Histidine 21 does not play a major role in diphtheria toxin catalysis.

It has been proposed that the histidine at position 21 (H21) of the diphtheria toxin A subunit (DTA) plays an important role in the ADP-ribosyltransferase (ADPRT) activity of the toxin. The region of DT encompassing H21 demonstrates sequence similarity with other toxins exhibiting ADPRT activity, is located along the catalytic cleft of DTA, and when H21 is chemically modified, ADPRT activity is abolished. H21 was mutagenized by a polymerase chain reaction-based system whereby all alternative amino acids were substituted in place of the histidine. The majority of the substitutions virtually abolished enzymatic activity, the exception being a mutant in which H21 was replaced with asparagine (DTA-H21N). This mutant demonstrated only a slight increase in Km and relatively small decreases in both reaction rate (kcat) and catalytic efficiency (kcat/Km). Asparagine is a sterically conserved substitution, but its side-chain is unable to replace the imidazole group of histidine in general acid-base mechanisms or to participate in electrostatic interactions. This suggests that H21 is important in maintaining a steric conformation required for catalysis rather than in participating in an electrostatic or acid-base type of exchange.

(e.g. G52, G128) were identified using nitrosoguanidine mutagenesis (Uchida et al., 1973). Another postulated critical site (E148) was identified by photoafinity labeling (Carroll and Collier, 1984) and was later confirmed by replacement (Tweten et al., 1985;Barbieri and Collier, 1987;Wilson et al., 1990;O'Keefe, 1992) and deletion mutagenesis (Emerick et al., 1985;Killeen et al., 1992). H21 has also been proposed to be a critical residue for the enzymatic activity of DT. This is based on a number of observations. 1) Analysis of the crystal structure of DT reveals that H21 is located within what is believed to be the enzymatic cleft (Choe et al., 1992).
2) The region around H21 has sequence similarity with other toxins that act as ADPRT, including Pseudomonas exotoxin A, pertussis toxin S1 subunit, cholera toxin, and Escherichia coli heat labile toxin (Gill, 1988;Carroll and Collier, 1988;Domenighini et al., 1991). 3) H21 in DT has been chemically modified with diethyl pyrocarbonate, resulting in an inhibition of toxicity (Papini et al., 1989). In order to better understand the structure-function relationships of DTA and the potential role of H21 in the enzymatic activity of DT, we initiated a series of studies aimed at the mutagenesis of this residue. All alternative amino acids were substituted for H21 to give a thorough understanding of the role that H21 plays in DT catalysis.
Construction of Plasmid Libraries Randomly Mutated at Residue 21-A diagrammatic representation of the cloning procedure used to generate the library of mutants is shown in Fig. 1. The numbering of amino acid residues is as defined by Greenfield et al. (1983). Genomic DNA was isolated from C. diphtheriae C7&3)t0x' as previously described , and the region of the toz gene encoding DTA was amplified in two halves, using two pairs of PCR primers; the first pair (1 and 2a) were designed to amplify the 5' portion of the gene (nucleotides encoding residues 1-30). Primer 1 included a BamHI site for cloning, in addition to a translation initiation codon in the optimum Kozak context immediately prior to the first nucleotide of the DTAgene. Random substitutions were generated at nucleotides 61-63 (encoding amino acid residue 21) using primer 2a, as shown in Fig. 1. T h e 3' portion of the DTA gene (nucleotides encoding residues 22-193) was amplified using the second pair of primers (2b and 3) which made no nucleotide substitutions, but generated a product which had 27 nucleotide homology at its 5' end with the 3' end of the primer 1-2a product

4349
(the stippled region in Fig. 1). Primer 3 was designed to include an EcoRI site for cloning, and to introduce tandem translation stop codons immediately after the third nucleotide of codon 193. The two halves of the gene were then spliced together using a PCR-based method as previously described Nicholls et al., 1993). Each round of PCR was comprised of 20 cycles. The corresponding region of the gene encoding CRM 197, a mutant of DT with no enzymatic activity, was also amplified using primers 1 and 3. Products were digested with BumHI and EcoRI, and treated with alkaline phosphatase, prior to cloning into the Bluescript K S ' vector in a region downstream ofthe T7 RNA polymerase promoter. The inserted region encodes for amino acids 1-193 and corresponds to the enzymatically active form of DTA (Williams et ul., 1990). The gene encoding DTA was cloned without DTB in compliance with the restrictions on cloning toxin genes (Federal Register, 1986). E. coli strain XL1-Blue was transformed with the recombinant library. Plasmid DNA from the resulting colonies was isolated and used directly as a template for transcriptiodtranslation, and for sequencing.
In Vitro Punscription I Panslation of DNA lkmplates-Plasmid templates were transcribed and translated acid using a rabbit reticulocyte lysate-based coupled transcriptiodtranslation system, according to the instructions of the manufacturer. ~-[~~S]Methionine-labeled proteins were quantitated by trichloroacetic acid precipitation and analyzed by SDS-PAGE. Radiolabeled proteins were also used for the determination of enzymatic activity.
ADP-ribosyltrunsferase Assay-~-[~~S]Methionine-labeled proteins were quantitated by trichloroacetic acid precipitation and the enzymatic activities of the various DTA mutants were determined as previously described (Johnson et al., 1988). The counts obtained from the trichloroacetic acid precipitation of H21M were adjusted to take into consideration the fact that this mutant has an extra methionine residue. %-Labeled proteins could be used in the ADPRT assay because counting 32P-labeled EF-2 by the Cerenkov method does not register the weaker emissions of 36S (data not shown).
Erpression of DTA Mutants in E. coli-DTA mutants were expressed in E. coli BL21(DE3), using the T7 RNA polymerase-driven expression vector, PET-lld (Studier et al., 1990). Genes encoding mutants of interest from the original library were modified by PCR to include a region encoding 6 histidine residues ((His)6) immediately following codon 197. After sequencing completely to verify the desired composition, constructs were used to transform the expression host BL21(DE3). A freshly transformed colony was grown, induced with 0.4 m~ isopropyl-1-thio-P-D-galactopyranoside for 3 h and the pellet frozen at -70 "C until required.
Purification of (His),-tagged DTA Mutants-(His)6-tagged proteins were purified under denaturing conditions, using immobilized metal chelate affinity chromatography, essentially according to the instructions of the manufacturer of Ni-NTA resin (Qiagen Inc., Chatsworth, CA). The recombinant proteins were dialyzed against phosphate-buffered saline, concentrated, quantitated, and analyzed by BCA assay (Pierce Chemical Co.) and SDS-PAGE, respectively.
Removal of (Hi& lbg with Immobilized Carboxypeptidase-Immobilized carboxypeptidase A (EC 3.4.17.1; Sigma) was used to remove the (His)6 tag from the carboxyl terminus of recombinant proteins. This enzyme preferentially cleaves COOH-terminal residues with aromatic or branched side-chains, and cleavage is terminated by basic residues. All DTA mutants contain 2 arginine residues at the COOH terminus (residues 192 and 193) so that proteins generated by this method have a COOH terminus identical to that identified in native DTA (Williams et al., 1990). 200-pg batches of protein were digested at 37 "C overnight with 0.1 units of enzyme in 0.025 M Tris-HC1, 0.5 M NaC1, pH 7.5. Complete digestion was confirmed by failure of carboxypeptidase A-digested proteins to bind to the metal chelating resin. The integrity of each protein was confirmed by SDS-PAGE prior to further analysis.
Kinetic Analysis of Enzymatic Actiuity-Initial rates of ADP-ribosylation of EF-2 by the toxin were monitored by the incorporation of the radiolabeled ADP-ribose moiety of NAD' into trichloroacetic acid-precipitable counts. [adenylafe-32PlNAD+ was diluted with unlabeled NAD' to a specific activity of 2 CUmmol (for H21H and H21N) or 0.5 CUmmol (for H21F, H21A, and H21G). EF-2 concentrations were kept constant and NAD' concentrations were varied from 1 to 20 p~ (for H21H and H21N)  The tox gene from wild-type C. diphfheriue genomic DNA corresponding to nucleotides coding for residues 1-193 was amplified in two halves. Totally random nucleotide substitutions (NNN, corresponding to nucleotides 60-63) were used in primer 2a thus enabling the replacement of H21 (encoded by CAC) with other amino acids. The two halves of the gene resulting from the first round of PCR are spliced together in the second or recombinant round. Primers 1 and 3 include a BamHI and EcoRI site, respectively, for use in cloning the PCR products. Each round of PCR was comprised of 20 cycles. camed out at 37 "C for a total of 3 min, precipitated with 10% trichloroacetic acid and Cerenkov counted. The kinetic parameters were obtained by analysis of Lineweaver-Burk plots from at least three separate experiments, each performed in duplicate.
SDS-PAGE-SDS-PAGE was camed out using 10-208 gels (Daiichi, Tokyo, Japan) as described by Laemmli (1970). For analysis by fluorography, the gels were fixed for 20 min in 10% methanol, 7% acetic acid, treated with Amplify (Amersham) for 30 min, and dried. Gels were exposed to X-Omat AR2 film (Eastman Kodak, Rochester, N Y ) in the absence of an intens&ing screen.
Other Methods--Trypsin digestion patterns of wild-type or mutant DTA were compared by premixing 4 pg of toxin with 200 p~ NAD+, digested with varying concentrations of trypsin (Boehringer Mannheim) for 1 h at 37 "C, and analyzing by SDS-PAGE. Dideoxynucleotide chain-termination sequencing of double-stranded DNA templates was performed using a Sequenase I1 kit (U. S . Biochemical Corp.), according to the manufacturer's protocol.

RESULTS
Mutagenesis of H21-H21 was mutagenized using a PCRbased method (Fig. 1). By introducing totally random nucleotide substitutions at positions 61 through 63, it was possible to generate a library of mutants encoding all possible amino acids at codon 21. PCR products were cloned into the Bluescript K S ' vector in a region downstream of the T7 RNA polymerase promoter. The presence of this promoter enabled initial toxin production using an in vitro coupled transcriptiodtranslation system, facilitating screening of a large number of DTA clones.
Equal volumes of radiolabeled products from the coupled transcriptiodtranslation system were analyzed by SDS-PAGE (Fig. 2). Under reducing conditions, wild-type DTA as well as the mutant forms of DTA produced a major band with molecular weight of about 21,000. In addition, a lower band of consistently lesser intensity ( M , about 19,000) is also visible. Investigation of the sequence of toz, the gene encoding DT, reveals a second in-frame ATG codon at nucleotide positions 40-42 (codon 14). This second potential translation initiation codon is also found in the optimum Kozak context (Kozak, 1991). The lower M , band may result from the ribosome failing to initiate translation at codon 1, with initiation instead starting at codon 14. It has been reported that rabbit reticulocyte lysate preparations frequently have impaired initiation fidelity (Kozak, 1990). It is interesting to note that the lower M, band is not evident when the proteins are expressed in E. coli (see Fig. 4). Further evidence to support the hypothesis that the lower M , DTA- band results from translation initiation a t codon 14 was obtained by modifying primer 1 (Fig. 1) so that it deleted the first 13 codons of the DTA chain and instead started at the ATG coding for residue M14. The protein produced ran as a single band after SDS-PAGE, migrating to precisely the same location as the lower M, band (data not shown).
Clones producing a translation product of the correct size were sequenced across the region spanning the mutagenized codon to determine the identity of the substituted amino acid. Clones containing all 20 possible amino acids at position 21 were isolated. The mutant genes were subjected to a total of 40 rounds of PCR, therefore, the potential for PCR-induced errors in regions that were not sequenced exists. However, only five unexpected mutagenic events were observed in approximately 13,000 bases sequenced; the probability of there being a single error in a given DTA clone is therefore approximately 1 in 4. Although there is a possibility that a PCR error in a site other than H21 could affect enzymatic activity, we have isolated multiple independent clones for the majority of the amino acid substitutions, and all exhibited identical activity for a given substitution.
Analysis of Enzymatic Activity-Radiolabeled translation products for wild-type DTA, CRM 197, and the DTA mutants were quantitated by trichloroacetic acid precipitation and comparable amounts of protein compared using a cell-free ADPribosylation assay (Fig. 3). A clone containing histidine at position 21 (DTA-H21H) was isolated and found to have comparable enzymatic activity to wild-type DTA generated by subjecting genomic DNA to 20 cycles of PCR amplification using primers 1 and 3 (Fig. l), cloning the PCR product into Bluescript KS' , and expressing in vitro. CRM 197, a mutant containing a single mutation a t amino acid position 52 (G52E) resulting in the inability to bind NAD' and a total lack of enzymatic activity, was also included as a negative control.
Multiple clones were isolated for the majority of the amino acid substitutions (all except V, Y, Q, and E). DTAmutants were analyzed in duplicate in three separate ADP-ribosylation assays and the results were plotted as the percent of wild-type enzymatic activity (Fig. 3). The majority of the amino acid substitutions resulted in mutants with extremely low levels of ADPRT activity; substitution with A, V, L, I, P, F, W, M, G, Y, Q, D, E, or K resulted in mutants expressing less than 5% of wild-type ADPRT activity under the conditions of the assay. Substitution of S, T, C, or R at position 21 resulted in enzymatic activity between 5 and 10% of wild-type. The one exception with significant activity was DTA-H21N which demonstrated 65% of the wild-type enzymatic level.
The data obtained from mutants produced in vitro was a valuable guide in the selection of candidates for further characterization. However, a more accurate determination of the kinetic parameters of the ADPRT reaction required larger quantities of pure protein at a known concentration. Since the yields obtained from in vitro expression systems are limited, selected mutants were expressed in E. coli to produce sufficient quantities of toxin to allow a more detailed characterization.
Expression of Selected H21 Mutants in E. coli"H21 mutants to be expressed in E. coli were selected on the basis of their relative enzymatic activity. DTA-H21N was chosen because it was the only mutant isolated with significant enzymatic activity. DTA-H21A was selected because it demonstrated low but measurable enzymatic activity (about 5% wild-type). DTA-H21F and DTA-H21G both displayed nondetectable enzymatic activity in a cell-free assay at the concentration generated in vitro. The phenylalanine and glycine mutants have very different sized side-chains at position 21, and may have significantly different effects on main-chain conformation, despite both essentially abolishing enzymatic activity. The effect of amino acid substitutions on the overall conformation of the toxin was an important consideration in this study; DTA mutants that are enzymatically inactive but retain the wild-type conformation could prove to be useful as potential vaccine candidates.
These four mutants together with wild-type (DTA-H21H) and CRM 197 DTA were selected for expression in E. coli BL21(DE3) using the PET-lld vector. The selected genes were modified by PCR to include a region encoding 6 histidine residues immediately following codon 197 to facilitate purification (Porath, 1992). All constructs were completely sequenced to verify the amino acid substitution and to confirm the remainder of the DTA chain sequence. The corresponding DTA proteins were produced in E. coli and purified to >95% homogeneity using immobilized metal chelate affinity chromatography (Porath, 1992) (Fig. 4).
To test whether the presence of the (His16 tail has any influence on enzymatic activity or conformation, toxins were digested with immobilized carboxypeptidase A to remove residues from the COOH terminus of the proteins. Cleavage terminated at Arg-193, the first basic residue encountered by the enzyme. Complete removal of the histidine tail was confirmed by the inability of carboxypeptidase A-digested proteins to bind to the metal chelating resin. The presence of the histidine tail did not alter the enzymatic or conformational proper-  . 4. Expression of selected H21 mutants in E. coli. Four of the H21 mutants as well as wild-type (DTA-H21H) and CRM 197 DTA were modified by PCR to include a region encoding 6 histidine residues immediately following codon 197 to facilitate purification. All constructs were fully sequenced to verify the required composition, prior to expression in E. coli BL2UDE3) using the PET-lld vector. The DTA proteins were purified using immobilized metal chelate affinity chromatography. CRM 197 DTA migrates with a reduced mobility relative to the other mutants as reported by others (Papini et al., 1987). ties of the toxins (data not shown).
Trypsin sensitivity of DTA-H21H was compared to that of the four mutants. The amino acid changes at position 21 caused no significant increase in sensitivity to trypsin implying that these substitutions produced little or no alteration in the overall protein structure (data not shown).
Enzymatic Analysis of Selected DTA-H21 Mutants-A comparison of the ADPRT activities of selected H21 mutants at varying toxin concentrations is shown in Fig. 5. Activities relative to wild-type were based on the concentration of enzyme required to ADP-ribosylate 50% of the available EF-2, under the conditions of the assay. The activity of DTA-H21H produced in E. coli was comparable to that of commercially purchased DT (data not shown). DTA-H21N was approximately 30-fold less enzymatically active than native DTA-H21H, and DTA-H21A was approximately 230-fold less active. Full ADPRT activity curves could not be generated for DTA-H21F or DTA-H21G due to limitations in toxin concentration. However, estimation of their relative enzymatic activities by extrapolation suggests that DTA-H21F was approximately 1500-fold and DTA-H21G approximately 4000-fold less enzymatically active than DTA-H21H. Supporting preliminary data generated by in vitro expression of these mutant A chains with wild-type DTB suggests that toxicity on a DT-sensitive cell line follows the same pattern as enzymatic activity (toxicity of DTAB-H21H > DTAB-H21N > DTAB-H21A > DTAB-H21F > DTAB-HBlG), as expected.2 A detailed analysis of the kinetic aspects of enzymatic activity was also carried out. K , values for NAD+ of the mutant and wild-type toxins in the ADPRT reaction were determined from initial rate data obtained at fmed EF-2 concentrations. The kinetic parameters obtained by analysis of Lineweaver-Burk plots of initial reaction velocities are summarized in Table I. The K , for wild-type DTA-H21H was 9 p~, corresponding to the value previously published for DT (Kandel et al., 1974). DTA-H21N and DTA-H21F both show a slight increase in K , (17 and 25 p~, respectively). In contrast, K , for DTA-H21Aand DTA-H21G was greatly increased (63 and 48 p~, respectively), suggesting that these mutants had significantly reduced affinity for. NAD+.
The kinetic parameters k , , and kcat/Km were also calculated for the wild-type toxin and the mutants ( Table I). The asparagine substitution at position 21 resulted in an enzyme with only slightly reduced catalytic rate (16-fold lower kcat) and efficiency (30-fold lower kcat/Km). The other substitutions, H21A, H21F, and H21G, produced more dramatic reductions in both the catalytic rate (67-, 457-, and 913-fold reduction, respectively, in V. G. Johnson and P. J. Nicholls, unpublished data.

DISCUSSION
Site-directed mutagenesis has proven extremely useful in analyzing structure-function relationships of DT (see Wilson and Collier (1992) and Nicholls and Youle (1992), for recent reviews). This paper demonstrates further development of such methods, facilitating substitution of all possible amino acids at the site of mutagenesis. The technique circumvents problems encountered in selecting the amino acid to replace the native residue and gives a more complete view concerning the role that a particular residue plays. Mutagenesis of DTA could potentially produce mutants with impaired enzymatic activity yet with the native conformation, and such mutants could prove to be useful vaccine candidates. In a vaccine context, it is becoming increasingly clear that the potential for reversion back to wild-type poses a serious safety problem , and it is therefore important to have a thorough understanding of the effect of all possible amino acid substitutions at the site of mutagenesis. The PCR-based method, together with in vitro expression of the resulting mutants and analysis by a cell-free ADPRT assay, allows an easy and rapid method of screening a large number of mutants. It also provides essential data required for safety assessment prior to large scale in vivo culture of toxins.
H21 was selected for mutagenesis, because this residue has been proposed to play an important role in ADPRT activity of DTA. Of the 16 histidine residues present in DT, it is the only one found within the A chain of the toxin. Analysis of the crystal structure of DT reveals that H21 is located within the enzymatic cleft of DTA (Choe et al., 1992). The histidine imidazole ring is believed to project to the inside of this cavity, thus facilitating its involvement in hydrogen bonding (Domenighini et al., 1991). The region encompassing H21 also demonstrates sequence similarity with other toxins expressing ADPRT activity, including Pseudomonas exotoxin A, pertussis toxin S1 subunit, cholera toxin, and E. coli heat labile toxin (Gill, 1988;Carroll and Collier, 1988;Domenighini et al., 1991). The corresponding residue of the S1 subunit of pertussis toxin (H35) has been mutagenized resulting in a marked reduction in ADPRT activity (Kaslow et al., 1989). Treatment of DTA with diethyl pyrocarbonate selectively modifies H21 and results in inhibition of NAD+ binding and ADPRT activity (Papini et al., 19891, suggesting that H21 is involved in the enzymatic activity of DT. Nevertheless, without site-directed mutagenesis studies directed at this residue, the possibility that these results are due to steric hindrance from the bulky chemical modification cannot be discounted (Miles et al., 1989).
In an attempt to clarify the role of H21 in catalysis, all alternative amino acids were substituted in place of H21 and the results of these mutations on ADPRT activity were analyzed. DTA-H21N was the only mutant that retained significant enzymatic activity. When DTA-H21N was assayed at varying toxin concentrations, it was found to be only 30-fold less enzymatically active than wild-type DTA. Similarly, its K,,, was increased only slightly, thereby indicating that its ability to bind NAD+ is not significantly impaired. The catalytic rate (kcat) and efficiency (kcat/K,,,) are also only reduced slightly relative to wild-type toxin. This is interesting considering that H21 also has been proposed to function in ADPRT activity through its interaction with E148 (Wilson et al., 1990;Wilson and Collier, 1992). A titratable group with a pK, of 6.2-6.3 (Papini et al., 1990) was identified within DTA and assumed to represent the ionization of the H2 1 side-chain imidazole. It was proposed that the carboxyl group of E148 strongly affects this titration, and maintains a particular active-site conformation possibly by confining the imidazole ring of H21 in the position required for catalysis. Alternatively, E148 may alter the nucleophilicity of the incoming diphthamide through acid-base interaction with the histidine. Asparagine is assumed to be a sterically conservative substitution for histidine, but the sidechain of asparagine can neither replace the imidazole group of histidine in general acidhase mechanisms nor participate in electrostatic interactions. The fact that asparagine can substitute for histidine at position 21 suggests that histidine is important in maintaining the required conformation rather than in participating in an electrostatic or acid-base type of exchange.
The majority of amino acids substituted at position 21 virtually abolish ADPRT activity. DTA-H21A and DTA-H21G both show significant increases in K, as well as decreases in both the catalytic rate (kcat) and efficiency (kcat/Km) ofADPRT. It is noteworthy that DTA-H21A has the highest K, , , value, although it is by no means the least efficient enzyme (the phenylalanine and glycine mutants are approximately 3-and 11-fold less efficient than DTA-H21A, respectively). Alanine does not have a side-chain beyond the /3-carbon and therefore is not believed to alter the main-chain conformation or impose extreme electrostatic or steric effects (Ashkena et al., 1990). Nevertheless, the H21A substitution results in an inability to maintain the ideal conformation required for catalytic activity. For DTA-H21F, K,,, was only slightly higher than for the asparagine mutant, yet catalytic efficiency was significantly lower than for DTA-H21N. This suggests that the bulky side-chain of phenylalanine does not cause a drastic decrease in ability to bind NAD+, but that the conformation of the bound substrate is not ideal with respect to the catalytic mechanism. These conformational changes appear to be subtle since the amino acid substitutions analyzed do not appear to result in gross structural changes as demonstrated by sensitivity to trypsin proteolysis.
It has recently been shown that the deletion of a n active site residue drastically reduces ADPRT activity, although activity can be partially restored by second-site mutations . Clearly, in the context of a potential vaccine candidate, reversion of a n inactive mutant to a n enzymatically active form would pose a serious safety problem. Therefore, multiple mutations, each independently detoxifying the molecule, may be required for the construction of a genetically inactivated toxin for use as a vaccine. A substitution at position 21 may prove useful as one of these mutations. H21G has greatly reduced enzymatic activity (rel. kcat/K,,, is reduced about 5000fold), yet based on trypsin sensitivity, is believed to have a conformation similar to that of the wild-type toxin. Work is in progress to further analyze this mutant and others as potential vaccine candidates.