A Conservative Amino Acid Substitution, Arginine for Lysine

THEJOURNAL OF BIOLOGICAL CHEMISTRY 0 1989 by The American Society for Biochemistry and Molecular Biology. Inc.

Vol. 264, No. 33, Issue of November 25, p p . 20082-20088,1989 Printed in 0: S.A.

A Conservative Amino Acid Substitution, Arginine for Lysine, Abolishes Export of a Hybrid Protein inEscherichia coli IMPLICATIONS FOR THE MECHANISM OF PROTEIN SECRETION* (Received for publication, May 17, 1989)

Richard G . Summers$, Chris R. Harris, and Jeremy R. Knowless From the Departmentsof Chemistry and Biochemistry, Harvard University, Cambridge, Massachusetts 02138

A hybrid protein that comprises the @-lactamasesignal peptide fused precisely to chicken muscle triosephosphate isomerase is not secreted into the periplasm of Escherichia coli. The protein can be secreted, however, if an arginine residue at position 3 of the isomerase is replaced by either a serine or a proline residue. In contrast, replacement of a neighboring lysine residue has no effect on secretion of the protein. Furthermore, if the arginine is removed from position 3 to generate a secreted protein, but is then reintroduced in place of the neighboring lysine, the blockade to secretion is re-established. The singular effect of the arginine residue on secretion does not result from the role this residue plays in the formation or stabilization of the native isomerase structure: mutational alterations remote from the N terminus of the isomerase that prevent the proper folding of the protein do notrelieve the block to secretion. The finding that an arginine residue prevents secretion while a lysine residue does not, suggests that basic residues near the mature N terminus of a secreted protein must be deprotonated if orderly export is to occur. This implies that the signal peptide along with the N-terminal portion of the mature protein partitions directly into the lipid bilayer in the course of the secretory process.

Three types of models have been proposed to explain the almost universal requirement for an N-terminal signal peptide in protein secretion. First, the signal hypothesis envisions that the signal peptide serves primarily as a recognition element. The peptide interacts with a specific receptor such as the signal recognition particle that recruits the protein to a secretory apparatus (1, 2). Once recruited, the protein is extruded linearly across the membrane through a proteinaceous channel (3). In contrast, the membrane trigger hypothesis speculates that thesignal peptide allows a secretable protein to undergo a series of conformational changes that propel the protein through the membrane (4). These conformational changes are mediated first by an interaction between the signal peptide and the nascent presecreted protein, then by an interaction of this species with the lipid bilayer, and last, by the proteolytic removal of the signal peptide on the distal side of the membrane. Each conformational change

* This work was supported by the National Institutes of Health and the National Science Foundation. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. $ Present address: School of Pharmacy, University of WisconsinMadison, 425 N. Charter St., Madison, WI 53706. § To whom correspondence should be addressed.

moves the protein along the secretory pathway by virtue of the different partitioning of the various conformations between aqueous and lipid phases. Finally, the direct transfer model (5),the loop model (6), and thehelical hairpin hypothesis (7) each suggests that thesignal peptide initiates protein secretion by dissolving directly in the lipid bilayer of the membrane. The free energy of insertion of the signal peptide into the membrane pulls the N-terminal residues of the mature protein into thebilayer. Transfer of the remaining residues of the protein into and through the membrane is then energetically more or less neutral, the entry of each residue (or group of residues) into thelipid bilayer being matched by the exit of another residue (or group of residues) on the opposite side of the membrane. In the preceding paper (8) we demonstrated that the /Ilactamase signal peptide alone is not sufficient to drive the normally cytoplasmic enzyme triosephosphate isomerase across the inner membrane of Escherichia coli, although the isomerase can be secreted if a few residues of the mature @lactamaseintervene between the signal peptide andthe isomerase. We were thus able to delineate a region of constrained properties that immediately follows the signal peptide, and thatencompasses about the first 14 residues of the mature protein. Such a region had been anticipated by those models of protein secretion that postulate the co-ordinate insertion of the signal peptide and thebeginning of the mature protein directly into the lipid bilayer ( i e . the direct transfer model, the loop model, and the helical hairpin hypothesis). Consistent with these findings, other studies in both prokaryotes (9-13) and eukaryotes (14, 15) have demonstrated that portions of the mature protein,especially near the N terminus, influence the efficiency of secretion. Here we examine secretion of triosephosphate isomerase in theabsence of sequences from the mature 8-lactamase in order to define further the characteristics of a mature protein that dictate its secretion competence in E. coli. We have considered two explanations for the failure of the @-lactamase signal peptide alone to direct the secretion of triosephosphate isomerase. First, when the isomerase is not secreted, it has adopted what appearsto be its native tertiary structure in the cytoplasm (8). Folding of the protein thus may preclude its subsequent secretion if translocation across the bilayer requires an unfolded or an alternative conformation. Indeed, Randall and Hardy (16) have correlated the folding of maltose-binding protein into itsnative structure in the cytoplasm with a loss of secretion competence, and have suggested that the signal peptide acts to delay the folding of the preprotein until interaction with the secretory apparatus has occurred. As a second possibility, we have considered that the positively charged amino acid residues that pepper the N terminus of the isomerase might interfere with secretion of

20082

Amino Acid Substitution Preuents Protein Secretion

20083

this enzyme. Proteins that are secreted from bacteria gener- similarly ligated into M13mpll RF DNA, appropriately linearized. The RF DNA circles that now carry the regions to be mutagenized ally have few basic amino acids near the N terminus of the were then used to transform E. coli strain JM101, and mutagenic mature protein (17), and the addition of arginine or lysine template was prepared from single clear plaques after growth on residues to the N termini of several secreted proteins has been JM101. Oligonucleotideswere allowedto hybridize to theappropriate template and wereused to prime in vitroDNA synthesis by the found to inhibit their export in prokaryotes (9, 10, 12, 13). The results reported here indicate that the secretion incom- Klenow fragment of DNA polymerase I ip the presence of 2”deoxypetence of triosephosphate isomerase is not due to its rapid cytosine 5”a-thiotriphosphate. Subsequent ligation incorporated the oligonucleotide into a covalently closed circle that was treated acfolding in the cytoplasm, The block to secretion is localized cording tothe commercial protocol andthen used to transform within the first 14 residues of the protein, and is relieved JM101. Typically, three plaques were chosen from the transformants when an arginine residue at position 3 (R3) is replaced by an that resulted and the DNA in the region of the desired mutation was uncharged residue. It is not j u s t the unit positive charge of sequenced by the SequenaseTMmethod (United States Biochemical the arginine residue at neutral pH that inhibits secretion of COT., Cleveland, OH). Except for the D226N mutation, a fragment the isomerase, however. If, having replaced the arginine at that carries the desired mutation, obtained by digestion of the approRF DNA with EcoRI and S a d , was ligated into plasmid pBllT, position 3 by proline (with the result that the protein is priate similarly linearized. The fragment that carries the D226N mutation secreted), an arginine residue is reintroduced i n place of the was obtained by digestion of the RF DNA with KpnI and PstI, and lysine at position 4, secretion of the protein is blocked anew. this was ligated into plasmid pBOT, analogously digested. Transfer secretion while of each of the mutations was verified by DNA sequencing. This finding,that an arginine residue inhibits Construction of pBOT (A22-69A), pBOT(R3P,D226N), pBTB, and an identically placed lysine residue does not, suggests that these residues must be deprotonated if export is to occur, and pBTB(R3P)”Plasmid pBOT, digested initially with SacI and KpnI, was treated with nuclease S1 followed by end repair with the Klenow implies that the signal peptide and the N terminus of the fragment of DNA polymerase I. Ligation formed the plasmid pBOT mature protein partition directlyinto the membrane early in (A22-69A)lacking a small fragment and this was used to transform the secretory process. E. coli strain DH1 totetracycline resistance (tetR).The deletion was verified by DNA sequencing. Plasmid pBOT(R3P,D226N) was constructed from plasmids EXPERIMENTALPROCEDURES pBOT(R3P) and pBOT(D226N). Fragments that carried the approBacterial Strains, Bacteriophage, Plasmids, and Media-E.coli priatemutations, obtained by digestion of both pBOT(R3P) and strains DH1 (endAI, hsdR17, supE44, thi-1, recAI, gyrA96, relAI) pBOT(D226N) with SacI and NheI, were isolated from low gel(18),DF502 (A(rha-pfkA-tpi),pfkBI, pyrD, strA, galB, sup, his-, edd-) ling temperature agarose. Following ligation, the DNA was used to (19), and JMlOl (Alacpro, supE44, thi/F, traD36, proAB, ladq, transform E. coli strain DH1 to tetracycline resistance. Assembly lacZAM15) (20), the bacteriophage Ml3mpll (20), and plasmids of bothmutations sequencing. Plasmids was verified byDNA pBOT and p B l l T (8), pX1 (19), and pAP5* have been described pBOT(R3K,D226N) and pBOT(K4Q,D226N) were constructed analelsewhere. Except when otherwise indicated, bacterial media were ogously. prepared as described in Maniatis et al. (22). Dry media were purTo form plasmids pBTB and pBTB(R3P), the coding sequence for chased from Difco Laboratories (Detroit,MI). The salts NaCl, the first 14 residues (alanine = 1) of chicken muscle triosephosphate Na2HP04, KH2P04, NH,Cl, MgCL, MgS04, andCaC12 wereobtained isomerase (either the wild-type codons or the R3P altered codons) from Sigma or Mallinckrodt Inc. a-Lactose was purchased from were inserted by oligonucleotide-directed mutagenesis between the Fisher Scientific Co. L-Amino acids and antibiotics were purchased codons for the (3-lactamase signal peptide and the codons for the from Sigma. mature p-lactamase of pAP5. The fragment that carried the region Enzymes and Chemicals-All restriction endonucleases were pur- to be mutagenized was obtained by digestion of plasmid pAP5 with chased from New England Biolabs (Beverly, MA), Bethesda Research EcoRI and PstI, andthis was ligated into M13mpllRF DNA, Laboratories (Gaithersburg, MD), or Boehringer Mannheim. Klenow similarly linearized. The RFDNA circles that now carry the regions fragment of DNA polymerase I and T4 DNA ligase were purchased to be mutagenized were then prepared as described, using the Amerfrom either New England Biolabs or Bethesda Research Laboratories. sham oligonucleotide-directed in vitro mutagenesis system. Following S1 nuclease was purchased from Boehringer Mannheim. Buffers for mutagenesis, fragments carrying the desired insertions were obtained these enzymes were either used as provided or prepared as described by digestion of RF DNA with EcoRI and PstI, and these fragments elsewhere (22). Low gelling temperature agarose was purchased from were returned to pAP5, analogously digested. Incorporation of each FMC BioProducts (Rockland, ME). Nucleoside triphosphates, o- of the insertions was verified by DNA sequencing. nitrophenyl-0-D-galactopyranoside,DL-glyceraldehyde 3-phosphate, f5S/Methionine Radiolabeling of Cell Cultures, Cell Fractionation, NADH, and protein A-Sepharose CL-4B were purchased from Sigma. and Proteinase K Digestions-Cultures of strain DH1 transformed Nitrocefin was purchased from BBL Microbiology Systems of Becton with pX1 or with the gene fusion plasmids were grown, radiolabeled, Dickinson and Co. (Cockeysville,MD). Proteinase K was purchased fractionated, and, when appropriate, treated with proteinase K as from Boehringer Mannheim. Glycerol-3-phosphate dehydrogenase described previously (8). was purchased from Boehringer Mannheim and was treated with Immune Precipitations, Denaturing Polyacrylamide Gel Electrophobromohydroxyacetone phosphate to remove triosephosphate isomer- resis, and Fluorography-Details of our technique have been reported ase activity (23). 2’-Deoxyadenosine 5’-[a-35S]thiotriphosphate and (8). Samples were treated with either rabbit anti-chickentriosephos~-[~‘S]methionine were purchased from Amersham Corp. Oligonucleo- phate isomerase serum or rabbit anti-p-lactamaseserum as appropritides were synthesized by the phosphoramidite method on a MilliGen ate and then incubated at 4 “C for 30 min. Immune complexes were 7500 DNA synthesizer. then collected and prepared for electrophoresis, as described. Finally, Oligonucleotide-directedMutagenesis-Unless otherwise indicated, samples were loaded onto 12% slab gels prepared using the stacking the techniques and conditions used for DNA manipulations were as system and buffers of Laemmli (24) and run at constant current (24 described by Maniatis et al. (22). The DNA molecules constructed mA). To prevent acidification of the gel and theresulting compression below are based on plasmid pBOT. Point mutants were produced of the protein bands,the buffer in the lower (positive) electrophoresis using the Amersham oligonucleotide-directed in vitro mutagenesis chamber was identical to that of the resolving phase of the gel (375 system (Amersham Corp.). Except for the construction of the D226N mM Tris-HC1, pH 8.8, containing sodium dodecyl sulfate (0.1%, mutation, a fragment that carried the region to be mutagenized was w/v)). Gels werefixed in water:methanol:acetic acid (5:41, by volume) obtained by digestion of plasmid pBOT or pBOT(R3P) with HindIII and thenenhanced with the fluorophore Enlightnine (Du Pont-New and S a d , and this was ligated into MlSmpll RF2DNA that had been England Nuclear), before drying in a heated apparatus under reduced analogously linearized. For the construction of the D226N mutation, pressure. Dried gels were exposed to Kodak X-Omat film at -70 “C. a fragment that carried the region to be mutagenized was obtained TriosephosphuteIsomerase Assay-Cultures of strain DF502 transby digestion of plasmid pX1 with HindIII and PstI, and this was formed with the appropriate plasmids were grown overnight in Y T medium containing tetracycline (20 pg/ml). These inocula were di’ A. Laminet and A. Pluckthun, EMBO J., submitted for publica- luted 1000-fold into M9 salts (5 ml) supplemented with lactose (0.4%, tion. w/v), casamino acids (0.25%, w/v), CaC12 (0.2 mM), MgSO, (1 mM), The abbreviation used is: RF, replicative form. thiamine hydrochloride (2 pg/ml), and appropriate antibiotics.

Amino Acid Substitution Prevents Protein Secretion

20084

Growth was allowed to continue at 37 "C until the Aemnmwas between 0.5 and 1.0. After the cultures had been chilled on ice for 10 min, the bacteria were isolated by centrifugation at 5000 X g for 10 min at 4 "C. Cell pellets were washed at 4 "C by suspension in chilled M9 salts (5 ml) and were reisolated by centrifugation (5000 X g for 10 min at 4 "C). Cell fractions were then generated as described previously (8).The periplasmic and soluble spheroplast fractions of each strain were assayed for the presence of triosephosphate isomerase enzymaticactivity.Triosephosphate isomerase activity was determined at pH 7.6 by the method of Plaut and Knowles (25) using DLglyceraldehyde 3-phosphate as the substrate.

methionine. As Fig. 1 shows, this change does not allow secretion of the isomerase into theperiplasm. The preprotein (with a very small amount of a lower molecular weight species) is observed in the spheroplast fraction at all time points. To show that theD226N mutation does disrupt the tertiary structure of the isomerase, the catalytic activityof the enzyme was assayed in an isomerase-minus host strain (E. coli DF502). The parentalenzyme (encoded by pBOT) is active in this strain,but the D226N mutant is not. Since aspartate 226 is far from the active site of triosephosphate isomerase and is not involved in catalysis, it is likely that theD226N preprotein RESULTS has not folded properly. This viewwas confirmed by the Mutations That Alter Tertiary Structural Stability Do Not susceptibility of this protein to digestion by proteinase K (Fig. Affect Secretion of Triosephosphate Isomerase-Plasmid 2 ) . The isomerase portion of the parentalprotein is resistant pBOT encodes a hybrid protein that comprises the P-lacta- to proteolysis after detergent solubilization of the spheromase signal peptide fused precisely to chicken muscle trioseplasts and treatment with the protease (Fig. 2 A , lane L9 phosphate isomerase. This hybrid protein is not secreted into K ) : only the pendant p-lactamase signal peptide is removed. the periplasm of E. coli, and the preprotein accumulates in In contrast, the D226N mutant is completely digested in the cytoplasm (8).To examine the possibility that the isosolubilized spheroplasts (Fig. 2B, lane LS + K). Clearly the merase is not secreted because it folds rapidly into a conformation that is not competent for export, we have introduced D226N mutation has disrupted the native tertiary structure structure-alteringmutations intothe isomerase and have of the isomerase, but thisdisturbance is not sufficient to allow determined the cellular location of the resulting constructs secretion of the protein. (The proteolysis experiment also confirms that theD226N mutant preprotein is trulyintracel(Table I). In mutant D226N, the aspartate residue at position 226 of lular since the preprotein is not digested by protease when the native isomerase has been replaced by asparagine. The the spheroplast membrane is intact (Fig. 2B, lane S K ) . ) Although the D226N mutation disrupts the normal strucrationale for this change derives from consideration of the ture of the isomerase, this change could be too subtle to primary and tertiary structure of triosephosphate isomerase. Aspartate 226 is a conserved residue in eight of the nine influence secretion of the protein. Consequently, a more drassequenced triosephosphate isomerases (26,27), and the crystal tic alteration was made that removes residues 22-69 of the structures of the chicken and yeast enzymes suggest that this isomerase, and replaces them with a single alanine residue residue forms a structurally important ion pair with a simi- (A22-69A). Since triosephosphate isomerase monomers norlarly conserved arginine residue at position 3 (R3) (28, 29). mally fold into reasonably self-contained domains, this brutal (In the one exceptional case, that of the triosephosphate deletion of47 residues should abolish all elements of the isomerase from trypanosomes, both R3 and D226 are replaced native tertiarystructure of the isomerase. Moreover, the deletion occurs early in the translated sequence and is exby uncharged residues (P and N, respectively: see Ref. 27). The coordinate replacement of these two conserved residues pected to interfere from the outset with the normal pathway supports the view that for most isomerases the ion pair of events that occur in the folding of this protein. Yet, as is between R3 and D226is structurally important.) We may shown in Fig. 3A, even this truncated protein is not secreted expect, therefore, that breaking this ion pair will destabilize into the periplasm. A single species, corresponding in size to the native structure of the isomerase. Accordingly, the cellular the precursor, is found in the spheroplast fraction. As exlocation of the protein encoded by pBOT(D226N) was exam- pected, this polypeptide is digested by protease in lysed spheined as a function of time after a short pulse with radioactive roplasts, indicatingthat thenative (protease-resistant) struc-

+

+

TABLEI Mutations engineered into the triosephosphate isomerase portion of the P-lactamase/triosephsphate isomerase hybrid protein encoded by plasmid pBOT and the effectof these mutations on thesecretion and protease sensitivity of the hybrid protein Mutation

Description

None D226N A22-69A 47 R3S R3P R3P,D226N

Wild-type Aspartate 226 to asparagine amino acid deletion Arginine 3 to serine Arginine 3 to proline Arginine 3 to proline and aspartate 226 to asparagine Lysine 4 to glutamine Lysine 4 to glutamine; aspartate 226 to asparagine Arginine 3 to proline and lysine 4 to arginine Arginine 3 to lysine Arginine 3 to lysine and asNot partate 226 to asparagine

Effect on secretion

Protease sensitivity

Enzymatic activity

Amino acid sequence of altered isomerase"

4

K4Q K4Q,D226N R3P,K4R R3K R3K,D226N

Active Not secreted Resistant Not active Not secreted Sensitive Not secreted Sensitive NDb ND Secreted Sensitive Ala-Pro-Ser-Lys ND Secreted Sensitive ND Secreted Sensitive Ala-Pro-Pro-Lys Not secreted Not secreted

Resistant ND Ala-Pro-Arg-Gln ND Sensitive

Not secreted

Sensitive

Not secreted Resistant secreted Sensitive

Differences from the wild-type sequence are in bold type.

* ND. not determined.

Not active ND ND Ala-Pro-Lys-Lys

13 2 226 Ala-Pro-Arg-Lys . . . Asp. . . Ala-Pro-Arg-Lys . . . Asn . . . Ala-Pro-Arg-Lys . . .Asp. . . . . . Asp.. . Ala-Pro-Pro-Lys . . . Asp.. . , . . Asn . . .

. . .Asp. . . .., A m . . . Ala-Pro-Pro-Arg . . . Asp. . . Ala-Pro-Lys-Lys . . .Asp. . . Ala-Pro-Arg-Gln

...A m . . .

Amino Acid Substitution Prevents Protein

Secretion

20085

we considered the possibility that some feature of the isomerase sequence could be the source of the secretion blockade. The N terminus of triosephosphate isomerase carries considerably more positive charge than that of a typical bacterial secreted protein (17), and it seemed likely that this region might contain the barrier to secretion. To test this possibility, we constructed a chimeric protein in which the first 14 resiFIG.1. Location of t h e proteins encoded by pBOT(D226N). dues of triosephosphate isomerase were interposed between Cellswere pulse-labeled for 30 s with [35S]methionineand chased the p-lactamase signal peptide and the mature p-lactamase. with an excess of unlabeled methionine for 30 s, 4 min, and 12 min. The samples were then separated into spheroplast-associated (S)and If some feature of these 14 residues of the isomerase prevents periplasmic ( P ) proteins. Triosephosphate isomerase species were export, then secretion of the p-lactamase in this chimera immune precipitated and analyzed by denaturing polyacrylamide gel should be blocked. Indeed, secretion of this protein (BTB) electrophoresis and fluorography. w-t,native triosephosphateisomer- into theperiplasm cannot be detected (Fig. 4), and theprotein ase encoded by pX1 (19). remains as the preprotein in the cytoplasm. The first 14 residues of triosephosphate isomerase thus encompass the BOT BOT secretion blockade. B (D226N) The Role of Basic Residues in the Inhibition of SecretionA W " Among the first14 amino acids of triosephosphate isomerase, s s LS s s LS positively charged residues fall at positions 3 (arginine), 4 (lysine), and 12 (lysine). There are no acidic residues in this K K K K *..~ region, so the N terminus of the isomerase carries a charge of +3 at physiological pH (Fig. 5). The role of these charged residues in the secretion of the isomerase has been assessed from the behavior of the mutants described below (summarized in Table I). When arginine 3 is changed to either a serine or a proline, FIG.2. Proteinase K sensitivity of the proteins encoded by pBOT and pBOT(D226N). Cells were pulse-labeled for 90 s with a fraction of each of the resulting mutant isomerases R3S and ["S]methionine and chased with an excess of unlabeled methionine R3P is secreted into the periplasm (Fig. 6, A and C). Aside for 4 min. Whole spheroplasts ( S ) and lysed spheroplasts (LS)were from the processed periplasmic species, a substantialportion generated from the sample and a portionwas treated with proteinase of each mutant is processed but retained with the spheroplast K (+K)for 5 min a t 20 "C. Triosephosphate isomerase species were fraction, and some preprotein is also found in the spheroplast immune precipitated and analyzed by denaturing polyacrylamide gel (Fig. 6, A and C). The processed species in the spheroplast electrophoresis and fluorography. A , proteins encoded by pBOT. B, fraction is a form that has beenmore or lesscompletely proteins encoded by pBOT(D226N). w-t, nativetriosephosphate translocated, but not released fromthe innermembrane. This isomerase. location is demonstrated by protease treatment of whole spheroplasts and detergent-solubilized spheroplasts, the reA w 1 30" " - 4' 12' B E ! sults of which are illustrated in Fig. 6, B and D.Behavior of S P S P S P S P s s LS + + this type (that is, translocation but not release) has been observed forother P-lactamaseltriosephosphate isomerase hybrid proteins (€9, for the wild-type p-lactamase secreted from cells grown at lower temperatures (30),and for several normally secreted proteins that have been altered by mutation (31-33).We presume that this results from improper folding FIG.3. Location and proteinase K sensitivity of the proteins of the protein that precludes release fromthe membrane. The A , location. Cells were pulse-labeled encoded by pBOT(A22-69A). for 30 s with [35S]methionineand chased with an excess of unlabeled protease treatment also reveals that the isomerase moiety of methionine for 30 s, 4 min, and 12 min. The samples were then the R3S and R3P preproteins has not adopted its native separated into spheroplast-associated ( S ) and periplasmic ( P ) pro- structure in the cytoplasm: both preproteins are completely teins. Triosephosphate isomerase species were immune precipitated digested by the protease when the spheroplast membrane is

+ +

+ +

and analyzed by denaturing polyacrylamide gel electrophoresis and fluorography. w-t,native triosephosphate isomerase. B, proteinase K sensitivity. Cells werepulse-labeled for 90 s with [35S]methionineand chased with an excess of unlabeled methionine for 4 min. Whole spheroplasts (S) and lysed spheroplasts (LS)were generated from the sample and a portion was treated with proteinase K (+K)for 5 min a t 20 "C. Triosephosphate isomerase species were immune precipitated and analyzed by denaturing polyacrylamide gel electrophoresis and fluorography.

BTB

BTB (R3P) P S P

" s w- t

+

ture of the isomerase has not formed (Fig. 3B,lane LS K). When the spheroplast membrane is intact the moleculeis protected from proteolysis (Fig.3B, lane S K ) : the preprotein is thus completely intracellular. We must conclude from these results that thesecretion incompetence of these hybrid proteins is not due to their rapid folding into a stable native structure. The Unsecretability of Triosephosphate Isomerase Derives from Residues Near the N Terminus-Since the failure of the isomerase to be secreted is not explained by its rapid folding,

+

FIG.4. Location of the @-lactamase/triosephosphate isomerase/@-lactamasechimericproteinsencodedbypBTBand pBTB(R3P). Cells were pulse-labeled for 90 s with [35S]methionine and chased with an excess of unlabeled methionine for 4 min. The samples were then separated into spheroplast-associated ( S ) and periplasmic (P)proteins. P-Lactamase species were immune precipitated and analyzed by denaturing polyacrylamide gel electrophoresis and fluorography. w-t,wild-type p-lactamase.


20086 -20

+

-ts

-to

-5

++

I

10

(5

+

+

20

~:BIQE=~VA_:P?=A~~'@,"W?AAPRKFFVGGNWKMNG~LKSLGTL-

FIG.5. DNA-derived amino acid sequence of the N terminus of the ,"ctamase/triosephosphate isomerase hybrid protein encoded by pBOT. The p-lactamase signal peptide is in open characters. Charged residues are indicated. A

w-1

30" P

12'

4'

" "

S

P

S

S

P

S

B

w-t

30"

12'

4'

" "

s

P ,S-P-S

P

s

P

LS

+K +K

R3S

C

A

w-1 -~

s s

P

D

c LS

+ + K

K

R3P

P

30" S P

..

4' S

P

S

B K !

12' P

s s

,

w-t

S

P

S

30" P

K

K

,-

12'

4' S

LS

t t

.ah7

K4Q

w-t -~

s s

w-1

S

P

S

P

K4Q, D226N

FIG.6. Location and proteinase K sensitivityof the proteins encoded by plasmids pBOT(R3S) and pBOT(R3P). A and B, location of the proteins encoded by pBOT(R3S). C and D, location of the proteins encoded by pBOT(R3P). A and C, location. Cells were pulse-labeled for 30 s with [35S]methionineand chased with an excess of unlabeled methionine for 30 s, 4 min, and 12 min. The samples were then separated into spheroplast-associated (S) and periplasmic (P)proteins. Triosephosphate isomerase species were immune precipitated and analyzed by denaturing polyacrylamide gel electrophoresis and fluorography. w-t, native triosephosphate isomerase. B and D, proteinase K sensitivity. Cells were pulse-labeled for 90 s with [35S] methionine and chased with an excess of unlabeled methionine for 4 min. Whole spheroplasts (S)and lysed spheroplasts (LS) were generated from the sample and a portion was treated with proteinase K (+K) for 5 min at 20 "C. Triosephosphate isomerase species were immune precipitated and analyzed by denaturing polyacrylamide gel electrophoresis and fluorography.

+

disrupted by detergent (Fig. 6, B and D, lunes LS K ) . It is likely, therefore, that theR3 alterations allow secretion of the isomerase by eliminating the primary arginine-basedblock to secretion and, possibly, by preventing formation of the enzyme's native structure. To explore the generality of the beneficial R3P change, this mutation was engineered into the P-lactamaseltriosephosphate isomeraselp-lactamase chimeric protein described in the previous section. Consonantwith the results for the complete isomerase, the R3P mutation dramatically relieves the blockade to secretion in this chimera (BTB, R3P) (Fig. 4). In contrast to replacement the of R3 by uncharged residues, a similar change in the lysine residue at position 4, as inK4Q, does not result in protein secretion (Fig. 7A). Proteolysis experiments suggest that the K4Q mutation does not fully disrupt the normal structure of the isomerase: the isomerase domain of the preprotein shows some resistance to proteolysis (Fig. 7 B ) . To eliminate any secondary effects due to folding that may mask the full effect of substitution for the lysine, the D226N change described earlier that blocks the folding of the isomerase was combined with the K4Q change. The double mutant (K4Q,D226N) is, however, still not secreted (Fig. 7C). Complete protease digestion after detergent solubilization of the spheroplasts demonstrates that the tertiary structure of the K4Q,D226N protein has been disrupted (Fig. 7 0 , lane LS K ) . (The D226N mutation itself does not prevent protein

+

E

w-t

30"

4' ..

12'

' .S

P 1

F .

w-1 -s s

R3P, K4R

LS

t

+

K

K

0FIG.7. Location and proteinase K sensitivity of the proteins encoded by pBOT(K4Q), pBOT(K4Q,D226N), and pBOT(R3P,K4R). A , location of the proteins encoded by pBOT(K4Q). Cells were pulse-labeled for 30 s with [35S]methionine and chased with an excess of unlabeled methionine for 30 s, 4 min, and 12 min. The samples were then separated into spheroplastassociated (S) and periplasmic ( P )proteins. Triosephosphate isomerase species were immune precipitated and analyzed by denaturing polyacrylamide gel electrophoresis and fluorography. w-t, native triosephosphate isomerase. B, proteinase K sensitivity of the proteins encoded by pBOT(K4Q). Cells were pulse-labeled for 90 s with [35S] methionine and chased with an excess of unlabeled methionine for 4 min. Whole spheroplasts (S) and lysed spheroplasts (LS)were generated from the sample and a portion was treated with proteinase K (+K) for 5 min at 20 "C.Triosephosphate isomerase species were immune precipitated and analyzed by denaturing polyacrylamide gel electrophoresis and fluorography. C, location of the proteins encoded by pBOT(K4Q,D226N). Samples were prepared and analyzed as in A . D, proteinase K sensitivity of the proteins encoded by pBOT(K4Q,D226N).Samples were prepared and analyzed as in B. E, location of the proteins encoded by pBOT(R3P,K4R). Samples were prepared and analyzed as in A . F, proteinase K sensitivity of the proteins encoded by pBOT(R3P,K4R). Samples were prepared and analyzed as in B.

secretion: the R3P,D226N double mutant is secreted slightly more rapidly than that of the R3P protein.) We must conclude, therefore, that R3 plays a different role from K4 in blocking the secretion process. Since the R3S and R3P mutant isomerases are secreted, whereas the K4Q mutant isomerase is not, it seemed that either theposition or thehigher pK, of the arginine could be responsible for the inhibition of protein export. To examine this question, an arginine residue was reintroduced to the secretable R3P protein inplace of its lysine residue at position 4. In the resulting double mutant R3P,K4R, the N terminal

Amino Acid Substitution Prevents Protein Secretion a brutal deletion that removes from the isomerase the 47 residues that follow glycine 21 enables the (3-lactamase signal peptide to guide secretion of the isomerase. Furthermore, the interposition of just the first 14 residues of the isomerase between the p-lactamase signal peptide and the mature 8lactamase blocks secretion of the p-lactamase, and thisblockade is dramatically eliminatedwhen the offending arginine is replaced by proline (see Fig. 4). At physiological pH, both lysine and arginine side chains are essentially fully protonated (thepK, of lysine’s €-ammonium group is about 10.5, and that of arginine’s guanidinium group is about 12.5), and each carries a charge of +l. Yet an arginine residue near the N terminus of the mature protein prevents export, while a similarly placed lysine residue does not. This finding indicates that the block to secretion does not derive from the interaction of the cationic group with the membrane potential (34), nor from a requirement that the N terminus of a secretable protein be bipolar (17), but rather from the primary difference between the two side chains, namely their pKn values. The singularly deleterious effect of an arginine residue near the mature N terminus of a protein that is to be secreted therefore suggests that the residues immediately following the signal peptide must be deprotonDISCUSSION ated at some point in the secretory process. Such a requireTo examine the features of a protein that influence its ment implies that these residues leave the aqueous phase and secretion into the periplasm of E. coli we have made a series enter a non-polar environment such as themembrane bilayer. of changes in anonsecreted hybrid protein that comprises the Indeed, the fact that monocarboxylic acids and monoamines p-lactamase signal peptide fused precisely to the complete (each having a pK, value about 3 units away from the pH of triosephosphate isomerase from chicken muscle. Some of the most physiological systems)can passively cross biological alterations have been designed to disrupt the tertiary struc- membranes, whereas phosphates and guanidinium comture of the isomerase while leaving the N-terminal sequence pounds cannot (35), is consistent with the difference between of the protein unchanged, while otheralterations have Lys and Arg noted here. The length of the mature N terminus that must be buried changed the basic amino acid residues that lie near the N in the bilayer is suggested by the results described in the terminus of the mature isomerase (Table I). Our results show that formation of the native structure of previous paper (8).There itwas shown that efficient secretion the isomerase is not responsible for the secretion incompe- of the P-lactamaseltriosephosphate isomerase hybrid proteins tence of the protein. Rather, the nature of the N-terminal requires intervention of the first 12 amino acids of mature presidues of the isomerase is the key factor in determining lactamase between the endof the signal peptide and the start whether the p-lactamase signal peptide can direct the isomer- of the triosephosphate isomerase. This places arginine 3 of ase intothe periplasm. Observations consistent with this the isomerase at position 15 of the mature hybrid protein. conclusion have been reported for the secretion of other Thus, at least the first 14 N-terminal residues of the mature proteins in prokaryotes (9-13) and eukaryotes (14, 15).It has portion of a bacterial secreted protein may have to become been suggested by von Heijne (17) that positive charge may buried in the membrane in order that protein secretion be not be tolerated near the N terminus of proteins that are properly initiated. (It is arguable whether this stepis an early secreted in bacteria, and it hasbeen shown that theintroduc- step, and evidence has been presented that the hydrophobic tion of basic residues into thisregion can inhibit secretion in core of the signal sequence hasits effect notduring the lower organisms (9, 10, 12). We have found here, however, attachment of the precursor protein to the membrane but that the basic side chains of arginine and lysine arenot rather during the subsequent translocation (36).)Outside this equivalent in termsof their effect on secretion. Thus, whereas critical region, charged residues seem not to interfere with replacement of arginine 3 with serine or proline allows secre- translocation across the membrane (8). Within this critical tion of the isomerase (in mutants R3S or R3P), replacement region, however, there is a limit to the hydrophilicity that is of the vicinal lysine at position 4 with a glutamineresidue (in compatible with entry into the secretion pathway. Thus we K4Q) has no beneficial effect. Moreover, if, after replacing find that 1 lysine residue near the N terminus of triosephosthe arginine residue at position 3 (thus generating the secreted phate isomerase is tolerable, but that 1 arginine or 2 lysine mutant R3P), an arginine is reintroduced in place of the residues, are not. Unsurprisingly in this context, it has been lysine residue at position 4 (to create R3P,K4R), the secretion observed that acidic residues can counterbalance basic resiblockade is re-established. The arginine residue evidently dues within the mature N terminus (8, 12, 37, 38), and we prevents secretion of the isomerase while an identically posi- may presume that the beneficial effect of counterbalancing tioned lysine residue does not. charges results from the formation of ion pairs that enables The beneficial effect of removing the arginine residue at these charged residues to enter the membrane without the position 3 does not derive from the role of this residue in the unfavorable energetic consequences of (de)protonation. Such formation or stabilization of the native structure of the ion pairs will be of particular benefit when they involve an isomerase, since structure-altering mutations remote from the arginine residue, since the energetic cost to deprotonate this N terminus do not allow the enzyme to be secreted. Neither residue is so high. Indeed arginine residues are never observed a point mutationthat breaks a structurallyimportant ion pair near the mature N terminus of secreted bacterial proteins in between arginine 3 and aspartate226 at thedistal residue nor the absence of nearby acidic residues (for a compilation of charge is the same as that of the secreted R3P protein, and this charge is located at the same position in the primary structure of the isomerase. The only difference is that this charge is now carried by an arginine residue rather than a lysine residue. As Fig. 7 E shows, this arginine for lysine substitution completely abolishes secretion of the hybrid protein, which now remains as the preprotein in the cytoplasm. The preprotein is protected from proteolysis in intact spheroplasts, and is digested when the spheroplast membrane is disrupted by detergent (Fig. 7F). The molecule is therefore intracellular and has not adopted the native structure of the isomerase. As expected from its sensitivity to proteolysis, this polypeptide is not enzymatically active. The finding that a lysine residue a t position 4 does not block secretion of triosephosphate isomerase whereas an arginine residue does, led us toask whether the isomerase might be rendered secretable by simply replacing the arginine residue at position 3 with a lysine. The resulting R3K mutant is, however, not secreted (data not shown). The secretory mechanism can therefore tolerate a single lysine residue at the N terminus of the isomerase, but either an arginine or 2 lysine residues in this region is unacceptable.

20088


secreted protein sequences, see Ref. 39). The results of the present work are difficult to reconcile with the initial events postulatedeither by the signal hypothesis (1,2) or by the membrane trigger hypothesis (4). Neither model envisions the obligate insertion of the early part of the mature protein into a nonpolar environment, nor is there another basis in these theoriesfor any discriminationbetween arginine and lysine residues near the N terminus of the mature protein. On the other hand, the direct transfer model (5), the loop model (6), and thehelical hairpin hypothesis (7) are each based on the central tenet that the N terminus of a secreted protein partitions directly into thelipid bilayer. These models each propose that theprincipal function of a signal peptide is to provide the free energy necessary to pull the first residues of the mature protein into the membrane. Indeed, we should predict that proteins having a marked imbalance of charged residues a t their N termini will be unsecretable, because the free energy gained upon insertion of the hydrophobic core of the signal peptide into themembrane will not be sufficient to drive the coordinate insertion of the attached N terminus of the mature protein.We therefore believe that theearly events in themechanism of protein secretion in bacteria occur much as has been postulated by these direct insertion models. Our data do not, however, address the mechanism of subsequent translocation of the protein across the bilayer. Genetic evidence suggests that a proteinaceous apparatus is ultimately involved in this process (40-42). Nevertheless, it seems that one barrier to protein secretion in bacteria derives from the direct partitioning of the N terminus of the preprotein into a nonpolar phase such as the lipid bilayer, and thatone role of the signal peptide is to provide the driving force for this event. Interestingly, in contrastto secreted proteins of prokaryotes, proteins secreted from eukaryotic cells are notnecessarily characterized by having few (or balanced) charges near the N terminus of the mature protein (17, 39). Indeed, several secreted eukaryotic proteins such as rat, bovine, and human serum albumin (43, 44), bovine and human parathyroid hormone (45, 46), human apolipoprotein A-11 (14), and human &-macroglobulin (21) have extremely basic stretches that immediately follow the signal peptide. On the basis of our present results, we predict that these proteins willbe unsecretable in E. coli. In fact, humanP2-macroglobulinis not secreted in E. coli: only if an acidic linker is placed at the beginning of the mature protein is secretion observed (37). This emphasizes a difference between early eventsin the secretory process of prokaryotes and eukaryotes. In summary, we have shown that the residues near the N terminus of a mature protein influence its secretability, and that this influence most likely derives from the requirement that these residues partition into a hydrophobic phase. Charged residues appearparticularly harmful to secretion because of the unfavorable energetic consequences of neutralization, although the formation of ion pairs between side chains of opposite charge may abate this effect (8,12,37,38).

These findings are consistent with those models of protein secretion that postulate the directinsertion of the signal peptide and a portion of the mature protein into the membrane such as the direct transfer model (5), the loop model (6), and thehelical hairpin hypothesis (7). Acknowledgments-We thank Drs. A. Minsky, J. Hermes, D. Wiley, and G. Guidotti for useful discussions. REFERENCES Blobel, G., and Dobberstein, B. (1975) J. Cell Biol. 6 7 , 835-851 Walter, P., Gilmore, R., and Blobel, G. (1984) Cell 3 8 , 5-8 Gilmore, R., and Blobel, G. (1985) Cell 42,497-505 Wickner, W. (1979) Annu. Reu. Biochem. 48.23-45 von Heijne, G., and Blomberg, C. (1979) Eur. J. Biochem. 9 7 , 175-181 Inouye, M., and Hale oua, S (1980) CRC Crit. Reu. Biochem. 7,339-371 Engelman, D. M., a n f Steitz; T.A. (1981) Cell 23,411-422 Summers, R. G., and Knowles, J. R. (1989) J. Biol. Chem. 2 6 4 , 2007420081 9. Abrahmsbn, L., Moks, T., Nilsson, B., Hellman, U., and Uhlbn, M. (1985) EMBO J. 4,3901-3906 10. Liss, L. R., Johnson, B. L., and Oliver, D. B. (1985) J. Bacteriol. 164,9251. 2. 3. 4. 5. 6. 7. 8.

92A

11. &mussen, B. A,, and Silhavy, T.J. (1987) Genes Deu. 1,185-196 12. Li, P., Beckwith, J., and Inouye, H. (1988) Proc. Natl. Acad. Sci. U. S. A. 86, 7685-7689 13. Yamane, K., and Mizushima, S. (1988) J. Eiol. Chem. 263,19690-19696 14. Folz, R. J., and Gordon, J. I. (1987) J. Biol. Chem. 262,17221-17230 15. Andrews. D. W.. Perara. E., Lesser.. C... and Linaappa,. V. R. (1988) J. Biol. Chem.'263,15791-15798 16. " Randall. and Hardv. S. (1986) ~L. .~L.. ~ ~ S. J.~ ~ Cell 46. ~ 921-928 ~ , 17. von Heijne, G: (1986) j.Mol. Biol. 192,'287-290' 18. Hanahan, D. (1983) J. Mol. Biol. 166,557-580 19. Straus. D.. and Gilbert., W. (1985) Proc. Natl. Acad. Sci. U. S. A. 82.2014. 2018 ' 20. Messln J (1983) MethodsEnzymol. 101,20-78 21. Suggs, V., Wallace, B., Hirose, T., Kawashima, E. H., and Itakura, K. (1981) Proc. Natl. Acad. Sci. U. S. A. 78,6613-6617 22. Maniatis, T.,Fritsch, E.F., and Sambrook, J. (1982) Molecular C h i Laboratory Manual, Cold Spring Harbor Laboratory, Cold Spring Ha%; I

"

~

g.

'hTV I I .

23. de la Mare, S., Coulson, A. F. W., Knowles, J. R., Priddle, J. D., and Offord, R. E. (1972) Biochem. J. 129,321-331 24. Laemmli, U. K. (1970) Nature 227,680-685 25. Plaut B. and Knowles J. R. (1972) Biochem. J. 129,311-320 26. M a r c h i o k , M., and Gilbert, W. (1986) Cell 4 6 , 133-141 27. Swinkels, B. W., Gibson, W. C., Osinga, K. A,, Kramer, R., Veeneman, G. H., van Boom, J. H., and Borst, P. (1986) EMBO J. 6, 1291-1298 28. Banner, D. W., Bloomer, A. C., Petsko, G. A., Phillips, D. C., Pogson, C. I., Wilson, I. A., Corran, P. H., Furth, A. J., Milman, J. D., Offord, R. E., Priddle, J. D., and Waley, S. G. (1975) Nature 2 6 6 , 609-614 29. Alber T.,Banner, D. W., Bloomer, A. C., Petsko, G. A,, Philli s, D., Rivers, P. b., and Wilson, I. A. (1981) Phil. Trans. R. SOC.Lomf B Biol. Sei. 2 9 3 , 159-171 30. Minsky, A,, Summers, R. G., and Knowles, J. R. (1986) Proc. Natl. Acad. srLr I s ~. A. 8x4180-4184 ", 31. Ito, K., and Beckwith, J. (1981) Cell 26,143-150 32. Koshland, D., and Botstein, D. (1982) Cell 30,893-902 33. Hengge, R., and Boos, W. (1985) J. Bacteriol. 162,972-978 34. Wickner, W. (1988) Biochemistry 27,1081-1086 35. Davson, H., and Danielli, J. F. (1970) The Permeability of Natural Membranes, Hafner Publishing Co., Inc., New York 36. Thom, J. R.,and Randall, L. L. (1988) J . Bacteriol. 170,5654-5661 37. Parker, K. C., and Wiley, D. C. (1989) Gene (Amst.),In press 38. Summers, R. G. (1988) Pb.D. Thesis, Harvard University 39. Watson, M. E. E. (1984) Nucleic Acds Res. 12,,5145-5164 40. Oliver, D. B., and Beckwith, J. (1982) J. Bacterrol. 160,686-691 41. Akiyama, Y., and Ito, K. (1985) EMBO J . 4,3351-3356 42. Gardel, C., Benson, S., Hunt, J., Michaelis, S., and Beckwith, J. (1987) J. Bacteriol. 169,1286-1290 43. MacGillivray, R. T.A,, Chung, D. W., and Davie, E. W. (1979) Eur. J . Biochem. 98,477-485 44. DuEaiczyk, A., Law, S. W., and Dennison, 0.E. (1982) Proc. Natl. Acad. Sci. U. S. A. 79.71-75 45. Habener, J. F., Rosenblatt, M., Kemper, B., Kronenberg, H. M., Rich, A., and Potts, J. T. (1978) Proc. Natl. Acad. Sci. U. S. A: 76, 2616-2620 46. Hendy, G. N., Kronenberg, H. M., Potts, J. T., and Rlch, A. (1981) Proc. Natl. Acad. Sci. U. S. A. 78,7365-7369 ~~~