Mutant Membrane Protein Toxicity*

This report describes an extensive mutational analysis of the most carboxyl-terminal membrane-spanning sequence ofEscherichia coli lac permease (TM12). In addition to identifying residues important for lactose transport function, the analysis revealed that numerous mutations made lac permease highly toxic to cells. In the most extreme cases, production of such proteins at very low steady-state levels reduced cell viability greater than 104-fold. Both frameshift and missense mutations led to toxicity, with the frameshift mutations having the strongest effects observed. The toxic missense mutations corresponded to changes in TM12 expected to interfere with membrane insertion or folding, such as the introduction of charged residues or prolines in the putative helix. The results suggest that cellular toxicity may be a relatively common consequence of mutations altering integral membrane protein folding. An analogous toxicity might contribute to the pathogenesis of several degenerative diseases caused by mutant membrane proteins, such as retinitis pigmentosa, Charcot-Marie-Tooth syndrome, and Alzheimer’s disease.

Several diseases caused by mutant integral membrane proteins show dominant inheritance and are associated with tissue degeneration. These diseases include autosomal dominant retinitis pigmentosa, Charcot-Marie-Tooth disease, and Dejerine-Sottas syndrome (1)(2)(3). A simple hypothesis to help account for the pathogenesis of such diseases is that they reflect a general phenomenon in which mutations render membrane proteins directly toxic to cells (4).
This report describes a mutational analysis of the 12th transmembrane segment (TM12) 1 of Escherichia coli lac permease (5,6). Our studies identify a set of TM12 residues that tolerate a variety of substitutions without loss of transport activity. These residues show an ␣-helical periodicity and may correspond to a side of the TM12 helix which faces the lipid bilayer. A number of missense and frameshift mutations were also identified which render lac permease toxic to cells. The toxic missense changes did not cluster on either the tolerant or sensitive face of TM12 and corresponded to changes expected to cause misfolding of the mutant proteins.

MATERIALS AND METHODS
Bacteria and Plasmids-The strains and plasmids used in this study are listed in Table I. Plasmid pCS112 is a pBR322-based plasmid derived from pCM472 (7) which carries a modified lacY with a UGA nonsense codon corresponding to codon 386 and a StyI restriction site at the NH 2 -terminal end of the sequence encoding TM12 (Table I). To construct pCS112, a SspI-SspI fragment that included phoA was first deleted from pCM472. Second, the StyI site in the tet gene was removed by StyI cleavage, followed by DNA polymerase I Klenow fragment and T4 DNA ligase treatments. Third, a StyI cleavage site was introduced at a position corresponding to the NH 2 -terminal end of TM12, and a UGA termination codon (codon 386) and SpeI cleavage site were introduced within the sequence corresponding to TM12 by site-directed mutagenesis. The new StyI and SspI restriction sites are unique in pCS112, and the lacY it carries is inactive because of the termination codon. Plasmid pCS111 is identical to pCS112 except that it encodes active lac permease, with a glycine codon at position 386.
Cassette Mutagenesis-A degenerate mixture of 80-nucleotide oligomers corresponding to LacY TM12 was synthesized (5Ј-ggt ttc caa ggc GCt TAT CTg GTg CTg GGt CTg GTg GCg CTg GGc TTC ACc TTA ATT TCc GTg TTC ACg ctt agc ggc cc-3Ј, with 2.5% contamination of an equimolar mixture of all 4 bases at the positions shown in capital letters). This oligomer spans two restriction sites in pCS112: the engineered StyI site (CCAAGG), corresponding to the NH 2 -terminal end of TM12, and a naturally occurring BlpI site (GCTTAGC), corresponding to the COOH-terminal end of TM12. A double-stranded DNA cassette corresponding to the 80-mer mixture was synthesized essentially as described by Lim and Sauer (9). The resulting DNA was purified by elution from an 8% nondenaturing polyacrylamide gel and digested with StyI (15 units) and Bpu1102 (7.5 units) or BlpI (30 units). This fragment was mixed with pCS112 that had also been digested with StyI and BlpI and treated with T4 DNA ligase. The mixture was then digested with SpeI to render uncut parental pCS112 linear (and therefore poorly transformable), and the DNA was then precipitated and transformed by electroporation into CC815. Transformants were selected by growth on TYE agar containing 100 g/ml ampicillin and 0.2% glucose. (The glucose is included to minimize expression of lacY.) Colonies were replica printed onto MacConkey lactose agar containing ampicillin to assess LacY activity and the efficiency of mutagenesis (cells receiving parental plasmids formed white colonies, whereas cells receiving active mutant plasmids formed red colonies) and onto TYE agar containing ampicillin and 40 g/ml 5-bromo-4-chloro-3-indolyl-␤-D-galactopyranoside (X-gal) and 1 mM isopropyl-1-thio-␤-D-galactopyranoside (IPTG) to detect toxic phenotypes associated with induction of LacY synthesis.
Sequence changes in TM12 were identified by the dideoxynucleotide termination DNA sequencing of plasmid DNA using a primer (5Ј-TAA GCA ACT GGC GAT GA-3Ј) hybridizing at the 5Ј-side of the sequence encoding TM12.
Quantitation of Lactose Permease Transport Activity-Radioactive lactose uptake assays were carried out using the ⌬(lacZ-lacY) strain CC819 carrying mutant plasmids. The lacZ mutation prevents metabolism of transported lactose. CC819 maintains the lacY plasmids at reduced copy number because of a pcnB mutation, leading to a level of lac permease production such that uptake of [ 14 C]lactose under assay conditions is linear with respect to the amount of protein/cell (10). 2 Uptake of [ 14 C]lactose (Amersham Pharmacia Biotech) was assayed as described (10). Cultures were grown overnight at room temperature, diluted to A 600 ϳ0.1 and grown at 37°C to A 600 ϳ0.9. IPTG was then added (2 mM final), and the cells were harvested after a 45-min incubation at 37°C. Cells were then rinsed twice in buffer (100 mM potassium phosphate, pH 7.0, 10 mM MgSO 4 ), resuspended at A 600 ϳ5.8, and [ 14 C]lactose uptake into 50-l cell aliquots measured as a function of time at room temperature.
SDS-Polyacrylamide Gel Electrophoresis and Immunoblotting-Cells were harvested from cultures grown as described above for [ 14 C]lactose uptake assays. Whole-cell protein was prepared by centrifuging 1.5 ml of culture (5 min at 5,000 ϫ g) followed by resuspension of the cell pellet in 0.25 ml of sample buffer (62.5 mM Tris⅐HCl, pH 6.8, 10% glycerol, 2% SDS) and heating to 55°C for 10 min. Protein samples were diluted 1/20 in distilled water and quantified by Lowry assay (Sigma). Immediately before loading samples for SDS-polyacrylamide gel electrophoresis, ␤-mercaptoethanol was added to 5%, and the samples were again heated at 55°C for 10 min.
Equal amounts of total protein from each sample (usually 50 g) were subjected to SDS-polyacrylamide gel electrophoresis (9 -12% gel with 3.5% stacking gel) and transferred to nitrocellulose or Westran polyvinylidene difluoride membrane (Schleicher & Schuell). Blots were incubated in 5% Carnation dry milk in TBST (100 mM Tris⅐HCl, pH 8.0, 1.5 M NaCl, 0.05% Tween 20) for 1 h and then exposed to an antibody recognizing the COOH-terminal cytoplasmic sequence of lac permease. This antibody can detect protein levels that are approximately 2% that of wild-type. 3 Frameshift mutations were subcloned into pCM701, which encodes LacY with a 31-codon insert at a position corresponding to amino acid 38 (11). To do this, the 1.4-kilobase BstXI-PstI fragment of the pCS112-derived mutants was joined to the 3.7-kilobase BstXI-PstI fragment of pCM701 (not shown). The production of LacY frameshift products was then analyzed by Western blotting using an antibody recognizing the 31-residue inserted sequence (11). The anti-insert antibody can detect protein levels that are approximately 4% that of wild-type. 3 A secondary antibody (anti-rabbit) complexed to alkaline phosphatase (Boehringer Mannheim) was used with both primary antibodies.
In some cases, Western blotting was carried out as above except that blots were incubated in TNT (10 mM Tris⅐HCl, pH 7.4, 0.9% NaCl, 0.2% Triton X-100), the secondary antibody was linked to horseradish peroxidase, and the blot was developed with ECL reagents (Amersham Pharmacia Biotech).
Assays of Mutant lac Permease Toxicity to Cells-Two assays of toxicity to cells were used: one based on the relative efficiency of colony formation under conditions of lac permease expression, and the other based on the segregation of lacY Ϫ cells during growth on indicator agar-inducing lac expression. The assay of colony-forming efficiency under inducing conditions was carried out by plating dilutions of cells (grown to stationary phase in LB containing ampicillin) after centrifugation and resuspension in buffer (10 mM Tris⅐HCl, pH 7.0, 10 mM MgCl 2 ) on M63-glycerol supplemented with IPTG (1 mM) and cAMP (2.5 mM) (for plasmids in CC1006) or IPTG alone (1 mM) (for plasmids in CC1351) and unsupplemented M63-glycerol. Colonies were counted after a 2-day incubation at 42°C. The second assay depended on the fact that cells expressing toxic plasmid-borne mutant lacY genes segregate faster growing plasmid-free cells. Such cells are LacY Ϫ and can be distinguished from LacY ϩ cells on media containing X-gal. For the tests, cells were streaked on TYE containing IPTG and X-gal and scored after a 2-day incubation at 37°C for the presence of light blue sectors and papillae that had arisen from dark blue primary colonies.

Mutant Isolation-
The sequence corresponding to the TM12 of lac permease ( Fig. 1A) was mutagenized using an efficient "cassette" method ( Fig. 1B) (9). The lac permease gene (lacY) gene was present in a small plasmid (pCS112) and carried a mutation creating a termination codon (UGA) corresponding to a site within TM12 (codon 386). The nonsense codon inactivates the lac permease, and pCS112 transformants of a lacZ ϩ lacY Ϫ strain therefore produce white (LacY Ϫ ) colonies on MacConkey lactose indicator agar. To generate mutations, double-stranded oligonucleotide fragments corresponding to TM12 and derived from a degenerate mixture were inserted into pCS112 plasmids, replacing the pCS112 sequence. The mixture was transformed into CC815, a lacZ ϩ ⌬(lacY) strain that maintains pCS112 at reduced copy number because of a pcnB Ϫ mutation (12). CC815 produces about three times as much lac permease as a strain containing a single chromosomal copy of lacY (10). 3 Transformant cells that formed LacY ϩ (red) colonies on indicator agar (MacConkey lactose) had acquired plasmids in which the inactive pCS112 TM12 sequence had been replaced by functional sequences. Transformant cells that formed LacY Ϫ (white) colonies on the indicator agar carried either mutagenized plasmids encoding nonfunctional TM12 sequences or pCS112 parent plasmid that escaped mutagenesis. The parent plasmid DNA could be distinguished from mutant by the presence of a restriction site (SpeI), and cleavage by SpeI before transformation reduced the amount of parent plasmid in the mixture (Fig. 1B). Using these procedures, 35 different LacY ϩ and 28 different LacY Ϫ mutant plasmids were identified for analysis (Tables II and III).
Transformants resulting from the cassette mutagenesis of pCS112 were also screened for toxicity associated with mutant lac permease synthesis (see "Materials and Methods"). Transformant colonies (on MacConkey lactose or TYE agar) were replica printed onto TYE agar supplemented with IPTG and X-gal to induce high level lac expression. This medium allows the detection of LacY Ϫ cells that arise during further growth because such cells do not hydrolyze X-gal as efficiently as do LacY ϩ cells and therefore form light blue sectors and papillae. Toxicity was detected as decreased growth accompanied by abundant papillation of LacY Ϫ cells. 18 mutants exhibiting growth inhibition associated with lac induction were identified (Table IV).
Plasmid DNAs were sequenced to identify TM12 changes, and representatives of all mutant classes were assayed to determine rates of [ 14 C]lactose uptake (Tables II-IV).
Nontoxic Mutants That Retain lac Permease Activity-The 3 C. Stewart, unpublished results. TM12 sequences of the mutants exhibiting significant lac permease activity are listed in Table II. 25 of these mutants carry single amino acid substitutions, and 10 carry changes at two sites. Most of the changes are relatively conservative. These mutants formed red colonies on MacConkey lactose agar, grew significantly better than a ⌬(lacY) strain on minimal lactose agar, and (when assayed) exhibited at least 54% of the wildtype rate of [ 14 C]lactose transport. These mutants thus carry sequence changes that are tolerated without major effects on the membrane insertion, folding, or transport activities of the proteins. From the collection of active substitutions identified, it appears that some sites in TM12 tolerate a greater variety of substitutions than do others (Table II). When TM12 is diagrammed as a helical wheel, the residues tolerating the greatest variety of substitutions cluster on one face (centered on Ile-395) (Fig. 2). Opposite this face is a cluster of 10 residues (centered on Thr-393) for which either no active substitutions or only relatively conservative active substitutions were identified. (The only exception was Y382C.) The angle of the net variation moment for the helical wheel sequences corresponds approximately to the angle of the net hydrophobic moment (Fig. 2) (13). By analogy to studies of the Rhodobacter photosynthetic reaction center (see "Discussion"), we suspect that the substitution-tolerant, hydrophobic side of TM12 faces membrane phospholipids in folded lac permease.
Nontoxic Mutants with Reduced LacY Activity-The TM12 sequences of substitution mutants exhibiting reduced lac permease activity are shown in Table III. Although the majority of mutants that lacked all detectable lactose transport activity carried frameshift or deletion mutations, such mutants were not studied in detail. 4 The substitution mutants ranged from those that appeared to be completely inactive (i.e. indistinguishable from a ⌬lacY strain in lactose utilization on different media and in lactose uptake activity) to those exhibiting considerable activity (e.g. up to 19% of the rate of [ 14 C]lactose uptake of wild-type control cells). Nearly all of the missense mutants that expressed greatly reduced activity had suffered changes that either altered intolerant residues or introduced charged residues (Table III). Toxic Mutants-The TM12 sequences of the mutants that were toxic to cells are shown in Table IV. To quantify the growth defects of different mutants, the efficiencies of colony formation under lac-inducing conditions were measured. The tests were carried out using strains maintaining plasmids at normal (PcnB ϩ ) or reduced (PcnB Ϫ ) copy number. (The pcnB mutation reduces the pBR322-based plasmid copy number approximately 10-fold (12).) Both LacY ϩ and LacY Ϫ toxic mutations were identified (Table IV). Again, most of the completely inactive mutants carried frameshift mutations. The recovery of frameshift mutants was a surprise because the cassette mu-4 M. Jurica and C. Stewart, unpublished data. cassette mutagenesis strategy. A degenerate oligonucleotide mixture encoding TM12 and spanning two restriction sites (StyI and BlpI) was first converted to a double-stranded "cassette" form (see "Materials and Methods"). The resulting mixture of fragments and pCS112 (which carries a lacY gene with a nonsense codon corresponding to a site in TM12) was then digested with StyI and BlpI, mixed, and joined using T4 DNA ligase. The resulting population of plasmids was digested further with SpeI to linearize parental plasmids (rendering them poorly transformable), and the mixture was transformed into the lacZ ϩ lacY Ϫ strain CC815. A plasmid that has undergone mutagenesis will have lost the UGA nonsense codon present in the parent plasmid and will complement the lacY Ϫ host mutation if the new TM12 sequence is functional for lactose transport.
tagenesis was designed to yield missense mutants (Fig. 1B). We assume that the frameshift mutations were generated by contaminating oligonucleotides present in the mixture used for cassette mutagenesis.
Three classes of toxic mutants could be distinguished based on the degree of toxicity (Table IV). The most toxic class (all of the frameshift mutants except FS12) reduced the efficiency of colony formation dramatically (10 1 to Ͼ10 4 -fold) under inducing conditions when expressed at low copy number in PcnB Ϫ strains. The second class of mutants (G386A V388E, L390P F392Y, MQD, YREMVQ, FS12) showed greatly reduced efficiency of plating (10 1 to 10 5 -fold) when the corresponding plasmids were carried in PcnB ϩ strains but not in PcnB Ϫ strains. The third and weakest class of mutants (L383R, L385P, G391V I395S, L394P T399F, LEF, PEP) formed small colonies and/or showed increased papillation and sectoring of faster growing cells on X-gal indicator agar containing IPTG in PcnB ϩ strains but did not show greatly decreased efficiencies of colony formation.
Levels of Toxic and Nontoxic Proteins-The steady-state cellular levels of missense mutant proteins were measured by Western blotting using an antibody recognizing a COOH-terminal epitope of lac permease (Fig. 3A; Tables III and IV) (see "Materials and Methods"). The amounts of protein detected from toxic missense mutants varied from being comparable to wild-type to undetectable (i.e. less than approximately 2% of wild-type) ( Fig. 3A and Table IV). The results imply that the COOH-terminal epitope is susceptible to cellular degradation in some of the mutants; it is possible that the rest of the protein is degraded in such mutants as well. Because the antisera used to detect missense proteins could not recognize frameshift mutant proteins due to the altered COOH-terminal sequence, toxic frameshift mutations were subcloned into a plasmid (pCM701) that encodes an epitope-tagged lac permease (11). The tag is a 31-amino acid insertion in the most NH 2 -terminal periplasmic segment of the protein, and the insert does not significantly reduce lactose transport or the steady-state cellular level of the protein (11). Using antisera recognizing the epitope, we observed that all of the toxic frameshift products were either undetectable or were detected at a greatly reduced level relative to the parent (Fig. 3B and Table IV). Nontoxic frameshift mutant proteins were also undetectable in analogous tests (not shown). Earlier studies of COOH-terminal truncation mutants using a method that did not involve antibody binding showed that loss of TM12 residues resulted in cellular proteolysis of the mutant proteins (14, 15).  C]Lactose uptake was determined for mutant plasmids in strain CC819 as described under "Materials and Methods." An isogenic ⌬(lacY) mutant showed 1% the lactose transport activity of the lacY ϩ strain. ϩϩϩ, estimated to correspond to Ͼ54% wild-type activity based on the behavior of mutants on MacConkey lactose and M63 lactose media. Dots correspond to positions with wild-type residues.

DISCUSSION
This report describes an extensive mutational analysis of the most COOH-terminal transmembrane sequence (TM12) of E. coli lac permease. There were two principal findings of the study. First, about half of the sites in TM12 tolerated a variety of residues without loss of transport activity. These tolerant sites were arranged with a helical periodicity and may make up part of the lipid-facing surface of the protein. Second, a number of mutations made the altered lac permeases highly toxic to cells. The frameshift and missense changes leading to toxicity are expected to cause protein folding defects, a result suggesting that integral membrane protein misfolding may, in many cases, inhibit growth.
A number of TM12 substitutions had small effects on lac permease transport activity. When TM12 is represented as an ␣-helix, the sites most tolerant of substitutions cluster on one face, which is also the most hydrophobic face of the helix (Fig.  2). Studies of sequence polymorphisms of the Rhodobacter photosynthetic reaction center have shown that residues facing the lipid bilayer are less conserved and more hydrophobic than residues internal to the folded protein (13). By analogy, we assume that the substitution-tolerant, hydrophobic side of TM12 faces the surface in folded lac permease. Studies of lac permease TM8 (16) and TM10 (17) also identified substitutiontolerant sites showing helical periodicity which were proposed to face the surface of the protein.
Substitutions in TM12 which reduced lactose transport activity affected both tolerant and intolerant sites in the putative helix (Table III). The substitutions affecting tolerant sites were generally highly nonconservative (such as the introduction of charged residues in V384E and G391D) and are expected to cause generalized defects in membrane insertion and/or folding. Most of the inactivating substitutions markedly decreased the steady-state recovery of protein (Table III), suggesting that defective folding frequently leads to cellular proteolysis. Few inactivating changes altering TM12 have been identified previously (18 -20). The earlier studies found that L385R, G386C, and L400C inactivated lac permease, whereas all of the other residues of TM12 could be converted individually into cysteine without decreasing transport greatly (19,20).
The most surprising finding of this study was the number and variety of mutations affecting TM12 which made lac permease toxic to cells. The toxic effects were in some cases quite dramatic, reducing plating efficiencies of cells carrying mutant plasmids at low copy number greater than 10 4 -fold. In addition, many of the toxic mutant proteins were present in cells at very low levels (undetectable in Western blotting), implying very high toxicity per molecule. The most toxic changes were frameshift mutations at sites corresponding to the sixth periplasmic domain (P6) of lac permease. Because nonsense mutations (and some frameshift mutations) corresponding to P6 did not cause dramatic growth inhibition (Table IV and data not shown), it appears that the frameshift sequence synthesized in place of TM12 contributes to whether the resulting mutant protein is toxic or not. The substitution mutations that were toxic corresponded to changes expected to favor unfolding of the protein:   the introduction of charged residues or proline, changes of glycine, or substitutions at "intolerant" sites (Table IV). Many of the toxic substitution mutant proteins retained significant transport activity (Table IV), suggesting that in some cases, toxicity is a more sensitive measure of misfolding than is loss of transport activity.
Several previous studies have documented instances of mutant membrane protein toxicity. For example, it was observed that a number of hybrid lac permease-alkaline phosphatase proteins inhibited growth, with toxicity increasing with the length of the hybrid (7). Substitution and in-frame insertion mutations of lac permease and the E. coli serine chemoreceptor have also been observed previously to inhibit growth (11, 18,  Table II are shown. The hydrophobic moment (H) and variation moment (V) are indicated. The hydrophobic moment was calculated according to Eisenberg et al. (26) using the hydrophobicity scale of Engelman et al. (27). The variation moment was calculated according to Rees et al. (13) .   FIG. 3. Cellular levels of toxic lac permeases. Panel A, missense mutants. Whole-cell protein (ϳ80 g) derived from the PcnB ϩ strain CC1351 carrying pCS111 (encoding wild-type LacY), pBR322 (vector plasmid), or pCS112 encoding the indicated substitution mutants was analyzed by Western blotting using antibody directed against the COOHterminal region of LacY. Panel B, frameshift mutants. Whole-cell protein (ϳ80 g) derived from the PcnB Ϫ strain CC1006 carrying pCM701 (encoding epitope-tagged wild-type LacY), pBR322 (vector plasmid), or pCM701 with the indicated frameshift mutations was analyzed by Western blotting using antibody directed against the 31-amino acid epitope. *, an unidentified protein that cross-reacts with the epitope antibody. High copy no. Low copy no.
6 ϫ 10 Ϫ3 0.2 a As determined by Western blot analysis using antibody directed against the COOH-terminal region of lac permease (see "Materials and Methods").
b [ 14 C]Lactose uptake was determined for mutant plasmids in strain CC819 as described under "Materials and Methods." An isogenic ⌬(lacY) mutant showed 1% the lactose transport activity of the lacY ϩ strain.
c Determined as the ratio of colony-forming units on minimal glycerol agar containing cAMP ϩIPTG/ϪIPTG (see "Materials and Methods"). The assays were carried out for both cells maintaining lacY plasmids at high copy number (pcnB ϩ ) and low copy number (pcnB Ϫ ). Residues of TM12 at which active substitutions were not recovered (Fig. 2) are shown in bold type in the wild-type sequence. *, small colonies on M63 glycerol containing cAMP and IPTG and sectoring/papillation on 1/4 ϫ TYE containing cAMP and IPTG; **, normal sized colonies on M63 glycerol containing cAMP and IPTG and sectoring/papillation on 1/4 ϫ TYE containing cAMP and IPTG. Dots correspond to positions with wild-type residues. 21), as have overproduced substitution mutants of yeast proton ATPase (22).
A simple model for mutant membrane protein toxicity is that it results from disruption of the lipid bilayer by polypeptides that insert into the bilayer but are unable to fold correctly. The resulting breakdown of the permeability barrier would kill cells or slow their growth. The effect might be analogous to that observed when cells are exposed to natural or synthetic amphipathic peptides that disrupt membranes (4,23,24).
Several degenerative diseases in humans are known to be caused by mutant integral membrane proteins, including retinitis pigmentosa and Charcot-Marie-Tooth syndrome (1-3). Many of the mutations associated with these diseases are analogous to those that make lac permease toxic. For example, mutations in the rhodopsin gene which introduce charged residues or prolines in transmembrane segments or cause frameshifts may lead to retinitis pigmentosa (1,2). Recent studies have also suggested that frameshift derivatives of ␤-amyloid protein may contribute to Alzheimer's disease (25). These similarities suggest that the pathogenesis of such degenerative diseases could include toxic effects analogous to those observed with mutant lac permeases.