Functional Analysis of Domains 11, Ib, and I11 of Pseudomonas Exotoxin*

Pseudomonas exotoxin is composed of three struc- tural domains that are responsible for cell recognition, membrane translocation, and ADP-ribosylation. The substitution of the cell recognition domain (domain Ia) with a growth factor such as transforming growth factor a (TGFa), creates a cell-specific cytotoxic agent, TGFa-PE40, which kills cells bearing epidermal growth factor (EGF) receptors. We have used TGFa- PE40 to define the role of sequences in domains 11, Ib, and 111. Various mutations were made in these domains and mutant forms of TGFa-PE40 expressed in Esche- richia coli. Mutant proteins were then tested for their ADP-ribosylation, EGF receptor-binding, and cell- killing activities. Additionally, the amino boundary of domain 111, which contains the ADP-ribosylation activity, was determined by deletion analysis. Data indicate that (i) the functional amino terminus of domain I11 is near amino acid 400; (ii) deletion of various regions in domain I1 or conversion of cysteines 265 and 268 to serines results in a loss of cytotoxicity which ranged from 10-fold to more than 150-fold, indicating that domain I1 is essential for full expression of cytotoxic- ity; (iii) deletion of the amino terminus of domain Ib results in a molecule with somewhat increased cyto- toxic activity, indicating that domain Ib is not essential for the cytotoxic effect of TGFa-PE40; and (iv) TGFa- PE40, produced by denaturing and refolding of insoluble Protein Synthesis Inhibition Assay-The cytotoxic activity of TGFa-toxin fusion proteins was determined by assaying the inhibition of protein synthesis using A431 cells (12). Protein synthesis was assayed by determining the incorporation of [’H]leucine in trichlo-roacetic acid precipitates of the cells 17).

TGFa recognizes and binds to cells that display EGF receptors (8). The fusion protein TGFa-PE40 is cytotoxic to EGF receptor-bearing cells. T o study the function of residues within PE40 which might be involved in the translocation process, we have modified the TGFa-PE40 molecule by creating mutations in domains I1 and Ib. The cytotoxic activity of the TGFa-PE40 fusion proteins was assessed using A431 human epidermoid carcinoma cells that contain large numbers of EGF receptors. The binding activities of TGFa-PE40 and two of the TGFa-PE40 mutant proteins were assessed by determining the amount of these proteins needed to displace lz51-EGF when added to A431 cells.
Mutations in the region adjoining domain I11 could affect ADP-ribosylation activity in a negative way and thereby decrease cytotoxicity. Therefore, it was necessary to determine the functional boundary between domains Ib and 111. To do this, various portions of the Pseudomonas exotoxin gene adjoining and within domain I11 were deleted and the amino boundary of domain 111 defined. Our studies indicate that the functional boundary required for full ADP-ribosylation activity starts near amino acid 400 in Pseudomonas exotoxin.
Based on results of cytotoxicity assays, mutant forms of TGFa-PE40 were placed into one of four groups. Two groups have greatly reduced cytotoxicity: one due to reduction in ADP-ribosylation activity, and one due to mutations in domain 11. A third group shows about a 10-fold decrease in cytotoxicity, and a fourth group shows a slight increase in cytotoxicity.

MATERIALS AND METHODS
Enzymes and chemicals were purchased from standard chemical sources. A cDNA coding for TGFa was a gift from Allan Oliff of Merck, Sharp and Dohme Research Laboratories (West Point, PA). lz5I-EGF was purchased from Du Pont-New England Nuclear.
Bacterial Strains, Plasmids, and Cell Lines"HB101 from Bethesda Research Laboratories was used for the propagation of the plasmids. BL21 (XDE3), which carries an inducible T7 RNA polymerase gene on a X prophage (5, 9 ) , was used as the host for fusion protein expression. Oligonucleotide-directed mutagenesis (for pCS8) was carried out using the method of Kunkel (10). Mutants were identified by DNA sequencing (11). The plasmids pVC17, pVC8, and pVC7 (Fig. 1) were described previously (12). The plasmids pCSll and The abbreviations used are: TGFa, transforming growth factor a; EGF, epidermal growth factor. Deletion mutants are shown, and numbers indicate amino acids present in PE40. The OmpA signal sequence is located directly after the T7 promoter in all the constructs. The cDNA for TGFa codes for amino acids 1-50 of all the constructions. Dotted lines in Pseudomonas exotoxin (PE) represent disulfide bonds. pCSl2 ( Fig. 1) were created from pVC4 (12). The plasmids pCS1-pCSl0 ( Fig. 2) were created from pyx3205 or pVC3875 and carry an OmpA signal sequence (13). The plasmids pyx3205 and pVC3875 will be described elsewhere? Briefly, pyx3205 contains the human TGFa cDNA sequence (Merck, Sharp and Dohme Research Laboratories) ligated to the 5' end of the gene encoding PE43 carried on pVC20 (12). pVC3875 contains the human TGFa cDNA sequence ligated to the 5' end of the gene encoding PE40 (12). The plasmids pVC3875, pCS2, and pCSl0 were also modified to encode fusion proteins without the OmpA signal sequence.
Expression of ~e c o m &~~n~ Fmion Proteins-Expression of Pseudomonas exotoxin containing fusion proteins was done using the host BL21 ODES) as described previously (5,7,12). Cells were incubated for 90 min following induction, and the periplasm fraction was C. B. Siegall prepared (12) for the mutant proteins expressed from pCS1-10. For proteins expressed from plasmids pVC17, pVC8, pVC7, pCS11, and pCS12, the total cell pellet was used for analysis. The modified plasmids pCS2, pCS10, andpVC3875, in constructions without OmpA signal sequence, were expressed and fusion proteins purified from inclusion bodies (7) for use in the EGF receptor-binding displacement assay.
Gel Electrophoresis and Immunoblotting-Sodium dodecyl sulfatepolyacrylamide gel electrophoresis was performed on 12.5% gels using the method of Laemmli (14). For immunoblotting, samples were transferred from gels to nitr~ellulose paper, and antibodies to Pseudomonas exotoxin (2) and Vecta stain kits (Vector Laboratories, Burlingame, CA) were used to visualize Pseudomon~ exotoxin fusion proteins. Quantitation of fusion proteins was done by comparison with standards (purified PE40) on Western blots.
A~P -r~~s y~t i o n Assay-ADP-ribosylation activity of protein samples was measured by following the procedure of CoIIier and Kandel (15) using wheat germ extract enriched in elongation factor 2.
Binding Assay-"'I-EGF-binding assays were performed using A431 cells plated at 8 X lo3 cells/ml. Competitions were done using binding buffer containing 2 ng of '251-EGF/ml with or without the potential inhibitors exactly as described (16). Protein Synthesis Inhibition Assay-The cytotoxic activity of TGFa-toxin fusion proteins was determined by assaying the inhibition of protein synthesis using A431 cells (12). Protein synthesis was assayed by determining the incorporation of ['H]leucine in trichloroacetic acid precipitates of the cells (5, 17).

P r e~r a~i o~ of T~F~-P E~O a d Other Mutants-To ex-
press TGFa-PE40, PE40, and derived mutant proteins, Escherichia coli BL21 (XDE3) were transformed with the appropriate plasmid, cultured, and induced with isopropyl p-Dthiogalactoside. PE40, TGFa-PE40, and its mutant forms, expressed using constructs containing OmpA signal sequences, were directed to the periplasm in a soluble form and were isolated as described under "Materials and Methods." Mutant proteins composed of domain I11 and nearby amino acids remained within the cell and were isolated from sucrosewashed spheroplasts. TGFa-PE4O and its mutant forms were also expressed from constructs without OmpA signal sequences and were isolated and purified from inclusion bodies (7). Toxin-containing fractions were subjected to sodium dodecyl sulfate-polyacrylamide gel electrophoresis and analyzed with a rabbit polyclonal antibody to Pseudomonas exotoxin.
Several typical examples are shown in Fig. 3, A and B. For each construction, the major band reacting with antibodies to Pseudomonas exotoxin migrated at the size expected for the gene fusion protein. The amount of recombinant toxin present was estimated visually by comparison of serial dilutions with known standards (purified PE40) on immunoblots using an antibody to P s~~o m~n a s exotoxin. Amino-terminal Boundary of Domain 111-Before making mutations in domains I1 and Ib, it was essential to determine how much of the amino terminus of domain I11 was necessary for full ADP-ribosylation activity. Therefore, the ADP-ribosylation activities of toxin proteins beginning at amino acids 381, 400, 405, and 408 were measured and compared with PE40 (amino acids 253-613) ( Table I). Full ADP-ribosylation activity was present in the proteins beginning at positions 381 and 400. The deletion of five additional amino acids to 405 resulted in loss of more than 50% of the ADP-ribosylation activity. This result is in apparent conflict with a previous report in which a protein believed to contain amino acids 405-613 retained no ADP-ribosylation activity (2). Further examination of the plasmid pJH7, which was used previously to express this truncated Pseudomonas exotoxin protein, revealed that it actually contained amino acids 406, 408-613, and was missing amino acids 405 and 407. We conclude, therefore, that functionally the domain responsible for ADPribosylation activity begins close to amino acid 400.

ADP-ribosylution Actiuity of Domains ZZ and Zb Mutants-
Next the ADP-ribosylation activity was determined for TGFa-PE40 and various proteins with mutations in domains I1 and Ib ( Fig. 2 and Table 11). Most of the mutant proteins containing deletions retained full or almost full (over 75%) ADP-ribosylation activity. A protein with a deletion of amino acids 365-399 encoded by pCS9 retained 47% of the activity of TGFa-PE40. This was somewhat unexpected because a protein expressed from pCSll (Fig. l ) , which lacks the same amino acids but has no TGFa, had full activity and suggested that TGFa was interfering with access of some substrate to a region of domain I11 important in catalysis. A protein with deletion of amino acids 253-407, which deleted a portion of domain I11 (pCS3), retained less than 2% of ADP-ribosylation activity. Thus, all the mutants except for one (pCS3) had high ADP-ribosylation activity (Table 11).
Cytotoxic Activity of Mutant Proteins-To determine the effect of the mutations in domains I1 and Ib on cytotoxicity, protein synthesis inhibition assays on A431 cells were performed using a secreted form of the protein which was readily prepared from the periplasm. A431 cells contain about 2 x lo6 EGF receptors/cell and are very sensitive to the secreted form of TGFa-PE40 with an IDso of about 1 ng/ml ( Fig. 4 and Table 11). The mutant proteins fell into four groups. The first group, which had very low cytotoxic activity, contained deletion mutants that either removed a large part of the amino end of domain I1 (amino acids 253-308), all of domain I1 (amino acids 253-380), or a large portion of the carboxyl end of domain I1 (amino acids 337-362 or 337-380). The second group consisted of one mutant encoded by pCS3 (deletion of amino acids 253-407). It had very little cytotoxicity as well as very little ADP-ribosylation activity because the mutation extended into domain 111.
A third group of mutants had about 10% of wild-type cytotoxicity. One mutant in this group had a deletion of amino acids 365-399 (pCS9). Another had a deletion of amino acids 254-263 (pCS7), and the third had cysteine residues at positions 265 and 288 converted to serines (pCS8).
The fourth group was different and displayed a small increase in cytotoxic activity as compared with wild-type TGFa-PE40. This group contained pCS6 (deletion of amino acids 361-380) and pCSl0 (removal of amino acids 364-380). These deletions remove large parts of domain Ib, which begins at amino acid 364 and ends at 400 and contains one disulfide bridge.
The low activity of the domain I1 mutants could have been due to a decrease in entry of the toxin into the cytosol after endocytosis or a decrease in binding to the EGF receptor. To examine the binding activity of the secreted forms of the chimeric toxin, several different fusion proteins were isolated from the periplasm and tested for their ability to displace 1251-EGF from A431 cells using standard binding conditions (16). Unexpectedly, neither TGFa-PE40 nor the mutants showed a substantial displacement of lZ5I-EGF, although excess EGF was able to compete for the cytotoxic activity of secreted form of TGFa-PE40 (data not shown). This result suggested that the TGFa portion of these fusion proteins was not folded correctly and therefore reacted poorly with the EGF receptor. Because TGFa-PE40 itself bound poorly to the EGF receptor, there was no way to tell if the binding of the mutant forms was diminished relative to TGFa-PE40.
To prepare TGFa-PE40 with more normal binding prop-

ADP-ribosylation and cell-killing activity of various mutant f o r m of TGFa-PE40
The amount of soluble fusion protein produced from each construct was approximately 1.2 pg/ml of original culture. For ADP-ribosylation activity assays, 10 pl of 100-fold dilution of periplasm (containing soluble protein) was used in a 250-pl reaction mixture containing 2.4 mM [''CINAD (6 X lo4 cpmj and 10 pl of wheat germ extract. IDs0 is the concentration of toxin which is required to inhibit protein synthesis by 50% as measured by ~~Hlleucine incorporation. Leucine incorporation into cellular protein was measured as described previously (11). Results are expressed as percent of control cells not exposed to TGFa-PE40. TGFa-PE40 (W), TGFa-PE40A253-308 ( 0 " 4 ) , TGFa-PE40A254-263 (U), and TGFcu-PE~OA365-380 ( U ) .

TABLE 111
Cell-killing activity of refolded fusion proteins Protein was determined by the Bradford assay. The amount of purified refolded fusion protein produced for each construct was approximately 1.5 pg/ml of original culture.

Se-
Ratio of renatured to secreted natured creted TGFn-PE408365erties, we utilized a r~naturation protocol that was successfully used previously with several other chimeric toxins (6, 7). In this procedure, the expression plasmid does not contain an OmpA signal sequence, and the chimeric toxin accumulates in inclusion bodies that are isolated, dissolved in 7 M guanidine, and renatured by rapid dilution. The renatured material was sequentially purified by Mono Q anion exchange and TSK-250 gel filtration columns attached to Pharmacia LKB ~i o t e c~n o l o~ Inc. fast protein liquid chromato~aphy. TGFa-PE40 and two mutant forms, one with a deletion of amino acids 253-380 representing the largest deletion that retains full ADP-r~bosylation activity (Table 111), and a second with a deletion of amino acids 365-380 representing a domain Ib deletion, were purified and studied.
The cytotoxic activities of the chimeric toxins made from inclusion bodies are shown in Table 111, and these activities are compared with the activities of the chimeric toxins isolated from the periplasm. In each case, the renatured material was more cytotoxic to A431 cells than was the secreted ma-teriaI, and the mutant toxins were consistently less active than TGFa-PE40.

Displacement of "'1-EGF from EGF Receptors Using TGFa-PE40 and Mutant
Forms-Since the renatured toxins were more active, it seemed likely that they would bind to the EGF receptor better than the secreted forms of the toxin. This prediction is confirmed in the binding assays shown in Fig. 5, The refolded chimeric proteins (normal and mutant) competed for "'1-EGF binding equally well. Comparison of the competition data on a molar basis with native EGF (6 kDa) demonstrates that the refolded TGFa-fusion proteins bind to the EGF receptor approximately 10-fold less well than native EGF.

DISCUSSION
In this study, we have used a chimeric protein composed of TGFa fused to mutant forms of Pseudomonas exotoxin to study the role of domains I1 and Ib in the cytotoxic activity of this molecule. To determine the regions involved in the translocation process, we created a series of deletion mutants in domains I1 and Ib. To ascertain that the mutants preserved other biological activities, they were tested for ADP-ribosylation and EGF-binding activities. We also partially determined which region of domain I11 was necessary for full activity. Previous studies have indicated that His-426, Tyr-481, and Glu-553 in domain 111 were essential for the full ADP-ribosylation activity of pseudo^^ exotoxin (3,(18)(19)(20). Our results indicate that full ADP-ribosylation activity was found in a protein containing residues 400-613 of Pseudomonas exotoxin, partial activity in a protein containing residues 405-613, and no activity in a protein containing residues 408-613 (Table I). Therefore, the functional region of domain I11 is simiIar to the structural region deduced from x-ray crystallography, which consists of amino acids 404-613 (1). Using these data, we constructed and tested a series of mutations in domains I1 and Ib using the chimeric molecule TGFa-PE40 (Table 11). The mutants fell into four activity groups based on their cytotoxicity for EGF receptor-containing A431 cells. The first two groups had very low cytotoxic activity (less than 2% of TGFa-PE40), the third had about 10% of the cytotoxic activity of TGFa-PE40, and the fourth had a slight increase in activity.
The proteins in group 1 had various portions of domain I1 deleted. Removal of the amino-terminal portion of domain I1 (residues 253-308) or internal portions of domain I1 (residues 337-380 and 337-362) dramatically reduced cytotoxicity although retaining full ADP-ribos~lation activity ( Table 11). Removal of all of domain I1 and a portion of Ib up to amino acid 380 also produced a molecule with extremely low cytotoxicity but full ADP-ribosylation activity ( Table 11). The fact that these molecules had intact binding domains and full ADP-ribosylation activities indicates that domain I1 is required for a step after binding and internalization and prior to ADP-ribosylation. Since unlike native P~e~o m o~s exotoxin, TGFa-PE40 does not require treatment with urea and dithiothreitol for activation, it is very likely that these mutants are defective in the translocation step. This conclusion is in agreement with previous results in which a portion of domain I1 was deleted from Pseudomonas exotoxin and produced a m u~n t toxin with very low cytotoxic activity (2). The second group contained only one member (deletion of 253-407) which contained extremely low ADP-ribosylation activity and, as would be expected, was not cytotoxic to A431 cells.
The third group consisted of three proteins with considerably different mutations. Two of the proteins, one missing residues 254-263 and the other in which Cys-265 and Cys-288 were changed to serines, both contained full ADP-ribosylation activity but showed a 10-fold reduction in cytotoxicity (Table 11). From these data, we conclude that small changes in domain I1 will dramatically reduce the cytotoxic activity of TGFa-PE40 fusion proteins. Once again, this indicates the requirement of domain I1 for the translocation step. A third protein in this group, missing residues 365-399, contained 47% of full ADP-ribosylation activity, which in part accounted for its reduced cytotoxic activity (Table 11). The reduced ADP-ribosylation activity in this mutant may be a result of the proximal location of domains I1 and 111, which can potentially interfere with each others structure. Deletion of sequences including 365-399 (pCS11) in a molecule containing neither TGFa nor domain I1 (Table I) did not reduce ADP-ribosylation activity.
In the fourth group of mutant proteins, we removed the amino portion of domain Ib (pCS10, missing residues 365-380) in addition to a few amino acids in domain I1 (pCS6, missing residues 361-380). Unexpectedly, these proteins resulted in a slight increase in c~otoxicity although retaining full ADP-ribosylation activity (Table 11). These results clearly indicate that most of domain Ib is not essential for the cellkilling activity of TGFa-PE40. Since domain Ib is closely associated with domain Ia, it may only be required for cytotoxicity when domain Ia is present and may act with domain Ia to help the full length molecule but not the chimeric molecules carry out its cytotoxic action. Alternatively, domain Ib may not have any role in the cytotoxic action of Pseudomonas exotoxin but may be required for secretion of Pseudomonas exotoxin by Pseudomonas aeruginosa or serve yet another function.
Initially, we studied soluble secreted forms of TGFa-PE40 and mutants with alterations in domain 11. Since these proteins did not bind well to the EGF receptor, we isolated TGFa-PE40 expressed by plasmids without signal sequences in an insoluble form and purified the protein by denaturation and renaturation. Renatured TGFa-PE40 bound reasonably well to the EGF receptor (Fig. 5) and was generally 10-fold more cytotoxic to A431 cells than the secreted form (Table 111). TGF~-PE40 that was prepared by renaturation and two representative mutants showed no differences in their binding to the EGF receptor. Altogether our data indicate that deletions within domain I1 reduce the cytotoxic activity of TGFa-PE40 by altering entry of the mutant toxins into the cytosol.