Cell-mediated cleavage of Pseudomonas exotoxin between Arg279 and Gly280 generates the enzymatically active fragment which translocates to the cytosol.

Pseudomonas exotoxin (PE) is a three-domain toxin which is cleaved by a cellular protease within cells and then reduced to generate two prominent fragments (Ogata, M., Chaudhary, V. K., Pastan, I., and FitzGerald, D. J. (1990) J. Biol. Chem. 265, 20678-20685). The N-terminal fragment is 28 kDa in size and contains the binding domain. The 37-kDa C-terminal fragment, which translocates to the cytosol, contains the translocation domain and the ADP-ribosylation domain. Cleavage followed by reduction is essential for toxicity since mutant forms of the toxin that cannot be cleaved by cells are nontoxic. Previous results with these mutants suggest that cleavage occurred in an arginine-rich (arginine residues are at positions 274, 276, and 279) disulfide loop near the beginning of the translocation domain, but the exact site of cleavage was not determined. Since very few molecules of the 37-kDa fragment are generated within cells it was not possible to determine the site of cleavage by performing a conventional N-terminal sequence analysis of the 37-kDa fragment. Two experimental approaches were used to overcome this limitation. First, existing amino acids near the cleavage sites were replaced with methionine residues; this was followed by the addition of [35S]methionine-labeled versions of these toxins to cells. The pattern of radioactive toxin fragments recovered from the cells indicated that the toxin was cleaved either just before or just after Arg279. Second, [3H]leucine-labeled toxin was produced and added to the cells. Sequential Edman degradations were performed on the small amount of radioactive 37-kDa fragment that could be recovered from toxin-treated cells. A peak of radioactivity in the fifth fraction indicated that leucine was the 5th amino acid on the C-terminal side of the cleavage site. This result confirmed that cleavage was between Arg279 and Gly280.

toxin that kills mammalian cells by gaining entry to the cytosol and inactivating protein synthesis (1,2). The pathway of toxin action includes binding to a surface receptor (3,4), internalization via coated pits and endosomes (5,6), proteolytic processing, reduction of disulfide bonds, and finally the translocation of an enzymatically active 37-kDa C-terminal fragment to the cytosol (7). Translocation to the cytosol requires a specific sequence (REDLK) at the C terminus of this fragment which appears to function as an endoplasmic reticulum retention sequence (8,9). Once in the cytosol, the fragment inhibits protein synthesis by the ADP-ribosylation of elongation factor 2 (10). Structural studies have shown that PE is composed of three distinct domains (11). Domain I at the N terminus binds to surface receptors, domain I1 in the middle of the protein has the translocating activity and is cleaved by a cellular protease, and domain I11 at the C terminus has the ADP-riboslyating activity and the endoplasmic retention sequence (7, 8,12).
Diphtheria toxin and ricin are also single chain protein toxins that kill cells by translocating an enzymatically active fragment to the cytosol and inhibiting protein synthesis (13)(14)(15). Although all three toxins have similar functional properties, it has been easier to identify the active fragments of DT and ricin since these fragments can be generated before the toxin is added to cells. After DT is synthesized and secreted, a large proportion of the toxin is proteolytically nicked in the growth medium (14). When nicked toxin is reduced, distinct A and B fragments are produced. The toxin is nicked within an arginine-rich region bounded by a disulfide loop formed between cysteines 186 and 201 (16). The A fragment has the ADP-ribosylating activity and is the fragment which translocates to the cytosol. Since there are 3 arginines in this loop, it is possible to generate A fragments with different C termini (16). Olsnes and colleagues (17) have suggested that the fragment produced by cleavage after ArglgO is the only one that translocates to the cytosol. Murphy and colleagues (18) have made recombinant derivatives of DT with mutations in this region. Their results suggest that the presence of an arginine at residue 193 is most important for the activity of DT-derived chimeric toxins. However, the amino acid composition of the A fragment generated from these constructs was not determined.
For ricin, a 12-amino acid linker peptide (linking the A and B domains) has to be removed before the toxin can exhibit any cytotoxic activity (19). This linker is proteolytically excised within germinating castor bean seeds leaving the A and B fragments held together by the disulfide bond formed from cysteines 259 in the A fragment and 4 in the B fragment (20).
Again it is the A fragment of the toxin which is translocated to the cytosol. It is presumed that the entire A fragment is 25396 N Terminus of Exotoxin Fragment Which Translocates to Cytosol 25397 translocated to the cytosol, although this has never been addressed directly. Since PE is secreted as a single chain protein and is not cleaved until it enters target cells, determining the precise composition of its active fragment has been more difficult. Previously this was addressed by following the cellular entry and metabolism of various mutant forms of 3H-labeled PE (7). The results indicated that the fragment recovered from the cytosol was 37 kDa in size and that its N terminus was derived from the arginine-rich loop (arginines are at residues 274, 276, and 279) located near the beginning of domain 11. The exact amino acid sequence at the C terminus was not determined, but because the fragment extended for 37 kDa from the arginine-rich loop, the C-terminal end of the fragment was thought to include most of the residues found at the C terminus of PE itself. In addition, deletion analysis of residues at the C terminus of PE has shown that if more than 13 residues are removed, there is complete loss of ADPribosylating activity (8). Thus the active fragment from PE begins near residues 275-280 and extends past residue 600 and probably extends all the way to the last amino acid at residue 613. Here we determine the sequence at the N terminus of this fragment.
Since we could not, by immunoprecipitation or other biochemical means, recover enough of the 37-kDa fragment from cells to determine its N-terminal sequence by conventional microsequencing techniques, alternative approaches had to be used. All of the naturally occurring methionine residues in PE are found near the N terminus of PE in domain Ia (21). Therefore, when PE, metabolically labeled with [35S]methionine, is added to cells, radioactivity can be detected in immunoprecipitates containing either unprocessed toxin or the 28-kDa fragment (or small fragments derived from the 28-kDa fragment). No radioactivity is associated with the Cterminal 37-kDa fragment. To localize the site of cleavage we replaced amino acids on either side of the site of cleavage with methionine and studied the nature of the labeled fragments produced when the toxin is added to cells. In addition, PE metabolically labeled with [3H]leucine was added to cells, the radiolabeled 37-kDa fragment recovered by immunoprecipitation, and the distance of leucine residues downstream from the site of cleavage determined by repeated cycles of Edman degradation.
Both types of experiments indicate that within cells PE is cleaved between arginine 279 and glycine 280.

MATERIALS AND METHODS
Plasmids and Strains-Plasmids were propagated in HBlOl or DH5a (Bethesda Research Laboratories). For expression of proteins, BL21(XDE3) was transformed with the appropriate plasmid and isopropyl 1-thio-6-D-galactopyranoside added for 90 min to induce the production of T7 polymerase (22). The structural gene for PE was located immediately downstream for the T7 promoter. Isopropyl 1-thio-P-D-galactopyranoside was added when the cells had grown in broth culture to the desired absorbance at 650 nm (usually 0.3-0.5 h).
Native PE with an OmpA leader sequence was produced from the plasmid pVC45f+(T). PE with no ADP-ribosylating activity, because i t lacked glutamic acid at residue 553 (23), was produced from the plasmid pVC45Df+(T). In native PE, all six naturally occurring methionines are located between amino acids 28 and 176 (21). The plasmids pMOA3, pMOA4, pMOA6, and pMOA8, encoding PE with a 7th methionine (replacing the wild type amino acid a t residues 275, 277, 281, and 282, respectively) were constructed by site-directed mutagenesis using a single-stranded DNA template derived from the phagemid pVC45f+(T). As well as the desired change in amino acid composition, the oligonucleotide introduced either a Sac11 or a PuuII site (see Fig. L4). Miniprep DNAs were initially screened with these enzymes for the introduction of the new restriction site.
Because of the very high G-C content found in this region of the toxin gene some mutants were difficult to construct by site-directed mutagenesis. To overcome this problem, changes were made in the pVC45 vector (see Fig.   1B). The intermediate vector, pMOAlA2VK352, was constructed by site-directed mutagenesis (see Fig. 1B). This vector encoded the same amino acids as pVC45f+(T), but the unique XhoI site at position 1646 was eliminated and moved to position 981, and the BspMI site at 1251 was eliminated. Thus the intermediate vector had unique BspMI and XhoI sites within the disulfide loop region of domain I1 (residues 271-284). pMOIIA7 and pMOIIA5 with methionine replacements at residues 278 and 280, respectively, were then constructed by ligating synthetic oligonucleotide duplexes into the vector after cutting with BspMI and XhoI. Ligation of the relevant duplex was checked by the introduction of an MhI/SacII site (pMOIIA5) and a MluI site (pMOIIA7). In addition the sequence of plasmids pMOA4, pMOA6, pMOIIA7, and pMOIIA5 was checked by double-stranded dideoxy sequencing using Sequenase and the kit supplied by U. S. Biochemical Corp.
Oligonucleotides were produced on an Applied Biosystems DNA synthesizer. The trityl group was left on, and the oligonucleotides were purified on OPC cartridges.
Biosynthesis of Radiolobeled PE and PE Mutants-PE and PE mutant proteins were metabolically labeled as described (7). Briefly, Escherichia coli BL21(XDE3) was transformed with plasmids encoding PE or PE mutants, and the synthesis of proteins was induced by the addition of 1 mM isopropyl 1-thio- 6 After a 90-min induction period, cells were harvested by centrifugation, the periplasm recovered by osmotic lysis, and radiolabeled proteins were purified either by ion exchange chromatography (Mono Q) alone or ion exchange chromatography plus gel filtration chromatography (TSK-250 BioSil). In some experiments, crude periplasm was added to L929 cells. Since PE comprised 30-40% of the periplasmic protein and PE-related proteins could be recovered from cells by immunoprecipitation, this approach did not give results substantially different from those using purified radioactive proteins. The specific activity of [aH]leucine-labeled toxins ranged from 4,000 to 9,000 dpm/ng and of [35S]methionine-labeled toxins from 2,000 to 4,500 dpm/ng.
Toxicity of Mutant PE molecules-All mutant forms derived from PE were checked for cellular toxicity. This was done by adding various dilutions of each mutant to L929 cells for 20 h. At the end of this incubation, [3H]leucine was added to cells, and the level of protein synthesis was determined. Results were expressed as percent of control protein synthesis compared with cells that received no toxin. Relative activity was determined by comparing the ICs0 values of mutant toxins with those of native PE. The ADP-ribosylating activity of mutant toxins was assessed using procedures described previously (24). All mutants had 100% of this activity (data not shown).
The Addition of f5S]Methionine-labeled Toxin to Cells-3sS-Labeled PE was added to L929 cells for 2.5 h at 37 "C. Cells were then washed and lysed with radioimmune precipitation buffer. Labeled SDS-PAGE and radioautography (7). fragments were recovered by immunoprecipitation and analyzed by Sequencing of the 37-kDa Radiolabeled Fragment-Mouse L929 cells (-5 X loa, total) were grown to confluence in Dulbecco's modified Eagles's medium supplemented with 5% fetal bovine serum and penicillin-streptomycin. The medium was removed and replaced with 10 ml of fresh medium containing 10 pg of [3H]PE553D (34.2 X lo6 cpm)/plate. Plates were placed on rocker platform at 37 "C for 6 h. At the end of the 6 h, the medium was removed and replaced with 50 ml of fresh medium, and plates were incubated for 30 min more at 37 "C. This was repeated one more time followed by two washes with 100 ml of phosphate-buffered saline. Five milliliters of lysing buffer (radioimmune precipitation buffer) was added to each plate to lyse the cells. The lysed cells were removed, pooled, and 100 pl of packed protein-A Sepharose CL-4B (Pharmacia LKB Biotechnology Inc.) was added to diminish nonspecific binding in the sample. The mixture was placed on rotator at 4 'C for 45 min. The Protein A was removed by centrifugation and discarded. Protein A-Sepharose preloaded with rabbit anti-PE was then mixed with the precleared lysate and incubated at 4 "C overnight. The protein A-antibody complex was centrifuged from the lysate, and the pellet was washed four times with 1 ml of Tween/phosphate-buffered saline. Sample buffer was added to the pellet and the sample boiled for 4 min. The sample, which N Terminus of Exotoxin Fragment Which Translocates to with an asterisk. Panel C, plasmids pMOIIA5 and pMOIIA7 were contained 1-2 X 10' cpm, was loaded on a 12% polyacrylamide gel and run under standard conditions. The gel was then transferred to polyvinylidene difluoride membranes (Bio-Rad) at 225 mA for 3 h. The polyvinylidene difluoride membranes containing the transferred proteins were placed in 50% methanol for 15 min followed by 30 min in Enlightning (Du Pont-New England Nuclear) for autoradiography. The membrane was allowed to dry overnight at room temperature before placing against preflashed XAR-2 film. The film was exposed for 3 days. The 37-kDa fragment was cut out of the membrane for amino acid sequencing. Amino acid analyses were performed in the Protein Analysis Core Laboratory of the Cancer Center of Wake Forest University. After each round of Edman degradation, fractions were collected and counted for either 5 or 10 min in a scintillation counter.

Strategy to Replace Residues near the Site of Cleavage with
Methionine-Previously we determined that PE is cleaved in an arginine-rich loop near the beginning of domain I1 to produce a 28-kDa N-terminal fragment (and smaller fragments derived from it) and a 37-kDa C-terminal fragment (7). After proteolytic cleavage, the 37-kDa fragment which contains the ADP-ribosylating activity is separated by reduction and later translocates to the cytosol where it inhibits protein synthesis. Because all methionine residues are located in domain Ia of PE, the 37-kDa fragment does not contain this amino acid (21). To help determine the site of cleavage methionine was substituted for existing residues. Only if methionine residues are substituted on the C-terminal side of the cleavage site will the 37-kDa fragment contain methionine. To use this strategy to determine the site of cleavage, radiolabeled PE mutants were generated by growing E. coli in chemically defined media containing [36S]methionine (see below).
Construction of Methionine Replacement Mutants-With the exception of arginine residues at 276 and 279, previously shown to be essential for toxicity (25), all amino acids from histidine 275 to glutamic acid 282 were individually replaced with methionine. PEmet275, PEmet277, PEmet281, and PEmet282 were made by site-directed mutagenesis (Fig. hi), whereas PEmet278 and PEmet280 were made by oligonucleotide duplex-directed mutagenesis (Fig. 1, B and C). Plasmids were constructed as outlined under "Materials and Methods" and their composition confirmed by dideoxy sequencing.
Bioactivity of Methionine Mutants-Since proteolytic processing of PE is required for toxicity and these substitutions were near the cleavage site, it was first necessary to determine if any of these substitutions altered the cytotoxic activity and/ or processing of the toxin. Plasmids encoding each methionine substitution were transformed into BL21(XDE3) and the mutant proteins were expressed. These toxins were added to L929 cells and their activity determined by measuring inhibition of protein synthesis (Table  I). Two substitutions showed reduced toxicity, the others had apparent wild type activity. The substitution of methionine for either proline 278 or tryptophan 281 reduced toxicity by 100-fold. Possible reasons for loss of activity are discussed below.
To examine the cellular processing of the various mutant forms, toxins were radiolabeled metabolically with [3H]leucine and added to L929 cells for 2.5 h. Full sized toxin molecules and processed fragments were recovered from cells by immunoprecipitation and the degree of processing determined by radioautography. Unlike the situation with PEgly276 and 279, which could not be processed appropriately, cells processed all the methionine mutants into fragments of 28 and 37 kDa. constructed by digesting pMOAIA2VK352 with BspMI and XhoI and replacing this fragment with the oligonucleotide duplexes as shown. The sites used for screening are underlined.
A representative experiment is shown in Fig. 2. Routinely, the substitution of methionine for proline 278 showed slightly reduced processing, whereas the substitution of methionine for tryptophan 281 showed slightly enhanced processing.

Recovery of ''5'S-l-..aheled PE Fragments Locates Cleavage Site hetween and
Gly2*"-To determine the site of cleavage, [:"S]methionine-labeled PE mutant proteins were incubated with cells for 2.5 h. Full-length toxin and fragments were recovered by immunoprecipitation and analyzed by SDS-PAGE and autoradiography. Results indicated that when histidine 275, glutamine 277, or proline 278 was replaced the 28-kDa fragment and smaller fragments derived from it were recovered as a radiolabeled species, but the 37-kDa fragment was not (Fig. 3). However, when glycine 280, tr-yptophan 281, or glutamic acid 282 was replaced by methionine, both 37and 28-kDa fragments contained the radiolabel. glycine 280. Since arginine 279 w a s shown previously to he essential for cleavage at this site, this amino acid was not replaced with methionine.
Sequencing the .'H-l,ahclcd 97-kI)a Fmgmrmt S h o u~ That Cleavage Occurs hetn.cen Ark"" and Gl?"Since it was not possible to recover enough of the 37-kDa fragment to perform conventional microsequencing, [~"H]leucine-laheled PE was prepared and added to cells. Leucine was chosen as the radiolabeled amino acid since, after proteolysis, it was expected to he the 5th or 6th amino acid in from the N terminus of the 37-kDa fragment (if cleavage occurred hetween k " 2 7 * and Gly'"'). [:'H]PE at a concentration of 1.0 pg/ml was added to cells for 6 h. If native toxin was used for this analysis all of the cells would be killed in this time frame; therefore ['HI PE553D, which had no ADP-rihosvlating activity. was used in its place. This form of PE is not cytotoxic because it has no ADP-ribosylating activity hut is like native PE in every other respect and was shown previously to be processed to 28and 37-kDa fragments like PE itself (7). At the end of the fih incubation cells were washed to remove unbound toxin and the cell-associated toxin allowed to process for a further hour. Immunoprecipitation was then used to recover whole toxin and toxin fragments. These were separated by SDS-PAGE and transferred to polyvinvlidene difluoride memhranes. The membranes were fluorographed to determine the location of the 37-kDa hand (Fig. 4). The radioactive band was cut out and subjected to repeated cycles of Edman degradation. To determine the location of radioactive leucine, the product from each cycle was analyzed for its content of radioactivity. In two separate experiments leucine was located primarily in the fifth cycle (Fig.  5 ) . A lesser peak was seen in the 15th cycle, consistent with the location of the next leucine in the 37-kDa fragment. Recovering ,"H-labeled leucine in the 5th and 15th cycles indicated that cleavage had occurred after arginine 279 and before glycine 280. Thus the N terminus of the fragment which translocates to the cytosol hegins with the amino acids GWEQLEQCGYPVQRL etc.

DISCUSSION
The composition of the active fragment derived from PE, which translocates to the cytosol and ADP-ribosylates elon-66kD 37kD N Terminus of Exotoxin Fragment Which Translocates to Cytosol PE553D terminal residue of the native toxin.
Because PE is a very active enzyme, only a few molecules of the active fragment need to reach the cytosol to inhihit protein synthesis and cause cell death. Although the potency of this toxin is impressive from a biological point of view, the small numher of molecules involved makes biochemical identification and analysis of the active fragment very difficult. Here we describe two methods that were used to overcome this prohlem. First, to locate the site of cleavage, we took advantage of the fact that the methionine residues within PE are not uniformly distributed. All naturally occurring methionine residues are found near the N terminus of PE. Therefore one can suhstitute methionine for existing amino acids near the site of cleavage. Since the 37-kDa fragment produced from native PE has no methionine, the site of cleavage can he identified as the location where one can first introduce a methionine into this C-terminal fragment. The analysis of .'"'S-laheled mutant proteins indicated that cell-mediated cleavage occurred between residues 278 and 280. From previous studies. using mutant toxins with suhstitutions for either Ar8" or Arg2?", we knew that cleavage within cells prohahly occurred in the arginine-rich loop region at the heginning of domain I1 (residues 265-2873, hut we did not know the exact location. Thus the methionine replacement strategy served to narrow down the location of the site to only one or two peptide bonds. Because the methionine replacement strategy could not distinguish between a cleavage after 278 and one after 279, the amino acid composition of the N terminus of the 37-kDa fragment needed to he determined. T o do this, the toxin was + -+

FIG. 4. Radiosequencing of [sHH]leucine-labeled 37-kDa fragment recovered from cells.
['H]PESS9D was added to cells for F h. Cells were washed and allowed metaholize cell-hound toxin for a further hour. Cells were then lysed and the 37-kDa fragment recovered by immunoprecipitation. The immunoprecipitate (containing approximately 1 X lo" cpm) was analyzed first by SDS-PAGE and then transferred to Immobilon membranes and fluorographed. Rands at 66 and 37 kDa were seen. The exposed x-ray film indicated the location of the 37-kDa hand which was cut out of the membrane for sequencing.  A and panel H).

Two separate experiments were performed (panel
Results for each fraction are shown R R cpm after 5-10 min of counting. No background subtraction has been performed. The deduced Nterminal amino acid sequence is shown below the fractions (panel C ) .
gation factor 2, has been determined. Its N terminus begins with glycine 280 and includes the remainder of domain 11, Ih, and most, if not all, of domain 111. The C-terminal end of the active fragment must extend a t least to residue 600 since deletions upstream of this residue eliminate ADP-rihosylating activity (8). Since there is no evidence of a second cleavage within the fragment, it is likely that the C-terminal amino acid sequence extends to residue 613, which is also the C-radiolaheled intrinsically with ['Hlleucine, taking advantage of the fact that there is a leucine residue close the site of cleavage on the C-terminal side. Analysis of the fractions obtained after each cycle of Edman degradation clearly showed that leucine was in the fifth cycle. A lesser peak was noted at position 15. This result confirmed that cleavage produced a fragment beginning at glycine 280.
The results obtained in this study would not have heen possible without the production of rnetaholically laheled toxin to high specific activity. The addition of ,'"S-labeled PEmet280-282 to cells generated radiolaheled 37-kDa fragments with strong enough signals to detect by radioautonaphy despite the fact that there was only 1 methionine/fragment available for labeling. The addition of ['HHJPE to approximately 5 X IOR cells for 6 h at a concentration of 1 pg/ ml (120 pg, total) generated a radiolabeled 37-kDa fragment. However, only 10-20 ng of the 37-kDa fragment was recovered from 5 X IOR cells. Sequence information could only he ohtained after immunoprecipitation, transfer to polyvinvlidene difluoride memhranes, Edman degradation of eluted sample, and detection of radioactivity in each of the fractions. These sensitive techniques, although by themselves are not new, may play an important role in future toxin research since only a few toxin molecules are usually needed to produce a hiological effect.
In addition to revealing the site of cleavage, the methionine suhstitutions indicated that proline 278 and tryptophan 281 are important for the toxicity of PE. The reasons why these mutations reduced toxin activity are unknown at the present hut include the possibilities that they interfered with a step after cleavage or they changed the structure of the loop to cause cleavage at another amino acid. Since the arginine-rich loop resides between two prominent cr-helices ( l l ) , proline at 278 may he needed in this location to interrupt the helical pattern and produce a prominent loop structure. If this were true, one would expect that substituting methionine (or other amino acids) for proline might change the shape of the loop and either reduce the amount of proteolysis or change the site of cleavage to one of the other arginines. This possibility is currently under investigation. The replacement of tryptophan with methionine did not interfere with the cleavage of the toxin. However, the resulting 37-kDa fragment is apparently not able to translocate to the cytosol. Preliminary data from recent studies indicate that tryptophan is important for a step in toxin action that occurs after proteolysis and before ADPribosylation in the cytosol? However, not every amino acid at the N terminus of the 37-kDa fragment is important for toxicity since PEmet280 and PEmet282 had apparent wild type activity.
As mentioned, the replacement of glycine 280 with methionine produced a mutant form of PE with full biological activity. Fortuitously, this substitution allows for the production of a recombinant form of 37-kDa fragment that begins with methionine (in place of glycine 280) and then continues with tryptophan and the other amino acids that make up the 37-kDa fragment (see 26). The production of large quantities of this active fragment will allow for the development of in vitro assays for toxin translocation that would not be possible if the only source of the fragment was the minute amount of protein that can be recovered from toxin-treated cells.
Defining the composition of toxin fragments that translocate to the cytosol has not been widely reported. In part this is because of the lack of sensitive techniques that can recover sufficient material for analysis. Also, other toxins tend to have their enzymatic domains at the N terminus of the protein. Thus, the critical amino acids for analysis are at the C terminus of the fragment, and C-terminal sequencing is difficult and inefficient. One report that tries to identify directly the composition of an active fragment is by Olsnes et al. (17) in which they show that only one of three possible cleavage products in the arginine-rich loop of DT can insert into membranes. Murphy and colleagues (18) have identified 1 of 3 arginine residues that is important for the activity of a chimeric toxin between DT and interleukin-2. Their analysis showed that proteolytic processing is needed for toxicity and that arginine at 193 is essential for the action of the intact fusion toxin.