The role of charged residues in determining transmembrane protein insertion orientation in yeast.

The first 79 residues of the yeast Ste2p G protein-coupled pheromone receptor, including the negatively charged N-terminal domain, the first transmembrane segment, and the following positively charged cytoplasmic loop, has been fused to a Kex2p-cleavable β-lactamase reporter. Insertion orientation was determined by analysis of cell-associated and secreted β-lactamase activities and independently corroborated by analysis of membrane association and glycosylation patterns. This fusion inserts with exclusively N terminus exofacial (Nexo) topology, serving as a model type III membrane protein. Orientation is unaffected by removal of all three positively charged residues in the cytoplasmic loop or by deletion of all but 12 residues from the N-terminal domain. The residual −2 N-terminal charge apparently provides a signal sufficient to determine Nexo topology. This is entirely consistent with the statistically derived rule in which the charge difference, Δ(C-N), counted for the 15 immediately flanking residues, is the primary topology determinant. Mutations altering Δ(C-N) to zero favors Nexo insertion by 3 to 1, whereas increasingly negative values cause increasing inversion of orientation. All results are consistent with the charge difference rule and indicate that whereas positive charges promote cytoplasmic retention, negative charges promote translocation.

The first 79 residues of the yeast Ste2p G proteincoupled pheromone receptor, including the negatively charged N-terminal domain, the first transmembrane segment, and the following positively charged cytoplasmic loop, has been fused to a Kex2p-cleavable ␤-lactamase reporter. Insertion orientation was determined by analysis of cell-associated and secreted ␤-lactamase activities and independently corroborated by analysis of membrane association and glycosylation patterns. This fusion inserts with exclusively N terminus exofacial (N exo ) topology, serving as a model type III membrane protein. Orientation is unaffected by removal of all three positively charged residues in the cytoplasmic loop or by deletion of all but 12 residues from the Nterminal domain. The residual ؊2 N-terminal charge apparently provides a signal sufficient to determine N exo topology. This is entirely consistent with the statistically derived rule in which the charge difference, ⌬(C-N), counted for the 15 immediately flanking residues, is the primary topology determinant. Mutations altering ⌬(C-N) to zero favors N exo insertion by 3 to 1, whereas increasingly negative values cause increasing inversion of orientation. All results are consistent with the charge difference rule and indicate that whereas positive charges promote cytoplasmic retention, negative charges promote translocation.
The structure and function of integral membrane proteins is determined by the topology with which their peptide backbone is threaded through the phospholipid bilayer. This topology depends, in turn, on the insertion orientation of each transmembrane (TM) 1 segment, the N termini of which can be either cytoplasmic (N cyt ) or exofacial (N exo ). With very few known exceptions (1), membrane proteins adopt a single unique topology. The TM segments of multispanning membrane proteins necessarily alternate in their insertional orientation so that their orientation appears to be dictated by the insertion of the first; downstream TM segments, however, may have independent topogenic determinants (2). In eukaryotes, excluding components of mitochondria, other plastids, and peroxisomes, almost all TM proteins are inserted at the endoplasmic reticulum (ER) and topology is normally determined by the events accompanying insertion.
With the exception of the ␤-barrel proteins found in the outer membrane of Gram-negative bacteria, the vast majority of TM segments are ␣-helices of 18 or more mostly hydrophobic amino acid residues. Proteins of this type, whose mature forms span the membrane once, are commonly categorized into three major groups: Type I proteins contain a cleavable signal sequence at the N terminus. The signal recognition particle binds to this signal and targets the nascent polypeptide chain to the ER, where signal function initiates transfer of the mature N terminus of the protein across the ER membrane; a second hydrophobic domain acts as an anchor or stop-transfer sequence so that the mature protein adopts an N exo topology. Type II proteins contain an uncleaved signal/anchor domain and adopt the opposite (N cyt ) topology. Type III proteins also contain an uncleaved anchor sequence, but the proteins insert with N exo orientation. All three classes utilize the Sec machinery for insertion at the ER. Type IV proteins have a C-terminal signal/ anchor domain and adopt the N cyto topology but apparently use a different entry mechanism (3).
The topogenic signals that dictate the orientation of TM segment insertion consist principally of charged residues within the amino acid sequences flanking the TM segment (4). In prokaryotes, statistical analysis of sequences and extensive analyses of model proteins has led to formulation of the "positive inside rule"; positive charges impede translocation and cause retention of the adjacent end of the TM segment in the cytoplasm (4,5). Arginine and lysine are of principal importance; histidine plays little part because of its low average degree of ionization at physiological pH. Negatively charged residues, aspartate and glutamate, also appear to have little effect. The dominance of positive charges is thought to reflect enhancement of ionization of amino groups and suppression of ionization of carboxylate groups on insertion into the membrane (6). This topogenic signal has no discernible sequence conservation, suggesting an electrostatic mechanism. The receptor for this signal is, at least in part, the TM potential at the bacterial cytoplasmic membrane (7,8).
In eukaryotes, empirical analyses of the topology of TM proteins indicate a similar reliance of insertion orientation on adjacent charge, frequently interpreted as indicating compliance with the positive inside rule; in particular, analysis of the paramyxovirus HN protein, a model type II TM protein, has suggested a dominant effect for positively charged N-terminal residues (9 -12). Orientation correlates best, however, with ⌬(C-N), the difference in charges within an arbitrarily chosen window of 15 residues flanking the TM segment on either side (13). This "charge difference" rule differs from the positive inside rule in giving equal weight to positive and negative charges; the terminus with the most positively charged adjacent segment is retained in the cytoplasm so that positive and negative values of ⌬(C-N) correlate with N exo and N cyt orientation, respectively. Statistically, a net charge difference of zero is also found to favor N exo orientation (13). As in bacteria, the lack of sequence conservation in these topogenic signals implies an electrostatic mechanism. There is, however, no detectable potential across the ER membrane, implying the existence of some other electrostatic receptor. In addition to charged residues, there is a statistically significant difference in the content of aromatic amino acids, cysteine, and alanine between the exofacial and cytoplasmically disposed segments of integral membrane proteins, probably reflecting selection for function in different folding environments (14,15).
Since insertion at the ER is cotranslational for most eukaryotic proteins, only the N-terminal domain preceding the first TM segment is potentially free to fold before translocation. In type III membrane proteins and in other proteins with an N exo N terminus, the cytoplasmic folding stability of this domain is presumably constrained so that it can unfold during translocation. In contrast, no such constraint need exist for the N cyt N-terminal domains of type II membrane proteins. In consequence, inversion of N cyt to N exo topology (II to III) is complicated by folding considerations, whereas inversion of N exo to N cyt insertion (III to II) should provide a less biased test of the role of charged residues as topogenic determinants (16,17). Only partial inversion of the asialoglycoprotein receptor, from N cyt to N exo , occurred when ⌬(C-N) was changed from Ϫ3 to ϩ5. Stable folding of its normally cytoplasmic N-terminal domain was probably responsible, since truncation or disruption of this domain aided insertion, whereas insertion of even more stable domains prevented inversion (17). In contrast, in vitro studies of cytochrome P-450, a type III protein with a very short N exo N terminus, showed complete inversion to N cyt orientation when ⌬(C-N) was changed from ϩ1 to Ϫ2 (18). Although topogenic signals have been studied in such model type III TM proteins, including the influenza A virus M2 protein, no systematic analysis has been published, and most analyses have been performed in vitro, where conditions may not fully reflect the in vivo situation (11,12).
We have developed an in vivo approach in yeast (Saccharomyces cerevisiae) based on S79, the N-terminal 79 residues of Ste2p, the ␣ mating pheromone receptor. We had previously confirmed that this receptor adopts the seven-TM segment topology common to G protein-coupled receptors, with an N exo N terminus and cytoplasmic C terminus ( Fig. 1A) (19). S79 includes only the N-terminal domain, the first TM segment, and the short first cytoplasmic loop of Ste2p. We now show that an S79-PB fusion, in which S79 is fused to a ␤-lactamase reporter, PB, adopts an exclusively N exo orientation (Fig. 1C), acting as a model type III protein. Effects of alterations in flanking charged residues on S79-PB orientation are entirely consistent with the charge difference rule and demonstrate that whereas a net positive charge does retard translocation of the adjacent end of the TM segment, negative charges are of similar importance and appear to promote translocation.
Expression Vectors-Fusion constructs are designated as either PB, ␣B, or SPB derivatives. P is a processing fragment from the K 1 preprotoxin encoding one internal and one C-terminal Kex2p cleavage site and three N-glycosylation sites (23). ␣ is a fragment of prepro-␣-factor containing one C-terminal Kex2p cleavage site and two N-glycosylation sites. S is the signal sequence from K 1 preprotoxin, and B is the mature sequence of ␤-lactamase (23). In p21-PGK-SPB, SPB is inserted between the PGK promoter and transcription terminator fragments in the multiple cloning site of the episomal URA3 vector YEp352 (21). Expression results in translocation of PB; following cleavage of PB by Kex2p, mature ␤-lactamase is secreted, preceded by the tripeptide Asp-Pro-Gly. pGAL-SPB is identical to p21-PGK-SPB, except that the PGK promoter is replaced by the galactose inducible/glucose repressible GAL1 promoter (21). p32-SPB is derived from p21-PGK-SPB by elimination of the PstI site in B, the BglII and SalI sites flanking the PGK terminator, the vector NsiI sites, and the BamHI and BstEII sites between P and B. Following cleavage by Kex2p, mature ␤-lactamase is secreted, preceded by the tripeptide Asp-Leu-Arg; expression is unaffected. p34nSPB is derived from p32-SPB by insertion of a unique NsiI site in place of the XhoI site at the N terminus of S (AT GCA TCC ATG) and restoration of the SalI site following the PGK terminator. The NsiI site is in the same reading frame as the unique PstI site at the N terminus of P, facilitating interconversion of SPB and PB fusions.
DNA Manipulations-The first 250 bp of STE2, encoding the first 79 amino acid residues of Ste2p, were amplified by PCR using the N-and C-terminal oligonucleotide primers 5Ј GGGCTCGAGAAAATGTCT-GATGCGGC and 5Ј GGGCTGCAGTCGGCGTTTTTCTGCT and pDTG-Ste2 (19) as a template; the Ste2p initiation codon is shown in boldface. The product was cloned as an XhoI/PstI fragment into the corresponding sites of p32-SPB. Sequence analysis confirmed that the resultant wild-type Ste2 79 -PB construct, pS79-PB, contains the STE2 gene fragment fused, in frame, to the P fragment and the downstream mature ␤-lactamase sequence. pS79-PB was subjected to site directed mutagenesis (U.S.E. mutagenic kit, Pharmacia Biotech, Inc.) using the mutagenic primers 5Ј ATGTTTGGTGTCATATGTGGTGCAG, which created an NdeI site and a single amino acid mutation, R58I in the TM region of Ste2p, and 5Ј GACTATTTTTCCATGGAG GGCACAG, which removed the NcoI site within the URA3 gene. The product is the pS79a-PB construct. All other mutants are derived from pS79a-PB; sequences of all mutants were confirmed.
Construction of C-terminal Mutants (URA3 Derivatives)-Using pS79a-PB as a template, the previously described N-terminal STE2 oligonucleotide primer (XhoI site) and the degenerate C-terminal primer 5Ј GGGCTGCAGTCGGCGTTKTTSTGCTTSTCGATGTC 3Ј (where K ϭ G or T and S ϭ G or C) a mixed pool of full length 250-bp PCR products were obtained and cloned as XhoI/PstI fragments into the corresponding sites of pS79a-PB. The following C-terminal S79-PB mutants, all of which contain the NdeI site marking the R58I mutation, were identified by sequence analysis: S79b, S79c, S79d, and S79e, containing the mutations R74T, R76T, K77T, and R74T/R76T respectively. Using pS79b-PB as a template, the same N-terminal oligonucleotide primer, along with the degenerate C-terminal primer 5Ј GGGCTGCAGTCGGCGTTKTTSTGCTTGTCGATGTCATCC 3Ј, and mutants pS79f-PB and pS79g-PB (R74T/K77T and R74T/R76T/K77T, respectively) were constructed and identified by sequence analysis. All of these C-terminal mutants have the wild-type 50 residue Ste2p Nterminal domain preceding the TM segment.
Construction of ␣B Derivatives-The P fragment of constructs pS79a, S79b, S79f, and S79g-PB was replaced by ␣, a fragment of prepro-␣factor with a C-terminal Kex2p processing site, to eliminate the potential contribution of charges within P to topological signals. The ␣ fragment was amplified by PCR using the oligonucleotide primers 5Ј GGGCTGCAGTTTCCAACAGCACAAATAAC 3Ј and 5Ј GGGTGGC-CACCCCTTCTTCTTTAGCAG 3Ј with p17␣F (23) as template DNA. The 93-bp PCR product was digested with PstI and MscI and ligated into the corresponding sites within pS79a, S79b, S79f, and S79g-PB constructs. The resulting constructs are pS79a-, S79b-, S79f-, and S79g-␣B, respectively.
Construction of N-terminal and Double Mutants (LEU2 Derivatives)-To facilitate cloning, LEU2 derivatives of both pS79a-PB and pS79g-PB were constructed as follows: the LEU2 yeast expression vector YEp351 was digested with HindIII/AatII. The 3.4-kbp fragment containing the LEU2 selectable marker and 2 micron DNA was cloned into the corresponding sites in pS79a-PB and pS79g-PB, producing pS79a-PB(L) and pS79g-PB(L). Using pS79a-PB(L) as a template and the degenerate N-terminal oligonucleotide primers 5Ј GGGCTC-GAGAATGGATRATRAGTTGCAAGGTTTAGTTAAC or 5Ј GGGCTC-GAGAATGAAGAAKRAGTTGCAAGGTTTAGTTAAC (where R ϭ A or G and K ϭ T or G), with the C-terminal oligonucleotide primer 5Ј CCGACTGCAGCCATACTC, a mixture of N-terminally truncated 120-bp PCR products were obtained and cloned as XhoI/PstI fragments into the complementary sites of pS79a-PB(L). In these mutants the initiator methionine is immediately followed by Asp or Lys, the 50residue Ste2p N-terminal domain preceding the TM segment is truncated to 12 residues, and the Ste2p fragment is shortened from 79 to 42 residues. Mutants were identified by sequence analysis in which the first three amino acids after the methionine are either DDE (pS42h-PB), DNE (pS42i-PB), DNK (pS42j-PB), KNE (pS42k-PB), KKE (pS42L-PB), KNK (pS42m-PB), or KKK (pS42n-PB).
Cell Fractionation and ␤-Lactamase Activity Assay-Cells were grown at 30°C in SC (synthetic complete) Ura Ϫ or Leu Ϫ medium buffered with 50 mM BisTris (pH 7.0) and harvested in early postexponential growth phase (2 ϫ 10 7 to 3 ϫ 10 7 cells/ml), and the supernatant was removed for direct ␤-lactamase assay. ␤-Lactamase activity is stabilized in culture supernatants above pH 6.0. The cells were then washed once with Buffer A (25 mM Tris⅐HCl, pH 7.6, 50 mM NaCl, 20 mM NaN 3 , 10 mM KF, 2 mM EDTA) containing 0.2 mM phenylmethylsulfonyl fluoride and 2 mg/ml pepstatin A and resuspended in the same buffer. The cells were broken by vortexing in the presence of glass beads (0.45 m), and large cell debris were removed by centrifugation at 330 ϫ g for 5 min, producing a low speed supernatant (LSS). The ␤-lactamase activities in culture supernatants and LSS extracts were determined spectrophotometrically using a synthetic substrate, PADAC, and a microtiter plate assay, as described previously (21).
Density Gradient Fractionation-Cells were grown at 30°C in SC Ura Ϫ or Leu Ϫ medium and harvested in early postexponential growth (2-3 ϫ 10 7 cells per ml). Cells (2 ml) were then washed once with 1 ml of Buffer A, once with 1 ml of TE buffer (50 mM Tris⅐HCl, 1 mM EDTA, pH 7.5; containing the same protease inhibitors) and then suspended in 1 ml of TE buffer. The cells were broken by vortexing in the presence of glass beads, and the LSS was isolated as described above. 0.5 ml of this LSS was combined with 0.5 ml of 76% Renografin in TE buffer and placed at the bottom of a SW 50.1 centrifuge tube. A step gradient of decreasing percentage of Renografin was prepared by layering onto this initial layer consecutive 1-ml layers of 34, 30, 26, and 22% Renografin solutions (prepared by diluting the 76% Renografin with TE buffer). The gradients were centrifuged in an SW 50.1 rotor at 100,000 ϫ g for 20 h at 4°C. 14 fractions (350 l) were collected from the top of each gradient and prepared for SDS-polyacrylamide gel electrophoresis.
Cell Fractionation and Immunoblotting-To identify expressed ␤-lactamase fusion proteins and ␤-lactamase-containing fragments, an LSS was prepared from cells grown in SC Ura Ϫ or Leu Ϫ as described above. The LSS was mixed with either 10% Triton X-100 or 1 M Na 2 CO 3 , pH 12.4 (9:1 by volume; final concentrations, 1% and 0.1 M, respectively). Mixtures were incubated for 20 min on ice and samples (150 l) were centrifuged in an Airfuge for 5 min at 26 psi (ϳ148,000 ϫ g) at 4°C (A-100 rotor, Beckman Instruments, Inc.). Supernatants (S100) were removed and the pellet fractions (P100) were resuspended in a Buffer A (150 l). 15-l samples of Triton X-100-solubilized P100 material were treated with 10 l endoglycosidase H (Boehringer Mannheim) in 50 mM K 2 P0 4 buffer (pH 7.0, 35 l) for 6 -8 h at 37°C. Samples, with and without prior endoglycosidase H treatment, were solubilized in SDS sample buffer at 100°C for 2 min and fractionated by SDS-polyacrylamide gel electrophoresis (12-15%) using Low Range Prestained Standards (Bio-Rad) as markers. Proteins were transferred to nitrocellulose membrane by standard electroblotting procedures. Filters were blocked and all antibody incubations were conducted in 5% nonfat dry milk in Tris-buffered saline (10 mM Tris⅐HCl, pH 7.6, 150 mM NaCl) containing 1% Tween 20. Filters were incubated for 1 h with a 1:1000 dilution of polyclonal rabbit anti-␤-lactamase primary antibody (5 prime-3 prime, Inc.) followed (after three washes) by a 20-min incubation with a 1:5000 dilution of anti-rabbit IgG peroxidase conjugate (Sigma). Detection of filter-bound antibodies was obtained by the enhanced chemiluminescence (ECL) system according to the manufacturers instructions (Amersham Corp.).

RESULTS
We previously confirmed (19) that the Ste2p receptor adopts the predicted seven TM segment topology common to G proteincoupled receptors, with an exofacially disposed N exo N terminus and a cytoplasmically disposed C terminus (Fig. 1A). Analysis involved the in vivo expression of C-terminal fusions of Ste2p fragments, starting after the second TM segment, to a reporter, PB. P encodes a fragment of the K 1 killer preprotoxin that contains two sites efficiently cleaved by the Kex2p protease (23); B, the mature form of ␤-lactamase, is downstream of the second Kex2p site ( Fig. 2A). Kex2p is located in the late Golgi with its active site in the lumen (24) so that cleavage of the PB reporter and secretion of B occurs only if the fusion site is luminal (exofacial). Secreted ␤-lactamase activity accumulates stably (half-life, about 8 h) in media buffered at pH 6.5-7 (19). Analysis of the ratio of secreted to cell-associated ␤-lactamase activity, therefore, can be used to determine whether the fusion site is cytoplasmic or exofacial. Exofacial location of the Ste2 98 -PB fusion was demonstrated in this manner ( Fig. 1B; The reliability of interpretation of this data depends first on the relative stabilities of secreted ␤-lactamase (from N cyt fusions) and of N exo fusions where the PB reporter is cytoplasmic and unprocessed; second, on the kinetics with which fusions with an exofacial PB reporter are translocated to the trans-Golgi and processed by Kex2p; and third, on the efficiency with which processed ␤-lactamase is released into the medium. Reliable correction factors for these factors are now provided by analysis of the fate of control secreted and N cyt PB fusions. Although the specific activities of the various fusions have not been determined directly, signal strengths in Western blots in these and previous studies (19) are always consistent with measured activity, when compared to that of a purified ␤-lactamase standard.
␤-Lactamase Secretion Efficiency and Cytoplasmic Stability-p34nSPB (Fig. 3A) is a control construct in which the PB reporter is expressed from the PGK promoter and is preceded by S (Fig. 2C), a 33-residue N-terminal fragment of the K 1 killer preprotoxin incorporating an efficient 26-residue secretion signal (23). Following secretion into the ER and processing by signal peptidase, PB is translocated to the Golgi and processed by Kex2p to produce secreted, mature ␤-lactamase preceded by the residues DLR ( Fig. 2A). Total activity accounts for about 0.5% of total cell protein, and 85% of this is found in the culture supernatant (Fig. 4A). An additional 5-7% is solubilized on spheroplast production and so is secreted but trapped in the cell wall. We previously showed (21) that about 30 min is required for ␤-lactamase secretion, suggesting that much of the residual 10% is in transit. Western blot analysis (21) indicates  98. B, the Ste2 98 -PB fusion in N exo orientation, illustrating Kex2p processing of the exofacial PB reporter, releasing mature ␤-lactamase. C, the S79-PB fusion in N exo orientation. D, the nascent SPB control construct in N cyt orientation, prior to processing by signal peptidase. E, the nascent S79-SPB fusion in N exo orientation with exofacial PB reporter, prior to processing by signal peptidase. that secretion and Kex2p processing is effectively complete. The distribution data in Fig. 4 are corrected for the 20% underestimate of secreted activity represented by supernatant activity.
The control construct cPB is essentially identical to p34nSPB except that S is replaced by the first 8 residues of prepro-␣-factor. All ␤-lactamase activity from cPB expression is cellassociated and is found in the S100 supernatant from Airfuge fractionation of broken cells and so is cytoplasmic. Activity accumulates to the same level as secreted activity in cells expressing p34nSPB (Table I), so cytoplasmic ␤-lactamase has similar stability.
The S79-PB Fusion Is Inserted Exclusively with N exo Orientation-Although exofacial location of the Ste2 98 -PB fusion site implies that TM 1 , the first transmembrane segment of Ste2p, inserts into the bilayer with an N exo orientation, this was not independently demonstrated (19). We constructed the S79-PB fusion for this purpose. The S79 fragment comprises the first 79 residues of Ste2p, which include an N-terminal fragment of approximately 50 residues, TM 1 , and the 8-residue first cytoplasmic loop ( Fig. 2A). This loop (TSRSRKTP) carries three positive charges, predicted to favor N exo insertion of the S79-PB fusion (Fig. 1C). If insertion is efficient, it should provide an easily manipulable model for the insertion of the first TM region of G protein-coupled receptors (15) and of Type III (N exo )-oriented proteins in general. P has three N-glycosylation sites ( Fig. 2A), and in secreted PB, all three are core-glycosylated in the ER; elongation in the Golgi occurs, but is obvious only in the absence of Kex2p function (23). Full-length Ste2p has four potential N-glycosylation sites but only two of these, at Asn 25 and Asn 32 , in the exofacial N terminus, lie sufficiently distant from TM regions to be accessible to the glycosylation machinery (25). The predominant species seen in Western blots are consistent with coreglycosylation at either one or both sites (22). In N exo -inserted S79-PB, therefore, not only should the reporter be cytoplasmically disposed, giving exclusively cell-associated ␤-lactamase activity, but processing of P should be prevented, leaving an intact, membrane-associated fusion protein with a glycosylated Ste2p N terminus ( Fig. 2A). Any fraction inserted in the opposite N cyt orientation will be glycosylated on P and, if also translocated to the Golgi, will be processed by Kex2p to produce secreted, mature ␤-lactamase. Orientation, therefore, can be confirmed by Western blot analysis to detect the intact fusion protein and to determine its state of glycosylation. Size predictions are given in Table I.
Expression of the S79-PB fusion produced only cell-associ- Other charged residues are also shown in boldface. Potential sites of N-glycosylation are indicated by the Y mark. The S79 fragment includes the N-terminal N exo domain (residues 1-50), TM 1 (residues 51-71), and the first cytoplasmic loop (residues 72-79) of Ste2p. The R58I mutation near the center of TM 1 that differentiates S79a-PB from S79-PB is indicated as R/I. The P fragment (dashed box; residues 80 -142) contains two KR sites for cleavage by Kex2p and three sites for N-glycosylation. Mature ␤-lactamase (HPETL . . . ) starts at residue 143. B, the ␣B reporter. P is replaced by a 34-residue fragment of pro-␣-factor that includes two sites for N-glycosylation and a terminal KR site for cleavage by Kex2p. C, the S secretion signal included in SPB and S79-SPB.  Table I ated activity (Fig. 4B), suggesting that the fusion is exclusively oriented with an N exo topology. Cell-associated activity, however, may also result from failure in membrane insertion. To distinguish between these possibilities, the low speed supernatant from cell breakage was fractionated by Airfuge at 148,000 ϫ g into pellet (P100, membrane) and supernatant (S100, cytoplasm) fractions and analyzed by Western blot. As shown in Fig. 5A, lanes 1-6, anti-␤-lactamase detected fusion protein as a mixture of two adjacent bands of 47-50 kDa, as predicted for the fusion protein with one or two core N-glycosyl groups, respectively (Table I), together with a weak band at 44 kDa probably representing unglycosylated protein. 95% of the fusion protein species were found in the membrane fraction (Fig. 5A, lanes 1 and 2); most remained in the membrane after treatment with 0.1 M Na 2 CO 3 , pH 12.4 (Fig. 5A, lanes 5 and 6), whereas all were solubilized by 1% Triton X-100 (Fig. 5A, lanes  3 and 4), showing them to be integral membrane proteins. Antisera to both ␤-lactamase and the Ste2p N terminus detected the same 47-50-kDa species in the Triton-solubilized membrane fraction, and these were reduced to a single band of 44 kDa by treatment with endoglycosidase H (Fig. 5, C and D,  lanes 1 and 2), consistent with the deduced glycosylation pattern. The 30-kDa species seen only after endoglycosidase H treatment and detected only with anti-␤-lactamase (Fig. 5C, lane 2) comigrates with mature ␤-lactamase (Fig. 5C, lane 9) and probably results from artifactual cleavage at the Kex2p sites during the prolonged incubation with endoglycosidase H. The corresponding 8.7-kDa Ste2 fragment is too small to be retained on the gel shown in Fig. 5D. The S79-PB fusion, therefore, is efficiently inserted into the membrane in an N exo orientation and appears to be a mixture of species glycosylated on either one or both of the sites in the Ste2p N terminus.
Membrane association of the fusion protein was also demonstrated by Renograffin gradient fractionation of disrupted cells. Following fractionation, three fractions can be distinguished using markers identified by specific antisera: low density membranes, corresponding to vacuole, Golgi, and ER, are marked by the Ost1p Golgi oligosaccharoyl transferase (major bands at 40 and 50 kDa); the denser plasma membrane is marked by the Pma1p proton ATPase (98 kDa); and the cytoplasm, in the two densest fractions, is marked by ␤-lactamase expressed from cPB (30 kDa; Fig. 6A). In cells expressing S79-PB, most of the fusion protein, detected with anti-␤-lactamase, comigrated with Ost1p (Fig. 6A). A small fraction overlapped the plasma membrane fractions, whereas only traces of smaller species, probably resulting from proteolysis, were detectable in the cytoplasmic fractions. The S79-PB fusion, therefore, is almost entirely located in internal membranes, in contrast to functional Ste2p, which is located in the plasma membrane.
Because failure in the processing of S79-PB by Kex2p could reflect failure in translocation to the late Golgi, expression was repeated in the presence of ER-Kex2p. ER-Kex2p is a derivative of Kex2p lacking the C-terminal TM region, which is responsible for its normal localization in the late Golgi; instead, it has a C-terminal HDEL sequence that causes it to locate primarily in the lumen of the ER, where it becomes active (26). When coexpressed with ER-Kex2p, S79-PB still produced only cell-associated activity (data not shown). None of this activity, therefore, appears to result from protein-inserted N cyt that has failed to translocate to the Golgi. In conclusion, all of the detected S79-PB fusion protein is a TM protein, inserted into the ER exclusively in an N exo orientation.
The S79-SPB Fusion Is also Inserted Exclusively with N exo Orientation-The topology of insertion of the Ste2 79 fragment was further analyzed in the S79-SPB construct in which the S secretion signal is placed between the fusion site and the PB reporter (Figs. 1E and 3B). Expression of S79-SPB from either the PGK or GAL1 promoter resulted in secretion of 80 -85% of the ␤-lactamase activity. In both cases, total activity was about 25% of the p34nSPB control ( Fig. 4A; only data for the PGK

TABLE I Predicted sizes of the fusion protein components shown in Figs. 2 and 3
The increases in size due to core glycosylation are estimates.  4. ␤-Lactamase activity data for the indicated control constructs and fusions. Secreted activity, A sec , and cell-associated activity, A ca (ng of ␤-lactamase/10 7 cells), are corrected for the lag in secretion and for the turnover of cell-associated fusion protein using the formula: cytoplasmic PB/luminal PB ϭ N exo /N cyt ϭ 1.25 A sec /4(A ca Ϫ A sec /4). See "Results." All data shown are averages from at least two independent transformants, each analyzed at least three times. ⌬(C-N) is the charge difference calculated according to Hartmann et al. (13). The charged residues contributing to this signal that flank the TM segment (box) are indicated in single-letter code with positive charges in boldface and negative charges in outline font. The total and secreted activities and the inferred N exo /N cyt distribution are indicated. A, control constructs and the S79-SPB fusion; B, S79-PB fusions and mutants in the C-terminal RSRK motif; C, ␣B fusions; D, S42 fusions retaining the RSRK motif; E, S42 fusions lacking the RSRK motif. promoter is shown), presumably reflecting lower messenger stability or translational efficiency in the Ste2p fusion, but showing that expression levels are comparable and much higher than in fusions to longer Ste2p fragments (19). When cells expressing the GAL-driven fusion were shifted from galactose (inducing) to glucose (repressing) medium, an additional 10% of the total activity was chased into the culture supernatant; the remaining 5-10% was solubilized on spheroplast formation. The activity that is cell-associated at steady state is, therefore, in transit within the secretory pathway or trapped in the cell wall, indicating that all of the PB reporter in the S79-SPB fusion is translocated into the ER lumen. This was confirmed by Western blot analysis of the transiently cell-associated material. Anti-␤-lactamase serum detected products only in the S100 fraction (Fig. 5A, lanes 7-12). Since the intact SPB fusion would have been readily detected in the membrane fraction, efficient translocation and cotranslational cleavage by signal peptidase at the S-PB junction is indicated. Signal function in this fusion, therefore, is not compromised by the preceding S79 fragment. The major species of about 48 kDa collapsed to 40 kDa on treatment with endoglycosidase H (Fig.  5C, lanes 3 and 4), as predicted for luminal, soluble, coreglycosylated PB (Table I). Mature ␤-lactamase (30 kDa) is seen only after endoglycosidase H treatment and probably results from artifactual cleavage at the Kex2p sites during incubation with endoglycosidase H.
Although translocation of the PB reporter is consistent with N exo insertion of the Ste2 fragment, with S acting as an efficient signal sequence (Fig. 1E), two alternate insertion events could lead to PB translocation. First, S could perform its normal secretion signal function while the Ste2 fragment failed to insert. Second, we previously showed that S functions poorly as a stop-transfer segment (19), so PB would remain luminal even if the fusion inserted N cyt . These possibilities were also distinguished by Western blot. If translocation of the S79 N terminus is efficient, the S79-S fragment released by signal peptidase action should remain in the membrane fraction; antiserum to the Ste2p N terminus detected products only in the membrane fraction (Fig. 5D, lanes 3 and 4). No signal corresponding to the intact fusion was visible; two species of apparent size 19 and 16 kDa were seen, and these collapsed to a single species of 13 kDa after treatment with endoglycosidase H. These are the sizes predicted for the S79-S fragment carrying either two or one core N-CHO fragments and their endoglycosidase products (Table I). These results confirm efficient N exo insertion of the Ste2 N terminus and efficient cleavage by signal peptidase in this S79-SPB fusion (Fig. 1E) to give the secreted PB product and a transmembrane S79-S fragment. A clear indication of the core glycosylation pattern in the Ste2 79 N-terminal fragment is also provided, consistent with mobilities seen in the S79-PB fusions and presumably reflecting that seen in the intact receptor (22).
Relation of the Distribution of ␤-Lactamase Activities to the N exo /N cyt Ratio-As shown above, culture supernatant ␤-lac- tamase activity, A sec , underestimates translocated PB from both p34nSPB and S79-SPB by about 20%. A sec , therefore, is multiplied by 1.25 to estimate the total secreted fraction due to luminal PB. The half-life of the cell-associated S79-PB fusion, determined by expression from the GAL1 promoter and shift from galactose to glucose medium, was only 2 h compared to the 8 h for A sec . As a consequence, accumulated cell-associated ␤-lactamase activity (A ca ) from S79-PB was 4-fold lower than A sec from the S79-SPB fusion. To calculate the true ratio of cytoplasmic and luminal reporter in PB fusions, therefore, we used the following formula: cytoplasmic PB/luminal PB ϭ N exo / N cyt ϭ 1.25 A sec /4(A ca Ϫ A sec /4) (Fig. 4). Because of these factors, measured activities are very sensitive to a small fraction of luminal reporter when the reporter is predominantly cytoplasmic (N exo insertion) but are insensitive to a small fraction of cytoplasmic reporter when the reporter is predominantly luminal (N cyt insertion).
Elimination of Arg 58 in Ste2 TM 1 Affects neither Stability nor Orientation of the PB Fusion-Charged amino acid residues within TM segments of integral membrane proteins, particularly when near the center, can result in the retention and degradation of these proteins within the ER (27). We therefore tested the effect of replacing Arg 58 in the S79-PB fusion, near the center of Ste2p TM 1 , with Ile 58 , producing S79a-PB ( Fig.  2A). Neither stability nor orientation were affected (Fig. 4B), suggesting that Arg 58 does not destabilize this fusion in yeast. Western blot analysis confirmed that as for the S79-PB fusion, all of the activity was cell-associated and all was in the form of an integral membrane protein of unaltered size and apparent glycosylation pattern (Fig. 5A, lanes 13-18), as confirmed by the results of endoglycosidase H treatment and detection with either anti-␤-lactamase or anti-Ste2p (Fig. 5, C and D, lanes 5  and 6). We used the R58I mutant to construct all subsequent fusions to eliminate any possible effects of the Arg 58 charge on TM 1 insertion orientation.
Neither Elimination of the Three Positive Charges following TM 1 nor Increasing the Separation Between TM 1 and the First Charges of the Reporter Affects Orientation of the PB Fusion-Apart from the positively charged N-terminal methionine, the N-terminal N exo domain of the Ste2 79 fragment contains only negative charges, and only two of these, Asp 39 and Glu 40 , fall within the arbitrary 15-residue window to either side of TM 1 used by Hartmann et al. (13) in calculating the charge difference, ⌬(C-N) ( Fig. 2A). Since the first cytoplasmic loop following TM 1 contains three positive charges in the RSRK motif, ⌬(C-N) for S79-PB and S79a-PB is ϩ5 (Fig. 4B). To test the importance of the positively charged residues in the RSRK motif, they were mutated, individually and in combination, to T (constructs S79b-PB to S79g-PB). All of these fusions had the same properties and data for three representative constructs, in which ⌬(C-N) is reduced to ϩ4, ϩ3, or ϩ2, are presented in Fig. 4B. In each case, the expression level was essentially the same as in S79a-PB, and essentially all of the activity was cell-associated. All of these cell-associated materials behaved as integral membrane proteins, as shown for S79f-PB and S79g-PB (⌬(C-N) ϭ ϩ2 and ϩ3, respectively) in Fig. 5A, lanes  19 -24 and 25-28, respectively. Mobilities of the major species indicated a glycosylation pattern unchanged from that in the S79a-PB fusion, and all data imply the same exclusive N exo insertion orientation. Renograffin fractionation of disrupted cells expressing the most extreme mutant, S79g-PB, showed essentially the same pattern as S79-PB (Fig. 6B); the activity cosedimented with the Ost1p membrane marker. It is clear, therefore, that the charges in the RSRK motif of S79-PB are a dispensable part of the topogenic signal.
Although these data suggest that the N-terminal negative charges in the S79g-PB fusion may be sufficient to determine insertion orientation, it remained possible that the receptor for this signal senses charges located further from TM 1 within the PB reporter. The closest charges are found in the effectively neutral KRSDTAE motif of P, starting 21 residues after TM 1 (Fig. 2A). To test its potential role, P was replaced with ␣, a fragment of prepro-␣-factor also terminating in a Kex2p processing site but in which the first charges, present in the KEE motif, are separated from TM 1 by 32 residues (Fig. 2B). Expression of these four representative ␣B fusions, S79a-␣B to S79g-␣B (Fig. 3D), produced exactly the same results as their PB versions; all of the activity remained cell-associated, consistent with an N exo orientation (Fig. 4C). Expression levels were higher, suggesting a modest increase in the in vivo stability of these fusions. All of the fusion proteins were integral membrane proteins (data not shown). It appears that insertion remains efficient in these fusions and that orientation is unaffected by the location and nature of the proximal charges in the PB and ␣B reporters. Elimination of the Charge Difference Leads to Partial Inversion, Exaggerated by Charge Inversion-The results described so far are entirely consistent with the charge difference rule, if equal weight is given to positive and negative charges; in addition, the results with S79g-PB and S79g-␣B indicate that the residual topogenic signal in these fusions promotes translocation and N exo insertion with high efficiency. According to the charge difference rule, this signal is the two negative charges on Asp 39 and Glu 40 , located about 10 residues upstream of TM 1 . However, it remained possible that insertion orientation in these fusions was affected by the more distant negative charges at Asp 3 and Asp 14 or reflected some unique property of the Ste2 N-terminal sequence in addition to its negative charges. To resolve this issue, we replaced Ste2p residues 1-38 with Met-Asp, producing a truncated, 42-residue Ste2 fragment called S42 ( Figs. 2A and 3E) with an MDDE N terminus. Since the positively charged N-terminal Met is now close to TM 1 , this construct has the same effective charge difference of ϩ5 as S79a-PB. The N-terminal domain upstream of TM 1 is truncated to 12 residues and should no longer be N-glycosylated. We constructed a series of S42-PB fusions that retained the three positive charges in the C-terminal RSRK motif and in which the MDDE motif was mutated by PCR to provide a net N-terminal charge that ranged from Ϫ2 to ϩ4. The fusions, in order of decreasing charge difference, are called S42h-PB to S42n-PB (Fig. 4D).
In the S42h-PB to S42k-PB series, in which ⌬(C-N) ϭ 5, 4, or 2, all of the ␤-lactamase activity remained cell-associated (Fig.  4D) and all of this activity was in the form of an unglycosylated integral membrane protein of the predicted size (41 kDa; Table  I; data not shown). In S42L-PB, in which ⌬(C-N) ϭ 1, about 8% inversion was observed; this was increased to 26 and 47%, respectively, in S42m-PB and S42n-PB, in which ⌬(C-N) ϭ 0 and Ϫ1, respectively (Fig. 4D). In these fusions, TM 1 is flanked by two groups of high positive charge; however, all of the cell-associated activity from these fusions remained in the P100 fraction, showing that this high charge density did not impede membrane insertion (data not shown). Thus, whereas a charge difference of ϩ2 provided a strong topogenic signal and near exclusive N exo orientation, a unit charge differences was more ambivalent, allowing some inversion. There appears to be a bias toward N exo insertion in these constructs, as this still predominated when ⌬(C-N) ϭ 0. Reversal of the unit charge difference lead to a 50:50 ratio of insertion orientation, also consistent with a bias toward N exo insertion.
The predominant charges in these fusions are positive; replacing the C-terminal RSRK motif in S42h-PB to S42n-PB with the neutral TSTT peptide from S79g produced the S42o-to S42u-PB fusions (Fig. 4E). The topogenic signal in S42o-PB and S42p-PB, in which ⌬(C-N) ϭ ϩ2 and ϩ1, respectively, results entirely from a net N-terminal negative charge, as in S79g-PB. Activity data indicates 92-94% N exo insertion (Fig. 4E), and this is entirely in the form of an unglycosylated integral membrane protein (Fig. 5B, lanes 1-12) of the expected 41-kDa size, unaffected by treatment with endoglycosidase H (Fig. 5C, lanes  7 and 8). This emphasizes the ability of negative charges, even in the context of the truncated N-terminal domain, to promote translocation of the TM 1 N terminus. A ⌬(C-N) of Ϫ1 due to the single net N-terminal positive charge in S42q-PB and S42r-PB caused reversal of insertion, more completely than in the highly charged S42n-PB (Fig. 4E). Addition of one, two, or three additional N-terminal positive charges led to near total inversion in S42s-PB and complete inversion in S42t-PB and S42u-PB (Fig. 4E). Although orientation appeared to be exclusively N cyt , our assay procedure would be insensitive to 5% or less N exo insertion. Complete inversion of orientation was, however, confirmed by Western blot analysis.
The S42s-PB, S42t-PB, and S42u-PB ␤-lactamase fusion proteins were located in the high speed pellet (P100 membrane) fractions from broken cells, virtually all as integral membrane proteins, solubilized by 1% Triton but not by 0.1 M Na 2 CO 3 (Fig.  5B, lanes 13-30). Most of the S42s-PB fusion and all of the S42t-PB and S42u-PB fusions migrated on SDS-polyacrylamide gel electrophoresis as a pair of 47-50-kDa proteins that collapse to a species of about 42 kDa on treatment with endoglycosidase H (Fig. 5C, lanes 10 -13). These species have the sizes predicted for the N cyt -inserted fusions, core N-glycosylated at two and three of the sites in the PB reporter (Table I), and must represent translocated fusion protein in transit to the Golgi prior to Kex2p cleavage. No signal is seen for mature ␤-lactamase in the sample prior to endoglycosidase H treatment, implying rapid kinetics for export following Kex2p cleavage. The ␤-lactamase seen after endoglycosidase H treatment is presumably an artifact of proteolysis. Only in S42s-PB is an unglycosylated 41-kDa species corresponding to N exo -inserted fusion visible (lanes [13][14][15][16][17][18]; it accounts for about 15% of the total signal, consistent with the activity data (Fig. 4E). The absence of this species in the S42t-PB and S42u-PB fusion products confirms their exclusive N cyt insertion.
These data show that the S79 fragment and its truncated S42 form have redundant topogenic signals: either the C-terminal positive charges or the N-terminal negative charges being sufficient to determine insertion in N exo orientation, strictly in accordance with the charge difference rule. DISCUSSION S79, the 79-residue N-terminal fragment of Ste2p, includes the first transmembrane segment and flanking charged topogenic sequences of this receptor. Fusion to the PB or ␣B Kex2pcleavable ␤-lactamase reporters produced model Type III membrane proteins that inserted in vivo in the N exo orientation native to the Ste2p N terminus and provided an assay system for quantitatively monitoring this insertional orientation. We have exploited this assay to test the effect of changes in flanking charge on orientation, allowing us to draw several conclusions concerning topogenic signals for eukaryotic membrane protein insertion, an issue that has previously been addressed only to a limited extent, using principally in vitro systems that may not fully reflect the intracellular environment. It was necessary, however, to first establish the validity of the assay technique.
In this fusion technique, the ratio of cytoplasmic and exofacial fusion sites for the PB reporter is deduced from the ratio of cell-associated and secreted ␤-lactamase activities. For integral transmembrane protein fusions, this ratio translates directly into the ratio of N exo and N cyt insertion orientation. Validity, therefore, depends on a demonstration that all of the cellassociated activity is indeed in the form of integral membrane proteins and on allowance for incomplete release of secreted ␤-lactamase into the medium and any differences in specific activity and turnover rates of cell-associated fusion proteins and secreted ␤-lactamase. Through analysis of appropriate controls, we have shown that all of the cell-associated fusions are integral membrane proteins, that culture supernatant activity underestimates secretion by 15-20%, that PB fusion proteins and free ␤-lactamase have similar specific activities, and that correction is needed for the 4-fold shorter half-life of membrane-associated PB fusion proteins. As a consequence, the confidence limits for detection of N cyt insertion in a construct that gives near exclusive N exo insertion are Ϯ2%, whereas those for N exo insertion in a construct that gives near exclusive N cyt insertion are only Ϯ 8%. Western blot analyses of glycosylation patterns, however, provide an independent test of insertion orientation, and results with both the S79-PB and S79-SPB fusions were entirely consistent with the exclusive N exo topologies deduced from activity distribution data. Besides confirming translocation of the N terminus of the Ste2p fragment, the glycosylation pattern of the SPB fusion confirmed the suspected modification pattern in the parent Ste2p receptor (22). We can conclude, therefore, that at least 98% of both fusions insert in N exo orientation. These observations effectively complete our previous in vivo topological analysis of Ste2p (19). The only remaining issue is bias introduced by the reporters. First, all components of the PB and ␣B reporters are normally secreted by yeast with high efficiency (23), so they are not likely to impede translocation. Second, reporter neutrality is implied by the identical orientations seen for PB and ␣B fusions, in which the separation of the TM segment from charged residues in the reporter is increased from 21 to 32 residues. Finally, neutrality is also implied by the inversion of orientation seen when the proximal flanking charge is reversed.
If the PB reporter is neutral, then the S79 fragment must contain all of the topogenic information necessary for determining its N exo orientation. It has been clear for some time that the major topogenic signal for determining TM segment insertion orientation is provided by flanking charged residues. In bacteria, the validity of the positive inside rule codifying the predominant role of positively charged residues in this signal has been firmly established (4,5). The relative suppression of the effects of negative charges has been attributed to the effects of the immediate membrane environment during translocation (6), and the major receptor for this signal has been identified as the transmembrane potential (7,8). In eukaryotes, published analyses of the effects of flanking charge on insertion orientation have focused principally on model type II membrane proteins. The use of type II proteins complicates interpretation because inverted N exo insertion requires translocation of the prefolded native, normally cytoplasmic N-terminal domain. Nterminal domains preceding the first TM segment of integral membrane proteins can fold in the cytoplasm before translocation commences. In type III proteins, such domains must be amenable to facile unfolding for translocation, whereas stable N-terminal domains in Type II proteins have been shown to impede translocation, favoring N cyt insertion (17). Results of analyses of mutant type II proteins have been interpreted as being consistent with the positive inside rule (10,11). However, statistical analysis of eukaryotic membrane proteins (13) indicates a strong correlation between orientation and the charge difference across the TM segment, ⌬(C-N), using an arbitrary window size of 15. This charge difference rule differs signifi-cantly from the positive inside rule in giving equal weight to positive and negative charges. Our analysis of the Ste2p-PB fusions in yeast gives strong support to the charge difference rule and emphasizes the potential of negative charges for providing dominant topogenic signals.
Before testing the consequences of such alterations, we first replaced Arg 58 , near the center of TM 1 , with Ile in S79a-PB and its derivatives. This affected neither stability nor orientation, even though charged residues in similar locations are known to labilize TM proteins to ER degradation (27). Since the fusion proteins were found to reside almost exclusively in internal membranes, their turnover is probably mediated by translocation to the vacuole.
Both the charge difference and the positive inside rules predict the observed exclusive N exo insertion of the S79-PB, S79a-PB, and S79-SPB fusions. The positive inside rule predicts cytoplasmic location for the RSRK motif C-terminal to TM 1 ; its ϩ3 charge difference is enhanced to ϩ5 by the two negative charges about 10 residues upstream of TM 1 . We have demonstrated that sequential elimination of all of the positive charges in the RSRK motif failed to affect orientation, demonstrating that these charges are a dispensable part of the topogenic signal. Deletion of all but the last 10 residues upstream of Ste2p TM 1 , producing the S42 fusion series, caused only a slight relaxation of insertion orientation, indicating that any charge-independent tendency of the Ste2p N-terminal domain to favor translocation was also dispensable. In a mutant form of the asialoglycoprotein receptor in which ⌬(C-N) is changed from Ϫ3 to ϩ5, resulting in partial inversion from N cyt to N exo insertion, length of the N-tail was not found to affect translocation (16). In the S42o-PB fusion, where ⌬(C-N) is ϩ2 and is entirely due to N-terminal negative charges as in S79g-PB, the orientation remained 94% N exo , indicating a primary role for these negative charges in the topogenic signal. This was confirmed by additional N-terminal mutations in the truncated S42 series in which these negative charges were balanced by increasing positive charge.
In the continued presence of the C-terminal ϩ3 RSRK motif, increasing the N-terminal charge so that ⌬(C-N) was reduced to ϩ1, 0, or Ϫ1 allowed 8, 26, and 47% inversion, respectively. In the absence of the C-terminal RSRK motif, where total flanking charge is much less, increasing the N-terminal charge so that ⌬(C-N) is reduced to Ϫ1, Ϫ2, Ϫ3, or Ϫ4 increased N cyt insertion to 60 -70, 85-90, near 100, and near 100%, respectively. Thus, all results are entirely consistent with the charge difference rule and indicate that whereas a net positive charge does retard translocation of the adjacent end of the TM segment, consistent with the positive inside rule, negative charges are of similar importance and appear to promote translocation.
Where the charge difference is small, our data indicate a bias toward N exo insertion orientation, as previously noted in statistical analyses (13). Thus, in the only effectively neutral construct, N exo insertion was 74%, increased to 92% when ⌬(C-N) ϭ ϩ1 and decreased to 27-53% when ⌬(C-N) ϭ Ϫ1. This bias could reflect inherent bias in the translocation machinery itself (for example, a bias in the signal receptor toward C-terminal charges) or more subtle signals in the translocated protein.
Such signals could be related to nonrandom amino acid distribution in the TM segment (27), a residual weak bias due to the reporter or the macrodipole on the TM ␣ helix. The fractional N-terminal positive charge of this macrodipole, however, should favor N cyt insertion.
Our studies have both confirmed the dominant role of the charge difference in the topogenic signal determining TM protein insertion orientation and established the utility of this particular in vivo model. This now provides us with the ability to dissect these topogenic signals in more detail. For example, although we have not attempted to systematically determine the effect of distance between the TM segment and charge on signal strength, our results are consistent with the 15-residue window chosen by Hartmann et al. (13), although a distancerelated gradient of effect can be anticipated. Identification of the receptor for this signal would clarify these issues (28). One can imagine that the signal recognition particle binds the hydrophobic core of nascent TM segments and presents this to the translocation machinery at the ER as a loop with its termini juxtaposed. The orientation receptor, a negatively charged patch on a component of the translocon adjacent to the pore, would respond to the sum of the charges adjacent to these termini, promoting translocation of the most negatively charged terminus.