Membrane Topology of the L-Rhamnose-H + Transport Protein ( RhaT ) from Enterobacteria *

The L-rhamnose-H+ symporter (RhaT) is a 344amino acid integral membrane protein, found in many Enterobacteria, which couples the uptake of the sugar L-rhamnose with the inward movement of protons. Based on its hydropathy profile and the application of von Heijne’s “positive inside” rule (von Heijne, G . (1992) J. Mol. Biol. 225, 487-494)’ a model of the Lrhamnose-H+ symport protein (RhaT) is proposed containing 10 transmembrane helices with the NHz and COOH termini in the periplasm. This model was tested by the creation of random B-lactamase (Bla) fusions. The data from 33 unique, randomly generated, RhaTBla fusions and from 5 site-specific fusions supported the proposed topology between transmembrane helices 2-10. However, the localization of the putative first hydrophilic loop and the NHz terminus was not possible because the B-lactamase fusions in this region were shown to be unreliable indicators of the topology of RhaT.

quence can act as a topological reporter molecule when fused to portions of a membrane protein . If the p-lactamase is fused to part of a membrane protein that is normally located in the periplasm, then the plactamase is translocated to the periplasm where it confers resistance to high concentrations of ampicillin. If the plactamase is fused to part of a membrane protein that is in the cytoplasm, then the p-lactamase remains in the cytoplasm, making the cells resistant only to low concentrations of ampicillin. The topology of a number of membrane proteins has been deduced from @-lactamase fusions (see for example, Edelman et al., 1987;Bowler and Spratt, 1989;Wang et al., 1991;Wu et al., 1992;Kampfenkel and Braun, 1993;Weiner et al., 1993). In summary, we find that the 8-lactamase fusion data support a model of RhaT comprising 10 transmembrane regions with the NH2 and COOH termini in the periplasm, different from the 12-helix model with NH2 and COOH termini in the cytoplasm proposed for most sugar-H+ symport proteins. [F'traD36, proA+B+, lucP, lacZAM151 (T. Gibson, Laboratory of Molecular Biology, Cambridge); E. coli strain BB4, hsdR514, hdM, supE44, supF58,lacYl or A(&IZY)6, galK2, galT22, metB1, trpR55, A(argF-lac) U169, [ F ' h P ,hZDM15,proAB,TnlO @etR)]. Bacterial strains were routinely grown in either 2TY or LB media (Maniatis et al., 1982). by the method of Witholt et al. (1976) and lysed by osmotic shock in Preparation of Bacterial Membranes-Spheroplasts were prepared ice-cold deionized water (Henderson and Macpherson, 1986). Membranes were recovered by centrifugation (40,000 X g for 20 min at 4 "C) and resuspended in deionized water by brief sonication (10 S, setting 2) and then resedimented by centrifugation (100,000 X g for 90 min at 4 "C). The membrane pellet was resuspended in deionized water and the protein concentration determined by the method of Schaffner and Weissman (1973).
DNA Manipulations-Plasmid DNA was prepared by the alkaline lysis method (Maniatis et al., 1982). Restriction digests were carried out according to the manufacturer's recommendations (Amersham Corp., Pharmacia LKB Biotechnology Inc., or New England Biolabs), and ligations were performed according to Maniatis et al. (1982). The CaC12 method for preparing competent cells was routinely used (Maniatis et al., 1982). Restriction fragments were isolated from agarose gels by excising the desired band and isolating the DNA by the "glass milk" method (Vogelstein and Gillespie, 1979).
DNA Sequencing-Single-stranded DNA was produced from derivatives of plasmids pJBS633 and pYZ4 using the helper phage R408 (Russel et al., 1986) and sequenced using Sequenase (U. S. Biochemical Corp.). The gene fusion point between the rhaT gene and the 6lactamase gene (bla') was sequenced using oligonucleotide BLAl (5'-CTCGTGCACCCAACTGA-3'), which is complementary to the coding strand of the p-lactamase gene 40 base pairs from the fusion point. Each fusion point was sequenced on only one strand.
Site-specific rhaT-bla gene fusions were created using PCR amplification of a portion of the rhaT gene; because of the possibility of errors introduced by the Vent DNA polymerase (New England Biolabs), the region of DNA which was amplified was resequenced on at least one strand. Sequencing was performed with the oligonucleotide primers used for PCR amplification (oligonucleotides RHA7, RHA8, RHA9, see below).
Separation of Proteins and Western Blotting-SDS-polyacrylamide gel electrophoresis was performed as described by Henderson and Macpherson (1986) and Western blotting as described by Burnette (1981). Proteins were transferred onto ECL-nitrocellulose (Amersham Corp.) and the blots developed using the ECL Western Blotting Kit (Amersham Corp.). The anti-&galactosidase antibody (Boehringer Mannheim) was used at a final concentration of 0.1 pg/ml and incubated with the blot for 2 h at room temperature. The anti-plactamase antibody (5 Prime 3 Prime Inc) was used at a final dilution of 1:5,000 and incubated with the blot at room temperature for 1 h. Secondary antibodies supplied with the ECL kit were used at a final dilution of 1:1,000. Blots were autoradiographed for 5-30 s.
Construction of a rhaT-bla COOH-terminal Gene Fusion-Initially, an rhaT-lacZ gene fusion was constructed, which was subsequently used to construct the rhaT-bla gene fusion. The lac2 gene was fused to a virtually full-length rhaT gene by using a BalI restriction site at the penultimate codon (Ala3") of the S. typhimurium rhaT gene.
Plasmid pJG2OO (Germino and Bastia, 1984) was used to make the rhaT-lucZ gene fusion. This plasmid contained a unique BamHI site that allows the in-frame fusion of genes with locZ via a collagenasesensitive linker. Upstream from the BamHI site is a X PR promoter and the ~1867 gene; thus, expression of the fusion gene is repressed at 33 "C but not at 42 'C. A BstEII/BalI restriction fragment from plasmid pJAR18 (Tate et al., 1992) which encompassed the S. typhimurium rhaT gene was purified from an agarose gel. The BstEII restriction site was 80 base pairs upstream from the initiation Met, and BalI cleaved the penultimate codon of the rhaT gene. The BstEII site was blunt ended using DNA polymerase Klenow fragment, which allowed the ligation of BamHI linkers (CGGATCCG, Pharmacia) onto both ends of the DNA fragment. After digestion with BamHI, the DNA fragment was separated from linkers on a Sepharose CL-6B 200 spun column (Sambrook et al., 1989). The BamHI-linkered DNA was ligated into BamHI-digested plasmid pJG2OO to produce plasmid pCGT17 (see Fig. 2), which contained the rhaT-lacZ gene fusion under the control of the X PR promoter. The correct orientation of the rhaT gene in the vector was checked by measuring the rate of uptake of L-["Clrhamnose in heat-induced TG2(pCGT17).
The h Z gene in the rhaT-lacZ gene fusion expressed from plasmid pCGT17 was replaced with the bla' gene to create an rhaT-bla fusion at codon in the S. typhirnurium rhaT gene. A BamHI restriction fragment that contained the rhaT gene in plasmid pCGT17 was ligated into BamHI-digested plasmid pYZ4 . The orientation of the gene was determined by an HincII/ Hind111 digestion; plasmid pCGT33 contained the rhaT gene in the correct orientation downstream from the lacUV5 promoter for gene expression. The BamHI restriction site proximal to the lacUV5 promoter was then destroyed; gel-purified plasmid pCGT33 that had been partially digested with BamHI was treated with DNA polymerase Klenow fragment to blunt end the DNA and was then religated and transformed into E. coli strain BB4. Plasmid pCGT36 contained a single BamHI site at the end of the rhaT gene.
The rhaT-bla gene fusion was made in a fashion analogous to that described by Broome-Smith et al. (1990). Plasmid pCGT36 was digested with BamHI, and the single-stranded DNA ends were re-moved with S1 nuclease (Pharmacia Nested Deletion Kit). The plasmid was then digested with EcoRI and ligated with a PuuII/EcoRI restriction fragment from plasmid pYZ5  which contained the truncated bla' gene. The ligation mix was transformed into strain TG1 and plated out on LB plates that contained 50 pg/ml kanamycin. Transformants that expressed an RhaT-Plactamase fusion protein were selected by picking colonies onto LB plates that contained 0.5 mM isopropyl 1-thio-@-galactopyranoside, 20 pg/ml ampicillin, and 50 pg/ml kanamycin. The precise junction of the gene fusion on plasmid pCGT38 (see Fig. 2) was determined by DNA sequencing using oligonucleotide primer BLAl (5'-CTCGTGCACCCAACTGA-3'), which is complementary to the coding strand of the 0-lactamase gene.
Construction of Plasmid pCGT15 and the Generation of Random rhaT-bla Fusions-Plasmid pCGT15 (see Fig. 2) was constructed from plasmid pJBS633 (Broome-Smith and Spratt, 1986) and from plasmid pCGT13,' which expresses two S. typhimurium rhaT genes under the control of the X P L promoter. To provide tight control of the X P L promoter, the CIS? gene was ligated into plasmid pJBS633; a 900-base pair EcoRIIBamHI restriction fragment from plasmid pJG2OO (Germino and Bastia, 1982) which carried the CIS, gene was ligated into EcoRI/BamHI-digested plasmid pJBS633, to form plasmid pCGT14. A 3.4-kilobase BgZII/BamHI restriction fragment from plasmid pCGT13, which carried the two rhaT genes under the control of the X PL promoter, was then ligated into BamHI-digested plasmid pCGT14. Two orientations of the insert in plasmid pCGT14 were obtained as determined by restriction enzyme digests using HincII, EcoRIIBamHI, and SphI. Plasmid pCGT15 (see Fig. 2) contained the rhaT genes in the correct orientation for the generation of rhaT-bla fusions.
BamHIISphI-digested plasmid pCGT15 was digested with exonuclease 111 (Nested Deletion Kit) to create unidirectional nested deletions in the rhaT genes. After the deletion reactions were completed, the DNA was blunt ended with S1 nuclease. Carrier tRNA (5 pg) was added to the DNA from each time point, and then the DNA was purified by phenol-chloroform extractions and ethanol precipitation. The DNA was digested with PuuII and then repurified before the DNA was recircularized with T4 DNA ligase. The ligation mixtures were transformed into CaCl'-competent TG1 cells and plated out onto LB plates that contained 50 pg/ml kanamycin and grown at 33 "C.
To determine which of the kanamycin-resistant transformants expressed p-lactamase activity, colonies were toothpicked onto LBkanamycin plates that contained 20 and 200 pg/ml ampicillin; colonies that expressed an RhaT-Bla fusion where the p-lactamase was in the cytoplasm grew only on plates that contained 20 pg/ml ampicillin, whereas colonies that expressed an RhaT-Bla fusion with the p-lactamase in the periplasm also grew on the plates that contained 200 pg/ml ampicillin (Broome-Smith and Spratt, 1986). This preliminary analysis was substantiated by defining the concentration of ampicillin which was required to kill single cells that expressed the RhaT-Bla fusions. Overnight cultures of strains expressing the fusion proteins were grown at 33 "C in 2TY that contained 50 pg/ml kanamycin. 5 p1 of a dilution of the overnight cultures was spotted onto LB plates that contained increasing amounts of ampicillin (2,3, 4, 5, 7.5, 10, 12.5, 15, 20, 25, 30, 35, and 50 pglml). To ensure reproducible results, the ampicillin plates were prepared freshly on the day of use and dried for 1 h at 37 "C before use. Ampicillin solutions were always made freshly from the solid and added to molten LB-agar that had been cooled to 45 "C. Expression of the fusion proteins was induced by growing the cells at 39 "C; full induction at 42 "C was lethal.
Generation of Site-specific rhaT-bla Fusions-Three site-specific rhaT-bla fusions were created in the S. typhimurium rhaT gene, and one site-specific fusion was created in the E. coli rhaT gene, using PCR-amplified DNA. Oligonucleotide primer RHA3 was homologous to the region downstream from the initiator Met codon but introduced an NcoI site at the initiator Met codon; this oligonucleotide was used in the construction of all of the site-specific rhaT-bla fusions. Oligonucleotides RHA7, RHA8, and RHA9 were used to construct RhaT-Bla fusions at amino acid residues Thr36, Gly%, and TyrS6 in S. typhimurium RhaT, respectively. Each of these PCR primers introduced a PuuII restriction site after the codon of interest. In addition, oligonucleotide primers RHA3 and RHA7 were used to create an rhaT-bla fusion at the codon Thr36 in the E. coli rhaT gene. The sequences of the oligonucleotides (5' + 3') were as follows.

RHAS: TTTTCAGCTGATAACGCATGGTCAGG
PCR was performed with Vent DNA polymerase using the manufacturer's protocol. Plasmid pJAR18 was used as a template for the S. typhimurium gene, and plasmid pCGT12 was used a template for the E. coli rhaT gene (Tate et al., 1992). After 15 rounds of amplification of 25 ng of template DNA, the PCR products were purified as described by Crowe et al. (1991). The DNA was then digested with NcoI and PuuII, repurified, and ligated into NcoI/EcoRI-digested plasmid pYZ4 in the presence of a gel-purified PuuII/EcoRI fragment from plasmid pYZ5 which contained the bla' gene . The ligation mixture was transformed into CaCl,-competent TG1 cells and plated out onto LB plates that contained 50 fig/ ml kanamycin. Transformants were toothpicked onto LB-kanamycin plates that contained 20 pg/ml ampicillin to identify the transformants that expressed RhaT-Bla fusions. Although the site-specific rhaT-blu gene fusions were expressed from a lacUV5 promoter under the control of the Lac1 repressor, it was found that sufficient expression occurred in the absence of inducer (isopropyl 1-thio-BD-galactopyranoside) for characterization of the fusions. The concentration of ampicillin required to kill single cells that expressed the RhaT-Bla fusions was determined by spotting 5 pl of a dilution of an overnight culture (grown in 2TY at 33 'C) onto plates that contained increasing concentrations of ampicillin, prepared as described above. Plates were incubated overnight at 37 "C. The DNA encoding rhuT from colonies that expressed RhaT-Bla fusions was completely sequenced on at least one strand.

Construction and Analysis of the COOH-terminal RhaT-LacZ and RhaT-Bla Fusion Proteins-A model of the
RhaT protein derived from a hydropathy plot and the location of positively charged amino acid residues in the cytoplasm suggested that the RhaT protein contained 10 transmembrane regions with both the NH2 and COOH termini in the periplasm (Fig. l). To provide initial evidence for this model, the location of the COOH terminus of RhaT in either the periplasm or cytoplasm was determined using gene fusions. A unique BalI restriction site in the penultimate codon of the S. typhimurium rhaT gene was used to construct an rhaT-lac2 gene fusion and an rhaT-bla gene fusion (see "Experimental Procedures").
The initial gene fusion constructed with the @-galactosidase gene lac2 was expressed under the control of the X PL promoter in plasmid pCGT17 (Fig. 2), but the levels of LacZ activity were intermediate between levels expected if the fusion point was in a periplasmic domain or a cytoplasmic domain (results not shown). In addition, the rhamnose uptake activity expressed in heat-induced strain TG2(pCGT17) was low. This could have been caused by either poor expression of the fusion protein or a perturbation in the structure of RhaT by LacZ, rendering the transporter less active. Western blot analysis of membranes derived from heat-induced strain TG2(pCGT17) showed that an RhaT-LacZ fusion protein (Mr 140,000) was expressed and that only a small proportion of it was degraded (results not shown). These results did not conclusively locate the COOH terminus of RhaT to either the periplasm or cytoplasm, so an RhaT-@-lactamase fusion was constructed.
Bla, unlike @-galactosidase (LacZ), is active when it is fused to a periplasmic domain of a membrane protein . The activity of the @-lactamase in uiuo is assessed by the amount of ampicillin which is required to kill single cells that express the fusion protein; less than 5 pg/ml ampicillin will kill cells that express @-lactamase fused to a cytoplasmic domain of a membrane protein, but @-lactamase  fused to a periplasmic domain confers resistance to much higher levels of ampicillin (Broome-Smith and Spratt, 1986;Edelman et al., 1987). The rhaT-bla gene fusion was constructed, using the rhaT-lac2 gene fusion as the source of DNA ("Experimental Procedures"), and expressed in plasmid pCGT38 under control of the lacUV5 promoter; the DNA sequence of the fusion point was confirmed (Fig. 3). The activity of the 8-lactamase portion of the RhaT-Bla fusion protein was tested by spotting 5 pl of a dilution of an overnight culture of strain TGl(pCGT38) onto a plate that contained 5 pg/ml ampicillin. The cells that contained plasmid pCGT38 grew, but control cells that expressed a fusion gene with p-lactamase fused to a putatively cytoplasmic portion of RhaT (fusion at T y P , see below) did not grow. Indeed, strain TGl(pCGT38) grew on plates that contained up to 300 pg/ml ampicillin in the presence of 0.5 mM isopropyl 1-thio-@D-galactopyranoside to induce gene expression. This confirmed that the COOH terminus of RhaT was in the periplasm.
The transport activity of the RhaT-Bla fusion protein was difficult to assess because of proteolysis of the fusion protein (Fig. 3). Western blot analysis of the fusion protein utilized an anti-8-lactamase antibody, so that the total amount of potentially functional RhaT (RhaT-Bla fusion and RhaT) could not be determined. A prominent RhaT-Bla degradation product (Mr 27,000) was observed to increase with the length of induction time, whereas the RhaT-Bla fusion protein (Mr 46,000) decreased with time (Fig. 3). In contrast to the decreasing amounts of RhaT-Bla fusion protein in the membrane, the L-rhamnose uptake activity of induced strain TGl(pCGT38) actually increased (Fig. 3). The implication of this observation was that the RhaT-Bla fusion protein was less active than the native protein; removal of the p-lactamase portion of the fusion protein by proteolysis may be responsible for the regeneration of active RhaT, resulting in the increase in uptake activity observed.
Isolation and Characterization of Randomly Generated a.  lanes 1 and 5.0 h; lanes 2 and 6, 1 h; lanes 3 and 7, 2  h; lanes 4 and 8 , 3 h. rhaT-bla Fusions-The localization of the COOH terminus of the RhaT protein to the periplasm was consistent with the orientation of RhaT proposed in Fig. 1; to determine the number of transmembrane regions present in RhaT, p-lactamase fusions were constructed throughout the S. typhimurium rhaT gene. All of these gene fusions were expressed under the control of the X P L promoter. Random deletions were made in the rhaT genes in plasmid pCGT15 with exonuclease I11 (see "Experimental Procedures"). The plasmid DNA was then blunt ended with S1 nuclease and digested with PuuII. Religation of the partially deleted rhaT genes with the bla' gene (see "Experimental Procedures") and transformation into TG2 resulted in more than 5,000 kanamycin-resistant colonies. T o determine which colonies expressed RhaT-Bla fusion proteins, 2,000 colonies were toothpicked on to LB plates that contained 20 pg/ml ampicillin and were grown overnight a t 37 "C. The high inoculum introduced onto the ampicillin plates with the toothpick ensured that cells expressing @-lactamase fused to a cytoplasmic portion of RhaT could grow, but cells that did not express any RhaT-Bla fusion protein could not grow. At 37 "C a small amount of expression from the X PL promoter, under the control of the cIa7 repressor, produced sufficient RhaT-Bla fusion protein to give ampicillin resistance to the cells. Of the 175 ampicillin-resistant colonies obtained, 66 were characterized further by sequencing the gene fusion point using the sequencing primer oligonucleotide BLA1. DNA sequencing showed that 49 of the plasmid constructs contained in-frame rhaT-bla gene fusions, and 7 of the plasmid constructs expressed cro-bla gene fusions; part of the cro gene is present downstream from the X PR promoter in the 900-base pair fragment from plasmid pJG200 which contained the cIW7 gene (Fig. 2). All of the cro-bla fusion genes expressed a protein in which the p-lactamase activity was predicted to be in the cytoplasm (results not shown). Of the 49 rhaT-bla gene fusions obtained, 33 were unique; 14 fusions were sequenced with a fusion point a t either the codon for Ser2I6 or Tyr217, presumably representing a region of DNA refractory to exonuclease I11 digestion.
The activity of p-lactamase in cells that expressed the rhaTbla fusions was determined by spotting 5 pl of a dilution of an overnight culture of the cells onto LB plates that contained increasing concentrations of ampicillin (see "Experimental Procedures"). The results shown in Table I were compiled from a single experiment in which all of the plates were poured at the same time. These results were consistent with individual determinations carried out during the sequencing of the fusion points (results not shown). The positions of the RhaT-Bla fusions were identified on the topological model of RhaT (Fig. 4) and were found to be mainly in the COOH-terminal half of the protein; only six fusions were obtained in the region encoding the first six putative transmembrane domains.
Isolation and Characterization of Site-specific RhaT-Bla Fusions-To improve the coverage of fusions in the NHa-terminal half of RhaT, three site-specific rhaT-bla fusions were constructed in plasmid pYZ4 under the control of the lacUV5 promoter (see "Experimental Procedures"). An additional site-specific fusion was constructed at the codon for Thr3'j in the E. coli rhaT gene (see "Discussion"). The DNA encoding part of the rhaT gene in the site-specific fusion constructs was resequenced on at least one strand to ensure that no mutations had occurred during the PCR amplification that was used to create the fusions. The gene fusions in plasmids pCGT40, pCGT44, and pCGT50 (see Table I) all contained an unchanged DNA sequence, except at the initiator Met

TABLE I P h m i d s expressing rhaT-bla fusions
Plasmids pCGT51-83 were made by exonuclease 111 deletions of pCGT15. Letters after a plasmid name designate individually isolated clones. Plasmids pCGT40, pCGT43, pCGT44, and pCGT50 were sitespecific fusions expressed in plasmid pYZ4. Plasmid pCGT43 contained part of the s. typhimurium (St) rhaT gene, and plasmid pCGT50 contained a fusion at the same position but in the E. coli (Ec) rhaT gene. All other constructs were derived from the S. typhimurium rhaT gene. The rhaT-bla fusion in plasmid pCGT38 was constructed using a BalI restriction site in the S. typhimurium rhaT gene. The last amino acid residue of RhaT before the fusion point is given, and the orientation of the @-lactamase moiety is indicated by P (periplasmic) or C (cytoplasmic). The concentration of ampicillin (Fg/ml) which causes cell death is given in the final column.

Plasmid
Fusion point 8-Lactamase Ampicillin concentration location causing cell death Tyr'" Ar2" codon where the insertion of the NcoI site required for cloning altered the protein sequence from MSNAI-to MGNAI-. In our experience, introduction of such an NcoI site did not alter the level of expression of the rhaT, galP, or araE sugar-H+ symporter genes.3 The DNA sequence at the initiator Met of the rhaT-bla gene fusion in plasmid pCGT43 was unexpectedly identical to the wild type rhaT gene sequence i.e. there was no NcoI site in this construct.
Determination of the Level of Expression of RhaT-Bla Fusion Proteins by Western Blot Analysis-To ensure that the levels of fusion protein expression were not affecting the topological analysis, membranes were prepared from some C. G. Tate, M. T. Cairns, and P. J. F. Henderson, unpublished data. strains expressing RhaT-Bla fusion proteins and analyzed by Western blotting with an anti-/3-lactamase antibody. Bacterial membranes were prepared from strains containing plasmids pCGT38, pCGT40, pCGT43, pCGT44, pCGT50, pCGT53, pCGT55, pCGT57, pCGT67, pCGT76, and pCGT82 (see Table I). This is a representative sample of all the rhaTbla fusions, each plasmid expressing a fusion protein with @lactamase fused to a different hydrophilic loop in RhaT. Membrane proteins were separated by SDS-polyacrylamide gel electrophoresis and blotted onto nitrocellulose (see "Experimental Procedures"). The blots were probed with an anti-P-lactamase antibody (Fig. 5). In each lane there was a single major band and a number of minor bands. The most intense band was assumed to be the intact fusion protein, and the less intense bands were assumed to be either oligomers of the RhaT-Bla fusion protein or proteolytic products. The relative mobilities of the RhaT-Bla fusion proteins on SDS-polyacrylamide gel electrophoresis was consistent with the increasing size of the fusion proteins, but the molecular masses determined from the gels were less than would be expected from the molecular mass of p-lactamase and the portion of RhaT fused to it. The RhaT-Bla fusion protein with @-lactamase fused at Ala343 (encoded by plasmid pCGT38) has an expected molecular mass of 66 kDa but has an apparent molecular mass of 46 kDa on SDS-polyacrylamide gels. This is probably a consequence of the abnormal mobility of the RhaT protein on SDS-polyacrylamide gel electrophoresis; RhaT has a molecular mass of 37 kDa but has an apparent molecular mass of 27 kDa on SDS-polyacrylamide gels' (Baldoma et al., 1990).
Despite the wide range in levels of expression of the RhaT-Bla fusion proteins, there seemed to be no correlation between the amount of protein that was expressed and the localization of the p-lactamase to either the periplasm or the cytoplasm. P-Lactamase fusions a t T h P , Ser'43 and (lanes 2, 6, and 7, Fig. 5a) were expressed at relatively high levels, similar to levels of protein expressed by fusions a t Gln3' and Ala343 (Fig. 5a, lane 3, and Fig. 5b, lane 2 ) . However, the in vivo assay for p-lactamase activity concluded that the p-lactamase moiety was in the cytoplasm for fusions at ThP', Ser243, and whereas fusions a t Gln31 and Ala343 resulted in the plactamase activity in the periplasm (Fig. 4). Problems could potentially arise if the fusion proteins were expressed at very low levels, as observed for fusions at Tyrg6, and GlyZo6, because there might be insufficient 6-lactamase expressed to protect the cell from lysis, thus giving an erroneous indication of topology. However, the low levels of protein expressed for fusions at and Gly206 gave cells resistance to 25 and 30 pg/ml ampicillin, respectively (Fig. 4), indicating that the plactamase moiety is in the periplasm. In contrast, the similar level of fusion protein expressed by plasmid pCGT40 (fusion at Tyrg6 ; Fig. 5b, lane 3 ) does not give resistance to high levels of ampicillin and was therefore proposed to be in the cytoplasm. We concluded from these Western blots that the amount of fusion protein expressed in this representative sample of fusions did not affect the topological analysis of RhaT. Weiner et al. (1993) used P-lactamase fusions to study the topology of the anchor subunit of E. coli dimethyl sulfoxide reductase (DmsC), and, in addition to using the in vivo 8lactamase assay used here, they also used a colorimetric assay for measuring P-lactamase activity. These colorimetric assays on the whole confirmed the analysis by the in vivo assay, but in some cases the colorimetric assays gave very low values because of proteolysis of the fusion protein. In these cases, the topological information was derived solely from the in vivo assay. 4. Positions of the 8-lacta-mase fusions on the topological model of RhaT derived from hydropathy analysis and charge distribution. The positions of the 8-lactamase fusions are indicated by circled or boxed residues. Circled residues indicate that the RhaT-Bla fusion conferred resistance to high concentrations of ampicillin (6-lactamase in the periplasm), whereas cells that expressed RhaT-Bla fusion proteins shown by the boxed residues were killed by low levels of ampicillin (6-lactamase in the cytoplasm). The concentration of ampicillin (pg/ml) required to kill single cells expressing fusion proteins is shown for each fusion in bold figures. The number of each amino acid residue is also shown. All fusions are with the S. typhimurium rhaT gene, except at T h P where a fusion was also made to the E. coli rhaT gene. The RhaT-Bla fusion at Alas3 was constructed from a BalI restriction site at the end of the S. typhimurium rhaT gene. Each lane contained 30 pg of protein from membranes prepared from E. coli strains by spheroplast lysis (see "Experimental Procedures"). Cells that expressed the rhaT-bla fusions from the lacUV5 promoter in plasmid pYZ4 were grown at 37 "C until they reached an A m of 3; the cells were then harvested and membranes prepared. Cells that expressed the rhT-bla fusions from the X PL promoter under the control of the cIu, repressor protein were grown overnight at 39 "C and harvested.  [Gly=]. The material in square brackets refers to the final amino acid residue in RhaT and its position in the RhaT protein before the fusion point with 8-lactamase. The letters underneath the lane numbers in the figure refer to the orientation of the p-lactamase part of the fusion proteins. P, periplasmic; C, cytoplasmic.

DISCUSSION
The topological model for S. typhimurium RhaT depicted in Fig. 4 was derived from the analysis of the hydropathic profile and the distribution of charged residues (Tate et al., 1992). The E. coli RhaT sequence is 91% identical to the S. typhimurium sequence and was predicted to have an identical topology. Although the NH2 terminus of RhaT is predicted to be in the periplasm, there is no NH2-terminal leader peptide sequence. The positions of all of the p-lactamase fusions on this model and the localization of the p-lactamase portion of the fusion proteins are shown in Fig. 4. In addition, the concentrations of ampicillin required to kill single cells that expressed RhaT-Bla fusion proteins are shown in Table I. Two RhaT-Bla fusions were obtained at the beginning of the RhaT protein (fusions a t Ser2 and Gly'). Since these fusions contained too few amino acid residues of RhaT to provide any means of crossing the membrane, the p-lactamase had to be located in the cytoplasm. Consistent with this conclusion was the observation that single cells that expressed these fusion proteins were killed by 2 or 3 pg/ml ampicillin. All other strains that expressed RhaT-Bla fusion proteins which were killed by less than 5 pg/ml ampicillin were assumed to express the p-lactamase portion of the fusion protein in the cytoplasm (Broome-Smith and Spratt, 1986). The concentration of ampicillin required to kill single cells that expressed the plactamase portion of the fusion proteins in the periplasm varied between 7.5 and 30 pg/ml. The generally low value of these figures compared with other fusion data (Broome-Smith and Spratt, 1986;Edelman et al., 1987;Wu et al., 1992) was probably caused by the low levels of expression from the X promoter or from the lacUV5 promoter. This was done deliberately to prevent high levels of RhaT expression which are known to be lethal to the cell.2 Indeed, cells that expressed RhaT-Bla fusion proteins under the control of the X promoter and the cIs7 repressor protein could not grow on LB plates at 42 "C. One reason why very few fusions were obtained in the NH2-terminal half of RhaT might be that fusions in this region are more toxic and therefore would be less likely to be isolated in the initial screening. Certainly the site-specific fusions created at Gly% and Tyr" resulted in cells that could be visibly distinguished from normal E. coli by their poor growth on LB plates, despite the absence of any inducer (i.e. isopropyl 1-thio-PD-galactopyranoside) for the lacUV5 promoter. T o ensure that the levels of fusion protein expression were not affecting the topological analysis, membranes were prepared from some strains expressing RhaT-Bla fusion proteins and analyzed by Western blotting with an anti-p-lactamase antibody (Fig. 5). The levels of expression of the RhaT-Bla fusion proteins varied enormously, but there seemed to be no correlation between the amount of protein that was expressed and the localization of the p-lactamase to either the periplasm or the cytoplasm. The fusion data in the COOH-terminal half of the protein are consistent with the model proposed from the hydropathy plot and charge distribution (Fig. 4). In addition, the smaller number of fusions in the hydrophilic loops between transmembrane regions 2, 3, 4, and 5 are consistent with the localization of these loops proposed by the model. The fusion in transmembrane region 3 at Thr7' suggested that this region should be in the cytoplasm; as 9-11 hydrophobic amino acids are thought to be sufficient to anchor the P-lactamase in the membrane, residue T h P was placed in the cytoplasmic leaflet of the membrane so that the transmembrane region started at residue 62 rather than 69 (compare Fig. l a with Fig. 6a). However, such conclusions must be treated with extreme caution as they are based only upon the assumption that 9-11 hydrophobic amino acid residues are required for the translocation of the P-lactamase across the membrane (Calamia and Manoil, 1990) and that no protein-protein interactions occur between transmembrane helices (see below). In addition, it is a tacit assumption that transmembrane regions are perpendicular to the plane of the membrane, although this is clearly not the case for bacteriorhodopsin (Henderson et al., 1990) and a plant light-harvesting complex (Kuhlbrandt and Wang, 1991).
A major discrepancy between the model and the P-lacta- mase fusion data for the S. typhimurium RhaT protein occurred in the first hydrophilic loop, between putative transmembrane regions 1 and 2. The model predicted that this region was in the cytoplasm, but the P-lactamase fusion data predict that this region is in the periplasm (Fig. 4). If the data from the P-lactamase fusions are considered alone, then a model for RhaT can be produced (Fig. 6), suggesting that the NH2 terminus is in the cytoplasm, the COOH terminus is in the periplasm, and that there are only nine transmembrane regions. However, this nine-helix model does not conform to the positive inside rule; in a recent analysis of 24 bacterial membrane proteins using this rule, 23 of the predicted topologies were identical to the topologies established by experiment (von Heijne, 1992). This problem with the nine-helix model is exacerbated when the protein sequence of the E. coli RhaT protein is considered. In the region between putative transmembrane regions 1 and 2 in the S. typhimurium RhaT sequence (Fig. 4) there are 2 Lys residues, whereas in the E. coli protein there are 4 Lys residues (both Gln residues, GlnZ8 and Gln31, in the S. typhimurium sequence are Lys in the E. coli sequence). All of the studies performed on altering the topology of leader peptidase suggest that a group of Arg or Lys residues will alter the topology of the protein so that they will be in the cytoplasm (Anderson et al., 1992), suggesting that the first hydrophilic loop in E. coli RhaT is most likely to be in the cytoplasm. An additional reason why the ninehelix model seems unfavorable is that a hydrophobic region (helix 2 in Fig. 6) is placed totally in the periplasm. Can the @-lactamase fusion data in the region of the first hydrophilic loop be reconciled with the 10-helix model for RhaT? As mentioned above, an assumption of the analysis of any gene fusion technique is that the fusion proteins behave in an identical fashion to the same region in the native protein. Alternatively, it is an assumption that there are no interactions between transmembrane helices required to maintain topology. This assumption may not be true for the B-lactamase fusions at Gln3' and T h P , i.e. the topology of helix 1 may be maintained by an interaction with, for example, helix 2, and, if this were the case, then the p-lactamase fusion data in this region would be an unreliable indicator of the topology of RhaT. The analysis of a P-lactamase fusion made at Thr3'j in the E. coli RhaT protein was performed to see if the localization of Thr3'j in the E. coli RhaT protein was predicted to be the same as for the S. typhimurium protein.
The only difference between the E. coli and S. typhimwium Thr36 p-lactamase fusions is that the E. coli fusion has 4 Lys residues in the region between the first putative transmembrane region and the p-lactamase, whereas the S. typhimurium sequence has only 2 Lys residues. The ampicillin concentration required to kill single cells that expressed the E. coli Thr3'j fusion was 4 pg/ml, compared with 15 pg/ml for the S. typhimurium protein (Table I). Thus, the p-lactamase fusions at Thr3'j predict a different localization of this residue in the RhaT protein sequence depending upon the source of the protein. Therefore, we have concluded that the P-lactamase fusion data in the first hydrophilic loop of RhaT are an unreliable indicator of the topology of RhaT and that they should not influence the final model of RhaT based upon this analysis. However, the RhaT-Bla fusions at Thr3'j created from the E. coli and S. typhimurium rhaT genes are both associated with the membrane (Fig. 5 ) , suggesting that the first hydrophilic region of the RhaT protein could be inserted in the membrane. This would be consistent with the 10-helix model in Fig. 4, with the NHz terminus in the periplasm. A different technique, such as chemical labeling, will be needed to assign definitively the location of the NH, terminus.