Vol. 169, No. 11 JOURNAL OF BACTERIOLOGY, Nov. 1987, P. 5140-5151 0021-9193/87/115140-12$02.00/0 Copyright X 1987, American Society for Microbiology High-Level Expression of a Proteolytically Sensitive Diphtheria Toxin Fragment in Escherichia coli WILLIAM R. BISHAI,l 2* RINO RAPPUOLI,3 AND JOHN R. MURPHY1 Biomolecular Medicine Section, Evans Department of Clinical Research and Department of Medicine, University Hospital, Boston University Medical Center, Boston, Massachusetts 02118'; Department of Microbiology and Molecular Genetics, Harvard Medical School, Boston, Massachusetts 021152; and Sclavo Research Center, 53100 Siena, Italy3 Received 1 May 1987/Accepted 25 August 1987 ABM508 is a recombinant fusion protein consisting of the N-terminal 485 amino acids of diphtheria toxin joined to a-melanocyte-stimulating hormone. When expressed in Escherichia coli under the control of the tox promoter and signal sequence, ABM508 is severely degraded. When overexpressed from a thermoinducible lambda PR promoter fusion, ABM508 is largely insoluble. We compared the expression of ABM508 (501 amino acids) to a full-length mutant form of the toxin (CRM197; 535 amino acids) and found that CRM197 showed minimal proteolysis. Thus, the removal of the C-terminal 50 amino acids of the toxin destabilizes the protein, making it a target for proteases. Proteolysis of ABM508 could be reduced by removal of the tox signal sequence (thereby directing the protein to the cytoplasm) and growth in Ion and htpR mutant strains of E. coli. We also showed that the solubility of tox gene products expressed in E. coli was directly related to the growth temperature of the culture. Thus, a fragment A fusion protein (223 amino acids), ABM508, and CRM197 were found in soluble extracts when expressed at 30°C but could not be released by the same procedures after growth at 42°C. On the basis of these observations, we fused the coding sequences for mature ABM508 to the trc promoter (inducible at 30°C by isopropyl-(B-D-thiogalactoside) and expressed this construct in a lon htpR strain of E. coli. This plasmid made 10 mg of soluble tox protein per liter of culture (7.7% of the total cell protein) or 14 times more than our previous maximal level. Extracts from lon htpR cells harboring this plasmid had high levels of ADP-ribosyltransferase activity, and although proteolysis still occurred, the major tox product corresponded to full-length ABM508. Diphtheria toxin (for a review, see reference 42) is a 535-amino-acid polypeptide (27, 32, 45) that is secreted in amounts as large as 500 mg/liter by toxinogenic strains of Corynebacterium diphtheriae grown under appropriate con- ditions (46). The toxin can be resolved into two fragments by mild proteolysis and disulfide reduction (17, 21). Fragment A (Fig. 1), the N-terminal 193 amino acids, is an ADP- ribosylating enzyme which catalytically inactivates eucary- otic elongation factor 2 (12, 26). Fragment B, the C-terminal 342 amino acids, performs two functions: translocation of fragment A across eucaryotic cell membranes (16, 34) and binding the toxin to a surface receptor on sensitive cells (31). The binding domain is believed to be localized to the C terminus of fragment B (55). Because diphtheria toxin can kill animals and humans at a dose as low as 100 ng/kg (20), there has been great interest in targeting the toxicity of the protein to particular types of cells. This has been attempted at the level of protein chem- istry by coupling whole toxin or toxin fragments to antibod- ies (23, 52), hormones (3), or lectins (22). More recently, attention has turned to making genetically engineered tar- geted toxins. This approach has the advantages that the junction between the toxin and the targeting ligand can be specified and that larger fragments of the toxin than cross- reacting material 45 (CRM45; a 386-amino-acid nonsense mutant of the toxin [19, 55] often used for making chemically linked hybrid toxins) can be engineered. We have reported the genetic construction and expression in Escherichia coli of two diphtheria toxin-hormone fusion proteins which contain the N-terminal 485 amino acids of the toxin fused to either a-melanocyte-stimulating hormone or * Corresponding author. interleukin-2 (39; D. Williams, K. Parker, W. Bishai, M. Borowski, F. Genbauffe, T. B. Strom, and J. R. Murphy, submitted for publication). The hybrid toxins are called ABM508 and ABI508, respectively, and both have been shown to have toxicity in vitro for cells that bear receptors for the particular hormone (human melanoma cells and activated T cells, respectively). Characterization of these targeted toxins has been ham- pered by low-level expression (less than 0.5 mg/liter of E. coli culture), proteolytic degradation, and insolubility of the recombinant hybrid toxins (6). In this study we analyzed two new fusions of the ABM508 gene to high-expression promot- ers, and we present a systematic analysis of variables which contribute to proteolysis and insolubility. MATERIALS AND METHODS Strains. The bacterial strains used in this study are listed in Table 1. SY327 is a recA strain and was used for DNA transformations and plasmid maintenance. JM105 and RB791 contain the lacIq mutation and were used for cloning genes under the control of the trc promoter. The SG strains are protease deficient and were used for expression studies. Plasmids. The plasmids used in this study are listed in Table 2. The parent vector for pDT201 is pUC8 (56), for pABM508 and pP197 it is pEMBL8 (14), for pDT1201, pABM1508, and pABM4508 it is pEMBLex3 (15), and for pABM6508 it is pKK233-2 (1; available from Pharmacia, Inc., Piscataway, N.J.). Construction of recombinant plasmids. Plasmid DNA was cut and modified with enzymes purchased from New En- gland BioLabs, Inc., Beverly, Mass., used as recommended by the manufacturer. Plasmids were introduced into E. coli by CaCl2 transformation, isolated by the alkaline lysis pro- 5140 on November 29, 2020 by guest http://jb.asm.org/ Downloaded from
Transcript
Vol. 169, No. 11JOURNAL OF BACTERIOLOGY, Nov. 1987, P. 5140-51510021-9193/87/115140-12$02.00/0Copyright X 1987, American Society for Microbiology
High-Level Expression of a Proteolytically Sensitive DiphtheriaToxin Fragment in Escherichia coli
WILLIAM R. BISHAI,l 2* RINO RAPPUOLI,3 AND JOHN R. MURPHY1Biomolecular Medicine Section, Evans Department of Clinical Research and Department of Medicine, University
Hospital, Boston University Medical Center, Boston, Massachusetts 02118'; Department of Microbiology and MolecularGenetics, Harvard Medical School, Boston, Massachusetts 021152; and Sclavo Research Center, 53100 Siena, Italy3
Received 1 May 1987/Accepted 25 August 1987
ABM508 is a recombinant fusion protein consisting of the N-terminal 485 amino acids of diphtheria toxinjoined to a-melanocyte-stimulating hormone. When expressed in Escherichia coli under the control of the toxpromoter and signal sequence, ABM508 is severely degraded. When overexpressed from a thermoinduciblelambda PR promoter fusion, ABM508 is largely insoluble. We compared the expression of ABM508 (501 aminoacids) to a full-length mutant form of the toxin (CRM197; 535 amino acids) and found that CRM197 showedminimal proteolysis. Thus, the removal of the C-terminal 50 amino acids of the toxin destabilizes the protein,making it a target for proteases. Proteolysis of ABM508 could be reduced by removal of the tox signal sequence(thereby directing the protein to the cytoplasm) and growth in Ion and htpR mutant strains of E. coli. We alsoshowed that the solubility of tox gene products expressed in E. coli was directly related to the growthtemperature of the culture. Thus, a fragment A fusion protein (223 amino acids), ABM508, and CRM197 werefound in soluble extracts when expressed at 30°C but could not be released by the same procedures after growthat 42°C. On the basis of these observations, we fused the coding sequences for mature ABM508 to the trcpromoter (inducible at 30°C by isopropyl-(B-D-thiogalactoside) and expressed this construct in a lon htpR strainof E. coli. This plasmid made 10 mg of soluble tox protein per liter of culture (7.7% of the total cell protein)or 14 times more than our previous maximal level. Extracts from lon htpR cells harboring this plasmid had highlevels of ADP-ribosyltransferase activity, and although proteolysis still occurred, the major tox productcorresponded to full-length ABM508.
Diphtheria toxin (for a review, see reference 42) is a535-amino-acid polypeptide (27, 32, 45) that is secreted inamounts as large as 500 mg/liter by toxinogenic strains ofCorynebacterium diphtheriae grown under appropriate con-ditions (46). The toxin can be resolved into two fragments bymild proteolysis and disulfide reduction (17, 21). Fragment A(Fig. 1), the N-terminal 193 amino acids, is an ADP-ribosylating enzyme which catalytically inactivates eucary-otic elongation factor 2 (12, 26). Fragment B, the C-terminal342 amino acids, performs two functions: translocation offragment A across eucaryotic cell membranes (16, 34) andbinding the toxin to a surface receptor on sensitive cells (31).The binding domain is believed to be localized to the Cterminus of fragment B (55).Because diphtheria toxin can kill animals and humans at a
dose as low as 100 ng/kg (20), there has been great interest intargeting the toxicity of the protein to particular types ofcells. This has been attempted at the level of protein chem-istry by coupling whole toxin or toxin fragments to antibod-ies (23, 52), hormones (3), or lectins (22). More recently,attention has turned to making genetically engineered tar-geted toxins. This approach has the advantages that thejunction between the toxin and the targeting ligand can bespecified and that larger fragments of the toxin than cross-reacting material 45 (CRM45; a 386-amino-acid nonsensemutant of the toxin [19, 55] often used for making chemicallylinked hybrid toxins) can be engineered.We have reported the genetic construction and expression
in Escherichia coli of two diphtheria toxin-hormone fusionproteins which contain the N-terminal 485 amino acids of thetoxin fused to either a-melanocyte-stimulating hormone or
* Corresponding author.
interleukin-2 (39; D. Williams, K. Parker, W. Bishai, M.Borowski, F. Genbauffe, T. B. Strom, and J. R. Murphy,submitted for publication). The hybrid toxins are calledABM508 and ABI508, respectively, and both have beenshown to have toxicity in vitro for cells that bear receptorsfor the particular hormone (human melanoma cells andactivated T cells, respectively).
Characterization of these targeted toxins has been ham-pered by low-level expression (less than 0.5 mg/liter of E.coli culture), proteolytic degradation, and insolubility of therecombinant hybrid toxins (6). In this study we analyzed twonew fusions of the ABM508 gene to high-expression promot-ers, and we present a systematic analysis of variables whichcontribute to proteolysis and insolubility.
MATERIALS AND METHODSStrains. The bacterial strains used in this study are listed in
Table 1. SY327 is a recA strain and was used for DNAtransformations and plasmid maintenance. JM105 andRB791 contain the lacIq mutation and were used for cloninggenes under the control of the trc promoter. The SG strainsare protease deficient and were used for expression studies.
Plasmids. The plasmids used in this study are listed inTable 2. The parent vector for pDT201 is pUC8 (56), forpABM508 and pP197 it is pEMBL8 (14), for pDT1201,pABM1508, and pABM4508 it is pEMBLex3 (15), and forpABM6508 it is pKK233-2 (1; available from Pharmacia,Inc., Piscataway, N.J.).
Construction of recombinant plasmids. Plasmid DNA wascut and modified with enzymes purchased from New En-gland BioLabs, Inc., Beverly, Mass., used as recommendedby the manufacturer. Plasmids were introduced into E. coliby CaCl2 transformation, isolated by the alkaline lysis pro-
JM105 thi rpsL endA sbcB15 hsdR4 59W(lac-proAB) [F' traD36proAB+ lacIqZA&M15]
RB791 W3110, lacIqL8 8SG931 F- lac(Am) trp(Am) S. Goff and A. Goldberg
pho(Am) supC(Ts)SG932 SG931, Alon-100 S. Goff and A. GoldbergSG934 SG931, htpR165(Am) Ts S. Goff and A. GoldbergSG936 SG931, AIon-100 S. Goff and A. Goldberg
htpRJ65(Am) Ts
cedure, and purified by CsCl density gradient centrifuga-tion (36). The 10-base oligonucleotide used to constructpABM6508 was made with a solid-phase, automated DNAsynthesizer (model 380A; Applied Biosystems, Foster City,Calif.), using phosphoramidite nucleotides and controlled-pore glass columns (Applied Biosystems). Before use it waspurified by polyacrylamide gel electrophoresis as describedpreviously (39).
Cross-reacting material 197 (CRM197) is a missense mu-tant of diphtheria toxin that is devoid of ADP-ribosyltrans-ferase (ADPRT) activity and is thus nontoxic (55). Its genecarries a single base substitution, resulting in the incorpora-tion of glutamic acid instead of glycine at residue 52 (19).p3197 contains a 1,875-base-pair HindIII-EcoRI restrictionfragment from corynebacteriophage 197 that carries the toxpromoter and signal sequence and the full-length CRM197gene. Work with E. coli harboring this plasmid was per-formed at the Sclavo Research Center, Siena, Italy.pABM4508 carries the ABM508 gene under the control of
the lambda PR promoter, but it lacks a functional signalsequence. It carries its own c1857 thermolabile lambdarepressor gene. It was constructed by digesting pABM1508with BamHI and ApaI, each of which cut this plasmiduniquely within the coding sequences for the signal se-quence. The double digestion deleted a 56-base-pair restric-tion fragment which encodes amino acids -26 to -7 of thepABM1508 signal sequence. The sticky ends were bluntedwith tnung bean nuclease and joined with T4 DNA ligase toproduce the pABM4508 in-frame fusion. This plasmnid codesfor the formylmethionine (fMet) of the PR promoter and thecro gene fused to amino acid -6 of the tox signal sequence(Fig. 1). The deletion inactivates the tox signal sequence andconcomitantly adds six amino acids (assuming removal ofthe fMet) to the amino terminus of the mature toxin.
pABM6508 carries the ABM508 gene under the control ofthe trc promoter; it lacks the entire tox signal sequence, withfMet fused to Gly1 of the mature toxin (Fig. 1). It wasconstructed as shown in Fig. 2. This construct has a ribo-some-binding site-ATG spacing of 8 base pairs, close to thewild type lacZ spacing of 7 base pairs.Growth of bacteria. For cloning experiments, bacteria
carrying recombinant plasmids were grown in Luria-Bertanimedium (36) supplemented with 100 ,ug of ampicillin per ml.For expression studies, strains from the SG series weregrown in M9 minimal medium (36) containing 0.2% glucose,0.2% Casamino Acids (Difco Laboratories, Detroit, Mich.),100 jig of ampicillin per ml, and 20 ,ug of L-tryptophan perml.For Western blots (immunoblots), constructs under the
control of Ptox were grown at 30°C to an A550 of 2.0 beforeharvesting. Ptrc constructs were grown at 30°C to an A550 of0.25, induced with 1 mM isopropyl-o-D-thiogalactoside(IPTG; Calbiocherm-Behring, La Jolla, Calif.), and harvestedabout 4 h later when the A550 reached 2.0. Constructs underthe control of PR were grown to an A550 of 0.5 at 30°C andinduced for 2.5 h at 42°C before harvesting.For radioimniunoassay (RIA) and ADPRT assay, 500-ml
cultures were grown at 30°C to an A550 of 0.5, at which timethe cultures were induced by the addition of IPTG or by atemperature upshift to 42°C. Samples (100 ml) were takenjust before induction and at 30, 90, and 180 min afterinduction.Western blots. Cells grown as described above were har-
vested by centrifugation and suspended in 0.5 ml of TEPbuffer (30 mM Tris hydrochloride [pH 8.0], 5 mM EDTA, 0.5mM phenylmethylsulfonyl fluoride) per 10 ml of culture perA550 unit. The cells were either lysed directly by the additionof an equal volume of 2x sodium dodecyl sulfate (SDS)reducing loading buffer (6) or fractionated by sonication onice with four 15-s pulses at 40 W (model 200 Sonifier;Branson Sonic Power Co., Danbury, Conn.) and centrifuga-tion at 16,000 x g for 10 min at 4°C. The resulting superna-tant was designated the soluble fraction and was mixed withan equal volume of 2x reducing loading buffer; the pellet,designated the insoluble fraction, was resuspended in 1volume of TEP and dissolved by the addition of an equalvolume of 2 x reducing loading buffer. Proteins wereseparated on 10% SDS-polyacrylamide gels, transferred tonitrocellulose, and stained with horse anti-diphtheria toxinantibodies (Connaught Laboratories, Swiftwater, Pa.) aspreviously described (6). With these antibodies, extracts ofE. coli carrying only the parent vectors for our tox plasrnidsshowed no appreciable immunoreactivity on Western blots.The molecular weight standards for the blots contained
TABLE 2. Plasmids used in this studyNo. Molecular Maximal Maximal
Plasmid Promoter Signal of Transcription M ar yield sp act (% Referencesequence amino terminator mDass (iig/liter of totalacidSa \ a, of culture) protein)
pDT201 Ptox + 223 - 24,576 35pDT1201 PR + 223 - 24,576 6pABM508 Ptox + 501 - 54,561 0.4 0.4 39pABM1508 PR + 501 - 54,561 0.7 1.2 38pABM4508 PR - 507 - 55,122 0.2 0.2 This studypABM6508 Ptrc - 501 + 54,561 10.0 7.7 This studyP1,197 Ptox + 535 +c 58,413 19a Assuming cleavage of the signal sequence (periplasmic constructs) or removal of the initial fMet (cytoplasmic constructs).b Deduced mass of mature forms.c A possible E. coli terminator occurs 10 bases after the tox gene UGA. This potential stem-loop structure is present in p0197.
6000 AG0MAAOACAG ATOCCN CSERIES 'frC fLust Sly ala
-I' I 2
4000 TAAGCACCTT=T ATC CCA OCT TCA CCC CAT CCC OCTSERIES fet pro pro ser a1 hOs ala gly ala
-7' -6 -5 -4 -3 -2 -1 1 2
TAAMCAOCTTCT ATO CAT CCC AGC AGA AAA CTG TET CCC TCA ATC TTA ATA CCC CCC CTA CTC CCC ATA COO GCC CCA CCT TCA CCC CAT GCA CCC OCTfuet asp pro sr arg lye lau ph. ala mer ile leu 11e gly ala leu leu gly lle gly a1a pro pro ser ala his ala gly ala-27'-26 -25'-24 -23 -22 -21 -20 -19 -18 -17 -16 -15 -14 -13 -12 -11 -10 -9 -8 -7 -6 -5 -4 -3 -2 -1 1 2
ACGCGATACGTT GTO AOC ACA AAA CTG mTT CGC TCA ATC TTA ATA COO CCC CTA CTG GGC ATA COO CCC CCA CCT TCA CCC CAT GCA CCC OCTtOX fuet ser arg lys lau ph* ala ar ila lou le gly ala leu leu gly Ole Sly als pro pro ser ala his ala gly ala
MT 0CT GTG CAT CC OCTG C CA ACT TAT ACT ATG GAO CAC TTC AGG TOO GCA AAC CCA GTA TAOangly val hi *la *a &ll lore tyrser t glu hBs ph arg trp Sly lys pro val b481 482 483 484 485 486'487'488'489'490'491'492'493'494'495'496'497'498'499'500'501'
M-508C-TERMINUS
TCT GCA 00 MCT COT GTC AGO CA TCC CCG GC T COT MT C OT CAT TOT TC CTG TOT CMA ATC CCC TC CM TTC CAC CA ACA TAC 0C CCC CM GCA TAcys &Is gly asa arg val arg arg ear pro gly Ile arg asn hls Sly hls er ey ph leu cy glu Ole val Ole ars *er lIn pbe his thr thr tyr glu pro glu ala oc6186 187 188 189 190 191 192 193 194 195'196'197'198'199'200'201'202'203'204'205'206'207'208'209'210'111'212'213'214'215'216'217'218'219'220'221'222'223'
DT-201C-TERMINUS
FIG. 1. Diagram of tox gene showing the relative positions of the 25-amino-acid signal sequence (SS), fragment A (193 amino acids) (A),fragment B (342 amino acids) (B), and the two disulfide loops. N-terminal modifications used in this study are shown above the gene. Aminoacids composing the signal sequence were given negative residue numbers, and foreign amino acids have apostrophes after their residuenumbers. The sequences of the two C-terminal modifications used in this study and the sequence of the CRM197 mutation are shown belowthe gene.
unnicked diphtheria toxin (List Biological Laboratories,Campbell, Calif.) mixed with nicked diphtheria toxin(Calbiochem-Behring) in the presence of ,B-mercaptoethanol.RIA. Portions (100 ml) of cultures grown as described
above were harvested by centrifugation and suspended in 10ml of 30 mM Tris hydrochloride-5 mM EDTA (pH 8.0). Thecells were lysed by sonication as described above, and aftercentrifugation (10 min at 7,500 x g at 4°C), the solublefraction was stored in aliquots at -70°C. To control forvarying degrees of proteolytic degradation, we used a mod-ification of the diphtheria toxin cross-reacting material RIAdeveloped by K. Parker and K. Campbell (Seragen, Inc.,Hopkinton, Mass.) by substituting fragment A-specificpolyclonal antibodies for the whole antitoxin serum. Withthis change, fragment A and nicked diphtheria toxin arerecognized in the assay with equal sensitivity (50% inhibi-tory concentration, 5 x 10-9 M). Because intact CRM45 isused as the tracer, intact CRM45 is recognized about threetimes more avidly (50% inhibitory concentration, 1.75 x10-9 M), with a lower limit of detection for intact CRM45 of0.75 x 10-9 M. In contrast, the same assay using unfraction-ated antitoxin serum is 10 times more sensitive for nickedwhole toxin and intact CRM45 than for fragment A. Purifiedanti-fragment A antibodies were prepared by passing hyper-immune horse antitoxin serum (Sclavo) over a fragment Aaffinity column. The purified antibody preparation, whichfails to recognize fragment B on Western blots, had an A280of 0.212 and was diluted 1:1,000 for use in the RIA.For the assay, 75 ,ul of sample diluted in assay buffer (10
mM sodium phosphate [pH 7.2], 0.9% NaCl, 0.1% bovineserum albumin) was mixed with 75 pLl of '25I-labeled CRM45(3 x 105 cpm/ml) in assay buffer and 75 ,ul of purified horseanti-fragment A (1:1,000) containing horse immunoglobulin
G (0.1 mg/ml; Cooper Biomedical, West Chester, Pa.) inassay buffer. After a 2-h incubation at 37°C, the antibodieswere precipitated by the addition of 750 p.1 of goat anti-horseimmunoglobulin G (0.3 mg of specific antibody per ml;Pel-Freez Biologicals, Rogers, Ark.) in 2.5% polyethyleneglycol 6000-8000 (Sigma Chemical Co., St. Louis, Mo.-10mM sodium phosphate (pH 7.2)-0.9% NaCl. After centrifu-gation at 16,000 x g for 5 min, the supernatants wereaspirated off and the radioactivity in the pellets was countedon a Beckman Gamma 5500 instrument (Beckman Instru-ments, Inc., Fullerton, Calif.). Under these conditions non-specific binding is typically 10 to 20% of total binding.Extracts from cells containing the parent vectors for ourconstructs produced no detectable fragment A immunoreac-tivity by this assay. Purified intact CRM45, prepared asdescribed previously (43, 44), was used as the standard.CRM45 and 125I-labeled CRM45 were kindly provided by K.Parker and K. Campbell (Seragen).ADPRT assay and protein determination. The samples
assayed by RIA were also tested for relative ADPRT activityby a method previously described (39). Protein extracts fromE. coli carrying the parent vectors for our tox plasmids weredevoid of ADPRT activity. Measurements of total protein inthe sonicated samples were made by the method of Bradford(7) and used in conjunction with the RIA determinations tocalculate specific activity.
RESULTS
Immunoblot analysis of our previous efforts to expressperiplasm-directed ABM508 from either the tox promoter(pABM508) or the lambda PR promoter (pABM1508) showedmany protein bands migrating between fragment A and
p C- AT GZ GCT GAT GAT PM?trc 0pl. 9 UICCCT ATC ACC ATC ATT ACC 1acz
signal*4Be mO TCAAGCC TC GCA GCC GCC GAT CAT toC3 C2A 1 2 3 4
FIG. 2. Construction of pABM6508. A 2.05-kilobase HaeII fragment carrying the ABM508 coding sequences was purified from pABM508and ligated to a self-annealing 10-base NcoI-HaeII linker oligonucleotide. After addition of the linker, the modified fragment was cut withNcoI and HindIlI and inserted into pKK233-2 as shown. The N-terminal sequence of pABM6508 is aligned with those of lacZ and tox forcomparison. The segments labeled 5S and rrnBTjT2 denote the 3' end of the gene encoding the 5S ribosomal subunit and the transcriptiontermination signals for the rrnB operon, respectively.
full-length ABM508 (6, 39). We attributed this to proteolyticdegradation rather than incomplete transcription or transla-tion because fragment A (a common proteolytic product ofwhole toxin) was a major product in our extracts. We alsofound that the recombinant protein was largely insoluble,since the spheroplast fraction after lysozyme-EDTA treat-ment of the harvested bacteria or the membrane fractionafter French press lysis contained most of the immunoreac-tive ABM508 protein. On the basis of these observations, webegan a series of experiments to identify variables thatcontributed to the proteolysis and insolubility.
Effect offragment length on proteolysis. In a previous studywith periplasm-directed proteins expressed from the lambdaPR promoter (the 1000 series; 6), we found that a fragment Afusion protein (pDT1201, 223 amino acids) was expressed ata high yield and was not degraded. In contrast, longer fusionproteins which included portions of fragment B (pABM1313,306 amino acids; pABM1402, 398 amino acids; and pABM1508, 501 amino acids) were heavily degraded and producedmultiple faint bands on Western blots. It appeared thatfragment A was stable to proteolysis within E. coli, but whenincomplete portions of fragment B were added, the moleculebecame a target for proteolytic attack. Thus, we decided totest whether a full-length diphtheria toxin-related poly-peptide would also be the target of proteolysis in E. coli.For these experiments we used the gene for CRM197, an
inactive missense mutant of diphtheria toxin that differs fromthe wild type by a single amino acid substitution (Fig. 1).
Although CRM197 is known to be less stable than wholetoxin (58), we found that when expressed from the toxpromoter with an intact signal sequence, p,197 producedlarge amounts of full-length protein (535 amino acids) in astrain with a wild-type protease background (Fig. 3, panelp,197, lane A). In contrast, all other versions of ABM508(501 amino acids) showed substantial proteolytic degrada-tion. It is noteworthy that in the wild-type strain, pABM508(Ptox driven, periplasmic) failed to direct the expression ofany full-length, 54-kilodalton (kDa) fusion protein (Fig. 3,panel pABM508, lane A), whereas p,B197 (also PiOx drivenand periplasmic) directed the expression of mostly full-length molecules. This shows that the deletion of 50 aminoacids and the addition of 16 foreign ones to the C terminus ofdiphtheria toxin changes the structure of the protein in a waythat makes it highly susceptible to E. coli proteases.
Deletion of the tox signal sequence. Our previous work withthe PR-driven, periplasm-directed fusion proteins (the 1000series) showed that the continued growth of the host bacteriawas inhibited when the PR promoter was induced by temper-ature upshift to 42°C. Because the effect was not seen withthe analogous low-level-expression, p,,x-driven, periplasm-directed constructs (the 0000 series) or with heat induction ofthe parent vector alone, we concluded that the lethalityresulted specifically from the high-level expression andattempted secretion of the hybrid toxin. Similar lethality hasbeen reported when periplasm-directed hybrid proteins suchas malE-lacZ fusions are overexpressed (5). The lethality is
FIG. 3. Westem blot analysis showing the effect of fragment length and host strain on expression. The five constructs indicated above thepanels were grown in four different strains of E. coli under the conditions shown and induced as described in the text. After centrifugationthe bacterial pellets were resuspended in volumes adjusted to give equivalent optical densities, lysed with SDS, and subjected to gelelectrophoresis under reducing conditions. The protein in each lane in the three left panels came from 0.3 ml of culture at an A550 of 1.0. Toavoid overloading in the panels for pABM6508 and pP3197, the lysed extracts were diluted fourfold relative to those in the other panels. Lanes:A, strain SG931 (wild-type protease activity); B, strain SG932 (lon); C, strain SG934 (htpR); D, strain SG936 (Ion htpR); E, diphtheria toxinstandards (whole toxin, 58 kDa; fragment B, 37 kDa; and fragment A, 21 kDa).
believed to result from jamming of the secretion apparatusby abnormally folded hybrid proteins.To test whether localization of ABM508 in the E. coli
cytoplasm would relieve the overexpression lethality asso-ciated with pABM1508 and perhaps result in higher yields offull-length fusion protein, we deleted the coding sequencesfor 18 amino acids of the tox signal sequence to producepABM4508, a PR-driven, cytoplasm-directed form of thehybrid toxin. We also wanted to examine the effect ofcytoplasmic localization on proteolytic degradation of thefusion protein. Although the cytoplasm is reputed to havegreater protease activity than the periplasm (50), most ge-netically engineered proteins have been directed to thecytoplasm, and several cytoplasmic protease mutants of E.coli have been developed to deal with the degradationproblems in this compartment.Our results show that deletion of the tox signal sequence
did indeed relieve the overexpression lethality that occursupon temperature induction of the periplasm-directed pAB3M1508 (see Fig. 7B and C). Shortly after the jump to 42°C, theculture harboring pABM1508 (periplasm-directed) stoppedgrowing, whereas the higher temperature actually enhancedthe growth rate of bacteria carrying pABM4508 (cytoplasmdirected). Despite the poor growth rate of the pABM1508culture, it did produce some full-length hybrid toxin in astrain with wild-type protease activity (Fig. 3, panelpABM1508, lane A. However, the cytoplasm-directed con-struct, pABM4508, which permits growth to higher densitiesafter temperature induction, failed to produce detectablelevels of full-length hybrid toxin in the same strain (Fig. 3,panel pABM4508, lane A). This observation is consistentwith the reported higher levels of protease activity in the E.coli cytoplasm.
Effect of host strain on proteolysis. Next we examined theeffect of lon and htpR mutant strains on the production ofhybrid toxin. The E. coli lon gene codes for protease La, a376-kDa, ATP-dependent, cytoplasmic protease that is oneof the E. coli heat shock proteins (25). The La protease hasspecificity for unfolded, abnormal proteins, and lon mutants
have been shown to have a reduced ability to degrade suchproteins. htpR is a heat shock regulatory gene which codesfor &2, an alternate sigma factor for RNA polymerase (29).cr3 iS made in increased amounts after exposure of cells totemperatures greater than 37°C, and it is believed to redirectRNA polymerase to the 17 or more heat shock genes in E.coli and increase their transcription (40). Because severalheat shock genes are known to encode proteases (e.g., theIon gene) or protein processing enzymes (4), htpR mutantshave reduced proteolytic activity, and the reduction isdetectable even at 30°C (24). htpR is an essential gene; theamber mutant we used is suppressed at low temperature bya supC(Ts) allele.Using Western blots, we assessed the levels of degrada-
tion of the ABM508 protein in four E. coli strains that areisogenic except at their Ion and htpR loci. The protease-deficient strains did little to improve the expression of theperiplasm-directed constructs, pABM508 (p.OX driven) andpABM1508 (PR driven) (Fig. 3). Moreover, the mutants didlittle to further stabilize the pP197 gene product (P,OX drivenand periplasmic), which showed little degradation in the firstplace. The mutants had more noticeable effects on thecytoplasm-directed pABM4508 construct (PR driven). Boththe lon strain (Fig. 3, panel pABM4508, lane B) and the htpRstrain (lane C) showed roughly equivalent accumulation ofthe full-length species. The effect seems to be additive, sirpcethe lon htpR double mutant (lane D) made even morefull-length hybrid toxin.These results show that directing ABM508 to the cyto-
plasm by deleting the tox signal sequence (as in pABM4508)leads to complete proteolysis in a protease-proficient strain.However, expression in lon and htpR mutants protected thecytoplasmic hybrid toxin and allowed the full-length speciesto accumulate. On the basis of this observation and the factthat bacteria expressing the cytoplasm-directed hybrid toxindid not display overexpression lethality and could thus begrown to high density during induction, we chose to pursuecytoplasmic accumulation in subsequent constructions.
Effect of temperature on proteolysis. High temperature is
OVEREXPRESSION OF A DIPHTHERIA TOXIN FRAGMENT 5145
known to induce protease activity in E. coli (24), and thehtpR gene serves as a high-temperature positive regulatoryelement for at least part of this inducible activity. Todetermine whether our hybrid toxin is a substrate for thistemperature-inducible activity, we grew some of our con-structs in a lon host (SG932) at three different tempera-tures. Growth at 30°C allowed for the greatest accumulationof full-length fusion protein with both pABM508 (piox driven,periplasmic) and pP197 (ptox driven, periplasmic) (Fig. 4). Inthese constructs degradation appears to increase with in-creasing temperature, suggesting that there may be a tem-perature-inducible protease activity which acts on themeither in the periplasm or while they are in transit to theperiplasm. It is noteworthy that cytoplasmic fusion proteinexpressed from pABM6508 (described below) did not showincreased degradation at higher growth temperatures andthus does not seem susceptible to the degradation acting onpABM508 and pP197.
Effect of temperature on solubiity. Attempts to purifyfull-length hybrid toxin from SG936 carrying pABM4508 (PRdriven, cytoplasmic) were unsuccessful because 90% of theimmunoreactive protein remained associated with the cellu-lar debris pellet after sonication or French press lysis andcentrifugation. At first we believed this resulted from thecytoplasmic accumulation of this protein. However, weobserved the same results with sonic extracts or lysozyme-EDTA-treated cells carrying pABM1508 (PR driven, peri-plasmic). This observation was a surprise since we (35) andothers (54) had shown that fragment A alone is expressedand secreted to the periplasmic space in a soluble form.
Thinking that this insolubility resulted from the presenceof much of fragment B in ABM508, we compared the
solubilities of fragment A (pDT201) and CRM197 (pP197)expressed under the same conditions as the hybrid toxinencoded by pABM1508 and pABM4508. We found thatneither the host strain nor the length of the tox fragmentinfluenced the solubility of the expressed proteins but thatincubation temperature did. Both our fragment A fusionprotein (24.5 kDa) and CRM197 (58 kDa) were not com-pletely released into the soluble phase when expressed at42°C but were almost entirely soluble when expressed at30°C (Fig. 5). Because both of our PR-driven constructscould only be expressed by temperature induction at 42°C,this finding explained our inability to isolate soluble toxprotein by using the PR system.
Analysis of pABM6508. Our findings described abovemade it clear that we needed a different high-expressionpromoter that could be induced at low temperature. Webelieved that the proteolysis problem could be overcomewith the Ion htpR double-mutant strain and that we couldtake full advantage of these mutations by localizing the newhybrid toxin to the cytoplasm. To test this hypothesis, weassembled pABM6508 (Fig. 2), in which the ABM508 gene isdriven by the trc promoter, a variant of the tac promoter (13)in which the spacing between the -35 and -10 consensussequences has been changed from 18 to 17 base pairs (10).The tac and trc promoters have been used by others toexpress recombinant proteins in E. coli at levels as high as25% of cell protein (2, 49). In addition to having a strong E.coli promoter, pABM6508 also carries two rho-independenttranscriptional terminators from the rrnB operon (9) at theend of the ABM508 gene.When introduced into the lon htpR strain, pABM6508
could be induced with IPTG to express much higher levels of
FIG. 4. Western blot analysis showing the effect of temperature on expression. The plasmids indicated were grown in strain SG932 (lon)to an A550 of 0.5 before the temperature was changed for an additional 2.5 h of incubation. For pABM6508, IPTG was added to 1 mM at thetime of the temperature change. Bacterial pellets were resuspended to give the same optical density, lysed with SDS, and subjected to gelelectrophoresis under reducing conditions. The protein in each lane came from 0.3 ml of culture at an A550 of 1.0. Lanes: A, cultures kept at30°C; B, cultures shifted to 37°C; C, cultures shifted to 42°C; S, diphtheria toxin standards. The numbers in the margin are molecular massesin kilodaltons.
FIG. 5. Western blot analysis showing the effect of temperature on solubility. The plasmids indicated were grown in strain SG932 (lon) at30°C to an A550 of 0.5, at which time the cultures were split and placed at both 30 and 42°C. At this stage the pABM6508 cultures were inducedwith 1 mM IPTG. The cells were harvested 2.5 h later, resuspended to give equivalent optical densities, and fractionated by sonication andcentrifugation as described in the text. Samples were run under reducing conditions; each lane contained the fractionated protein from 0.3ml of culture at an A550 of 1.0. Lanes: S, soluble fraction; I, insoluble fraction; St, diphtheria toxin standards. The numbers in the margin aremolecular masses in kilodaltons.
hybrid toxin than those expressed by previous ABM508vectors. This can be seen in the Coomassie blue-stainedSDS-polyacrylamide gel of extracts from cells carrying thedifferent constructs (Fig. 6). pABM6508 showed a proteo-lytic profile similar to that of pABM4508 (Fig. 3), with theprotease-deficient host strains showing greater accumulationof full-length product than did the protease-proficient strain.Surprisingly, growth at high temperature did not reduce theaccumulation of the pABM6508 gene product as significantlyas it did for the periplasmic constructs, pABM508 and p,197(Fig. 4). High temperature did, however, reduce the solubil-ity of the hybrid toxin expressed from pABM6508 (Fig. 5).We estimate that after growth at 30°C about 50% of thefull-length protein can be released into the soluble phase,whereas most of the full-length material is insoluble aftergrowth at 42°C.
Levels of accumulation. To determine how much hybridtoxin is produced by our various hybrid toxin genes, wegrew cultures of SG936 carrying the different ABM508vectors and harvested cells at four different times duringgrowth (Fig. 7). Soluble extracts were assayed for totalprotein, ADPRT activity, and fragment A immunoreactivity.We deliberately measured fragment A immunoreactivity andnot whole-toxin immunoreactivity so that C-terminal degra-dation products that might be present in various proportionsin the extracts would be recognized equally. The results are
shown in Fig. 7; the maximal values from the graphs arelisted in Table 2.As anticipated, pABM508 (Ptox driven, periplasmic) made
hybrid toxin in a constitutive fashion, with the specificactivity remaining constant during exponential growth and
then dropping toward the beginning of the stationary phase(Fig. 7A). Whereas fragment A immunoreactivity dropped atthe end of the exponential phase, ADPRT activity continuedto increase. This suggests that, like whole toxin, ABM508requires nicking within the disulfide loop joining fragments Aand B to be enzymatically active. In older cultures thehybrid toxin product is probably more nicked and thus moreenzymatically active even though synthesis levels havedecreased.Whereas ADPRT activity and fragment A immunoreactiv-
ity were not detectable in pABM1508 and pABM4508 beforeinduction, there was a rapid burst of synthesis by bothconstructs after the jump to 42°C (Fig. 7B and C). Asmentioned above, strain SG936 is temperature sensitive butnevertheless capable of several doublings when heated to42°C during the mid-exponential phase. The presence ofpABM1508, however, caused an overexpression lethalitythat was probably responsible for the precipitous drop inaccumulation levels 90 min after induction. pABM1508 andpABM4508 made 0.7 and 0.2 mg of hybrid toxin per liter,respectively, as assessed by fragment A immunoreactivity,which is close to the maximal level made by pABM508 (0.4mg/liter). This was unexpected because the tox promoter isconsidered a low-to-intermediate-strength E. coli promoter(33), whereas PR is a very strong promoter for E. coli. Webelieve that the pR-driven constructs actually produce higherlevels than these but that the protein goes undetectedbecause of high-temperature insolubility.At 3 h after induction, pABM6508 produced 10 mg of
hybrid toxin per liter of culture or 7.7% of the total cellprotein (Fig. 7D); this is 14 times higher than our previous
OVEREXPRESSION OF A DIPHTHERIA TOXIN FRAGMENT 5147
maximal level of expression. Unlike results with the otherthree constructs, the specific activity curve continued toincrease throughout the growth cycle, suggesting that evenhigher yields might be obtained by prolonged induction.Another factor that might improve the levels of productionby this vector is the introduction of a lacIq superrepressorallele in strain SG936. This strain carries a chromosomalwild-type lacI repressor gene which cannot completely re-press lac (or trc) transcription from multicopy plasmids. Thiswas borne out by the fact that even before induction withIPTG, pABM6508 made 1.0 mg of hybrid toxin per liter orabout 2.7% of the total cell protein. Although the plasmid isstably maintained without a lacIq allele, it may be possible toachieve more rapid growth to a higher preinduction densitywith this stronger repressor. Finally, it should be noted thatabout half of the hybrid toxin produced by pABM6508 at30°C was insoluble; thus, the actual level of synthesis may bewell above 10 mg/liter.A number of factors probably contributed to the high
levels of expression seen in pABM6508; these include thestrength of the trc promoter, the increased solubility gainedby growth at 30°C, the absence of overexpression lethalityresulting from deletion of the tox signal sequence, and thereduction in degradation provided by the lon htpR mutanthost. Two additional factors may also have played a role inthe 14-fold improvement in expression. First, all of ourconstructs except pABM6508 produce mRNA capable offorming a stem-loop structure with a free energy of -7.1 kcal(-29.7 kJ)/mol between the codons for Ser-4 and Asp+3(Fig. 8). Similar mRNA secondary structures around the siteof translation initiation have been reported to decreaseexpression levels in other systems (11, 57), and the removalof this potential stem-loop in pABM6508 may provide some
OD
CO
EHLU
...
94 .:it67..~~~~~~~~z:ss,is.... '..
43 .
30*
advantage. Second, pABM6508 contains two rho-indepen-dent transcription terminators after the ABM508 gene,whereas the other ABM508 constructs lack specific termina-tors. The presence of termination signals could stabilize thehybrid toxin message in pABM6508 and lead to increasedexpression (18).
DISCUSSION
There has been considerable interest in expressing recom-binant diphtheria toxin fragments in E. coli because of theirpotential use in making immunotoxins. Greenfield et al. (28)have successfully overexpressed fragment A to 7% of thetotal cell protein (by RIA) by deleting its signal sequence andfusing it to the lambda PL promoter. Like pABM1508 andpABM4508, this construct requires thermal induction at42°C. An analogous construct containing fragment B se-quences up to amino acid 383 also gives a high level ofsynthesis (5 to 10% of cell protein by SDS-polyacrylamidegel analysis), but its expression is not consistently high andit is not clear what proportion of the material is soluble. Toavoid proteolysis and insolubility, Zettlmeissl et al. (60) havefused the gene for the N-terminal 429 amino acids ofCRM228 (a full-length, inactive cross-reacting material ofdiphtheria toxin which contains four amino acid substitu-tions [32]) to the gene for P-galactosidase. When expressedfrom the lambda PR promoter by heat induction, this fusionprotein is soluble and can be purified in one step with affinitycolumns for ,B-galactosidase. We have successfully purified adiphtheria toxin-related interleukin-2 hybrid protein (ABI508) to near homogeneity from the periplasmic space of E.coli (Williams et al., submitted). Although this hybrid pro-tein is expressed at a low yield from the tox promoter, it issoluble when the cells are grown at low temperature and can
O0 0 0 0° LO LO L
10- 14- (C)
com mcl a- C
Q-0)
0~.
c\JI
fr()C\J
yQ
U)
0
* 58
*37
21
20O
FIG. 6. Coomassie blue-stained SDS-polyacrylamide gel of whole-cell extracts. Cultures of E. coli SG936 (Ion htpR) harboring theplasmids indicated were grown as described in the text. Cells carrying pABM1508 and pABM4508 were heat induced, and those containingpABM6508 and pKK233-2 were induced with 1 mM IPTG. The cells were resuspended to give equivalent optical densities, lysed with SDS,and subjected to electrophoresis under reducing conditions. Each lane contained the whole-cell lysate of 0.3 ml of culture at an A550 of 1.0.pEMBL8 is the parent vector for pABM508, and pKK233-2 is the parent vector for pABM6508. Abbreviations: MWS, molecular weightstandards; DTS, diphtheria toxin standards. The numbers in the margins are molecular masses in kilodaltons.
OVEREXPRESSION OF A DIPHTHERIA TOXIN FRAGMENT 5149
be purified away from the tox degradation products byaffinity chromatography with anti-interleukin-2 antibodies.ABM508 is a hybrid toxin analogous to ABI508; it is
comprised of the first 485 amino acids of diphtheria toxintranslationally fused to the 13-amino-acid peptide hormone,a-melanocyte-stimulating hormone. While it has been shownto have cytotoxicity for human melanoma cells in vitro,further studies of this toxicity have been hampered by acombination of low levels of synthesis, insolubility, andproteolytic breakdown to undesirable cross-reacting materials.
In this study we systematically considered variables thatcontribute to this low yield. We showed that fragments ofdiphtheria toxin intermediate in length between fragment A(193 amino acids) and whole toxin (535 amino acids) aresubstrates for proteolytic activity in E. coli, whereas frag-ment A alone and CRM197 expressed in E. coli are not.Indeed, our data show that the removal of as little as 50amino acids from the C terminus of diphtheria toxin preventsproper folding upon expression in E. coli and leads to anunstable structure with increased proteolytic sensitivity.
Pakula et al. (41) have studied a large number of missensemutations in the lambda Cro repressor, for which the three-dimensional structure is known. One class of mutationsdisrupts the structure of the protein and causes severeproteolytic degradation in E. coli that can be partiallyprevented in lon and htpR mutant strains. Although highlysensitive to E. coli proteases, many of these structuralmutants have specific activities close to that of wild-typeCro. Thus, while the protease sensitivity of hybrid toxinABM508 probably reflects altered folding of the polypeptide,this does not necessarily imply that the activity of fragmentB (namely, membrane translocation) has been compromised.Our results show that the htpR and lon mutations stabilize
cytoplasmic versions of the hybrid toxin. This is consistentwith the role of the htpR gene product as an activator ofdifferent proteases and the role of the lon gene in coding fora single cytoplasmic protease. ABM508 is thus a substratefor protease La and for other proteases regulated by htpR;however, even in these mutants there are other proteaseactivities which appear to cleave the ABM508 protein atspecific sites, giving rise to characteristic breakdown prod-ucts.When the ABM508 protein was directed to the cytoplasm
(pABM4508) in a protease-proficient host, degradation wasmore severe than for the periplasmic version under thecontrol of the same promoter (pABM1508). This agrees withthe work of Talmadge and Gilbert (51), who found thatpenicillinase is 10 times longer lived in the periplasm than inthe cytoplasm. However, the lon and htpR mutants relievedmuch of this proteolysis, allowing the cytoplasmic hybridtoxin to accumulate. Moreover, unlike the periplasm-directed construct (pABM1508), the cytoplasm-directedpABM4508 did not cause overexpression lethality uponinduction and thus could be grown to high density duringinduction. These factors led us to choose cytoplasmic accu-mulation for our p,rc-driven hybrid toxin gene in pABM6508.
I ala_G C -1U A
his-2 A Gc - G gly 1
c c +ala C- G
G - C ala+2A -U
ser4 C GU -A asp 3
/ \ +... AUG CCA CCU U GAU Guu u ...
fmet pro pro asp val val-7 -6 -5 +4 +5 +6
FIG. 8. Diagram of potential RNA hairpin that can form in all ofour tox constructs except pABM6508. The particular mRNA shownhere is from pABM4508. The stem-loop shown has a free energy of-7.1 kcal/mol at 25°C, as calculated by the method of Tinoco et al.(53).
An important finding in this study was that tox proteinswere largely insoluble when expressed at a high temperature(42°C). This phenomenon was independent of fragmentlength, since fragment A, ABM508, and CRM197 expressedin E. coli were each insoluble when expressed at 42°C. Theinsolubility was also independent of host strain and occurredwith both the cytoplasm-directed and periplasm-directedproteins. Since the optimal temperature for toxin productionfrom C. diphtheriae is 34°C, even though the organism iscapable of growth at higher temperatures (48), the tempera-ture-related insolubility probably reflected a true conforma-tional alteration of diphtheria toxin and its fragments at 42°Cand was not an artifact of expression in E. coli.Zhao and London (61) have shown that at 45 to 50°C
diphtheria toxin undergoes thermal denaturation to a confor-mation that is more hydrophobic than that at low tempera-ture and resernbles the low pH conformation of the toxin.Within endosomes, fragment B of the toxin penetrates theeucaryotic cell membrane when the pH drops below 5.5 (37,47). This process is believed to involve an acid-sensitiveconformational change which exposes hydrophobic domainswithin fragment B that are normally buried. Expression oftoxin-related proteins at high temperature might lead to asimilar denaturation phenomenon which exposes these hy-drophobic domains and causes aggregation. However, thehydrophobic character of fragment B at high temperature orlow pH cannot completely account for our findings becausefragment A alone expressed in E. coli showed high-temper-ature insolubility. Overexpression insolubility in E. coli hasbeen observed with other recombinant proteins and seems tobe related to abnormal polypeptide folding (25).
Using the results from the expression experiments, wedesigned and constructed pABM6508, a p,rc-driven cytoplas-mic version of the hybrid toxin. After induction with IPTG,our lon htpR strain made up to 10 mg of soluble hybrid toxinper liter (7.7% of cell protein by RIA), or 14 times more thanour previous maximal level. We believe that low-tempera-
FIG. 7. Levels of accumulation of four ABM508 constructs. Strain SG936 (lon htpR) harboring the plasmids shown was grown to an A550of 0.5 at 30°C before induction by heat shock (pABM1508 and pABM4508) or addition of IPTG to 1 mM (pABM6508). No alterations weremade in the culture conditions for pABM508. Samples were harvested just before induction and at 30, 90, and 180 min after induction. Solubleextracts were prepared for each time point and assayed for fragment A immunoreactivity, relative ADPRT activity, and total protein.Fragment A immunoreactivity values in milligrams per liter of culture were calculated from the molarity of the extracts, assuming a molecularweight of 54 kDa for ABM508; relative ADPRT values are given as percentages of that detected 180 min after induction of the pABM6508culture. Specific activity was calculated from fragment A immunoreactivity and the values for total protein. The vertical scales for ADRPTactivity, specific activity, and immunoreactivity in panel D are 10 times greater than in the other three panels.
ture growth, inducible, cytoplasmic expression, a strong E.coli promoter, and the use a protease-deficient host strain allcontributed significantly to this improvement in expression.The high levels of expression of soluble hybrid toxin from
pABM6508 should make it possible to isolate the full-lengthfusion protein at a high yield. The combined availability of ahigh-level expression system such as this and an efficientpurification scheme, which we are working to develop,should allow for rapid construction and characterization ofmany targeted forms of diphtheria toxin in the near future.
ACKNOWLEDGMENTS
We are grateful to Alfred Goldberg and Mark Ptashne for provid-ing us with bacterial strains, to Karen Parker and KathleenCampbell for sharing their RIA expertise and reagents, to MarianneBorowski for helpful discussions, and to Carol Wilson for antibodypurification.
This work was supported by Public Health Service grants AI21628from the National Institute of Allergy and Infectious Diseases andCA41746 and CA09031 (predoctoral support to W.R.B.) from theNational Cancer Institute and by a grant from Seragen, Inc.
LITERATURE CITED1. Amann, E., and J. Brosius. 1985. 'ATG vectors' for regulated
high-level expression of cloned genes in Escherichia coli. Gene40:183-190.
2. Amann, E., J. Brosius, and M. Ptashne. 1983. Vectors bearing ahybrid trp-lac promoter useful for regulated expression ofcloned genes in Escherichia coli. Gene 25:167-178.
3. Bacha, P., J. R. Murphy, and S. Reichlin. 1983. Thyrotropin-releasing hormone-diphtheria toxin-related polypeptide conju-gates. J. Biol. Chem. 258:1565-1570.
4. Baker, T. A., A. D. Grossman, and C. A. Gross. 1984. A generegulating the heat shock response in Escherichia coli alsoaffects proteolysis. Proc. Natl. Acad. Sci. USA 81:6779-6783.
5. Benson, S. A., M. N. Hall, and T. J. Silhavy. 1985. Geneticanalysis of protein export in Escherichia coli K12. Annu. Rev.Biochem. 54:101-134.
6. Bishai, W. R., A. Miyanohara, and J. R. Murphy. 1987. Cloningand expression in Escherichia coli of three fragments of diph-theria toxin truncated within fragment B. J. Bacteriol. 169:1554-1563.
7. Bradford, M. M. 1976. A rapid and sensitive method for thequantitation of microgram quantities of protein utilizing theprinciple of protein-dye binding. Anal. Biochem. 72:248-254.
8. Brent, R., and M. Ptashne. 1981. Mechanism of action of thelexA gene product. Proc. Natl. Acad. Sci. USA 78:4204-4208.
9. Brosius, J., T. J. Dull, D. D. Sleeter, and H. F. Noller. 1981.Gene organization and primary structure of a ribosomal RNAoperon from Escherichia coli. J. Mol. Biol. 148:107-127.
10. Brosius, J., M. Erfle, and J. Storella. 1985. Spacing of the -10and -35 regions in the tac promoter. Effect on its in vivoactivity. J. Biol. Chem. 260:3539-3541.
11. Buell, G., M.-F. Schulz, G. Selzer, A. Chollet, N. R. Movva, D.Semon, S. Escanez, and E. Kawashima. 1985. Optimizing theexpression in E. coli of a synthetic gene encoding somatomedin-C (IGF-I). Nucleic Acids Res. 13:1923-1938.
12. Collier, R. J. 1967. Effect of diphtheria toxin on protein synthe-sis: inactivation of one of the transfer factors. J. Mol. Biol.25:83-98.
13. DeBoer, H. A., L. J. Comstock, and M. Vasser. 1983. The tacpromoter: a functional hybrid derived from the trp and lacpromoters. Proc. Natl. Acad. Sci. USA 80:21-25.
14. Dente, L., G. Cesareni, and R. Cortese. 1983. pEMBL: a newfamily of single stranded plasmids. Nucleic Acids Res. 11:1645-1655.
15. Dente, L., M. Sollazzo, C. Baldari, G. Cesareni, and R. Cortese.1985. The pEMBL family of single-stranded vectors, p.101-107. In D. M. Glover (ed.), DNA cloning, vol. 1. A practicalapproach. IRL Press Ltd., Oxford.
16. Donovan, J. J., M. I. Simon, R. K. Draper, and M. Montal. 1981.
Diphtheria toxin forms transmembrane channels in planar lipidbilayers. Proc. Natl. Acad. Sci. USA 78:172-176.
17. Drazin, R., J. Kandel, and R. J. Collier. 1971. Structure andactivity of diphtheria toxin. II. Attack by trypsin at a specificsite within the intact toxin molecule. J. Biol. Chem. 246:1504-1510.
18. Gentz, R., A. Langner, A. C. Y. Chang, S. N. Cohen, and H.Bujard. 1981. Cloning and analysis of strong promoters is madepossible by the downstream placement of a RNA terminationsignal. Proc. Natl. Acad. Sci. USA 78:4936-4940.
19. Giannini, G., R. Rappuoli, and G. Ratti. 1984. The amino-acidsequence of two non-toxic mutants of diphtheria toxin: CRM45and CRM197. Nucleic Acids Res. 12:4063-4069.
20. Gill, D. M. 1982. Bacterial toxins: a table of lethal amounts.Microbiol. Rev. 46:86-94.
21. Gill, D. M., and A. M. Pappenheimer, Jr. 1971. Structure-activity relationships in diphtheria toxin. J. Biol. Chem.246:1492-1495.
22. Gilliland, D. G., R. J. Collier, J. M. Moehring, and T. J.Moehring. 1978. Chimeric toxins: toxic, disulfide-linked conju-gate of concanavalin A with fragment A from diphtheria toxin.Proc. Natl. Acad. Sci. USA. 75:5319-5323.
23. Gilliland, D. G., Z. Steplewski, R. J. Collier, K. F. Mitchell,T. H. Chang, and H. Koprowski. 1980. Antibody-directed cyto-toxic agents: use of monoclonal antibody to direct the action oftoxin A chains to colorectal carcinoma cells. Proc. Natl. Acad.Sci. USA 77:4539-4543.
24. Goff, S. A., L. P. Casson, and A. L. Goldberg. 1984. Heat shockregulatory gene htpR influences rates of protein degradation andexpression of the lon gene in Escherichia coli. Proc. Natl. Acad.Sci. USA 81:6647-6651.
25. Goldberg, A. L., and S. A. Goff. 1986. The selective degradationof abnormal proteins in bacteria, p. 287-314. In W. Reznikoffand L. Gold (ed.), Maximizing gene expression. Butterworths,Boston.
26. Goor, R. S., and A. M. Pappenheimer, Jr. 1967. Studies on themode of action of diphtheria toxin. III. Site of toxin action incell-free extracts. J. Exp. Med. 126:899-912.
27. Greenfield, L., M. J. Bjorn, G. Horn, D. Fong, G. A. Buck, R. J.Collier, and D. A. Kaplan. 1983. Nucleotide sequence of thestructural gene for diphtheria toxin carried by corynebac-teriophag, (B. Proc. Natl. Acad. Sci. USA 80:6853-6857.
28. Greenfield, L., H. F. Dovey, F. C. Lawyer, and D. H. Gelfand.1986. High-level expression of diphtheria toxin peptides inEscherichia coli. Bio/Technology 4:1006-1011.
29. Grossman, A. D., J. W. Erickson, and C. A. Gross. 1984. ThehtpR gene product of E. coli is a sigma factor for heat-shockpromoters. Cell 38:383-390.
30. Isberg, R. R., A. L. Lazaar, and M. Syvanen. 1982. Regulationof TnS by the right-repeat proteins: control at the level of thetransposition reaction? Cell 30:883-892.
31. Ittleson, T. R., and D. M. Gill. 1973. Diphtheria toxin: specificcompetition for cell receptors. Nature (London) 242:330-332.
32. Kaczorek, M., F. Delpeyroux, N. Chenciner, R. E. Streeck, J. R.Murphy, P. Boquet, and P. Tiollais. 1983. Nucleotide sequenceand expression of the diphtheria tox228 gene in Escherichia coli.Science 221:855-858.
33. Kaczorek, M., G. Zettlmeissl, F. Delpeyroux, and R. E. Streeck.1985. Diphtheria toxin promoter function in Corynebacteriumdiphtheriae and Escherichia coli. Nucleic Acids Res. 13:3147-3159.
34. Kagan, B. L., A. Finkelstein, and M. Colombini. 1981. Diphthe-ria toxin fragment forms large pores in phospholipid bilayermembranes. Proc. Natl. Acad. Sci. USA 78:4950-4954.
35. Leong, D., K. D. Coleman, and J. R. Murphy. 1983. Clonedfragment A of diphtheria toxin is expressed and secreted intothe periplasmic space of Escherichia coli K12. Science 220:515-517.
36. Maniatis, T., E. F. Fritsch, and J. Sambrook. 1982. Molecularcloning: a laboratory manual. Cold Spring Harbor Laboratory,Cold Spring Harbor, N.Y.
37. Marnell, M. H., S.-P. Shia, M. Stookey, and R. K. Draper. 1984.Evidence for penetration of diphtheria toxin to the cytosolthrough a prelysosomal membrane. Infect. Immun. 44:145-150.
OVEREXPRESSION OF A DIPHTHERIA TOXIN FRAGMENT 5151
38. Miyanohara, A., W. Bishai, J. Boyd, S. Nagle, and J. Murphy.1986. Genetic construction and expression of a diphtheriatoxin-related a-melanocyte stimulating hormone fusion gene, p.365-374. In P. Falmagne, J. E. Alouf, F. J. Fehrenbach, J.Jeljaszewicz, and M. Thelestam (ed.), Bacterial protein toxins.Gustav Fischer, Stuttgart, Federal Republic of Germany.
39. Murphy, J. R., W. Bishai, M. Borowski, A. Miyanohara, J.Boyd, and S. Nagle. 1986. Genetic construction, expression, andmelanoma-selective cytotoxicity of a diphtheria toxin-relateda-melanocyte-stimulating hormone fusion protein. Proc. Natl.Acad. Sci. USA 83:8258-8262.
40. Neidhardt, F. C., R. A. VanBogelen, and V. Vaughn. 1984. Thegenetics and regulation of heat-shock proteins. Annu. Rev.Genet. 18:295-329.
41. Pakula, A. A., V. B. Young, and R. T. Sauer. 1986. Bacterio-phage cro mutations: effects on activity and intracellulardegradation. Proc. Natl. Acad. Sci. USA 83:8829-8833.
42. Pappenheimer, A. M., Jr. 1977. Diphtheria toxin. Annu. Rev.Biochem. 46:69-94.
43. Rappuoli, R., M. Perugini, I. Marsili, and S. Fabbiani, 1983.Rapid purification of diphtheria toxin by phenyl sepharose andDEAE-cellulose chromatography. J. Chromatogr. 268:543-548.
44. Rappuoli, R., G. Ratti, M. Perugini, and I. Marsili. 1985.Production of large quantities of diphtheria toxoid CRM45.Bio/Technology 3:161-163.
45. Ratti, G., R. Rappuoli, and G. Giannini. 1983. The completenucleotide sequence of the gene coding for diphtheria toxin inthe corynephage omega (tox+) genome. Nucleic Acids Res.11:6589-6595.
46. Righelato, R. C., and P. A. van Hemert. 1969. Growth and toxinsynthesis in batch and chemostat cultures of Corynebacteriumdiphtheriae. J. Gen. Microbiol. 58:403-410.
47. Sandvig, K., and S. Olsnes. 1981. Rapid entry of nicked diph-theria toxin into cells at low pH. Characterization of the entryprocess and effects of low pH on the toxin molecule. J. Biol.Chem. 256:9068-9076.
48. Saragea, A., P. Maximescu, and E. Meitert. 1979. Corynebacte-rium diphtheriae: microbiological methods used in clinical andepidemiological investigations, p. 61-176. In T. Bergan and J. R.Norris (ed.), Methods in microbiology, vol. 13. Academic Press,Inc., New York.
49. Straus, D., and W. Gilbert. 1985. Chicken triosephosphateisomerase complements an Escherichia coli deficiency. Proc.Natl. Acad. Sci. USA 82:2014-2018.
50. Swamy, K. H. S., and A. L. Goldberg. 1982. Subcellulardistribution of various proteases in Escherichia coli. J. Bacte-riol. 149:1027-1033.
51. Talmadge, K., and W. Gilbert. 1982. Cellular location affectsprotein stability in Escherichia coli. Proc. Natl. Acad. Sci. USA79:1830-1833.
52. Thorpe, P. E., W. C. J. Ross, A. J. Cumber, C. A. Hinson, D. C.Edwards, and A. J. S. Davies. 1978. Toxicity of diphtheria toxinfor lymphoblastoid cells is increased by conjugation toantilymphocytic globulin. Nature (London) 271:752-755.
53. Tinoco, I., P. N. Borer, B. Dengler, M. D. Levine, 0. C.Uhlenbeck, D. M. Crothers, and J. Gralia. 1973. Improvedestimation of secondary structure in ribonucleic acids. Nature(London) New Biol. 246:40-41.
54. Tweten, R. K., and R. J. Collier. 1983. Molecular cloning andexpression of gene fragments from corynebacteriophage P en-coding enzymatically active peptides of diphtheria toxin. J.Bacteriol. 156:680-685.
55. Uchida, T., A. M. Pappenheimer, Jr., and R. Greany. 1973.Diphtheria toxin and related proteins. I. Isolation and propertiesof mutant proteins serologically related to diphtheria toxin. J.Biol. Chem. 248:3838-3844.
56. Vieira, J., and J. Messing. 1982. The pUC plasmids, anM13mp7-derived system for insertion mutagenesis and sequenc-ing with synthetic universal primers. Gene 19:259-268.
57. Wood, C. R., M. A. Boss, T. P. Patel, and J. S. Emtage. 1984.The influence of messenger RNA secondary structure onexpression of an immunoglobulin heavy chain in Escherichiacoli. Nucleic Acids Res. 12:3937-3950.
58. Yamaizumi, M., T. Uchida, K. Takamatsu, and Y. Okada. 1982.Intracellular stability of diphtheria toxin fragment A in thepresence and absence of anti-fragment A antibody. Proc. Natl.Acad. Sci. USA 79:461-465.
59. Yanisch-Perron, C., J. Vieira, and J. Messing. 1985. ImprovedM13 phage cloning vectors and host strains: nucleotide se-quences of the M13mpl8 and pUC19 vectors. Gene 33:103-110.
60. Zettlmeissl, G., M. Kaczorek, M. Moya, and R. E. Streeck. 1986.Expression of immunogenically reactive diphtheria toxin fusionproteins under the control of the PR promoter of bacteriophagelambda. Gene 41:103-111.
61. Zhao, J.-M., and E. London. 1986. Similarity of the conforma-tion of diphtheria toxin at high temperature to that in themembrane-penetrating low-pH state. Proc. Natl. Acad. Sci.USA 83:2002-2006.
