Redesign of Choline Acetyltransferase Specificity by Protein Engineering*

Since the development of site-directed mutagenesis techniques over 15 years ago (Zoller, M. J., and Smith, M. (1982) Nucleic Acids Res. 10, 6487–6500), it has been a goal of protein engineering to utilize the procedure to redesign existing enzyme structures to produce proteins with altered or novel catalytic properties. To date, however, the more successful achievements have relied exclusively on the availability of three-dimensional protein structure maps to direct the redesign strategies. Presently, such maps are unavailable for choline acetyltransferase and carnitine acetyltransferase, enzymes that catalyze the reversible transfer of an acetyl group from acetyl-CoA to choline and l-carnitine, respectively. A more empirical approach, based on cross-referencing substrate structure comparisons with protein alignment data, was used to redesign choline acetyltransferase to accommodate l-carnitine as an acceptor of the acetyl group. A mutant choline acetyltransferase that incorporates four amino acid substitutions from wild type, shows a substantial increase in catalytic efficiency (k cat/K m ) towardl-carnitine (1,620-fold) and shifts the catalytic discrimination between choline and l-carnitine by >390,000 in favor of the latter substrate. These dramatic alterations in catalytic function demonstrate that significant success in protein redesign can be achieved in the absence of three-dimensional protein structure data.

groups of carnitine serve to anchor the substrate to the enzyme. These proposals have been elaborated (2) to suggest that a charge-relay system involving an Asp-His couple on the enzyme serves to extract a proton from the C3 hydroxyl group of carnitine that allows for nucleophilic attack of the resulting oxyanion on the carbonyl of the acyl-CoA thioester. This model of CAT catalysis is supported by kinetic (3) and site-directed mutagenesis experiments (4) that tend to exclude a modified enzyme intermediate from the reaction pathway and by additional site-directed mutagenesis experiments that demonstrate essential catalytic roles for histidine and aspartate residues (2), and a role for an arginine residue (5) in the formation of a salt bridge with the carboxylate group of carnitine.
Choline acetyltransferase (ChAT) (EC 2.3.1.6) catalyzes a similar reaction to CAT, with the exception that the acetyl group from acetyl-CoA is transferred to choline rather than carnitine. Carnitine differs from choline by having a carboxymethyl group replace a hydrogen at C 1 of the latter compound. Relative to the trimethylammonium group of the substrate, acyl group esterification occurs at the same hydroxyl moiety in both enzyme reactions, on C 1 in choline and C 3 in carnitine. Like CAT (3), ChAT also follows a random-order ternary complex mechanism (6,7) and also has an essential histidine residue that is believed to act as the general acid/base catalyst (8). Thus, it seems likely that both enzymes follow similar chemical reaction mechanisms. Because the carboxymethyl group of carnitine may be involved strictly in binding (1,5), it is expected that CAT possesses a suitably constructed, positively charged pocket with which to accommodate and neutralize this group, and that this structure is absent in ChAT. Thus, the engineering of such a structure into ChAT might allow this enzyme to catalyze the acylation of both carnitine and choline.
Although the cDNA and deduced protein sequences of 13 carnitine acyltransferases and 5 ChATs have been reported (see Fig. 1 for references), a three-dimensional structure has not yet been described for any of these enzymes. It is a generally held belief that such structures are an essential requirement for any meaningful attempts at protein redesign (e.g. see Ref. 9). In the present study, however, this is shown not to be the case. By combining the information from substrate structure comparisons and protein alignment data, four amino acid residues in the primary sequence of ChAT were targeted for replacement. A modified ChAT enzyme, ChAT-R/TET, that contains these four substitutions shows a greater than three orders of magnitude increase in catalytic efficiency (k cat /K m ) for the acetylation of carnitine yet retains substantial choline acetyltransferase activity.
The complete rat brain ChAT cDNA of 2337 bp (supplied on vector pSPT18-rChAT-1; Ref. 10) was a kind gift from Drs. L. Houhou and J. Mallet (CNRS, Paris, France). With the exception of the initial transformation into Escherichia coli BMH 71-18 mutS (CLONTECH) during the site-directed mutagenesis protocol, all molecular biology experiments reported in this paper utilized E. coli DH5␣FЈIQ (Life Technologies, Inc.) as host.
Production of Recombinant Rat ChAT-A restriction site for NdeI encompassing the translational start codon and a HindIII site downstream of the termination codon were introduced into the cDNA encoding rat ChAT by using the polymerase chain reaction with plasmid pSPT18-rChAT-1 (10) as template and oligonucleotides oChAT5 (5Ј-T-GG⅐AGC⅐GAG⅐GCG⅐GCC⅐GCA⅐TAT⅐GCC⅐CAT⅐CCT⅐GGA⅐AAA⅐GGC⅐TC-C-3Ј) and oChAT3 (5Ј-CTG⅐ACT⅐CTC⅐GAG⅐GAT⅐CCA⅐AGC⅐TTC⅐AGT⅐G-GC⅐TGG⅐AGT⅐CAA⅐GAT⅐TGC-3Ј) as amplification primers (the underlined bases are those that differ from the target template). The amplified fragment was digested with NotI/XhoI (sites present in the primers only) and cloned between the NotI/XhoI sites of pBluescript KSϩ (Stratagene) to generate the intermediate vector pKS-rChAT. The sequences of the flanking regions of the amplified ChAT structural gene, i.e. lying upstream of the SphI site and downstream of the NcoI site, were sequenced to assess the integrity of the polymerase chain reaction process. An erroneous base within a subpopulation of the oChAT5 primer required replacement of the NotI/SphI fragment in pKS-rChAT with that obtained from a second polymerase chain reaction that proved to contain the correct sequence. The central 1,542-bp SphI/NcoI fragment within pKS-rChAT was then replaced with the SphI/NcoI fragment from the ChAT cDNA (obtained from pSPT18-rChAT-1) to generate pKS-rChAT*.
The rat ChAT structural gene was then assembled between the NdeI/HindIII sites of the E. coli protein expression vector pLENTY 2 by simultaneous ligation with the 737-bp NdeI/AatII and 1,200-bp AatII/ HindIII fragments from pKS-rChAT* to generate pLENTY-rChAT (this procedure was adopted because of the additional internal NdeI site within the ChAT sequence). E. coli DH5␣FЈIQ (Life Technologies, Inc.) transformed with pLENTY-rChAT was used for the production of recombinant ChAT. DNA sequence determinations and computer analysis of DNA sequence data were carried out as described previously (5).
Site-directed Mutagenesis of Rat ChAT-The replacement of the VDN tripeptide at residues 459 -461 of rat ChAT with the tripeptide TET (VDN 3 TET) and of asparagine at residue 514 with arginine (N514R) was carried out by oligonucleotide-directed mutagenesis according to the procedures of Deng and Nickoloff (11) by using the commercially available Transformer Site-directed Mutagenesis Kit (CLONTECH). The template for this procedure was pUC18 containing a blunt-ended (Klenow-treated) 610-bp NdeI/NcoI fragment from the central region of the rat ChAT coding sequence cloned into the SmaI site of the vector (NdeI end toward the HindIII site of the polylinker). Site-directed mutagenesis was carried out by using either oChAT-VDN 3 TET (5Ј-CGC⅐CGC⅐TTC⅐CAG⅐GAA⅐GGT⅐CGT⅐ACG⅐GAG⅐ACC⅐ATC⅐A-GA⅐TCA⅐GCC⅐ACT⅐CC-3Ј) or oChAT-N514R (5Ј-ACC⅐GGC⅐ATG⅐GCC⅐A-TC⅐GAT⅐AGG⅐CAT⅐CTT⅐CTG⅐GCA⅐CTG-3Ј) as mutagenic primer (the underlined bases are those that differ from the target template), and Trans Oligo NdeI/NcoI (5Ј-GAG⅐TGC⅐ACC⅐ATG⅐GGC⅐GGT⅐GTG⅐AAA⅐T-3Ј) as selection primer (the latter provided as part of the control components in the Transformer kit). The mutagenesis protocol followed that outlined by the manufacturers but with the single modification as described previously (5).
Mutants were identified by the presence of silent restriction sites for the enzymes BsiWI (VDN 3 TET mutagenesis) or ClaI (N514R mutagenesis) that had been introduced by the mutagenic primers. The inserts of selected clones were sequenced in their entirety to confirm the intended arrangements. The AgeI/SacI fragments from within the mutated inserts were used to replace the native AgeI/SacI sequence in pLENTY-rChAT. The ChAT-R/TET mutant was constructed by utilizing the unique NgoMI site that lies between the separate mutations to combine the two prior to transfer to pLENTY-rChAT. Mutant ChATs were produced in E. coli DH5␣FЈIQ and purified as described for the wild type enzyme.
Purification of Recombinant Rat ChAT-An inoculum of 0.25 ml of E. coli DH5␣FЈIQ harboring pLENTY-rChAT that had been grown overnight at 37°C and 300 rpm in 3 ml of Luria-Bertani medium containing 100 g/ml ampicillin was added to 20 ml of Luria-Bertani medium/ ampicillin and grown overnight at 30°C and 300 rpm. 10 ml of this culture was used to inoculate each of two 1-liter Luria-Bertani solutions without ampicillin, and growth continued at 30°C and 300 rpm. When the A 550 of the cultures reached 0.5-0.7, isopropyl-1-thio-␤-D-galactopyranoside was added to a final concentration of 0.1 mM, and growth continued overnight (12-18 h). Unless indicated otherwise, all subsequent procedures were carried out at 4°C. The cultures were centrifuged at 4,000 ϫ g for 10 min, the supernatant was discarded, and the whole cell pellet (about 8 g wet weight) was resuspended in 50 ml of 50 mM Tris-HCl buffer, pH 8.0, containing 5 mM EDTA (free acid) and 1 mM dithiothreitol (TED buffer) by using a Potter-Elvehjem homogenizer. The cells were centrifuged as before, the supernatant was discarded, and the pellet was resuspended in 25 ml of TED buffer. The cell suspension was sonicated on ice by using a microtip connected to a Sonifer Cell Disruptor W185 (Heat Systems-Ultrasonics) at setting 6 for 10 s on/10 s off and for a total time of 10 min. The suspension was centrifuged at 10,000 ϫ g for 10 min, and the supernatant was saved and placed on ice. The pellet was resuspended in 25 ml of TED buffer, and the suspension was sonicated and centrifuged as before. The supernatants were combined and centrifuged at 20,000 ϫ g for 20 min, and the pellet was discarded. The supernatant was dialyzed against 4 liters of 10 mM KH 2 PO 4 -KOH buffer, pH 7.0, containing 1 mM EDTA (free acid) (PE buffer) and 0.2 mM dithiothreitol for 4 h.
The dialyzed material was passed through a column (3.2 cm diameter ϫ 18 cm) of CM-Sepharose CL-6B (Amersham Pharmacia Biotech) that had been equilibrated with PE buffer at 1 ml/min, and the column was washed overnight with about 700 ml of PE buffer. ChAT was eluted by applying a linear gradient of 0 -0.75 M KCl (300 ml total volume) in PE buffer. A conservative pool of the peak fractions of ChAT activity was concentrated to approximately 5 ml by using an Amicon PM-10 membrane and dialyzed overnight against 1 liter of 20 mM KH 2 PO 4 -KOH buffer, pH 7.5, containing 1 mM EDTA (free acid), 1 mM dithiothreitol, and 50% (v/v) glycerol. The enzyme was stored at Ϫ 20°C and was dialyzed against 20 mM KH 2 PO 4 -KOH buffer, pH 7.5, prior to the determination of kinetic parameters. Recombinant soluble ChAT was produced at 2-3 mg⅐liter Ϫ1 of bacterial culture and was purified with a yield of approximately 50%. The concentration of purified ChAT was determined by using an extinction coefficient at 280 nm of 6.83 ϫ 10 4 M Ϫ1 cm Ϫ1 that was calculated based on its tryptophan and tyrosine content (12) (equivalent to A 280 1% ϭ 9.43). SDS-polyacrylamide gel electrophoresis was carried out as described previously (5).
Measurement of Enzyme Activity and Treatment of Kinetic Data-During the purification, ChAT activity was determined routinely at 30°C in 20 mM KH 2 PO 4 -KOH buffer, pH 7.4, and in the presence of 50 M acetyl-CoA and 1 mM choline in a total volume of 1 ml. The net decrease in the absorbance at 232 nm, due to the hydrolysis of acetyl-CoA, was used to determine ChAT activity by using an extinction coefficient of 4,440 M Ϫ1 cm Ϫ1 (13). The activity of ChAT in crude extracts of E. coli was corrected for the presence of an acyl-CoA hydrolase activity by measuring the rate of acetyl-CoA hydrolysis in the absence of choline. The mutant ChAT enzymes were assayed in a similar manner but with 10 mM choline. Kinetic measurements were carried out by using a Uvikon 810 spectrophotometer (Kontron Instruments).
The steady-state kinetic parameters for the various purified derivatives of ChAT listed in Table I were determined under conditions similar to those described above, but with the exceptions that the assay mixtures also contained 0.2 M KCl and the concentrations of both acetyl-CoA and the acceptor substrate, either choline or carnitine, were each varied as required. Carnitine was neutralized with KOH before use. When the assay concentration of choline (as choline⅐Cl) or carnitine (as carnitine⅐HCl-KOH) exceeded 10 mM, the KCl concentration was reduced by an equivalent amount to counteract fluctuations in ionic strength.
ChAT follows a random-order sequential reaction mechanism (6, 7), and thus, each experimental data set was fitted to the equation  (14). The K m values toward carnitine displayed by the wild type, ChAT-R, and ChAT-TET enzymes exceeded the practical concentration range of this substrate (100 mM). Therefore, the k cat /K m values of these enzymes toward carnitine were determined in the presence of 100 M acetyl-CoA, which should be sufficient to saturate the enzymes with regard to this latter substrate. The k cat /K m values were calculated by linear least-squares regression of the data using the computer program Enzfitter (Biosoft). The apparent K m and k cat /K m values of these enzymes toward acetyl-CoA were determined in the presence of 25 mM carnitine and were calculated by fitting the data to rectangular hyperbolas using Enzfitter. The relatively high K m values of these three enzyme forms toward carnitine necessitated adding greater amounts of the enzymes to the assay mixtures. In these instances, the measured activities were corrected for an observed carnitine-independent hydrolysis of acetyl-CoA. It was not determined whether this latter reaction was catalyzed by ChAT (such a reaction has been reported for chloramphenicol acetyltransferase; Ref. 15) or by contamination of the enzymes with a small amount of the acyl-CoA hydrolase activity that is present in E. coli.

RESULTS AND DISCUSSION
As described in the introduction to the text, the engineering of ChAT to utilize carnitine, in addition to its natural substrate choline, as an effective acceptor of the acetyl group from acetyl-CoA might be achieved by introducing a positively charged pocket into the enzyme with which to accommodate the additional negatively charged carboxymethyl group. Because it is expected that the amino acid residues that contribute to the formation of such a structure are already present in the carnitine acyltransferases, it was considered that these residues might be identified from protein alignment data as conserved substitutions between the choline and carnitine acyltransferase families. Specifically, such substitutions were expected to include an additional basic residue in the carnitine acyltransferases, coupled with a reduction of about 60 Å 3 in the side chain volume of one or more amino acid residues.
An alignment of the various choline and carnitine acyltransferase sequences showed that 30 amino acid residues are conserved identically among all proteins (data not shown). When the conservation of residues within each group was compared with the other, a number of distinct differences were identified, as illustrated in Fig. 1. First, Asn-514 in rat ChAT, conserved in four out of five ChAT sequences, is replaced exclusively in the carnitine acyltransferases by Arg. A recent site-directed mutagenesis experiment conducted in this laboratory with carnitine octanoyltransferase (5) showed that this Arg has a profound effect on the binding of carnitine. Thus, the introduction of an Arg at position 514 in rat ChAT was predicted to provide the ionic charge required to interact with, and neutralize, the carboxyl group of carnitine. Second, Val-459 and Asn-461 in rat ChAT, the former conserved in all ChAT sequences and the latter in four out of five ChAT sequences, are each replaced exclusively in the carnitine acyltransferases by Thr. In addi-tion to providing an alternative spatial arrangement of hydrogen bonding donors/acceptors, these substitutions create a side-chain volume decrease of approximately 33 Å 3 (16) in the carnitine acyltransferases. An Asp residue is found to occur between these two amino acids in all ChAT sequences, but is replaced conservatively by Glu in the majority of carnitine acyltransferases, although some CAT sequences retain Asp. Secondary structure prediction analyses (17) of these enzymes (data not shown) suggests that each of these tripeptide sequences occurs within a loop and not in areas of fixed secondary structure, whereas the Asn to Arg replacement occurs within a helical structure. Thus, while the latter residue may be spatially restricted, the tripeptide sequences may be afforded a degree of flexibility that is convivial to the accommodation and/or turnover of substrate. However, it does not necessarily follow that these tripeptide sequences actually interact with their respective substrates, since recent protein engineering experiments (18,19) have demonstrated that critical contributions to enzyme specificity and catalytic efficiency may be provided by loop structures that do not directly contact the substrate. Additional conserved differences between the two enzyme families that might contribute to their observed substrate preferences were not readily apparent, and it was felt The amino acid sequences of human, rat, pig, Caenorhabditis elegans, and Drosophila melanogaster ChAT, mouse, pigeon, C. elegans, Saccharomyces cerevisiae, and Candida tropicalis CAT, human and rat carnitine palmitoyltransferase-I (CPT-I), rat CPT-I-like, human, rat, and mouse carnitine palmitoyltransferase-II (CPT-II), and rat and cow carnitine octanoyltransferase (COT) were aligned using Clustal V (22). With the exception of CAT from C. elegans (23) and C. tropicalis (24), references to the above sequences may be found in Ref. 5. Two portions of the aligned sequences are presented that contain amino acids, indicated by the bullet points, that suggest conserved differences between the choline acetyltransferases and the carnitine acyltransferases. The residue numbers bordering each segment are indicated. An asterisk (*) identifies a residue that is identical in all sequences and a period (.) a residue that is conservatively substituted among all sequences.
that the substitutions described above may be sufficient to allow a modified ChAT to acetylate carnitine.
Thus, with the intention of producing a dual-specific choline/ carnitine acetyltransferase or a ChAT that is selective for carnitine, a variety of engineered derivatives of ChAT were produced. These included ChAT-R (Asn-514 replaced with Arg), ChAT-TET (the VDN tripeptide at residues 459 -461 replaced with TET), and ChAT-R/TET (containing each of the above mutations). Although CAT may contain either Asp or Glu at the position equivalent to residue 460 in rat ChAT (see Fig. 1), the majority of the carnitine acyltransferases contain Glu, and so the conservative replacement of this residue was included in the VDN 3 TET mutation. Native wild type ChAT and its engineered derivatives were produced in recombinant soluble form in E. coli. Advantage was taken of the relatively high isoelectric point of ChAT (7.97) to isolate the enzymes in nearhomogeneous form, as shown in Fig. 2, by a single chromatographic step through CM-Sepharose CL-6B (see "Experimental Procedures" for details).
The kinetic parameters determined for the purified wild type enzyme and for each of the three mutant ChATs with the substrate pairs of choline/acetyl-CoA and carnitine/acetyl-CoA are shown in Table I. In comparison to wild type, all three mutant ChATs show an improved catalytic efficiency (k cat /K m ) when carnitine is the acceptor of the acetyl group, with the greatest increase (1,620-fold) occurring with the ChAT-R/TET quadruple mutant. The data show that the separate mutations are approximately additive, with the greater improvement coming from the VDN 3 TET replacement. In the cases of the ChAT-R and ChAT-TET enzymes it was not possible to determine from these kinetic data whether the improvements in catalytic efficiency were derived from an increase in k cat or a lowering of the K m value toward carnitine, or a combination of both, since the K m values (Ͼ100 mM) toward carnitine exceeded the practical concentration range of this substrate, as it did also for the wild type enzyme. However, the results obtained for the ChAT-R/TET enzyme indicate that, in this case, the K m value has been substantially reduced (Ն21-fold) in comparison with the wild type enzyme (and the other enzyme derivatives). The increases in the catalytic efficiencies of the ChAT-R and ChAT-R/TET enzymes toward carnitine that results from the N514R mutation lends credence to the earlier proposal (5) that the Arg at the equivalent location in the carnitine acyltransferases forms a salt bridge with the carboxyl group of carnitine.
In contrast to the results obtained with carnitine, the mutations lead, in all cases, to an impairment in the catalytic effi- b Apparent value of k cat /K m acetyl-CoA used (the K m value toward carnitine was too great to obtain the true value). c In each of the reports involving the redesign of glutathione reductase, the data shown reflect the kinetic parameters determined at the pH optima, which are different for the wild type and mutant enzymes. ciency toward choline that is maximal in the ChAT-R/TET quadruple mutant with a reduction of 245-fold from wild type. Although the VDN 3 TET replacement elevates k cat by some 3to 4-fold for choline turnover, the reduction in the catalytic efficiency of the mutant enzymes toward this substrate may be accounted for primarily by an increase in the K m value. In each mutant, the increase in the K m value toward choline is mirrored by a similar increase in its dissociation constant (K d ), indicating that the loss in catalytic efficiency primarily reflects an impairment in substrate affinity.
In each enzyme derivative, the dissociation constant for acetyl-CoA is similar to that determined for the wild type enzyme, indicating that the mutations do not affect the affinity for the acyl group donor. Thus the modifications appear to have specific, and profound, effects on the binding and/or turnover of the acceptor of the acetyl group. In the case of the ChAT-R/TET enzyme, the net effect of these changes is a shift in the catalytic discrimination between choline and carnitine by a factor of Ͼ390,000 in favor of the latter compound, with the result that this enzyme acetylates both substrates with rather similar catalytic efficiencies (2.5-fold preference for choline; Table I).
Although the amino acid replacements introduced into the ChAT-R/TET enzyme lead to the acquisition of substantial carnitine acetyltransferase activity (k cat /K m ϭ 4.69 ϫ 10 2 M Ϫ1 s Ϫ1 ), the enzyme remains quite inferior to native CAT (for example, the data reported for human liver CAT in Ref. 20 suggest an approximate k cat /K m ratio of 8.8 ϫ 10 5 M Ϫ1 s Ϫ1 for carnitine turnover). Thus, while the residues targeted for replacement in the present study contribute substantially to the different substrate preferences displayed by the choline and carnitine acyltransferase families, it is clear that additional more subtle differences must exist between the two families that account for the remaining differences in the catalytic efficiencies displayed toward their respective natural substrates. Nevertheless, unlike the carnitine acyltransferases that have little ability to acetylate choline (carnitine octanoyltransferase acetylates choline with a k cat /K m ratio of 0.212 M Ϫ1 s Ϫ1 ; Ref. 5), the Chat-R/TET enzyme retains substantial choline acetyltransferase activity (k cat /K m ϭ 1.18 ϫ 10 3 M Ϫ1 s Ϫ1 ). Thus, the ChAT-R/TET enzyme represents a novel bifunctional carnitine/choline acetyltransferase.
The improvements engineered into the ChAT-R/TET enzyme for carnitine turnover compare very favorably with the more successful improvements in catalytic efficiency reported in previous redesigns of multisubstrate enzymes, some of which are listed in Table II, in which high resolution three-dimensional protein structure maps were utilized as "blueprints" for the redesign strategies. In the case of multisubstrate enzymes, it has been pointed out (21) that realistic measures of proteinengineered improvements in catalytic efficiency should take into account the properties of the redesigned enzyme with regard to the overall catalytic reaction, i.e. when k cat and the values of K m for each substrate are considered as a whole. For the ChAT-R/TET enzyme this "overall catalytic efficiency" (21) is defined as k cat /(K m acetyl-CoA ⅐K m carnitine ) and equals 1.65 ϫ 10 8 M Ϫ2 s Ϫ1 . When compared with the value for wild type, 1.44 ϫ 10 5 M Ϫ2 s Ϫ1 (k cat /K m for carnitine divided by the apparent K m for acetyl-CoA), the overall gain in catalytic efficiency for acetylation of carnitine is 1,150-fold (this gain may, in fact, be significantly greater since, by comparison with the data obtained for those reactions in which the true K m value for acetyl-CoA was determined, the true K m value for acetyl-CoA during turnover of carnitine by the wild type enzyme may well be higher). In Table II, only one of the previous studies with multisubstrate enzymes reported the effects of the redesign on each substrate in the reaction, and thus, the "overall catalytic efficiency" cannot be determined for the majority of studies. Therefore, the relative success of the present redesign may be more profound.