Family Shuffling of Expandase Genes To Enhance Substrate Specificity for Penicillin G

Deacetoxycephalosporin C synthase (expandase) from Streptomycesclavuligerus, encoded by cefE, is an important industrial enzymefor the production of 7-aminodeacetoxycephalosporanic acid frompenicillin G. To improve the substrate specificity for penicillinG, eight cefE-homologous genes were directly evolved by usingthe DNA shuffling technique. After the first round of shufflingand screening, using an Escherichia coli ESS bioassay, fourchimeras with higher activity were subjected to a second round.Subsequently, 20 clones were found with significantly enhancedactivity. The kinetic parameters of two isolates that lack substrateinhibition showed 8.5- and 118-fold increases in the k_cat/K_mratio compared to the S. clavuligerus expandase. The evolvedenzyme with the 118-fold increase is the most active obtainedto date anywhere. Our shuffling results also indicate the remarkableplasticity of the expandase, suggesting that more-active chimerasmight be achievable with further rounds.

INTRODUCTION

Top

Introduction

Cephalosporins, which constitute one of the major ß-lactamantibiotic classes, are produced by several microorganisms,such as filamentous fungi, actinomycetes, and gram-negativebacteria. The biosynthesis of cephalosporins in Streptomycesclavuligerus and Acremonium chrysogenum has been studied extensively(19). In S. clavuligerus, deacetoxycephalosporin C synthase(DAOCS) (expandase), encoded by cefE, catalyzes the oxidativeexpansion of the five-membered thiazolidine ring of penicillinN into the six-membered dihydrothiazine ring of deacetoxycephalosporinC (DAOC) (13) (Fig. 1). DAOC is sequentially converted to deacetylcephalosporinC (DAC) by deacetylcephalosporin C synthase (DACS) (hydroxylase),encoded by cefF (12). Both expandase and hydroxylase are ${alpha}$ -ketoglutarateand Fe(II) dependent oxygenases and share 55 to 60% amino acididentity. Interestingly, the enzyme from A. chrysogenum cancatalyze the direct conversion of penicillin N to DAC due topossession of both expandase and hydroxylase activity; hencethe encoded gene is termed cefEF (24).

View larger version (13K):
[in this window]
[in a new window]
FIG. 1. The biosynthesis of cephalosporins from penicillins catalyzed by the expandase (DAOCS) and hydroxylase (DACS).

7-Aminodeacetoxycephalosporanic acid (7-ADCA) is one of themost important semisynthetic cephalosporins, and its currentproduction includes the ring expansion of penicillin G intophenylacetyl-7-ADCA and the well-documented enzymatic removalof the phenylacetyl group. Due to potential pollution by thetoxic chemicals used for the ring expansion, Penicillium chrysogenumand A. chrysogenum have been engineered metabolically to produce7-ADCA derivatives (7, 27). However, the derivatives from thesestrains still need further complicated enzymatic processing.Therefore, the production of a penicillin G-specific expandasewould be a more straightforward solution to the problems of7-ADCA production.

Several expandases have been screened from actinomycetes, andthe S. clavuligerus enzyme has been well characterized. Theseenzymes have broad substrate specificity and can catalyze ringexpansion in penicillins with various side chains. However,they prefer their physiological substrate, penicillin N, ratherthan penicillin G (4). Random mutagenesis and structure-basedmutation analysis have improved the substrate specificity forpenicillin G but with only a limited degree of enhancement (2,3, 17, 18, 28). For example, Wei and colleagues have obtaineda more active S. clavuligerus expandase mutant but only withan increase in the k_cat/K_m value of 32-fold (28). In this study,we first screened several new expandase genes from actinomycetesand enhanced the substrate specificity for penicillin G usingDNA family shuffling. After two rounds, we have generated achimeric expandase with a significantly better k_cat/K_m value,118-fold higher than that of the S. clavuligerus enzyme.

MATERIALS AND METHODS

Top

Materials and Methods

Screening of new expandase genes from soil actinomycetes.
Various soil samples were collected from Fu-Shan Research Station,Ilan; Chatienshan, Taoyuan; and Smangus, Hsinchu, and GreenIsland, Taitung, in Taiwan. Soil suspensions were pretreatedwith sodium dodecyl sulfate (SDS)-yeast extract solution (6%yeast extract and 0.05% SDS) at 40°C for 30 min (9) or with1.5% phenol solution at 30°C for 30 min (10) before beingplated on humic acid-vitamin agar (8) and actinomycete isolationagar (Difco) with 20 µg of nalidixic acid/ml and 200 µgof cycloheximide/ml. The plates were incubated at 30°C for7 to 14 days. The actinomycete isolates were then streaked onISP (International Streptomyces Project) no. 4 medium platesand stored.

For the Escherichia coli ESS assay (4), a piece of agar ( $~$ 0.5cm³) containing each actinomycete isolate was placed on Luriabroth-penicillinase plates seeded with the E. coli ESS strain,a ß-lactam-supersensitive mutant, and the plates wereincubated overnight at 37°C. Formation of a clear zone indicatesthat the actinomycete isolate secretes non-penicillin-type antibiotics,causing bacterial lysis. Expandase-homologous genes from thosenon-penicillin-type strains were then identified by screeningwith two radiolabeled probes (nucleotides 157 to 305 and 487to 611 of S. clavuligerus cefE, generated by incorporation of[ ${alpha}$ -³²P]dCTP, using PCR amplification).

Construction and screening of genomic DNA.
The chromosomal DNA from the selected isolates were purifiedand partially digested with BamHI or Sau3AI. The recovered 2-to 7-kb DNA fragments were ligated into the BamHI site of the ${lambda}$ ZAP Express vector (Stratagene), and the recombinant phagemidswere packaged in vitro by the Gigapack III gold packaging extract(Stratagene) to construct a genomic library. Plaque hybridizationwas performed, and phagemids were excised from positive phageclones as recommended by the manufacturer.

DNA electrophoresis, Southern blotting, and general recombinantDNA techniques were carried out by standard procedures. PlasmidDNA was isolated from E. coli strains using a plasmid miniprepkit (Viogene). All restriction enzymes were purchased from Promega,and digests were purified by using a QIAquick PCR purificationkit (QIAGEN). DNA fragments from agarose gels were recoveredby using the QIAquick gel extraction kit (QIAGEN). Genomic DNAof actinomycetes was prepared as described by Kutchma et al.(15). The oligonucleotides were synthesized by Mission Biotech,Taipei, Taiwan, on an ABI 3900 high-throughput DNA synthesizer.DNA sequencing was performed using an ABI PRISM dye terminatorcycle sequence kit and an ABI 377 DNA sequencer by Mission Biotech.

Identification and phylogenetic analysis of isolates. Taxonomicidentification of actinomycete isolates was deputized to theBioresources Collection and Research Center (Hsinchu, Taiwan).Together with the cefE genes from S. clavuligerus (accessionnumber P18548), Streptomyces jumonjinensis (AAL09460), and Nocardialactamdurans (Q03047), and the cefEF gene from A. chrysogenum(P11935), phylogenetic analysis of our newly identified geneswas performed by the PHYLIP package using the neighbor-joiningmethod (23). Robustness of the tree (100 pseudo-replications)was assessed by the bootstrap and interior branch tests, andthe tree was drawn by using TreeView software (21).

Enzyme-directed evolution.
The previously identified cefE genes from S. clavuligerus (ATCC27064), N. lactamdurans (ATCC 27382), and S. jumonjinensis (ATCC29864), the cefF gene from S. clavuligerus, the cefEF gene fromA. chrysogenum (ATCC 11550), and our newly cloned genes wereisolated by PCR and cloned into the NdeI/HindIII-treated expressionvector, pET24a (Novagen). After amplification of the homologousgenes by PCR, DNase I was used to fragment the genes randomly,and fragments of around 100 bp in size were collected. The 100-bpfragments were added to a primerless polymerase reaction forreassembly, and external primers were added to amplify the recombinantexpandase genes (6, 25), which were then expressed in Tuner(DE3)(Novagen). Conditions for primerless and amplification PCR wereas follows: 5 min at 96°C and 40 cycles of 30 s at 95°C,30 s at 55°C, and 30 s at 72°C, followed by 10 min at72°C; 5 min at 98°C and 20 cycles of 1 min at 95°C,1 min at 55°C, and 1 min at 72°C, followed by 10 minat 72°C. Transformants were cultured in 96-well plates followedby an E. coli ESS bioassay as described previously (28), exceptfor use of a substrate concentration of 10 mM penicillin G.The Tuner(DE3)-based transformant YS16, containing the S. clavuligeruscefE gene, was used as a positive control. For the second shufflinground, the clones with a higher activity than that of YS16 wereused, and one of the clones, F42, was used as the positive control.Finally, the family-shuffled expandases with significantly enhancedactivities were purified by a fast protein liquid chromatographysystem, and their kinetic parameters were analyzed by a high-performanceliquid chromatography system using 20 µg of enzyme and0.1 to 10 mM penicillin G in a total volume of 200 µlof assay mixture as reported previously (28).

Nucleotide sequence accession numbers.
The newly identified cefE genes from S. ambofaciens and S. chartreusisand the cefF gene from isolate 65PH1 have been submitted toGenBank under accession numbers AY318742, AY318743 and AY318744,respectively.

RESULTS

Top

Results

Screening of new expandase genes from soil actinomycetes.
From 181 soil samples, we isolated 1,835 actinomycetes-likecolonies, and most of the isolates were obtained by using thesoil pretreatment of SDS-yeast extract method. Subsequently,133 isolates producing non-penicillin-type antibiotics, as assayedby E. coli ESS bioassay, were selected. Based on Southern blottinganalysis, 14 actinomycete isolates were found to carry expandase-homologousgenes, and these were distributed into four groups. One isolatefrom each group, 102SH3, 29SA4, 29SA13, and 65PH1, was selectedfor library construction. Isolates 102SH3 and 29SA4 were identifiedas Streptomyces ambofaciens and Streptomyces chartreusis, respectively.

Four homologous expandase genes, including three cefE genesand one cefF gene, were finally obtained by screening. The cefEgenes from the S. ambofaciens and S. chartreusis isolates contain936 nucleotides encoding 311 amino acids and have a G+C contentof 63.2 and 66.9%, respectively. The deduced protein sequencesof these two cefEs share 79% identity to each other and show81 and 76% identity with the S. clavuligerus enzyme, respectively(Table 1). Both Streptomyces strains have never been reportedto have the cefE gene. A phylogenetic tree of the cefE genesfrom S. ambofaciens, S. chartreusis, S. clavuligerus, N. lactamdurans,and S. jumonjinensis and the cefEF gene from A. chrysogenumwas then constructed (Fig. 2). Because the evolutionary splitbetween cefE and cefEF is probably ancestral to the evolutionof cefE in the Actinomycetales, we use the fungal cefEF as thetree root. This distance tree was compared to trees producedby PROTPARS and PROML from PHYLIP by using CONSENSE, and theconsensus tree was identical to the distance tree originallyobtained. The cefE genes share 67 to 79% protein sequence identityto each other (Table 1). Furthermore, the presence of the conservedresidues involved in iron ligation and substrate and cosubstraterecognition suggests that these two newly cloned cefE genespossess the expandase activity (Fig. 3).

View this table:
[in this window]
[in a new window]
TABLE 1. Percentages of sequence identities between shuffled genes

View larger version (19K):
[in this window]
[in a new window]
FIG. 2. A consensus tree constructed using the protein sequences of the five DAOCSs and one DAOCS/DACS, with the latter used as the root. Numbers at the nodes indicate statistical support for each branch as a percentile value. Scale bar, 0.1 substitution per site.

View larger version (81K):
[in this window]
[in a new window]
FIG. 3. Protein sequences of the six evolved clones and their parents. The shaded residues show the identity to the parents (S. clavuligerus, red; S. ambofaciens, yellow; S. chartreusis, blue), point mutations (green), and the regions of crossover events (gray). The conserved residues in several Fe(II)-dependent oxygenases (5, 22, 29) for iron ligation, H183, D185, and H243, and for interaction with the carboxylate group of the (co)substrate, R258 and S260, are shaded in magenta, while residues involved in substrate binding are in cyan.

The deduced amino acid sequences of the isolates 29SA4 and 29SA13show 99.6% identity to each other, strongly suggesting thatthey come from the same species. The new cefF-homologous genefrom the isolate 65PH1 has 939 nucleotides and 82% identityto the cefF gene from S. clavuligerus.

Family shuffling and screening.
For the first round of family shuffling, 6,578 clones were screened,and most of them had low or even no detectable activity. Onlyfour clones, F42, F58, F53, and F56, showed a higher activitythan the S. clavuligerus expandase, YS16. The sequences revealedthat F56 and F58 originated by one crossover event, F42 by twosuch events, and F53 by three events (Fig. 3). Sequence elementswere contributed by only three parental genes, namely, genesfrom S. clavuligerus, S. ambofaciens, and S. chartreusis, eventhough eight genes were used for the family shuffling. The constructionelements in F42, F53, and F56 were derived mainly from S. clavuligerus,with only 10 to 30% from S. ambofaciens. However, 80% of F58is from S. ambofaciens, and 20% is from S. chartreusis.

The four clones were reshuffled and rescreened to generate progenywith an activity higher than that of clone F42. During the secondround, 3,612 clones were screened and 20 clones with enhancedactivity were obtained. Two clones, FF1 and FF8, showed no substrateinhibition and were subjected to kinetic and sequence analyses.FF1 was derived mainly from F42 with some sequence elementsfrom S. ambofaciens and S. chartreusis, while FF8 was derivedmainly from F58 and F42 (Fig. 3). In addition, both FF1 andFF8 have five amino acid substitutions.

Kinetic parameters of selected expandases.
We purified the expandases from F42, F58, FF1, FF8, S. ambofaciens,and S. chartreusis and measured their kinetic parameters forpenicillin G expansion (Table 2). The recombinant proteins ofour two newly identified cefE genes showed weak activity towardspenicillin G. With comparison to the S. clavuligerus expandasewhen assayed with 1 mM penicillin G, the S. ambofaciens enzymedisplayed only 14% activity, the S. chartreusis counterparthad a fivefold higher K_m value, while F42 and F58 from the firstshuffling round showed 3.4-fold and a 5.7-fold increases, respectivelyin penicillin G binding. After the second round of shuffling,the kinetic parameters of clones FF1 and FF8 were improved further.FF1 and FF8 had an enhancement in the K_m value by 8.6- and 184-fold,respectively. To the best of our knowledge, the two-round-shuffledFF8 had the highest known k_cat/K_m value, 2,121 (M^–1 s^–1),for penicillin G, which is 118-fold higher than that for theS. clavuligerus expandase.

View this table:
[in this window]
[in a new window]
TABLE 2. Kinetic parameters for penicillin G conversion^a

DISCUSSION

Top

Discussion

After screening of 1,835 actinomycetes from soil, two isolatescontaining new cefE genes were found and identified as S. ambofaciensand S. chartreusis. S. ambofaciens has been shown to producethe macrolide antibiotic spiramycin (1), but no expandase-relatedgene has been reported. The production of cephamycins by S.chartreusis (11) implies the existence of an expandase genein this species. We could not identify promoter regions forthe newly discovered cefE genes; hence these genes might beexpressed as part of an operon, or they might exist as a silentgene within the chromosome. Although the two cefE genes sharehigh homology to S. clavuligerus cefE, their physiological functionsin the original strains need to be addressed further.

Although five cefE genes, two cefF genes, and one cefEF genewere used for DNA family shuffling, the most-active clones fromthe first round were constructed from the S. clavuligerus, S.ambofaciens, and S. chartreusis genes. The sequences of somerandomly selected clones, such as FS1 to FS9 in Fig. 4, demonstratedparts of cefEs from two other sources, indicating that the shufflingpool indeed consisted of five cefE sources. The first-roundclones F42 and F58 showed only a slight increase, less thanthreefold, in the k_cat/K_m value, relative to the S. clavuligerusexpandase. Nevertheless, the two clones, FF1 and FF8, obtainedin the second shuffling carried out with the most active fourclones from the first round showed a significant improvementfor penicillin G expansion. Therefore, the evolution of penicillinG-specific expandase has been successfully directed. These resultsalso indicate the remarkable plasticity of this enzyme, andthis suggests that more-active chimeras might be achievablewith further rounds.

View larger version (61K):
[in this window]
[in a new window]
FIG. 4. Distinct residues among the most active clone, FF8, and the clones FS1 to FS9 with little or even no detectable activity. The secondary structure elements ( ${alpha}$ A to ${alpha}$ E and ß1 to ß15) for the expandase are labeled. The solvent-accessible area for each residue is calculated based on the crystal structure of the S. clavuligerus expandase by using PROCHECK (16) and colored according to the degree of exposure from burial (blue) to exposure (white). The blanks indicate the disordered regions in the crystal structures. Obviously, the structural core is mainly maintained by the hydrophobic packing of ß-strands, and this hydrophobic core is conserved in this Fe(II)-dependent oxygenase superfamily (5, 22, 29).

During screening, most evolved expandases with higher activitiesshowed severe substrate inhibition, which would be a major disadvantagefor industrial applications. However, there is no significantcorrelation between the enzyme activity and the substrate inhibition,because some of the more-active evolved enzymes have no substrateinhibition. This implies that there are at least two penicillinG-binding sites: one is responsible for the expansion activity,whereas the other has the inhibition effect. Our observationof substrate inhibition and the low activities of some evolvedclones might be due to penicillin G preferring the inhibitorysite and thus attenuating activity. Therefore, the penicillinG concentration was increased to 10 mM in the screening mixtureto allow evolution of a higher affinity to the substrate-bindingsite instead of to the inhibitory site. Consistent with ourobservation, crystallographic studies indeed reveal two penicillin-bindingsites, which overlap with each other and clash with the cosubstrate ${alpha}$ -ketoglutarate (26) (Fig. 5).

View larger version (32K):
[in this window]
[in a new window]
FIG. 5. Stereo view of the penicillin G-binding sites. The substrate penicillin G (penG^S) is colored in black, whereas penicillin G in the inhibitory site (penG^I) is in orange. The iron is shown as an orange sphere, the conserved residues in several oxygenases are shown in magenta, and some other protein residues are shown in cyan. This figure was prepared using MOLSCRIPT (14).

Many inactive clones may be due to mutations at the essentialactive-site residues. For example, replacement of the conservedR162 with glutamine in FS1 results in no detectable activitybecause the R162Q mutant has been demonstrated to show a lossof penicillin binding (18) (Fig. 4 and 5). However, we do notknow why the chimera FF8 has such a high penicillin G-bindingaffinity, whereas most other evolved clones, such as FS2 toFS9 in Fig. 4, contain little or even no detectable activity.In order to identify some possible clues, we mapped the distinctresidues among expandases onto the enzyme structures. Theseinclude the naturally occurring substitutions between differentspecies and shuffled replacements in the crossover events andrandom mutations during family shuffling (Fig. 3 and 4). Mostof them are exposed at the protein surface, have no close contactswith other parts of the protein structure, and are far awayfrom the active-site cavity and hence might not significantlyaffect structural integrity and enzyme activity.

In contrast, several residues are buried (Fig. 4) and may involvecompactable side chains for the hydrophobic packing, contributingto part of the penicillin G-binding sites. The hydrophobic edgesof the substrate penicillin G face several hydrophobic regions,and these hydrophobic contacts contribute the major interactionsbetween penicillin G and the expandase (Fig. 5). The phenylacetylring makes close contacts with M73, L158, and F264, the ß-lactamring makes close contacts with M180, I305, and V262, and thethiazolidine ring makes close contacts with I192, L204, andV245. The I305L and I305M mutants, which show 3.9-fold and 3.4-foldreductions in K_m for penicillin G, respectively (28), suggestthat replacement with a leucine or methionine residue enhancesthe hydrophobic contacts. The N304K mutant with an 11.7-foldreduction in K_m (28) and the N304L and R306L mutants with significantlyenhanced activity (3) also suggest that the C-terminal tailis an attractive candidate for enhancement of penicillin G bindingas well as residues around other hydrophobic regions. An exampleof the latter is the residue M188, near the iron ligands H183and D185, which is buried in a hydrophobic pocket made up ofA134, V137, L138, F152, L153, V189, I229, A230, V233, and L265.Our M188V and M188I mutants contain a 3.4-fold and 1.3-foldreduction in K_m for penicillin G, respectively (Wei et al.,unpublished data). In addition, the residue A246, as well asseveral surrounding hydrophobic residues, L178, F201, V202,L204, and A247, is close to the iron ligand H243, which constitutespart of the active-site cavity. Replacement of this residuewith valine in FF8 might result in better steric complementarityand hydrophobic packing (Fig. 6) and therefore might enhancebinding of penicillin G to the substrate site in a manner similarto that for the R306L mutant. In short, it should be possibleto use these predictions in the future as a structural basisfor further mutational analysis.

View larger version (31K):
[in this window]
[in a new window]
FIG. 6. Predicted hydrophobic contacts surrounding V246 in FF8. Based on the enzyme structure, residue A246 (A) and the modeled V246 (B) are colored blue and cyan, respectively, while the surrounding hydrophobic residues are shown in gray. Replacement of A246 with valine might result in better hydrophobic packing. This figure was displayed by using GRASP (20).

In summary, we have obtained two new natural expandase genesfrom actinomycetes by screening and the most active penicillinG-specific expandase known to date by using family shuffling.In the next stage, we will attempt to further improve the enzymeactivity through a combination of DNA shuffling and mutagenesisbased on structural analysis.

ACKNOWLEDGMENTS

We thank A. L. Demain for kindly providing the E. coli ESS strainand R. Kirby for critical reading of the manuscript.

This work was supported by a grant from the Ministry of EconomicAffairs of the Republic of China (89-EC-2-A-17-0285-035) andby the Synmax Biochemical Co., Ltd., Hsinchu, Taiwan.

FOOTNOTES

* Corresponding author. Mailing address: Institute of Biochemistry, National Yang-Ming University, 155 Li-Nong St., Sec. 2, Pei-Tou, Taipei 11221, Taiwan. Phone: (886) 2-2826-7125. Fax: (886) 2-2826-4843. E-mail for S.-H. Liaw: shliaw{at}ym.edu.tw. E-mail for Y.-C. Tsai: tsaiyc{at}ym.edu.tw.

REFERENCES

Top