Misfolding of chloramphenicol acetyltransferase due to carboxy ...

Protein Engineering vol.11 no.12 pp.1211–1217, 1998

Misfolding of chloramphenicol acetyltransferase due to carboxy-terminal truncation can be corrected by second-site mutations

J.Van der Schueren1, J.Robben and G.Volckaert Laboratory of Gene Technology, Katholieke Universiteit Leuven, Kardinaal Mercierlaan 92, B-3001 Leuven, Belgium 1To

whom correspondence should be addressed. E-mail: [email protected]

Folding of chloramphenicol acetyltransferase (CAT) in Escherichia coli is hampered by deletion of the carboxyterminal tail including the last residue of the carboxyterminal α-helix. Such truncated CAT polypeptides quantitatively aggregate into cytoplasmic inclusion bodies, which results in absence of a chloramphenicol-resistant phenotype for the producing host. In this paper, a genetic approach is presented to examine this aggregation process in more detail. Random mutagenesis of inactive CAT followed by direct phenotypic selection for revertants with restored chloramphenicol resistance was used to isolate second-site suppressors of inactive truncation mutants of CAT. Two random mutagenesis procedures, independently of each other, yielded a unique substitution of Phe for Leu at amino acid position 145. This second-site mutation does not drastically affect the proteins’ stability under normal growth conditions of E.coli. Hence, the introduction of Phe at amino acid position 145 improves the ability of the protein to fold into a soluble, enzymatically active conformation. The conservative character of the Leu145Phe replacement indicates that limited changes at crucial positions can have important effects on protein folding in vivo. Keywords: enzyme evolution/inclusion bodies/protein folding/ protein stability/random mutagenesis

Introduction Protein folding in vivo is a very complex process that is influenced by many factors, including ribosomes, cofactors and prosthetic groups, cytoplasmic membranes, chaperones, intracellular ionic composition and temperature (Mitraki and King, 1989). Although high folding efficiencies are achieved in vivo, polypeptide misfolding does occur, particularly during heterologous gene expression or accelerated protein production. As a result, polypeptides may aggregate and accumulate intracellularly in inclusion bodies. The observation that polypeptides from inclusion bodies can be recovered into a functional conformation by refolding in vitro establishes that, at least in some cases, there is a problem of attaining the correct conformation (i.e. a folding problem) and not a problem of protein sequence or post-translational modification. Consequently, inclusion bodies are thought to derive from specific intermediates in intracellular protein folding pathways and not from native or fully unfolded proteins (Mitraki and King, 1989; Blum et al., 1992). Chloramphenicol acetyltransferase (CAT, EC 2.3.1.28) is a bacterial enzyme that confers resistance to the antibiotic chloramphenicol. Several naturally occurring CAT variants © Oxford University Press

have been described in the literature. Each CAT variant is composed of three identical subunits (Mr µ 25 000). CAT type I (CATI) and CAT type III (CATIII) can be overproduced in Escherichia coli in a soluble conformation up to 30% or more of total cell proteins (Murray et al., 1988; Robben et al., 1993a). Because the protein has a readily screenable phenotype (i.e. the ability to provide chloramphenicol resistance to the E.coli host cells), CAT is an interesting model for mutational studies of protein folding and stability in vivo. The shortest carboxy-terminal deletion mutant of CATI that is able to adopt a soluble, enzymatically active conformation is truncated by seven residues (Robben et al., 1993b; Van der Schueren et al., 1996). In the crystal structure of CATIII (Leslie, 1990), this minimal length corresponds to a protein with full-length α5helix at the carboxy terminal. Shorter polypeptides cannot fold into a soluble, active conformation, but aggregate quantitatively into inclusion bodies. In addition, a residue with hydrophobic side-chain at the last position of the α5-helix is required to obtain biologically active protein (Van der Schueren et al., 1996). These observations suggest a role of the carboxy terminal in proper CAT maturation through a hydrophobic interaction. However, it was not possible to deduce unequivocally whether this interaction acts at the level of CAT folding or structure stabilization. The question arose of whether the inclusion body formation of CAT due to carboxy-terminal truncation could be relieved by second-site mutations. From the analysis of such secondsite revertants, direct and indirect interactions between residues in a protein can be inferred (reviewed by Goldenberg, 1988). The CATIII variant—instead of CATI, which was the subject of our previous studies—was chosen for this mutagenesis experiment for two reasons. First, CATIII is the preferred variant for structural studies because of the availability of a high-resolution three-dimensional structure (Leslie et al., 1988; Leslie, 1990). This structural information is not available for the CATI variant. Second, it allowed us to check whether or not the structural effects due to modification of the carboxy terminal of CATI could also be observed in CATIII mutants. Materials and methods Bacterial strains and plasmids Vector constructions and expression experiments were performed in the E.coli strain WK6 {∆(lac-proAB) galE straA [F9lacIq Z∆M15 proA1B1]} (Zell and Fritz, 1987). After transformation (Mandel and Higa, 1970), cells were grown on LB plates supplemented, as required, with chloramphenicol (25 µg/ml), ampicillin (100 µg/ml) and/or 0.1 mM isopropyl β-D-thiogalactopyranoside. The E.coli mutator strain XL1-Red [endA1 gyrA96 thi-1 hsdR17 supE44 relA1 lac mutD5 mutS mutT Tn10(Tetr)] (Stratagene, La Jolla, CA) was used for in vivo mutagenesis experiments. Mutagenesis experiments were performed on the cat gene of plasmid pUCATIII[RS] (Figure 1), a derivative of plasmid 1211

J.Van der Schueren, J.Robben and G.Volckaert

Fig. 1. Construction of CATIII truncation mutants. Modification of the CATIII carboxy terminal was performed by exchanging a DNA cassette in the 39 end of the cat coding sequence. Synthetic cassettes both contained cohesive ends compatible with SacI/BglII restriction fragments of pUCATIII[RS] plasmid DNA. In cassette 2 multiple nonsense codons followed triplet 214 in order to reduce the chance that a single point mutation in the stop triplet would result in a protein with extended carboxyterminal tail. Recombinant clones were recognized by restriction analysis with TaqI. DNA sequence analysis was used to determine the specific mutations introduced in the 39 end of the cat gene. The specific DNA sequence of the 39 end of the cat gene of each of the five mutants is shown, with the name of the corresponding plasmid indicated below the sequence box. Amino acid residues, denoted in one-letter code, are written above the corresponding codons. Asterisks denote the carboxy terminus; the degenerate position W in cassette 1 represents the bases A or T; V and B in cassette 2 represent A, C or G and T, G or C, respectively.

pUCATIII. The latter was obtained by replacing the catI gene of vector pUCATI (Robben et al., 1995) by the catIII gene of vector pUC18:IM3:ClaI (Murray et al., 1988). In this way, the catIII gene was placed under control of the inducible tac promoter. The vector pUCATIII[RS] was obtained after introduction of a multiple cloning site in the 39 end of the catIII gene of pUCATIII. The carboxy-terminal fusion of the ArgSer dipeptide to CATIII had no significant effect on biological activity, protein production or thermal inactivation kinetics (see Results). These pUCATI-derived vectors contain the bla gene as second selection marker, allowing the isolation of CAT misfolding mutants by selecting for transformed cells in the presence of ampicillin. The minimal inhibitory concentration (MIC) of chloramphenicol was determined as described previously (Van der Schueren et al., 1996). Mutagenesis in vivo The mutator strain XL1-Red was used to introduce mutations randomly into the target DNA. The cultures of XL1-Red transformed cells, incubated for 1 day at 37°C with vigorous shaking, were subcultured three times by 40–400-fold dilution in ampicillin-containing liquid medium and incubation for 1 day. Subculturing increased the mutation frequency with only limited success (data not shown). Small quantities of chloram1212

phenicol (5–25 µg/ml) were added to the medium in order to enrich mutants that can provide chloramphenicol resistance. After amplification in XL1-Red, plasmid DNA was isolated and re-transformed to E.coli WK6 cells. A 5% fraction of these transformation mixtures, for the purpose of controlling mutagenesis efficiency, was plated on ampicillin-containing media; the remainder was plated on chloramphenicol-containing media to select for second-site suppressor mutants. DNA manipulations All DNA manipulations were carried out according to standard procedures (Sambrook et al., 1989). Small-scale plasmid preparations for clone analysis by restriction digestion were performed following Serghini et al. (1989). Plasmid DNA for sequencing and cloning purposes was prepared on Qiagen (Hilden, Germany) ion-exchange columns. Purification of DNA restriction fragments from agarose gel was performed with a QiaEx kit (Qiagen). Oligonucleotides were purchased from Pharmacia Biotech (Uppsala, Sweden). Unless stated otherwise, polymerase chain reactions (PCRs) were performed in 100 µl of a mixture containing 10 mM Tris–HCl (pH 9.0), 50 mM KCl, 1.5 mM MgCl2, 0.1% (v/v) Triton X-100, 0.2 mM of each dNTP, 0.01% (w/v) gelatine, 1 µM of each of the two primers, 0.01 ng of template DNA and 0.5 units of SuperTaq polymerase (H.T. Biotechnology, Cambridge, UK). This reaction mixture was covered with light mineral oil (Sigma Chemical, St Louis, MO) and submitted to 30 cycles of DNA amplification using a TRIO-Thermoblock (Biometra, Go¨ttingen, Germany): 95°C for 30 s (1 min in the first cycle), 56°C for 30 s and 72°C for 4 min (9 min in the last cycle). The yield of PCR product was estimated by comparing the fluorescence intensities of PCR fragments with those of known amounts of λ DNA after agarose gel electrophoresis and staining with ethidium bromide. This enabled us to estimate the real number of DNA duplications during PCR (Fromant et al., 1995). Amplification of a PCR product of 1079 bp from pUCATIII-derived vectors was achieved with the primers 59-GTAAAACGACGGCCAGT-39 and 59-GCGTCGATTTTTGTGATGCTCG-39. This PCR product contained the complete cat coding sequence, which could be excised using the appropriate restriction sites flanking the gene. In a first series of experiments that aimed at random mutagenesis of catIII(1–213)[R] in vitro, the ClaI–SacI fragment of pUCATIII(1–213)[R] (Figure 1) was replaced by the corresponding fragment from the PCR product. In this way, triplet 214 and the following nonsense triplets were not subject to mutagenesis, resulting in a small number of mutants with restored primary mutation. Because of low cloning efficiencies, however, subsequent experiments were carried out with AcsI and BglII restriction enzymes (Figure 1). As a consequence, the first six triplets of the cat gene were not comprised in the re-cloned PCR fragment, whereas triplet 214 and the following nonsense triplets were subjected to mutagenesis. All restriction enzymes were obtained from Boehringer (Mannheim, Germany). DNA sequences were analyzed on an Applied Biosystems (Foster City, CA, USA) Model 373A sequencing system using the ABI PRISM Dye terminator cycle sequencing protocol. Protein analysis To analyze the cytoplasmic solubility of mutant CAT proteins, cell pellets from cultures grown overnight in the presence of IPTG were resuspended in 50 mM Tris–HCl (pH 8.0), 10% sucrose and mixed with 50 mg/ml lysozyme in 25 mM Tris–

Suppressors of CAT misfolding mutants

HCl (pH 8.0), 25 mM EDTA to a final lysozyme concentration of 0.2 mg/ml. After a 15 min incubation at 30°C, cells were lysed by three freeze–thaw cycles and sonication (three bursts of 15 s with the microtip of a Vibra-Cell Model VC500 sonicator (Sonics and Materials, Dannbury, CT) set at the maximally allowed power output). Suspensions were kept on ice during sonication. Complete cell lysis was checked by phase contrast microscopy. Part of this cell lysate was kept as a sample containing total cell proteins. The rest of the lysate was centrifuged for 10 min at 15 000 g. The supernatant was saved as the soluble fraction. The pellet containing the insoluble protein fraction was washed and resuspended in 50 mM Tris– HCl (pH 8.0), 10% sucrose. SDS–PAGE (Lae¨mmli, 1970) was performed on equivalent amounts of total, soluble and insoluble fractions in a Mini-Protean II apparatus (Bio-Rad, Hercules, CA, USA). After staining of the proteins with Coomassie Brilliant Blue R-250, the percentage of CAT produced in a soluble conformation was determined using the BabyImager documentation system (Appligene, Illkirch, France) and Intelligent Quantifier software (B.I. Systems, Ann Arbor, MI). Protein concentrations were measured by the red pyrogallol molybdate method (Watanabe et al., 1986) using a SopaChem kit (Sopar-Biochem, Brussels, Belgium). Unknown protein concentrations were calculated from the absorption at 598 nm of a 1 mg/ml albumin–globulin standard (Sopar-Biochem). CAT activity assay CAT activity at 37°C was measured using the colorimetric assay described by Shaw (1975). The assay mixture contained 50 mM Tris–HCl (pH 7.5), 100 mM NaCl, 0.1 mM EDTA, 1 mM 5,59-dithiobis(2-nitrobenzoic acid), 0.1 mM chloramphenicol, 0.4 mM acetyl coenzyme A. CAT was added in a final concentration of ~5 ng/ml. The change in optical density at 405 nm was recorded with a THERMOmax microplate reader and analyzed using SOFTmax software (Molecular Devices, Menlo Park, CA). Determination of thermal inactivation kinetics The enzyme activity of CAT was used as a probe for correct protein conformation after heat treatment (Lewendon et al., 1988). The soluble fraction of crude cell extracts was prepared as described above. This solution, with a protein content of ~0.5 mg/ml, was diluted in assay buffer (50 mM Tris–HCl, pH 7.5, 100 mM NaCl, 0.1 mM EDTA) supplemented with bovine serum albumin (100 µg/ml). The latter prevented inactivation of CAT by adsorption on the wall of the reaction vessel, which may occur in diluted enzyme preparations (Suelter and DeLuca, 1983). Depending on the type of CAT mutant that was produced, the dilution factors varied between 10–5 (e.g. for CATIII and CATIII[RS]) and 10–1 [for CATIII(1–213)]. Samples of 350 µl were subjected to heat treatment in a water-bath for 10 min, immediately followed by chilling on ice. After centrifugation for 2 min at 15 000 g, the remaining CAT activity in these samples was compared with the activity in samples stored at room temperature during the same period. Results Primary substitutions in CATIII Based on the observations with CATI deletion mutants (Van der Schueren et al., 1996), inactive CATIII deletion mutants were designed and constructed (cassette 1 in Figure 1). If CATIII behaves structurally as CATI, the CATIII(1–214) mutant

Table I. Characteristics of CATIII truncation mutants Clone

MIC value (µg/ml)

WK6 WK6[pUCATIII] WK6[pUCATIII[RS]] WK6[pUCATIII(1–214)] WK6[pUCATIII(1–213)] WK6[pUCATIII(1–213)[G]] WK6[pUCATIII(1–213)[R]] WK6[pUCATIII(1–213)[S]]b

1 125 125 25 1 1 1 10

, , , , , , , ,

MIC MIC MIC MIC MIC MIC MIC MIC

ø ø ø ø ø ø ø ø

5 625 625 125 5 5 5 25

Soluble CAT (%)a – 100 100 20 0 0 0 0

aAll

CAT mutants accumulated at a high level in the E.coli cytoplasm (approximately 20% of the total amount of cell proteins). Consequently, reduced synthesis or proteolytic degradation of mutant CAT cannot account for the observed chloramphenicol resistance characteristics. bSmall amounts of soluble, active CAT are sufficient to provide a chloramphenicol-resistant phenotype. However, these limited amounts could not be visualized by SDS–PAGE analysis and staining with Coomassie Brilliant Blue R-250, resulting in an apparent 0% soluble CAT score in the last column. As % merely represents a detection level, 0% should not be interpreted as chemical concentration.

should have the required minimal length for productive folding, whereas CATIII(1–213), which is truncated by only one extra residue, should aggregate quantitatively into inclusion bodies. This was indeed observed when producing these mutants (Table I). Subsequently, Cys214 was replaced with a Ser, Arg or Gly residue in a CATIII(1–214) deletion mutant. These three substitutions were chosen because they had all scored negative for obtaining soluble, biologically active CAT in previous CATI randomization experiments (Van der Schueren et al., 1996). In addition, considering the effect on thermal inactivation kinetics of CATI mutants with substitutions at position 214, the Ser for Cys substitution caused the smallest effect, whereas Cys214Arg had one of the most pronounced negative effects among the 14 characterized mutants (Van der Schueren et al., 1996). The three substitutions could be introduced using a single synthetic DNA cassette with one degenerate base position (cassette 2 in Figure 1). Observations on these CATIII mutants (Table I) are in perfect agreement with those on CATI deletion mutants (Van der Schueren et al., 1996). Carboxy-terminal truncation of CATIII down to residue 213 indeed resulted in a loss of biological activity, whereas ~20% of the CAT polypeptides produced acquired an enzymatically active conformation when Cys214 was preserved. Based on the capacity to provide chloramphenicol resistance to the host, the Ser variant appeared to be the least deleterious among the truncation mutants with a unique substitution at position 214. This was expected considering the similarity of Ser and Cys side-chains (Taylor, 1986). In vivo random mutagenesis of catIII(1–213)[X] To introduce mutations randomly in the cat gene of truncated CATIII mutants, the respective plasmids were transformed to the E.coli mutator strain XL1-Red. Control experiments revealed an average mutation frequency of one base change per 3800 nucleotides (data not shown). This value of the mutagenic potency of XL1-Red is approximately half that described by Greener and Callahan (1994). However, 60% of the analyzed clones contained a mutation in triplet 214, which ought to be due to the chloramphenicol selection imposed during XL1-Red growth in liquid medium. After phenotypic selection for clones with restored chloram1213


phenicol resistance, again, a high mutation frequency at triplet 214 was observed (in 15 out of 22 sequenced cat genes). Reversion to Cys214 on the one hand and replacement of residue 214 by a hydrophobic (Trp, Val) or a Ser residue on the other were successful modifications to regain chloramphenicol resistance. Elongation of the carboxy terminus due to frameshifts was an alternative leading to repair of CAT activity. The only mutation that did not alter the 39 end of the cat gene, but nonetheless was able to restore chloramphenicol resistance, was an A to C transversion in triplet 145 of catIII(1–213)[G]. This base substitution corresponded to a Leu by Phe replacement at the protein level. The resulting CAT mutant was called CATIII(1–213)[G]:L145F. In vitro random mutagenesis of catIII(1–213)[R] To introduce several random mutations per cat gene and to reduce the background of revertants of the first-site mutation, a PCR-based method was used. Random mutagenesis by PCR relies on the inherent infidelity of Taq polymerase (Saiki et al., 1988; Tindall and Kunkel, 1988). We first performed PCR under standard conditions, but with a large number of PCR cycles, following the protocol of Zhou et al. (1991). The real number of DNA duplications in the PCR was estimated to be at least 12. The mutagenesis frequency, however, could not be reproduced and ranged between no observed mutations up to one base substitution per 650 nucleotides. Therefore, the PCR mixture was modified by mixing unequal concentrations of the four dNTPs, and/or adding Mg21 and Mn21 to the PCR buffer, in order to increase the misincorporation rate of Taq polymerase (Fromant et al., 1995). However, the changes in reaction conditions to decrease the reliability of Taq polymerase reduced the yield of amplification product (data not shown). The simultaneous addition to the standard PCR mixture of (i) a two-fold excess of dGTP, 0.2 mM Mn21, 1 mM Mg21 (low Mn21/Mg21) or (ii) a twofold excess of dGTP, 0.3 mM Mn21, 1.5 mM Mg21 (high Mn21/Mg21) gave sufficient amplification product for cloning. Sequence analysis revealed, on average, one mutation per 60 bp for the low Mn21/Mg21 conditions and one mutation per 40 bp for the high Mn21/Mg21 conditions. These misincorporation frequencies were similar to those observed by Fromant et al. (1995), although they used a 17-fold excess of one dNTP relative to the other three dNTPs. After cloning of PCR products, the DNA sequences of a 366 bp cat fragment of 15 recombinant clones derived from different amplification reactions were determined to analyze the kind of mutations that were introduced by PCR (Table II). Randomness in mutagenesis appeared to be obtained both at the positional level and at the level of mutation type. Out of the 80 observed mutations, more than 90% acted at different nucleotide positions (Table II). This clearly demonstrates that the mutations are well distributed along the target sequence. With respect to the nature of mutations, the four transitions and six of the eight possible transversions have been observed on the top strand of the cat gene (Table II). However, a preference of G-C for A-T substitutions was striking. This agrees with the already observed tendency of Taq polymerase to favor this kind of transition (Tindall and Kunkel, 1988; Fromant et al., 1995). A library of more than 30 000 recombinants was screened for survival on chloramphenicol-containing media. With the observed mutation frequencies, this means that every base substitution in the cat gene statistically was checked more 1214

Table II. Mutagenesis efficiency by PCR under non-optimal conditions Parameter

Number

Analyzed clonesa Analyzed triplets per clone Mutations Transitions T→C A→G G→A C→T Transversions A→C A→T T→A T→G G→T C→G G→C C→A Mutated nucleotide positions Silent mutations Mutations resulting in amino acid substitutions

15 122 79 base substitutionsb 1 1 deletion 28 25 4 1 6 5 4 3 2 1 0 0 74 33 43

aA

set of 15 clones obtained in the absence of chloramphenicol selection were subjected to DNA sequence analysis. Three clones with a total of four mutations in the sequenced fragment came from a PCR with increased number of amplification cycles. One clone with a single point mutation was obtained by PCR with a three-fold excess of dGTP relative to the other three dNTPs. Seven clones were derived from reaction mixtures with a twofold excess of dGTP and low Mn21/Mg21 concentrations. The four remaining clones together contained 33 mutations and were obtained after cloning of amplification products from PCR under high Mn21/Mg21 conditions, as described in the Results section. bAn A to G transition in triplet 148 was observed three times and a T to C transition in triplet 168 twice. All other 74 base substitutions were unique.

than 100-fold. However, the majority of base substitutions occurred in a background of, on average, 11 other mutations. This means that there was a fair chance that potent secondsite amino acid substitutions were obscured by lethal thirdsite mutations. Only two positive clones were obtained on chloramphenicolcontaining media. The first one contained nine point mutations, two of which had an effect at the amino acid level: TTA to TTT (Leu145Phe) and TTA to GTA (Leu187Val). The second clone harbored seven point mutations, with ATG to GTG (Met125Thr) and CGC to CTC (Arg214Leu), resulting in amino acid substitutions. Although not studied further, it seems likely that the regain of chloramphenicol resistance of the latter was due to restoration of the primary mutation at position 214 into a hydrophobic residue. Characterization of suppressor mutants The mutant CATIII(1–213)[R]:L145F:L186V contained two amino acid substitutions that may be responsible for regain of biological activity. These two mutations were separated from one another by exchanging SspI fragments (Figure 1) between pUCATIII(1–213)[R] and pUCATIII(1–213)[R]:L145F:L186V, resulting in plasmids pUCATIII(1–213)[R]:L145F and pUCATIII(1–213)[R]:L186V. The Leu145Phe substitution was sufficient for regain of chloramphenicol resistance (Table III). This was also apparent from the growth of the CATIII(1–213)[G]:L145F-producing clone on chloramphenicol-containing medium (see previous section). The Leu186Val substitution, in contrast, had no effect on chloramphenicol resistance characteristics and appeared


Table III. Effects of Leu145Phe substitution in carboxy-terminally truncated and full size CATIII mutants Culture

MIC value (µg/ml)

WK6 WK6[pUCATIII] WK6[pUCATIII(1–213)[R]] WK6[pUCATIII(1–D213)[R]:L145F:L186V]a WK6[pUCATIII(1–213)[R]:L145F] WK6[pUCATIII(1–213)[R]:L186V] WK6[pUCATIII[RS]] WK6[pUCATIII[RS]:L145F] WK6[pUCATIII(1–214)] WK6[pUCATIII(1–214):L145F] WK6[pUCATIII(1–213)] WK6[pUCATIII(1–213):L145F]

2.5 100 2.5 20 20 2.5 100 100 25 75 2.5 15

, , , , , , , , , , , ,

MIC MIC MIC MIC MIC MIC MIC MIC MIC MIC MIC MIC

ø ø ø ø ø ø ø ø ø ø ø ø

5 125 5 25 25 5 125 125 50 100 5 20

Soluble CAT (%) – 90 0 5 15 0 90 90 20 60 0 10

aAfter

a standard incubation time of 16 h, no visible colonies of WK6[pUCATIII(1–213)[R]:L145F:L186V] developed on a standard selective medium (25 µg/ml chloramphenicol). However, prolonged incubation for 24 h was sufficient for visible growth. Under the same conditions, no colonies developed after a 3 day incubation of pUCATIII(1–213)[R] transformants.

to influence negatively the recovery of biologically active CATIII(1–213)[R]:L145F:L186V. The soluble fraction of CATIII(1–213)[R]:L145F was three-fold higher compared with CATIII(1–213)[R]:L145F:L186V (Table III). This observation was confirmed by a three-fold higher CAT activity in crude cell extracts of pUCATIII(1–213)[R]:L145F-transformed cells relative to CATIII(1–213)[R]:L145F:L186V-containing cell lysates with equal total protein concentration (data not shown). To determine the effect of the Leu145Phe substitution on full-length CATIII and pure truncation mutants, the A to T transversion in cat triplet 145 was introduced in plasmids pUCATIII[RS], pUCATIII(1–214) and pUCATIII(1–213). As expected, possible improvements of biological activity or solubility in vivo by the Leu145Phe substitution in CATIII[RS] were not detected using our screening methods (Table III). A sufficiently large amount of full-length protein already adopts efficiently a correct conformation without the side-chain replacement at position 145. Conversely, the values of both MIC and in vivo solubility were increased for the two deletion mutants (Table III), proving that the side-chain replacement at position 145 can offset the negative effects of absence of crucial carboxy-terminal residues. SDS–PAGE analysis confirmed that all CAT mutants were produced in quantities of ~20% of the total amount of cell proteins. As a result, effects on chloramphenicol resistance were not caused by variations in protein synthesis, or by proteolytic degradation of mutant CAT. The observed effects can be attributed to differences in CAT fraction able to attain a soluble, biologically active conformation. Finally, to determine whether the observed higher solubility of CATIII mutants with the Leu145Phe substitution was due to higher stability of the native enzyme, on the one hand, or to effects on the folding pathway, on the other, the resistance against irreversible thermal inactivation was determined (Figure 2). Truncation of CATIII polypeptides down to residue 214 decreased the transition midpoint of irreversible thermal unfolding by ~6°C. In contrast, replacement of Leu145 by Phe increased the 10 min half-life of thermal inactivation by ~5°C. This was observed for both CATIII[RS] and CATIII(1–214),

Fig. 2. Thermal inactivation of CATIII mutants. Remaining enzymatic activity after 10 min heat treatment is given in arbitrary units, relative to the remaining activity obtained after a pre-incubation for 10 min at 60°C. The arrows represent the shift in thermal stability due to the Leu145Phe substitution.

suggesting that stabilizing effects caused by the Leu145Phe substitution were independent of destabilization due to protein truncation. It is therefore assumed that shortening of CATIII(1–214) by one additional residue down to CATIII (1–213) shifts the inactivation profile down by ~4°C. The addition of an Arg residue at the carboxy terminus of CATIII(1–213):L145F appeared to have no effect on the inactivation kinetics. It must be stressed, however, that all mutants quantitatively kept their functional structure after heat treatment at 60°C for 10 min. This shows that stability effects on native protein at the 37°C culture temperature of E.coli are not likely to be the reason for changes in recovery of soluble, mutant CAT. Discussion The importance of residue 214 in folding and/or structure stabilization of CAT was demonstrated previously by mutational analysis of the CATI carboxy terminal (Van der Schueren et al., 1996). To obtain more information on the role of the carboxy terminus on folding and structure stabilization of CAT, attempts were made to renature CAT deletion mutants from inclusion bodies, but so far without success (unpublished data). Controlled carboxypeptidase treatment in vitro of correctly folded purified CAT also failed. Therefore, a genetic approach was followed to investigate further the molecular mechanism of CAT aggregation due to protein truncation. We first demonstrated that mutations causing inclusion body formation of CATI have the same effect in CATIII (Table I). Although both variants have only 46% primary structure identity, it appeared that the structural information encoded at position 214 is important in CATIII also. This suggests a general characteristic within the entire CAT family. The same possibly holds true for variants of the catalytic domain of dihydrolipoyl transacetylase proteins, which are structurally related to CAT (Mattevi et al., 1992). 1215


Fig. 3. Stereo diagram (Kraulis, 1991) of the localization of Leu145 and Cys214 in the CATIII structure. Represented residues have side-chain atoms within 5 Å around Leu145 (amino acids in dark gray) or Cys214 (amino acids in light gray), based on the three-dimensional model of CATIII (PDB identifier: 3CLA).

In a subsequent step, mutations were randomly introduced in the gene of misfolding CATIII mutants to search for secondsite suppressors that circumvent the aggregation problem due to carboxy-terminal protein truncation. In vivo amplification of the target DNA in a mutator strain of E.coli had the disadvantage that second-site suppressor mutations were difficult to detect in the large background of mutants with reversion of the primary mutations. The relative lack of control of mutagenesis frequency was a second drawback. Although the misincorporation frequency during plasmid replication was very high for an in vivo situation, the method introduced only one base substitution per six cat genes. A PCR-based method is recommended when several mutations per gene-sized DNA fragment have to be introduced in a random manner. Because restored folding could easily be observed by selecting for transformants on chloramphenicol-containing medium, a large number of mutants could be screened. Only two real second-site suppressor mutants were isolated, one with each of the two mutagenesis methods. It appeared that the same mutation, Leu145Phe, was responsible for regained chloramphenicol resistance of the clones from both mutagenesis experiments. This indicates that Leu145Phe is one of the very few substitutions—if there is more than one—that can offset the negative effect of the primary missense mutation. This confirms the general view that the number of sites amenable to amino acid substitutions that result in improved protein folding and/or structure stabilization is limited. Note that in CATI, and also in eight other natural CAT variants, amino acid position 145 is already occupied by a Phe residue. However, a hydrophilic residue at the last position of CATI(1–214) does result in inclusion body formation (Van der Schueren et al., 1996). It remains an open question whether or not the global composition of amino acid residues can be optimized in CATI misfolding mutants, in the same way as has been possible for CATIII, so that second-site revertants have restored chloramphenicol resistance potency. Leu145 does not belong to the active site residues (Shaw and Leslie, 1991), so that observed effects can be attributed to structural modifications. The Phe for Leu substitution, although conservative, is able to have an important effect on the partitioning between productive folding on the one hand and misfolding and aggregation into inclusion bodies on the other. Similar effects have been observed with other proteins. For example, Wetzel et al. (1991) found multiple substitutions throughout the human interferon-gamma polypeptide that altered the distribution of heterologous expressed protein between soluble and inclusion body fraction of E.coli cell lysates. Vos et al. (1995) had to create a neighboring second1216

site mutation in the pore-forming domain of colicin A to alleviate the negative effects on correct folding caused by the creation of a hydrophobic cavity. Tsai et al. (1991) identified a global second-site suppressor mutation of temperature-sensitive mutants of human receptor-like protein tyrosine phosphatase that improves thermal stability and folding efficiency. Fane et al. (1991) found conservative substitutions in the tailspike protein of Salmonella typhimurium phage P22 resulting in suppression of temperature-sensitive folding mutations at many sites throughout the polypeptide chain. Leu145 and Cys214 both belong to the hydrophobic core of the protein (Figure 3). The nearest neighbor non-hydrogen atoms of the two residues (Cδ1 and Sγ, respectively) are separated by a 13.28 Å distance. This rather long distance in the native structure suggests that there is no direct interaction in the native structure between residues 145 and 214 that may explain the suppressor effects. Computer modeling studies on CATIII predict that the introduction of a benzyl group at position 145 does not cause major steric hindrance (data not shown). In addition, this substitution probably increases the number of close contacts within the hydrophobic micro-environment of residue 145, which may explain the increase in stability of the protein. However, although the Leu145Phe substitution improved the resistance against irreversible thermal denaturation of CAT, increased thermal stability of the native state cannot account for the observed biological effects. Formation in vivo of inclusion bodies occurs at temperatures ~40°C below those required to induce irreversible aggregation by heating native protein in vitro. Thus, the inability of E.coli transformants to grow on chloramphenicol-containing media when producing truncation mutants of CAT, on the one hand, and the observed increase in chloramphenicol resistance by the Leu145Phe substitution, on the other, can be fully attributed to changes in folding efficiency, resulting in a smaller and higher fraction, respectively, of polypeptides able to adopt a soluble, enzymatically active conformation in vivo. It can be speculated that the side-chains from residues 145 and 214 come close together in a folding intermediate with non-native structure, permitting suppression of folding defects by direct interaction. However, this seems not very plausible because the Leu145Phe substitution can relieve the negative effects of absence of the Cys214 side-chain, as observed in CATIII(1–213)[G]:L145F and CATIII(1–213):L145F (Table III). Alternatively, residues 145 and 214 may be important for initiation of protein folding. A correlation between residue conservation and its participation in the formation of a folding nucleus was recently suggested by Shakhnovich et al. (1996).


Both residues 145 and 214 and surrounding residues are functionally conserved among the natural sequences homologous to CATIII (data not shown). However, following the conservation parameters of CATIII residues, which are included in the database of homology-derived secondary structure of proteins (3CLA.HSSP; Sander and Schneider, 1991), neither amino acid position 145 nor position 214 meets the high conservation criterion of Shakhnovich et al. (1996). Nevertheless, unfavorable substitutions or residue deletions at position 214 increase aggregation, either by pure kinetic effects or by destabilization of an intermediate on the folding pathway. Our data cannot distinguish among these possibilities. Such negative effects of alterations at the carboxy terminal on proper folding can be offset by well defined, but minor changes at a distant site in the hydrophobic core, such as the Phe for Leu substitution at position 145. This suggests that the global information stored in folding intermediates of CAT (e.g. the total hydrophobic content) is more important to direct folding compared with local structural contributions of individual residues. In conclusion, we have identified residues in CATIII with a role in protein folding. Considering the competition in vivo between inclusion body formation and correct folding, the suppressor activity of the Leu145Phe substitution in CATIII misfolding proteins can be explained by favoring the latter pathway. At the molecular level, this can be caused by structural modification of aggregation-prone sites in folding intermediates or by purely kinetic effects making the aggregation-sensitive state less populated.

Sander,C. and Schneider,R. (1991) Proteins, 9, 56–68. Serghini,M.A., Ritzenthaler,C. and Pinck,L. (1989) Nucleic Acids Res., 17, 3604. Shakhnovich,E., Abkevich,V. and Ptitsyn,O. (1996) Nature, 379, 96–98. Shaw,W.V. (1975) Methods Enzymol., 43, 737–755. Shaw,W.V. and Leslie,A.G.W. (1991) Annu. Rev. Biophys. Biophys. Chem., 20, 363–386. Suelter,C.H. and DeLuca,M. (1983) Anal. Biochem., 135, 112–119. Taylor,W.R. (1986) J. Theor. Biol., 119, 205–218. Tindall,K.R. and Kunkel,T.A. (1988) Biochemistry, 27, 6008–6013. Tsai,A.Y.M., Itoh,M., Streuli,M., Thai,T. and Saito,H. (1991) J. Biol. Chem., 266, 10534–10543. Van der Schueren,J., Robben,J., Goossens,K., Heremans,K. and Volckaert,G. (1996) J. Mol. Biol., 256, 878–888. Vos,R., Engelborghs,Y., Izard,J. and Baty,D. (1995) Biochemistry, 34, 1734–1743. Watanabe,N., Kamei,S., Ohkubo,A., Yamanaka,M., Ohsawa,S., Makino,K. and Tokuda,K. (1986) Clin. Chem., 32, 1551–1554. Wetzel,R., Perry,L.J. and Veilleux,C. (1991) Bio/Technology, 9, 731–737. Zell,R. and Fritz,H.-J. (1987) EMBO J., 6, 1809–1815. Zhou,Y., Zhang,X. and Ebright,R.H. (1991) Nucleic Acids Res. 19, 6052. Received April 28, 1998; revised July 31, 1998; accepted August 17, 1998

Acknowledgements We gratefully thank W.V.Shaw (University of Leicester, UK) for providing the vector pUC18:IM3:ClaI. We thank B.Wroblowski and Y.Engelborghs (Katholieke Universiteit Leuven, Belgium) for help with the computer modeling work. J.V.D.S. is a postdoctoral fellow of the Fund for Scientific Research—Flanders (Belgium) (F.W.O.).

References Blum,P., Velligan,M., Lin,N. and Matin,A. (1992) Bio/Technology, 10, 301–304. Fane,B., Villafane,R., Mitraki,A. and King,J. (1991) J. Biol. Chem., 261, 11640–11648. Fromant,M., Blanquet,S. and Plateau,P. (1995) Anal. Biochem., 224, 347–353. Goldenberg,D.P. (1988) Annu. Rev. Biophys. Biophys. Chem., 17, 481–507. Greener,A. and Callahan,M. (1994) Srategies, 7, 32–34. Kraulis,P.J. (1991) J. Appl. Crystallogr., 24, 946–950. Lae¨mmli,U.K. (1970) Nature, 227, 680–685. Leslie,A.G.W. (1990) J. Mol. Biol., 213, 167–186. Leslie,A.G.W., Moody,P.C.E. and Shaw,W.V. (1988) Proc. Natl Acad. Sci. USA, 85, 4133–4137. Lewendon,A., Murray,I.A., Kleanthous,C., Cullis,P.M. and Shaw,W.V. (1988) Biochemistry, 27, 7358–7390. Mandel,M. and Higa,A. (1970) J. Mol. Biol., 53, 154–162. Mattevi,A., Obmolova,G., Schulze,E., Kalk,K.H., Westphal,A.H., de Kok,A. and Hol,W.G.J. (1992) Science, 255, 1544–1550. Mitraki,A. and King,J. (1989) Bio/Technology, 7, 690–697. Murray,I.A., Hawkins,A.R., Keyte,J.W. and Shaw,W.V. (1988) Biochem. J., 252, 173–179. Robben,J., Massie,G., Bosmans,E., Wellens,B. and Volckaert,G. (1993a) Gene, 126, 109–113. Robben,J., Van der Schueren,J. and Volckaert,G. (1993b) J. Biol. Chem., 268, 24555–24558. Robben,J., Van der Schueren,J., Verhasselt,P., Aert,R. and Volckaert,G. (1995) Protein Engng, 8, 159–165. Saiki,R.K., Gelfand,D.H., Stoffel,S., Scharf,S.J., Higuchi,R., Horn,G.T., Mullis,K.B. and Erlich,H.A. (1988) Science, 239, 487–491. Sambrook,J., Fritsch,E.F. and Maniatis,T. (1989) Molecular Cloning. A Laboratory Manual. 2nd edn. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY.

1217