-negative enteric bacteria is a membranous struct,ure which contains the lipopolysaccharide of the envelope in addition to phospholipid and protein (l-4). The recent development of techniques for separation of this outer membrane from the underlying cytoplasmic membrane (l-3) has opened a new approach to investigation of the mechanism of assembly of the outer membrane and the role of the cytoplasmic membrane in this process. The present studies were undertaken to determine whether synthesis of lipopolysacrharide occurs in situ in the outer membrane, or whether the polymer is synthesized in the cytoplasmic membrane and only subsequently inserted into the outer membrane structure.
The results of pulse-chase esperiments in vivo and localization of biosynbhetic enzyme activities in isolated membrane fractions strongly support the latter hypothesis.
The data indicate that synthesis of the internal core region of the polysaccharide and the O-antigen, as well as attachment of O-antigen chains to the core, occurs exclusively in the cytoplasmic membrane, and that newly synthesized lipopolysaccharide is then rapidly translocated into the outer menbrane.
The mechanism of the t,ranslocation profess from cytoplasmic: to outer membrane remains to be established.
Evidence suggesting that translocation is not readily reversible was obtained from studies with conditional mutants in lipopolysacchnride synthesis, which showed that the incomplete lipopolysaccharides synthesized under nonpermissive growth conditions could not be completed during subsequent growth under permissive conditions.
The results indicate that lipopolysaccharide which is integrated into the outer membrane is not available to biosynthetic enzymes in the cytoplasmic membrane, arid suggest that outer membrane lipopolysaccharide is not in equilibrium with the nascent polymer at the sites of synthesis. 1 1 I Man-Rha-Gal ;Core Lipopolysaccharide FIG. 1. Pathway of biosynthesis of O-antigen. P-GCL represents phosphoglycosyl carrier lipid (undecnprenyl-P), and PP-GCL, the pyrophosphoryl derivative.
Total membrane fractions were isolated from spheroplast lysates as described in the preceding paper (1); unless otherwise specified spheroplasts were lysed by sonication.
Minor modifications of the spheroplasting procedure were employed in pulsechase experiments and are described in the legend to Fig. 2. Membranes were separated into cytoplasmic and outer membrane fractions by isopycnic sucrose density gradient centrifugation as described (1). The SW 41 rotor was employed for small scale radioactive preparations, and the SW 27 rotor for separation of membrane fraction for enzyme assays. In the latter case centrifugation was for 20 hours, and visible membrane bands were collected by pumping from the top of the gradient tube. Purity of the fractions was routinely monitored by determination of DPNH oxidase activity (1).
The specific activities of cytoplasmic membrane fractions (L1 and Lz) were always greater than 2 pmoles per min per mg, and the specific activity of the outer membrane band, H, was always less than 5% that of L1.

Preparation of Glycosyl Carrier Lipid and O-Antigen Intermediates
Glycosyl Carrier Lipid (Undecaprenyl Phosphate)-Details of the purification procedure will be presented elsewhere. ' Briefly, the method involves alkaline hydrolysis of a crude lipid extract from strain G-30 and chromatography of the alkali-stable fraction on DEAE-cellulose.
The sample was applied in CHCll-methanol (1: 1) to a 2.5 x 20 cm column of DEAE-acetate, prepared by the method of Rouser et al. (14). The column was washed with 200 ml of CHCl,-methanol(1: 1) and 600 ml of 99% methanol and P-GCL2 was then eluted with 600 to 800 ml of 10 mM ammonium acetate-5 InM acetic acid in 99% methanol (15). Peak fractions were pooled and washed with H20 to remove salt. Purity at this stage was approximately 50 y0 on the basis of total phosphate, as judged by thin layer chromatography and enzymatic assay in the galactose-PP-GCL synthetase system. After rechromatography on a column (0.9 x 10 cm) of DEAE-cellulose under the same conditions, the product gave a single spot on thin layer chromatography in three systems, and was the preparation employed in the present studies. Enzymatic conversion to galactose-PP-GCL was 75 to 80% based on total phosphate. llfan-Rha-Gal-PP-GCL-Enzymatic synthesis of the trisaccharide-lipid intermediate ( Fig. 1)  Purified P-GCL (150 nmoles) in CHC13 was evaporated to dryness under a stream of Nz, taken up in 0.12 ml of methanol plus 0.42 ml of 0.57, (v/v) Alfonic 1012-6 (Conoco Petrochemicals, Rockville, Maryland), and dispersed by vigorous mixing on a Vortes mixer. To this were added 600 nmoles of MgC12, 750 nmoles each of UDP-galactose and TDP-rhamnose, 600 nmoles of GDP-[%]mannose (4,500 to 11,000 cpm per nmole) and 1.5 ml of enzyme (1.65 mg of protein).
After incubation at 25" for 2 hours, the reaction mixture was extracted with 18 ml of CHC&-methanol (2:ll).
The product was recovered in the organic phase, which was washed three times with 5 ml of methanol-0.1 M KC1 (1: 1). The washed organic phase was evaporated to dryness under a stream of Nz in the cold, and dissolved in 2 ml of 0.1% (v/v) Alfonir 1012-6. The yield of Man-Rha-Gal-PP-GCL was approximately 110 nmoles. The product was used as substrate for O-antigen polymerase without further purification, and could be stored at -70" for up to 10 days without significant degradation.
Polysaccharade-PP-GCL-Enzymatic synthesis of the GCLlinked polymer of trisaccharide repeating units was carried out as follows.s The incubation mixture contained 0.18 M Trismaleate buffer, pH 6.0, 12.5 IllM MgClz, 0.8 ml of [14C]mannoselabeled trisaccharide-PP-GCL in O.lv, Alfonic 1012-6 (40 to 50 nmoles) and cell envelope (18 mg of protein) in a total volume of 3.25 ml. The cell envelope fraction was obtained from strain  in order to prevent transfer of the polymeric product to endogenous lipopolysaccharide during the course of the reaction (16). After incubation for 2 hours at 25", the reaction mixture was extracted with 12 ml of CHCls-methanol (2: I), and the phases separated by centrifugation.
The product was recovered in the insoluble material at the interface.
The aqueous and organic phases were carefully removed and the residue was suspended in 2 ml of 0.257, (v/v) Alfonic 1012-6. The suspension was sonirated in an ice bath for 3 30-s periods, keeping the temperature below 15". Insoluble material was removed by centrifugation at 105,000 X g for 1 hour at 2-5". Recovery of polysaccharide-PP-GCL in the soluble fraction was 60 to 80%. The preparation was free of P-GCL and trisaccharide-PP-GCL and was used as substrate for O-antigen ligase without further purification.
The preparation was stable to storage at -70" for several days.
UDP-sugar Hydrolase-Hydrolysis of UDP-glucose was determined by the method of Glaser et al. (17).
After incubation for 15 min at 25", reaction was stopped by addition of 2 ml of CHCla-3 ill. Singh and M. J. Qsborn, manuscript, in preparation. methanol (2: 1). The mixture, which gave a single phase, was permitted to stand at room temperature for 5 min and then extracted three t.imes with 0.5 ml of CHCl~-methanol-Hz0 in 0.17, Alfonic 1012-6 as described above (approximately 0.5 nmole, 11,000 cpm per nmole), and membrane fraction (5 to 30 pg of protein).
After incubation for 90 min at 25", the reaction was stopped by heat,ing in a boiling Hz0 bath for 3 min. The entire reaction mixture was streaked on a sheet of Whatman No. 40 paper and developed in Solvent V for 18 to 24 hours. Polymeric products remained within the first 2 inches of the chromatogram. This portion of the paper was cut out and counted as described under "Counting Procedures." Values obtained were corrected for blanks ('ontaining boiled enzyme. O-Antigen Ligase-Transfer of O-antigen chains to lipopolysaccharide was measured by conversion of isolated "C-po1ysaccharide-PP-GCL to an alkali-stable product.3 Acceptor lipopolysaccharide was prepared by incubating a mixture of 20 ~1 of of TV119 lipopolysaccharide (11.2 pmoles of heptose per ml), 10 ~1 of 0.5 M Tris-maleate, pH 6.0 and 10 ~1 of 0.5oj, (v/v) Alfonic 1012-6 for 30 min at 37". Fifty microliters of '4C-polysaccharide-PP-CCL (2500 cpm) and 10 ~1 of membrane fraction (5 to 30 pg of protein) were then added, and the mixture incubated for 1 hour at 25". At the end of the incubation period 1 ml of 2 N NH40H-5 mM MgC12 was added and the mixture was incubated for 2 hours at 37" in order to hydrolyze the alkali-labile glycosyl-PP-GCL linkage (19). After this treatment lipopolysaccharide is retained by a Millipore filter, while the hydrolyzed substrate passes through the filter.
Samples were collected on a Millipore filter (type HA), washed three times with 5 ml of 0.1 N acetic acid, dried, and counted.
iZfter incubation for 20 min at 37", 2 ml of cold 5% trichloroacetic acid were added, and the product was collected on a Millipore filter (type HA, 0.45 k), washed, dried, and counted (1). Addition of the nonionic detergent to the reaction mixture had no inhibitory effert on the activity of either (;lucosyltrar~sferase Z-The procedure of Miiller et al. (21) was modified as described above for the &actosyltransferase.

Analytical Procedures
Protein-Determinations were made by the method of Lowry (22) Hepfose-The cysteine-HqS04 procedure was used as previously described (23).
Radioactivity was localized by scanning: wit,h a Packard radiochromat.ogra.m scanner model 7201. Counfing Procedures-These were as described earlier (I ) For counting radioactivity on paper, strips were placed in the counting vials, moistened with 0.5 ml of H,O, and allowed to stand for 1 hour prior to addition of 6 ml of toluene-liquifluor-Kosolv scintillation fluid (1).

Site of Synthesis of O-Antigen
In Vivo Pulse and Chase Experiments--It is now well estab- membrane marker (l), and UDP-glucose hydrolase, which is present in both membranes (l), are included for comparison. Galactose-PP-GCL synthetase activity closely paralleled that of DPNH oxidase and was virtually absent from the outer membrane band, H. sequential addition of glucose and galactose according to Reactions 1 and 2: The specific activities of galactose-PP-GCL synthetase, O-antigen polymerase (Reaction 5, Fig. 1) and O-antigen:lipopolysaccharide ligase (Reaction 6, Fig. 1) in the isolated fractions are summarized in Table II. The activibies of the latter two enzymes were determined independently of preceding steps in the pathway by use of isolated trisaccharide-PP-GCL as substrate for the polymerase reaction, and isolated polymer-PP-GCL plus exogenous acceptor lipopolysaccharide as substrates for the ligase reaction.
The enzymes are firmly membranebound, and all the activity of the lysate was recovered in the total membrane fraction.
The distribution of the three activities was similar, and all were clearly localized in the cytoplasmic membrane fractions, Li and Ls. The specific activities in Li were somewhat higher than Lz, as previously observed (1) with DPNH oxidase and other cytoplasmic membrane enzymes. The difference is thought to reflect a somewhat higher contamination of LZ by outer membrane material (1). Some activity was also present in the minor Fraction M, which has been identified as unseparated cell envelope fragments (l), but the residual activity in the outer membrane fraction, H, was less than 10% that of the cytoplasmic membrane.
The values observed were consistent with the levels of residual DPNH osidase activity in this fraction. The distribution of these two activities in isolated cytoplasmic and outer membrane was determined using a modification of the usual assay procedure which permits assay of membrane-bound transferase activity independently of the presence or nature of endogenous lipopolysaccharide.
Reactions were carried out in the presence of exogenous acceptor lipopolysaccharide (added as a lipopolysaccharide-phosphatidylethanolamine complex (20)) and nonionic detergent (see "Experimental Procedures" for details).
In contrast to the enzymes of O-antigen synthesis which are firmly membrane-bound and were associated exclusively with cytoplasmic membrane, the core glycosyltransferase activities are partially soluble (20, al), and significant activity was found in both cytoplasmic and outer membrane (Table III).
However, the distribution of activity between soluble and total membrane fractions, and between cytoplasmic and outer membrane, was dependent on the method employed for preparation of the membranes. When spheroplasts were lysed by sonication, only 20% of the galactosyltransferase and 38% of the glucosyltransferase activity was membrane-bound, and in such preparations the specific activities of cytoplasmic and outer membrane were similar.
After lysis by osmotic shock, 70% of each activity was recovered in the washed total membrane fraction5 and in this case the specific activities in the cytoplasmic membrane fractions (Li and LJ were 3-to 6-fold greater than that of outer membrane.
The results ob-5 Although little or no dissociation of enzyme activity from these membranes was encountered during the washing procedure prior to sucrose gradient centrifugation, partial dissociation of both activities occurred under the conditions of the sucrose gradient. As much as 3570 of the galactosyltransferase activity of the total membrane fraction, and 15 to 207, of the glucosyltransferase, was found in the soluble enzyme fraction at the top of the gradient. by guest on March 24, 2020 http://www.jbc.org/ Downloaded from tained with gentle lysis suggested the cytoplasmic membrane as the major site of (sore synthesis, but the significance of the relatively high activity in the outer membrane, which was not consistent with the localization of enzymes of O-antigen synthesis, remained to be assessed. Soluble galactosyltransferase has been shown by Rothfield and co-workers to bind to lipopolysaccharide-phospholipid complexes in bulk dispersion (27) and in monolayer films (28). This suggested that enzyme activity associated with isolated outer membrane might arise by secondary binding of soluble enzyme released during spheroplast lysis. This possibility was tested by experiments in which partially purified galactosyltransferase I was added during lysis of spheroplasts obtained from an amber mutant in galactosyltransferase I (SL 3657). Spheroplasts were lysed either by sonication or osmotic shock, and membrane fractions isolated by the usual procedure.
Two concentrations of soluble enzyme were employed, the lower of which (14 units per 140 ACOO units of culture) corresponded to the total galactosyltransferase content of an equivalent lysate of strain G-30. The results (Table IV) showed that added soluble enzyme was indeed bound to the membranes, and the binding of the exogenous enzyme was very similar to that of endogenous galactosyltransferase in G-30. Activity was recovered in both cytoplasmic A culture of SL3657 in 550 ml of Medium A was harvested in midexponential phase (A600 = 0.75). Spheroplasts were prepared as usual, and the spheroplast suspension divided into three equal parts. Partially purified soluble galactosyltransferase was added during lysis as follows. In Experiment I, the spheroplast suspensions were sonicated for 15 s, soluble enzyme was then added in the amounts indicated, and sonication continued for two additional 15-s periods. DTT (0.2 mM) was added prior to sonication.
In Experiment II, the spheroplast suspensions were poured with stirring into 4 volumes of cold 1 rnM EDTA-0.25 mM DTT, enzyme was then immediately added where indicated, and the lysates were stirred for 10 min at 0". Membranes were isolated and separated by the usual procedure.
DTT (0.2 mM) was present throughout. Purity of the isolated fractions was checked by assay of DPNH oxidase activity.
Galactosyltransferase activity was determined as described under "Experimental Procedures." Units of activity are expressed as nanomoles per min. The total amount of galactosyltransferase activity in comparable lysates prepared from strain G-30 was 13.2 to 15 units. a Values in parentheses represent specific activity (nanomoles per min per mg of protein). and outer membrane fractions, and the specific activities of the isolated fractions were comparable to those seen in G-30. Residual galactosyltransferase activity observed in the amber mutant in the absence of added enzyme is presumably due to the presence of galactosyltransferase II (a, 6-transferase, see Fig.  8). Entirely comparable results were obtained in similar experiments employing partially purified glucosyltransferase I and an amber mutant in this enzyme (SL 3656). The results of these experiments indicated that secondary binding of soluble core transferases released during lysis could account for essentially all of the activity recovered in the membrane fractions, and the observed distribution of activity was therefore not interpretable in terms of localization in viva. The observed distribution of core transferases was similar to that previously found (1) for endonuclease I and RNAse I. The latter enzymes belong to the group of periplasmic proteins which are partially or totally released into the medium in the osmotic cold shock procedure of Heppel and co-workers (29). The possibility that the core transferases were also periplasmic was therefore tested (Table V) . Although 90% of the known periplasmic enzyme, cyclic phosphodiesterase, and over 40% of the cytoplasmic enzyme, phosphoglucose isomerase were released under the conditions of the experiment, only 10% of the galactosyltransferase activity was found in the shock fluid.
If core transferase activities were normally present in the outer membrane of intact cells, it might be expected that the enzymes would be available to exogenous substrate.
Incorporation of [Wlgalactose from externally added UDP-[WJgalactose into endogenous galactose-deficient lipopolysaccharide was therefore measured in intact cells of a mutant lacking TJDP- A culture of strain G-30 in Medium A (200 ml) was harvested in mid-log phase (Aso = l.O), washed three times with 10 ml of 10 rnM NaCl in the cold, and finally suspended in 10 ml of 20°10 sucrose (w/w)-33 mM Tris, pH 7.3-2 mM EDTA at room temperature.
The suspension was stirred for 10 min at room temperature and centrifuged for 10 min at 12,000 rpm and 25". The supernatant solution (sucrose supernate) was saved in the cold. The pellet was rapidly suspended in 10 ml of cold HZO, stirred 10 min in the cold, and centrifuged at 4" for 10 min at 12,000 rpm. activity was abolished by addition of nonradioactive galactose-I-P, and probably resulted from hydrolysis of UDP-galactose by UDP-sugar hydrolase present in outer membrane (1) and subsequent uptake of galactose-l-l' or free galactose. The results suggested, but did not prove, that functional core t,ransferase activity was absent frorrl the outer membrane. Pulse-Chase Experiments in Viva-More convincing evidence for localization of core biosynthesis in the cytoplasmic membrane was obt.ained by pulse-chase experiments in viuo. In order to monitor synthesis of the core region only, a double mutant, rfb and 7XP-galactose-4-epimerase negative, was employed. In t,he absence of esogenous galac%ose the double mutant produces the gnlactose-deficient, inc~ompletJe ('ore lipopolysa~charide chnracteristicn of epimeraseless mutants (Fig. 8). On addition of galactose to the medium, t,he full core region is synthesized, but maruler. The distribution of radioactivity in the sucrose gradient is shown in Fig. 5, and the recoveries in cytoplasrnic and outer membrane are summarized in Table VI The results tlesrribrd abovr strotlglg support the c~onc*lusion that lipopulysaccharide is initially assembled in the cyto~~lasrriic: membrane and subsequently transferred to the outer membrane by a trnnslocation process of unknown mechanism. Information relevant to the question of the reversibility of translocstion was obtained in experiments designed to test the availability of the lil)opolysaccllaricle of t,he outer membrane to biosynthetic enzymes. The following q uestion was posed. In a conditional mutant in lipopolysacc~hnritle synthesis, can the incomplete lipopol?-sa(~cllaride synthesized durin g growth under nonpermissire conditions be completed following a shift to permissive conditions?
.\vnilability of phospholllalltlose isomerase and UDPgnlactose epimerase mutants, conditionally blocked in biosynthesis of GDP-mannose and UDP-galactose, respectively, made possible test of the question.
Since O-antigen chains and the core region of liI)opolvsaccllaride are assembled independently and by different mechanisms, separate experiments were carried out to determine the ability of pre-existing core lipopolysac-(allaride to accept newly synthesized O-antigen chains, and the ability of pre-existing incomplete core to be completed.
Addition of Newty Synthesized O-Antigen to Pre-existing Core Lipopolysaccharide-The mutant lacking phosphomannose isomerase was used for this esperiment, and was shifted from conditions in which only core lipol,ol?-sitc~(~haride was formed to ronditions permitting O-antigen synthesis and attachment by addition of esogeuous mannose to the Inediunl.
The design of the esperimeut was based on the difference in size between the core poly-snc~rharide and the complete polysacscharide carrying O-antigen chaius, which permits separatiou of the two polymers by gel filtration.
The experimental plan is summarized in Fig. 6. A culture growing in the abseltce of mannose was exposed to [3H]glucose for one generation in order to introduce 3H into the core lil")I,ol!-snc:c:haritle.
The culture was then harvested, and allowed t,o grow for oue additional generation in fresh medium containing nonmdioacti\-e glurose (but no mannose) in order to minimize further incorporation of 3H into lipopolysaccharide during the subsequent shift to permissive conditions. The cells were recovered by rent,rifug:rt,ion and split into t,hree parts. Culture I served as zero time rontrol, and was frozen without further incubation.
Cultures II and III were suspended in fresh medium in the absence (Culture II) and presence (Culture III) In addition, hybrid pal\-sn(,c~llaritle chains containing [W]O-antigen attached to 31-I-labeled rare would be present in III if, and only if, the old Sl%-labeled core were completed by attachment of new O-antigen (Fig. 6). This species of polysaccharide chain could be recognized in gel filtration by a shift of 3H from low molecular weight core polysaccharide into high molecular weight complete polysaccharide. Accordingly, lipopolysacrharide was purified from the cell ellvelope fraction of each culture, and the polysaccharitle chains liberated from the lipid portion by hydrolysis with dilute aretic arid (23). The lipid-free polvsaccharides were purified by rhromatography on DEAE-cellulose prior to gel filtration on Sephndes G-50. Recoveries of radioactivity at each step of tile procedure are summarized in Table VI I. Loss of "1 I during purification was due t,o nonsperific incorporation of radioact i\.it>from [3H]glucose into iioilli~)ol,ol~sar(~~iaride c~omponents. 'l'he purity of the Wlabelrd polysarc~harides from Cultures I :\lltl I I was assessed by tot,al acid hydrolysis and c~lnoltlatogral)l~~ ill Xolvents I alid III.
Four major radioart,ive c~onipoiuitls \rere deeI,, with PF values corresponding to the major const,itueuts of tile core polysarcharide (glucose , g:ilart,ose, gluc~osnmine, an;1 heptose). All of the 14C of the polysncrharide from Culture III was recovered as ['%']mallllose after total hydrolysis, and partial arid hydrolysis yielded [W]oligosarcharides rorresponding to the known O-antigen fragments, h' ~~lac~tos~l-rn:~ii~ios~l-r~i~~~~i~i~~~~ :ilitl mannosyl-rhamnose (31). Gel filtration of the polysacrharides on Sephades C:&W is shown in Fig. 7. The bulk of the 14C emerged in the excluded volume, although about 157; (Table VII) appeared in a second, partially included peak. The distribution of l4C calosely pnralleled that of authentic complete polysaccharide isolated from wild type cells. The latter also showed the minor peak of partially included material, which may represent a population of polysaccharide chains containing abnormally short 0-:llltigen One milliliter of each sample was taken for gel filtration on Sephadex G-50 (Fig.  7). The remainder of polysaccharides I and II and 1 aliquot of III (10,000 cpm) was hydrolyzed in 1 N HCl for 5 hours at loo", and radioactive sugars identified by paper chromatography in Solvents I and III. Radioactive spots were located by cutting the chromatograms into l-cm pieces and counting.
The rest of polysaccharide III was subjected to part,ial acid hydrolysis and chromatography, as previously described (31)   The structure of the core polysaccharides formed in the absence and presence of galactose is illustrated at the top of the figure, which also shows the origin of the melibiose unit. The isotopic composition of the expected core structures and the melibiose derived from them is indicated schematically.
See text for details. LPS, lipopolysaccharides.
described above. The culture was exposed to [3H]glucose for one generation, chased with nonradioactive glucose for one generation, and then divided into three parts. One part was harvested immediately (Culture I), one was grown for one generation in the absence of galactose (Culture II), and the third grown for one generation in the presence of [i4C]galactose (Culture III).
Lipopolysaccharides were isolated and the lipidfree polysaccharides prepared as described above. The 3H content of the full core polysaccharide obtained from Culture III was again employed as an index of the degree of completion of the pre-existing galactose-deficient lipopolysaccharide.
In this case, however, gel filtration failed to separate adequately the full core polysaccharide from the galactose-deficient core. The 3H content of the completed core chains was therefore determined by isolation of melibiose after partial acid hydrolysis of the polysaccharide.
The core region contains two galactosyl residues, linked o-l,3 and a-l,6 to the proximal glucose residue (Fig. 8). This glucose residue is present in the incomplete core produced under nonpermissive conditions, and was labeled during gro\vth in [3H]glucose. hielibiosyl units (a-galactosyl-1 ,6-glucosyl) arising by gnlactosylation of the old 3H-labeled lipopolysaccharide would therefore be expected to contain [3H]glucose as well as [%]galactose, while those derived from de nova synthesis of lipopolysaccharide should contain only nonradioactive glucose. Thus, the ratio of [%]glucose to [i4C]galactose in the isolated melibiose provided a measure of the fraction of old chains which had been completed following addition of galactose.
Melibiose was isolated from the polysaccharide of Culture III as follows.
The preparations from Cultures I and II were carried through the entire procedure as controls as described in the legend of Table VIII. The lipid-free polysaccharide was subjected to partial acid hydrolysis and the oligosaccharide fractions isolated by filtration through Sephadex G-10. Melibiose was purified from the crude oligosaccharide fraction by paper chromatography.
Followin g chromatography in Solvent I (Fig. 9a), the melibiose area was eluted and rechromatographed in Solvent II (Fig. 9b). The recoveries of 3H and r4C at each step of the purification are given in Table VIII, and the theoretical and observed ratios of 3H :i4C in the isolated melibiose are summarized in Table IX An exponentially growing culture of GB-1 in Medium A was exposed first t,o [3H]glucose and then to nonradioactive glucose (see Table  VII for condit,ions Aliquots of the polysaccharides (8 to 9 X 10" cpm) were hydrolyzed in 1 ml of 1 E HCl for 5 hours at lOOa, and hydrochloride was removed by extraction with dioctylmethylamine in CHCl, The data presented here provide strong evidence that the lipopolysaccharide of the outer membrane of X. typhimurium is synthesized in the cytoplasmic membrane and secondarily translocated to its final position in the outer membrane.
The assignment of the site of synthesis and attachment of O-antigen chains is consistent both with the observed distribution of the hiosynthetic enzymes in isolated membrane fractions and the results of pulse-chase experiments in viva. Definitive localization of the site of synt,hesis of the internal core region of lipopolysaccharide proved to be more difficult, since in vitro distribution of core glycosyltransferase activities was found to be ambiguous as a criterion of in vivo localization of these enzymes. In contrast to the enzymes of O-antigen synthesis, the core glycosgltransferases were partially soluble, and activity was recovered in both outer and cytoplasmic membrane fractions.
However, recoIistruction experiments showed that exogenous soluble core trnnsferases, added during spheroplast lysis, Lvere bollrid t,o both membranes in a manner qualitati\rely and quant,itatively similar to that of the endogenous activities.
These findings raise some question as to the extent to whirh these nrtivities are membranebound in the intact cell and the nature of their presumed asso& tion with menlbrane in vivo. It should be emphasized, however, that lysis of spheroplasts and isolation of mernbranes were carried out at low ionic st,rength and in the prcsenre of EDTA. It is possible that these conditions favor solubilization of core glucosyl-and galactosyltransferase activities. Romeo et al. (28) have reported that &I; cl++ is required for interaction of purified galactosyltransferase I with monolayer films containing lipopolpsaccharide plus phosphatidyl ethanolamine, and repeated washing at low ionic strength is known to result in release of the membrane AT&se and other prot,eins in Streptococcus faecalis (32). At present t,he conclusion that core synthesis is localized exclusively in the cytoplasmic membrane rests primarily on the results of pulse-chase experiments in vivo, which showed that over 90% of the gnlactose incorporated into the core region of lipopolysaccharide during a short pulse was recovered in the cytoplasmic membrane fraction.
The failure of the conditional mutant in core biosynthesis to complete the core region of pre-esisting lipopolysaccharide after a shift to permissive conditions also argues that the outer membrane does not participate in lipopolysaccharide synthesis in viva.
The mechanism of translocation of lipopolysaccharide to the outer membrane and the mechanism of assembly of this menbrane are unknown.
The data presented here, and kinetic studies to be described elsewhere4 indicate that translocation of lipopolysaccharide is rapid and unidirectional.
Other studies have shown that the enzymes of phospholipid synthesis are also specifically localized in the cytoplasmic membrane in S. typhimurium (33) and E. coli (33,34), and it is therefore probable that incorporation of phospholipids into the outer membrane requires translocation from sites of synthesis in cytoplasmic mem brane. Although information is lacking on the site of origin of specific out,er membrane proteins, it appears likely that some type of translocation occurs between synthesis of the polypeptides and assembly into the outer membrane structure.
Two extreme types of model of translocation and outer membrane assembly might be envisioned: assembly of a lipopolysaccharidephospholipid-protein