Rapid Evolution of β-Glucuronidase Specificity by Saturation Mutagenesis of an Active Site Loop

Abstract

Protein engineers have widely adopted directed evolution as a design algorithm, but practitioners have not come to a consensus about the best method to evolve protein molecular recognition. We previously used DNA shuffling to direct the evolution of Escherichia coli β-glucuronidase (GUS) variants with increased β-galactosidase activity. Epistatic (synergistic) mutations in amino acids 557, 566, and 568, which are part of an active site loop, were identified in that experiment (Matsumura, I., and Ellington, A. D. (2001) J. Mol. Biol. 305, 331–339). Here we show that site saturation mutagenesis of these residues, overexpression of the resulting library in E. coli, and high throughput screening led to the rapid evolution of clones exhibiting increased activity in reactions with p-nitrophenyl-β-d-xylopyranoside (pNP-xyl). The xylosidase activities of the 14 fittest clones were 30-fold higher on average than that of the wild-type GUS. The 14 corresponding plasmids were pooled, amplified by long PCR, self-ligated with T4 DNA ligase, and transformed into E. coli. Thirteen clones exhibiting an average of 80-fold improvement in xylosidase activity were isolated in a second round of screening. One of the evolved proteins exhibited a ∼200-fold improvement over the wild type in reactivity (kcat/Km) with pNP-xyl, with a 290,000-fold inversion of specificity. Sequence analysis of the 13 round 2 isolates suggested that all were products of intermolecular recombination events that occurred during whole plasmid PCR. Further rounds of evolution using DNA shuffling and staggered extension process (StEP) resulted in modest improvement. These results underscore the importance of epistatic interactions and demonstrate that they can be optimized through variations of the facile whole plasmid PCR technique. Protein engineers have widely adopted directed evolution as a design algorithm, but practitioners have not come to a consensus about the best method to evolve protein molecular recognition. We previously used DNA shuffling to direct the evolution of Escherichia coli β-glucuronidase (GUS) variants with increased β-galactosidase activity. Epistatic (synergistic) mutations in amino acids 557, 566, and 568, which are part of an active site loop, were identified in that experiment (Matsumura, I., and Ellington, A. D. (2001) J. Mol. Biol. 305, 331–339). Here we show that site saturation mutagenesis of these residues, overexpression of the resulting library in E. coli, and high throughput screening led to the rapid evolution of clones exhibiting increased activity in reactions with p-nitrophenyl-β-d-xylopyranoside (pNP-xyl). The xylosidase activities of the 14 fittest clones were 30-fold higher on average than that of the wild-type GUS. The 14 corresponding plasmids were pooled, amplified by long PCR, self-ligated with T4 DNA ligase, and transformed into E. coli. Thirteen clones exhibiting an average of 80-fold improvement in xylosidase activity were isolated in a second round of screening. One of the evolved proteins exhibited a ∼200-fold improvement over the wild type in reactivity (kcat/Km) with pNP-xyl, with a 290,000-fold inversion of specificity. Sequence analysis of the 13 round 2 isolates suggested that all were products of intermolecular recombination events that occurred during whole plasmid PCR. Further rounds of evolution using DNA shuffling and staggered extension process (StEP) resulted in modest improvement. These results underscore the importance of epistatic interactions and demonstrate that they can be optimized through variations of the facile whole plasmid PCR technique. Adaptive evolution is arguably the most fundamental biological process, although it remains poorly understood. A better understanding of molecular adaptation would facilitate the engineering of enzymes with novel specificities. Theoreticians have proposed that a relatively small number of catalytically inefficient, broad specificity proto-enzymes diversified through gene duplication and adaptive evolution into the multitude of efficient and specialized catalysts present in the contemporary biosphere (2Kacser H. Beeby R. J Mol. Evol. 1984; 20: 38-51Crossref PubMed Scopus (113) Google Scholar, 3Lazcano A. Diaz-Villagomez E. Mills T. Oro J. Adv. Space Res. 1995; 15: 345-356Crossref PubMed Scopus (19) Google Scholar, 4Jensen R.A. Annu. Rev. Microbiol. 1976; 30: 409-425Crossref PubMed Scopus (817) Google Scholar, 5Ycas M. J. Theor. Biol. 1974; 44: 145-160Crossref PubMed Scopus (160) Google Scholar, 6Waley S.G. Comp. Biochem. Physiol. 1969; 30: 1-11Crossref PubMed Google Scholar). This supposition cannot be proven nor refuted through experimentation, but recapitulation of the diversification process in the laboratory would support its feasibility and reveal possible underlying structural mechanisms. The glycoside hydrolases exemplify enzyme diversification. Natural selection has matched the great structural diversity of carbohydrates with a multitude of enzymes that selectively catalyze their cleavage. These enzymes have been classified into 91 families based on amino acid sequence similarity (7Bourne Y. Henrissat B. Curr. Opin. Struct. Biol. 2001; 11: 593-600Crossref PubMed Scopus (359) Google Scholar). The GH-A clan (sometimes called the 4/7 superfamily) comprises >3000 GenBank™ sequences from Families 1, 2, 5, 10, 17, 26, 35, 39, 42, 53, 59, 72, 79, and 86 (8Henrissat B. Davies G. Curr. Opin. Struct. Biol. 1997; 7: 637-644Crossref PubMed Scopus (1400) Google Scholar). All GH-A enzymes retain the same (β/α)8-barrel fold and catalytic mechanism (9Henrissat B. Callebaut I. Fabrega S. Lehn P. Mornon J.P. Davies G. Proc. Natl. Acad. Sci. U. S. A. 1995; 92: 7090-7094Crossref PubMed Scopus (512) Google Scholar) and are thought to have diverged from a common ancestor (10Henrissat B. Callebaut I. Fabrega S. Lehn P. Mornon J.P. Davies G. Proc. Natl. Acad. Sci. U. S. A. 1996; 93: 5674PubMed Google Scholar). Consider the specific example of two well characterized glycoside hydrolases: human β-glucuronidase (GUS, 1The abbreviations used are: GUS, β-glucuronidase; XYL, β-xylosidase; PSQ, β-glucuronidase mutants containing Pro557/Ser566/Gln568; PAF, β-glucuronidase mutants containing Pro557/Ala566/Phe568; pNP-gluc, p-nitrophenyl-β-d-glucuronide; pNP-xyl, p-nitrophenyl-β-d-xylopyranoside; StEP, staggered extension process; LB, Luria broth.1The abbreviations used are: GUS, β-glucuronidase; XYL, β-xylosidase; PSQ, β-glucuronidase mutants containing Pro557/Ser566/Gln568; PAF, β-glucuronidase mutants containing Pro557/Ala566/Phe568; pNP-gluc, p-nitrophenyl-β-d-glucuronide; pNP-xyl, p-nitrophenyl-β-d-xylopyranoside; StEP, staggered extension process; LB, Luria broth. Family 2) and Thermoanaerobacterium saccharolyticum β-xylosidase (XYL, Family 39). These distant homologs catalyze the hydrolysis of similar substrates that differ only in their C5 substituents (carboxylate for β-glucuronides, hydrogen for β-xylosides), but overlap very little in specificity. Their amino acid sequences are too divergent to align, but six conserved active site residues are superimposable (Fig. 1) (11Yang J.K. Yoon H.J. Ahn H.J. Il Lee B. Pedelacq J.D. Liong E.C. Berendzen J. Laivenieks M. Vieille C. Zeikus G.J. Vocadlo D.J. Withers S.G. Suh S.W. J. Mol. Biol. 2004; 335: 155-165Crossref PubMed Scopus (63) Google Scholar, 12Jain S. Drendel W.B. Chen Z.W. Mathews F.S. Sly W.S. Grubb J.H. Nat. Struct. Biol. 1996; 3: 375-381Crossref PubMed Scopus (201) Google Scholar). Our goal is to understand how enzymes like GUS and XYL evolved to become so specific. X-ray crystallography is essential to understanding the structural basis of enzyme specificity, but the complexity of protein structures necessitates the development of complementary functional approaches. We randomly mutate genes or parts of genes, and express the resulting libraries in populations of microorganisms. High throughput screening enables the functional evaluation of thousands of sequence variants in parallel. Iterative cycles of mutagenesis, point mutant recombination, and high throughput screening lead to the accumulation of beneficial mutations in selected clones (13Rowe L.A. Geddie M.L. Alexander O.B. Matsumura I. J. Mol. Biol. PubMed Scopus Google Scholar). evolution has been used to of enzymes specificity in T. Curr. Opin. 2001; PubMed Scopus Google Scholar, H. Chen Curr. Opin. PubMed Scopus Google Scholar, C. 2001; PubMed Scopus Google Scholar, Curr. Opin. Biol. PubMed Scopus Google Scholar, S.W. 2001; PubMed Scopus Google Scholar). The Escherichia coli GUS is a for the of adaptive the human GUS protein has been the amino acid sequences of the E. coli and human GUS homologs are with conserved active S. Drendel W.B. Chen Z.W. Mathews F.S. Sly W.S. Grubb J.H. Nat. Struct. Biol. 1996; 3: 375-381Crossref PubMed Scopus (201) Google Scholar). The E. coli β-glucuronidase gene is to directed it can be high in E. coli, the of catalytic activities in reactions with and We have previously directed the evolution of GUS variants with increased β-galactosidase activity. We a library of randomly in a coli exhibiting increased β-galactosidase activity on containing were two rounds of DNA shuffling recombination, Proc. Natl. Acad. Sci. U. S. A. PubMed Scopus Google Scholar) and clones exhibiting in β-galactosidase activity were All clones were and the fittest 13 We the of these mutations through mutagenesis and that amino acids in an active site for the in specificity I. J. Mol. Biol. 2001; PubMed Scopus Google Scholar). Here we a mutagenesis J. Biochem. PubMed Scopus Google Scholar, B. J. D. M. Protein 15: PubMed Scopus Google Scholar, J. Mol. Evol. PubMed Scopus Google Scholar) to direct the evolution of GUS variants with increased XYL activity. We residues that were identified in by saturation mutagenesis called mutagenesis, J. Mol. Evol. PubMed Scopus Google Scholar, D.J. Nat. 20: PubMed Scopus Google Scholar, G. B. G. H. Chen P. J. PubMed Scopus Google Scholar, J. Mol. Biol. 1996; PubMed Scopus Google Scholar, S. A. R. A. S. E.C. J. Nat. 20: PubMed Scopus Google Scholar, PubMed Scopus (63) Google Scholar). This in a round of directed The selected clones were evolved through mutagenesis and point recombination and high throughput screening of evolution The library in a for round of The average selection in over the for in whole xylosidase We of the average of clones that exhibited improvement over its and by the xylosidase activity of the wild of selection shuffling in a of the the were previously (13Rowe L.A. Geddie M.L. Alexander O.B. Matsumura I. J. Mol. Biol. PubMed Scopus Google Scholar). The (Fig. 2) were by the from The the and from amino acids 557, 566, and were by whole plasmid PCR 1997; PubMed Scopus Google Scholar) using a of and S. C. R. Proc. Natl. Acad. Sci. U. S. A. PubMed Scopus Google Scholar) and and is or G. The long PCR reactions of the of and The reactions were with and to in a for a of the and the were to for by cycles of The for and The and the PCR with and for to the The PCR using the PCR as directed by the with the in of The DNA with enzyme and the PCR using a as directed by the The of DNA of we of DNA and high of T4 DNA of PCR were with of T4 DNA and of T4 DNA in reactions containing and for The T4 DNA for and DNA by Google Scholar). E. coli were transformed by as by Res. PubMed Scopus Google Scholar). The transformed were into and in high throughput M.L. L.A. Alexander O.B. I. M. 2004; PubMed Scopus Google Scholar) using as the with the were and characterized in a whole activity with high xylosidase activity in were were and to saturation in a of the in of in a and the for in and the average were The activity of to that of the 14 plasmids selected in the round of evolution were pooled, and as using and The with enzymes and the in the and The similar to that that we used a of T4 DNA of T4 DNA and and a of E. coli transformed with the plasmid and the were as DNA isolated in the second round of evolution were and amplified in a PCR using and of and of The with and for and for 2 The resulting PCR to the PCR and randomly by DNA shuffling Proc. Natl. Acad. Sci. U. S. A. PubMed Scopus Google Scholar). We the PCR with of in for 2 The DNA in and The were in a with The products were amplified in the PCR as and using overlap PCR as B. 1996; PubMed Scopus Google Scholar). The and part of the gene were amplified in a long PCR S. C. R. Proc. Natl. Acad. Sci. U. S. A. PubMed Scopus Google Scholar) with and the mutants and the were in a long overlap PCR using and The resulting PCR and E. coli transformed with the plasmid and the were for clones exhibiting increased XYL activity. from clones isolated in the round of evolution were and of the were amplified in PCR that cycles with extension were This that can by H. Nat. PubMed Scopus Google Scholar). The of the gene amplified with and the amplified with and and the amplified with and The products and amplified in an overlap PCR using and The amplified with and The and were in a long overlap PCR using and The resulting library and E. coli were transformed with the resulting and the were for clones with XYL activity. DNA evolved were using the the for and The of evolved using the and were in their using the and Protein and GUS protein to an and the enzymes and were to by using The protein using the protein to of protein were with of from to in and the of the in a The were by the to the I. Nat. PubMed Scopus Google Scholar). of the in the is an average of of the evolved wild type wild in a mutants in and using the to direct the evolution of GUS variants with XYL activity. The plasmid for these a for a (13Rowe L.A. Geddie M.L. Alexander O.B. Matsumura I. J. Mol. Biol. PubMed Scopus Google Scholar). We gene site saturation mutagenesis to the sequences of 557, 566, and 568, which were identified in a directed evolution experiment I. J. Mol. Biol. 2001; PubMed Scopus Google Scholar). We all amino acids in a library containing which is about the throughput of The library by whole plasmid PCR 1997; PubMed Scopus Google Scholar, PubMed Scopus Google Scholar, A. PubMed Scopus Google Scholar) A. and I. using containing the is or The PCR and transformed into E. coli E. coli and its not XYL activity Curr. Microbiol. 1997; PubMed Scopus Google Scholar). We GUS mutants in a high throughput M.L. L.A. Alexander O.B. I. M. 2004; PubMed Scopus Google Scholar). We used a to transformed into well that of the an average of in of a were with transformed with the The were with and in an for The to saturation these and the GUS proteins were high The were and of in were to The were a that the into the of the for The XYL activity with of the in a that all XYL activity with the than the that the A the of the containing the most XYL activity. The of the GUS mutants exhibited XYL activity than the (Fig. We were not all of the clones were to mutations in active site The fittest clones were characterized in a to demonstrate that the in xylosidase activity were of were in a in with of the of the with for The activity of by the of The improvement exhibited by of selected clones not the corresponding plasmids mutations that GUS to become These clones were The average XYL activity of the 14 clones 30-fold higher than that of the The best improvement in and we that mutagenesis of the whole gene mutants with only in β-galactosidase activity I. J. Scholar). The higher in the support the of saturation mutagenesis of mutants isolated in round of of isolated in the round of screening in its is to a that evolved to and is a of selection by with the of by in a and by the wild of isolated in the round of screening in its is to a that evolved to and is a of The of isolated in the round of screening in its is to a that evolved to I. J. Mol. Biol. 2001; PubMed Scopus Google and is a of The selection by with the of by in a and by the wild type in a of evolved The fittest clones from of rounds of directed evolution in were isolated from and in a of on of of the were with in for of the were by the so that the wild-type on would be The of of the 14 selected genes were This the and about of the catalytic S. Drendel W.B. Chen Z.W. Mathews F.S. Sly W.S. Grubb J.H. Nat. Struct. Biol. 1996; 3: 375-381Crossref PubMed Scopus (201) Google Scholar). The 14 selected clones of amino acids 557, amino acids 566, and amino acids These that a of amino that differ in and can be in the active site The relatively of the mutations in the fittest that are epistatic in We mutations in than 557, 566, and This to a similar to the for long PCR reactions using the and S. Res. 1995; PubMed Scopus Google Scholar). a of 14 and is to the with increased β-galactosidase activity that evolved in laboratory I. J. Mol. Biol. 2001; PubMed Scopus Google Scholar). in the library very a of a We not and not from the not the and only mutations in 557, 566, and rounds of evolution and beneficial mutations in residues than 557, 566, and 568, so we randomly the selected genes by PCR. The amplified in a using a of and S. C. R. Proc. Natl. Acad. Sci. U. S. A. PubMed Scopus Google Scholar) and The PCR with T4 DNA and self-ligated with T4 DNA E. coli transformed with the plasmid library and into The were and with for a the clones exhibiting the most improvement in XYL activity were in with the in the fittest from the round of and 13 (Fig. these 13 clones isolated in the second round of evolution an 80-fold improvement in xylosidase activity over its wild-type We the of of the selected and in to round 1, that only two sequence from the second round of called and of 13 round 2 isolates PAF, and the All of the sequence selected in the round of evolution were to by the and (Fig. The not the round of but most the 13 clones selected in the second The round and cannot be into with a so the clones through recombination and during whole plasmid PCR. during C. Res. PubMed Scopus Google Scholar) and long S. Microbiol. PubMed Scopus Google Scholar) PCR through of The by a isolated in the round of screening but mutations in residues than 557, 566, and 568, in the of the gene of and two of second round not of the We that the mutations were and that the round 2 clones were products of two recombination events during long PCR. all 13 of the clones selected in the second round were through recombination than The sequence analysis of the clones selected in the second round the of recombination in laboratory evolution to that of mutagenesis (13Rowe L.A. Geddie M.L. Alexander O.B. Matsumura I. J. Mol. Biol. PubMed Scopus Google Scholar, Proc. Natl. Acad. Sci. U. S. A. PubMed Scopus Google Scholar). The recombination with PCR, is relatively modest C. Res. PubMed Scopus Google Scholar). We higher recombination and DNA shuffling Proc. Natl. Acad. Sci. U. S. A. PubMed Scopus Google Scholar) to the mutations 557, 566, 568, and in the catalytic We used overlap PCR to the library into the B. 1996; PubMed Scopus Google Scholar). The library transformed into E. coli and in as and clones were for activity in reactions with The that clones exhibited XYL but on average the improvement over the second round relatively modest improvement over the with an 80-fold average improvement round 2 and The DNA shuffling of point mutations is with a Proc. Natl. Acad. Sci. U. S. A. PubMed Scopus Google which to an average of mutations the we randomly the mutations using the H. Nat. PubMed Scopus Google Scholar). The isolated in the round of evolution were and were using with a extension (Fig. The products were and a amplified in an overlap PCR using to the The library into the by overlap PCR as The library transformed into E. coli. The library and for clones with XYL activity. The that exhibited ∼200-fold improvement over the wild the were in to isolated in the round (Fig. of selected the most clones for and the fittest and the fittest a with activity to the average of all and the ancestor of the two These clones and the were in E. coli and to by of the enzymes with of or The second which is a of the of of by an enzyme into a J. Scopus Google Scholar) of the fittest exhibited a ∼200-fold improvement in in reactions with a for over pNP-gluc, in to the of the wild-type enzyme for the mutations in which is wild a in specificity with to the wild-type The variants were and their in reactions with and were like a for over pNP-gluc, its in reactions with is 80-fold higher than the wild-type the fittest in the of the whole but the corresponding enzyme xylosidase activity (kcat/Km) than enzyme These results demonstrate that is not the of in the whole and all the but and have has the of all enzymes but the has two and and mutations six and These mutations in are with to in the whole and the in xylosidase so it is recombination and high throughput screening led to their The of enzymes and the and over rounds of The of in reactions with is higher than that of but its is The two enzymes are similar in specific activity the is which the in the high throughput we used higher it that the mutations would have become we used the would have its into the and of mutations 557, 566, and can in XYL activity round All and PSQ, were into in the second round of We to the of these occurred through or through adaptive the all of the were similar in and the and were amplified by the adaptation the and were selected they were beneficial and in We these by using mutagenesis to the of and of wild-type E. coli these mutant proteins were that the protein not but enzyme exhibited XYL activity. These results in that the and The GH-A glycoside hydrolases are to a common (β/α)8-barrel fold (9Henrissat B. Callebaut I. Fabrega S. Lehn P. Mornon J.P. Davies G. Proc. Natl. Acad. Sci. U. S. A. 1995; 92: 7090-7094Crossref PubMed Scopus (512) Google Scholar). The active site is the by the of the and is of residues from of the that the to the amino acids in the active site can be (Fig. The and substrates differ only in their C5 so it is to that a amino acid in GUS would be to it into a of enzyme specificity have been R. H. T. PubMed Scopus Google but the functional of the GH-A enzymes The of the active site that the superimposable residues 2, and are divergent is the most divergent part of the active site and is most for the C5 specificity of GH-A The residues that we are part of that the of the hydrogen with on and 1) in the wild-type GUS I. J. Mol. Biol. 2001; PubMed Scopus Google Scholar). The of the T. saccharolyticum β-xylosidase (11Yang J.K. Yoon H.J. Ahn H.J. Il Lee B. Pedelacq J.D. Liong E.C. Berendzen J. Laivenieks M. Vieille C. Zeikus G.J. Vocadlo D.J. Withers S.G. Suh S.W. J. Mol. Biol. 2004; 335: 155-165Crossref PubMed Scopus (63) Google Scholar) interactions C5 of the and and A of the E. coli β-galactosidase Family 2) to a A. J.D. Withers S.G. 2001; PubMed Scopus Google Scholar) that the with and a of the and mutations with structural the active The the second round of and we that it the of We that in a with C5 of pNP-xyl, and that the to a The is We that and a hydrogen the of These mutant residues not with so the of the with is higher than the corresponding for the The of the active results in a higher The most efficient enzymes and not mutations than residues 557, 566, and We that the mutations that in and are or in We previously whole gene mutagenesis and DNA shuffling to direct the evolution of E. coli GUS variants with increased β-galactosidase activity. of clones isolated in the round of screening were in of 557, or and of the mutations or in a in β-galactosidase activity The is in but a in in the of I. J. Mol. Biol. 2001; PubMed Scopus Google Scholar). we 557, 566, and and isolated 14 clones that were on average 30-fold in the of whole plasmid PCR of the resulted in recombination and the of variants on average than the have mutagenesis and screening to residues that specificity J. Biochem. PubMed Scopus Google Scholar, B. J. D. M. Protein 15: PubMed Scopus Google Scholar) or J. Mol. Evol. PubMed Scopus Google Scholar). These residues were and high throughput screening of the resulting libraries led to the of novel sequence These led to relatively modest in the from randomly and a exhibiting a in activity A the resulting library but of the second clones improvement over the B. J. D. M. Protein 15: PubMed Scopus Google Scholar). a the of randomly Sequence variants that on a novel were The fittest exhibited a improvement in with a novel residues were and clones were identified in the of these exhibited catalytic than J. Biochem. PubMed Scopus Google Scholar). evolution of proteins cannot of be We that the of site saturation mutagenesis experiment is the of The residues we to in were based on from rounds of mutagenesis and DNA shuffling I. J. Mol. Biol. 2001; PubMed Scopus Google Scholar). is very that epistatic mutations in a round of whole gene we selected residues based round of evolution 557, or we would not have to for saturation mutagenesis be selected through the analysis of an that has been by Proc. Natl. Acad. Sci. U. S. A. 2001; PubMed Scopus Google Scholar, PubMed Scopus Google Scholar, J. M.L. J. A. Proc. Natl. Acad. Sci. U. S. A. PubMed Scopus Google Scholar). This the practitioners of directed The and mutations have been present in the site saturation mutagenesis but we not these the second and rounds of The of adaptation round of in that mutations that are The most active remains (kcat/Km) than the wild-type T. saccharolyticum β-xylosidase D.J. J. Withers S.G. PubMed Scopus Google Scholar). saturation mutagenesis and screening the to the so in xylosidase activity site saturation We that the of and residues in lead to the evolution of a This for the and the of the active to specificity and catalytic We are We and the of the Matsumura for and for of the We are to for with the molecular and to for with the