Pathway engineering by designed divergent evolution Yasuo Yoshikuni1,2 and Jay D Keasling1,2,3,4 Designed divergent evolution is a proposed protein engineering methodology to redesign enzyme function. The methodology was developed on the basis of the theories of divergent molecular evolution: (i) enzymes with more active and specialized functions have evolved from ones with promiscuous functions; (ii) this process is driven by small numbers of amino acid substitutions (plasticity); and (iii) the effects of double or multiple mutations are often additive (quasi-additive assumption). Thus, in many cases the impact of multiple mutations can be calculated by first determining the effects of a mutation at a single position and subsequently summing these effects using the quasi-additive assumption. In this way, the shape of the fitness landscape of a particular enzyme function can be estimated. The combinations of mutations predicted to yield global optima for desired functions can then be selected and introduced into the enzymes. The methodology has been demonstrated to be very powerful to redesign enzyme function. The use of multiple redesigned enzymes in novel or reconstructed metabolic pathways will enable the production of natural and unnatural products that will find use as pharmaceuticals, agrochemicals and many other applications. Addresses 1 UCSF/UCB Joint Graduate Group in Bioengineering, University of California at Berkeley, Berkeley, California 94720 USA 2 Synthetic Biology Department, Physical Biosciences Division, Lawrence Berkeley National Laboratory, Berkeley, California 94710 USA 3 Department of Chemical Engineering, University of California at Berkeley, Berkeley, California 94720 USA 4 California Institute for Quantitative Biomedical Research (QB3), University of California at Berkeley, Berkeley, California 94720 USA Corresponding author: Keasling, Jay D (
[email protected])
Current Opinion in Chemical Biology 2007, 11:233–239 This review comes from a themed issue on Biocatalysis and biotransformation Edited by Bernhard Hauer and Manfred T Reetz Available online 13th March 2007 1367-5931/$ – see front matter # 2007 Elsevier Ltd. All rights reserved. DOI 10.1016/j.cbpa.2007.02.033
Introduction Over the past decade, significant advances in molecular genetics, an accumulated wealth of sequence information of genes encoding various enzymes and genomes specifying their organization, and innovative technologies to deduce, manipulate and analyze metabolisms have inspired us to design and reconstruct metabolic pathways www.sciencedirect.com
in genetically tractable organisms such as Escherichia coli. Others and we have recently demonstrated successful reconstruction of metabolic pathways enabling one to use microorganisms to synthesize valuable natural products such as terpenoids [1,2,3], polyketides [4] and nonribosomal peptides (NRPs) [5]. Those natural products are produced in a variety of organisms. However, their production suffers from relatively low yields, unculturable or slow growing host organisms, and the lack of genetic tools to manipulate the native producer. Because of their structural complexity, it is often very difficult and costly to synthesize these compounds using traditional chemical synthesis approaches. Hence, given its rapid growth rate, the variety of genetic tools and its capability in producing heterologous proteins, genetically modified E. coli might eventually be proven to be an ideal host for their production. However, the use of genetically engineered E. coli for the synthesis of complex natural products is currently limited to those natural products for which there are known enzymes that produce them. Thus, there is a need to either find these enzymes or better yet design enzymes that will catalyze the production of the desired molecule. Protein engineering is a widely accepted methodology to redesign enzyme function (e.g. substrate specificity, reaction selectivity, thermal stability and several other properties). Currently, the most successful methodology is probably directed evolution [6]. However, the use of directed evolution is limited by the availability or the quality of high-throughput screens (HTSs). In addition, it is often very difficult or impossible to exploit HTSs to find the desired enzymatic properties. With this limitation in mind, alternative protein engineering approaches are needed. Hence, we recently proposed a new designed divergent evolution approach that makes use of a systematic remodeling algorithm [1]. This approach is based on divergent molecular evolution of enzymes, and the methodology does not require HTSs. Here, we review the theories of divergent molecular evolution on which this methodology is based and the application of these theories to redesign enzyme function. Because nature has already created enormous numbers of enzymes responsible for producing natural products with high efficiency and specificity, it is essential to learn how, at the molecular level, large chemical diversities were created through divergent molecular evolution of enzymes.
Promiscuous enzyme function is an important determinant of divergent molecular evolution It is widely believed that enzymes with more active and specialized function evolved divergently from enzymes Current Opinion in Chemical Biology 2007, 11:233–239
234 Biocatalysis and biotransformation
with promiscuous function, and this process is thought to be closely associated with the evolution of metabolic pathways [7–9]. Although several models have been proposed, recent studies showed that the recruitment of single enzymes from other metabolic pathways might significantly drive the evolution of both enzymes and metabolic pathways [7]. It is thought that enzymes with promiscuous functions are initially shared by two distinct metabolic pathways. Although it might not be as efficient as having specific enzymes for each distinct pathway, enzymes with promiscuous function might give organisms novel metabolic capabilities and, thus, render them adaptable to different environments. When the gene is duplicated, one enzyme is free to abandon the role it has in one pathway and to specialize its function for the other pathway, and vice versa, resulting in divergent molecular evolution of enzymes and a mosaic or patchwork evolution of metabolic pathways. If multiple steps in a metabolic network were catalyzed by a series of promiscuous enzymes, although inefficient, this network would be able to produce a large library of natural products. If any of these products were to be captured by positive selection, the metabolic network could then converge on the pathway through divergent evolution of each component enzyme. The group of enzymes related by divergent evolution is called a superfamily [10,11]. Among those, the plant terpene synthase superfamily is an interesting example to consider for divergent molecular evolution, their use in constructed metabolic pathways, and the application of their products as medicines, agrochemicals, fragrances and flavors. Terpene synthases, which share a strikingly similar active site scaffold comprising several a-helices [12], catalyze the formation of various terpene olefins from several different classes of prenyl diphosphates through wide varieties of carbocation rearrangements (Figure 1) [13]. Terpene synthases are classified into subfamilies depending on the length of the hydrocarbon tail of the substrate: geranyl (C10), farnesyl (C15) and geranylgeranyl (C20) diphosphates are enzymatically converted to mono-, sesqui- and di-terpene olefins, respectively. Terpene synthase subfamilies within angiosperms are more closely related to each other than are members in the same subfamily from gymnosperms. In each subfamily, terpene synthases from the same or related species are more closely related to each other than are ones from different species with similar catalytic mechanisms [14,15]. These observations indicate that divergent evolution of terpene synthase subfamilies occurred after angiosperms and gymnosperms separated and that of terpene synthases within each subfamily occurred after the series of subsequent speciation events. Interestingly, almost all terpene synthases show promiscuous function. Among those, g-humulene and dselinene synthases are very promiscuous sesquiterpene synthases that are constitutively expressed in Abies grandis, each catalyzing the formation of at least 52 and Current Opinion in Chemical Biology 2007, 11:233–239
36 sesquiterpenes, respectively [16]. In addition, these enzymes can use geranyl diphosphate as a substrate and catalyze the formation of monoterpenes. Although the specific roles of these enzymes have not been identified, it is thought that they might create chemical libraries that are important in general defense against microbial invasion. By contrast, many other terpene synthases have highly specialized functions and are often found to have very specific roles in the formation of bioactive metabolites. For example, (+)-d-cadinene, vetispiradiene and 5-epi-aristolochene synthases catalyze the first reaction step of phytoalexin (anti-fungal agents) production in various plant species and yield their respective sesquiterpenes with more than 98, 90 and 70% selectivity, respectively [2,17,18]. In addition, these sesquiterpene synthases are known to be expressed upon elicitation [19–21]. Similarly, ent-kaurene synthase produces almost exclusively ent-kaurene, the first intermediate in the synthesis of the diterpene plant hormone gibberellin [22,23]. Thus, it has been suggested that terpene synthases with specific functions have evolved from ones with promiscuous function after having been captured by positive selection. Another similar example is the cytochrome P450 (CYP) superfamily. CYPs comprise a large superfamily of oxidative enzymes that share strikingly similar overall structure [24]. CYPs catalyze monooxygenations in addition to related reactions involving oxidative C–C or C–N bond cleavage, dehydrations, isomerizations and reductions. CYPs are found in all lineages of life and are ubiquitously involved in cellular metabolism, such as the catabolism of carbon sources, detoxification of xenobiotics and biosynthesis of steroid hormones and defensive chemicals, to name a few. CYPs also catalyze regio- and stereo-selective oxidation of terpenoids. Like terpene synthases, these CYPs are generally more closely related within one or related species and, hence, they often catalyze the formation of unique terpenoids in related species. In addition, many CYPs are known to have both broad substrate specificity and reaction selectivity [24]. For example, Taxol, the effective anti-cancer drug, and its related diterpenoids are produced in the Pacific Yew and related yew species. Several CYPs have been isolated from yew species and characterized [25,26]. The primary sequences of these CYPs are closely related, sharing approximately 75% or higher homology. Interestingly, some CYPs derived from yew species show broad substrate specificity. Both taxoid 2a- and 7b-hydroxylases can hydroxylate (+)-taxusin at its C2a and C7b positions, respectively. The products, 2aand 7b-hydroxytaxusins, can be hydroxylated, albeit at a lower rate, by 7b- and 2a-hydroxylases, respectively, to yield 2a,7b-dihydroxytaxusins [25].
Single amino acid substitutions often drive the evolution of enzyme function How is a particular mutation incorporated into enzymes during divergent molecular evolution, selected and fixed www.sciencedirect.com
Pathway engineering by designed divergent evolution Yoshikuni and Keasling 235
Figure 1
The biosynthetic pathways for various terpenoids. Isopentenyl diphosphate (C5; IPP) and dimethylallyl diphosphate (C5; DMAPP), universal precursors for all terpenoids, are synthesized by either the mevalonate pathway or non-mevalonate pathway. One, two or three molecules of IPP are then condensed with DMAPP to yield geranyl (GPP), farnesyl (FPP) and geranylgeranyl diphosphate (C20; GGPP), and these prenyldiphosphates are readily converted to mono-, sesqui- and di-terpenes, respectively, by different classes of terpene synthases. Each terpene is further modified by enzymes, such as cytochrome P450s, to yield many terpenoid products. One such sesquiterpenoid is Artemisinin, an effective anti-malarial drug. Taxol, an effective anti-cancer drug, is an example of a diterpenoid.
in a population? The fixation probability of each mutation is dependent on its fitness effects. The fitness effect of a particular mutation can be advantageous (positively selected and probably fixed in a population), neutral or nearly neutral (genetic drift) or deleterious (negatively selected and probably excluded from a population) [27,28]. For enzymes whose functions are to be optimized, the fitness effects of mutations are likely advantageous. In principle, the larger the advantageous fitness effect of a mutation is, the higher the fixation probability of the mutation becomes. Thus, it is implied that mutations with the largest fitness effects are likely to be integrated first into the enzyme in a population. In fact, many biochemical analyses have shown that a few amino acid substitutions significantly drive the evolution of enzyme function (plasticity). We refer to those amino acid residues primarily involved in alteration of enzyme function as plasticity residues. www.sciencedirect.com
For example, HisA and TrpF catalyze similar reactions in the biosynthesis of the amino acids histidine and tryptophan, respectively. These enzymes share a similar (ba)8barrel structure. It has been shown that a single amino acid substitution of Asp127 to Val in HisA significantly altered its function to TrpF [29]. L-Ala-D/L-Glu epimerase (AEE), muconate lactonizing enzyme II (MLE II) and o-succinylbenzoate synthase (OSBS) are also members of (ba)8-barrel fold-containing enolase superfamily. A single mutation of Asp297 in AEE and Glu323 in MLEII to Gly significantly enhanced their promiscuous OSBS activity (104–106-fold) [30]. Another example is human serum paraoxonase (PON1), which has broad substrate specificity and hydrolyzes aryl esters to organophosphates such as paraoxon. The promiscuous activity of PON1 was significantly improved for four different substrates with a few amino acid substitutions [31]. More recently, we demonstrated that single amino acid Current Opinion in Chemical Biology 2007, 11:233–239
236 Biocatalysis and biotransformation
substitutions can dramatically alter the reaction selectivity of the sesquiterpene g-humulene and (+)-d-cadinene synthases [1,2]. In general, this class of enzymes is known to be extremely plastic [13].
The construction of a fitness landscape of enzyme function with quasi-additive assumption provides an effective strategy to redesign the enzyme functions The cumulative effect of two single mutations to an enzyme’s function is often described as its energetic additivity: DDGX;Y ¼ DDGX þ DDGY þ DDGI where DDGX,Y, DDGX, and DDGY represent the free energy differences between the functions of a wild-type enzyme and enzymes with mutations X and Y, X, and Y, respectively. DDGI represents the change in the free energy of the interaction between X and Y. Thus, the effects of double mutations are determined on the basis of the interaction between X and Y, such as charge–charge, charge–dipole, dipole–dipole, and repulsive and attractive van der Waals interactions between those mutations, relative to the effects contributed by single mutations. The effect of the two mutations can be antagonistic, silent, partially additive, additive or synergistic. The interactions of the double mutations will most likely decrease with the distance between the two mutations, because these intra- and inter-molecular interactions decrease with the distance r between the mutations (by 1/r, 1/r4, 1/r6, and 1/r12 and 1/r6, respectively). In fact, several previous studies have shown that the effects of two or more single mutations to an enzyme are likely additive or nearly additive (as much as 90%) when the mutations are more than 4 A˚ apart in the three-dimensional enzyme structure [32,33]. Thus, it is reasonable to assume that the effects of double mutations or even more mutations are additive (quasi-additive assumption). The fitness landscape model has been proposed to explain adaptive molecular evolution. The fitness can be defined as the objective enzyme properties in a particular experimental condition as a function of protein sequences. Thus, the fitness given by different protein sequences shapes the surface of landscape, where an adaptive walk via mutations and subsequent selection enables one to search for global optima of the desired function [34]. However, it is generally very difficult to draw the shape of the fitness landscape because the possible sequence space is too large. For proteins comprising N amino acid residues, the possible sequence space is 20N. (If N = 100, the sequence space is 20100 10130.) The quasi-additive assumption enables us to approximate the shape of the fitness landscape from a significantly smaller number of data points (Figure 2). On the basis of this idea, we previously redesigned the function of the sesquiterpene ghumulene synthase and constructed seven specific sesquiCurrent Opinion in Chemical Biology 2007, 11:233–239
terpene synthase variants. The fitness landscape for each promiscuous function of g-humulene synthase was calculated on the basis of the mutagenic profiles of 19 amino acid residues (some of which were plasticity residues) located in its active site. The fitness landscape of the g-humulene synthase function successfully enabled us to identify the combinations of mutations to make the promiscuous enzyme function specific [1]. Another interesting example is isopropylmalate dehydrogenase (IMDH). IMDH is involved in leucine biosynthesis and catalyzes the conversion of 3-isopropylmalate to 2-isopropyl-3-oxosuccinate using nicotinamide adenine dinucleotide (NAD) as a co-enzyme. Six amino acids (Asp236, Asp289, Ile290, Ala296, Gly337 and Arg341) are thought to determine the co-enzyme specificity. The substitution of five residues (Asp236Arg, Asp289Lys, Ile290Tyr, Ala296Val and Gly337Tyr) significantly altered the co-enzyme specificity of IMDH from NAD to nicotinamide adenine dinucleotide phosphate (NADP) by a factor of 20 000. The fitness landscape of co-factor use in this enzyme was successfully predicted on the basis of the quasi-additive assumption from a limited number of data points [35]. The quasi-additive assumption should also be useful in redesigning other proteins such as antibodies and receptors, because many mutations involved in protein–protein interactions, DNA–protein interactions and protein stability also additively affect their function [33].
Limitations of pathway engineering by the designed divergent evolution approach Although the designed divergent evolution approach has been shown to be a very powerful protein engineering methodology, several limitations still remain. First, to redesign enzyme function, it is necessary to start with an enzyme that has a detectable level of the desired function. However, this is also the case for other protein engineering methodologies, such as directed evolution. Second, although the fitness landscape can be relatively accurately estimated using the quasi-additive assumption, epistasis (non-additive interaction between mutations) has a very important role at the molecular level. Although both partially additive and synergistic effects are somewhat predictable, it is more difficult to predict silent and antagonistic effects from mutagenic studies. It might be possible to improve the model and, thus, the predictability of non-additive effects of mutations by using structural approaches that incorporate physically realistic atomic force fields. Another potential limitation is pleiotropy. Each mutation affects not only the enzyme function but also many other properties of the enzyme: solubility, foldability, stability, resistance to proteases, efficiencies of transcription and translation, and many others. In addition, unnatural intracellular concentrations of substrates and products might www.sciencedirect.com
Pathway engineering by designed divergent evolution Yoshikuni and Keasling 237
Figure 2
The use of the quasi-additive assumption to estimate the fitness landscape of given sequence spaces. The fitness landscape (in this case, the difference between designed and desired enzyme function) can be estimated from the effects of single amino acid substitutions, xA,n (where A and n represent the type and the position of an amino acid in the protein), on a particular enzyme (this can be wild-type or engineered enzymes). The effects of single amino acid substitutions to a particular enzyme are first measured and profiled. Based on the quasi-additive assumption, the effect of each single mutation can be summed by a fitness function, and the difference between the designed and desired enzyme function can be estimated to draw the fitness landscape. Finally, the combination of mutations representing the global optimum is introduced to the wild-type enzyme.
cause severe growth arrest for the producer. For example, five amino acid substitutions to the TEM family of blactamases are known to jointly increase resistance to cefotaxime (a third generation cephalosporin b-lactam) by a factor of 100 000. The number of possible combinations of mutations is 5! = 120. However, 85% of all possible trajectories (combinations of mutations) were shown to be inaccessible via positive selection [36]. Thus, consideration of pleiotropy resulting from each mutation and incorporation of compensatory mutations might improve the strategy. www.sciencedirect.com
Conclusion The reconstruction of metabolic pathways in genetically tractable organisms provides an alternative strategy to synthesize useful natural products that would be otherwise difficult to produce using traditional methodologies. However, many of the enzymes from the natural producers either do not function in a heterologous host or do not produce the desired product. One method to alter or improve an enzyme’s function is designed divergent evolution, a protein engineering methodology developed on the basis of the theories of divergent molecular Current Opinion in Chemical Biology 2007, 11:233–239
238 Biocatalysis and biotransformation
evolution. The redesigned enzymes can be introduced into the heterologous host so that it can produce the various natural and unnatural products. Although several limitations remain because of the relative difficulties to identify plasticity residues and to predict the epistasis and pleiotropy of mutations, the methodology has been shown to be very powerful. In addition, successful redesign of enzyme function based on these ideas help us to better understand the mechanisms of divergent molecular evolution.
References and recommended reading Papers of particular interest, published within the period of review, have been highlighted as: of special interest of outstanding interest 1. Yoshikuni Y, Ferrin TE, Keasling JD: Designed divergent evolution of enzyme function. Nature 2006, 440:1078-1082. This paper shows the concept of designed divergent evolution and the first practice of this concept to design enzyme function. g-Humulene synthase (HUM) was used as a model enzyme. HUM is known to be the most promiscuous sesquiterpene synthase and catalyzes the formation of at least 52 different sesquiterpenes. On the basis of the homology model of HUM, several plasticity residues located in the active site were identified by site-directed saturation mutagenesis. The mutations were recombined on the basis of the mathematical model, and seven specific sesquiterpene synthases were successfully designed and constructed. These results provide the evidence that the current theories of divergent molecular evolution are feasible in nature. 2.
Yoshikuni Y, Martin VJJ, Ferrin TE, Keasling JD: Engineering cotton (+)-d-cadinene synthase to an altered function: Germacrene D-4-ol synthase. Chem Biol 2006, 13:91-98.
3.
Martin VJJ, Pitera D, Withers S, Newman J, Keasling JD: Engineering a mevalonate pathway in Escherichia coli for production of terpenoids. Nat Biotechnol 2003, 21:796-802.
4.
Pfeifer BA, Admiraal SJ, Gramajo H, Cane DE, Khosla C: Biosynthesis of complex polyketides in a metabolically engineered strain of E.coli. Science 2001, 291:1790-1792.
5.
Watanabe K, Hotta K, Praseuth AP, Koketsu K, Migita A, Boddy CN, Wang CCC, Oguri H, Oikawa H: Total biosynthesis of antitumor nonribosomal peptides in Escherichia coli. Nat Chem Biol 2006, 2:423-428.
6.
Powell KA, Ramer SW, del Cardayre SB, Stemmer WPC, Tobin MB, Longchamp GW, Huisman GW: Directed evolution and biocatalysis. Angew Chemie Int Ed Engl 2001, 40:3948-3959.
7.
Schmidt S, Sunyaev S, Bork P, Dandekar T: Metabolites: a helping hand for pathway evolution? Trends Biochem Sci 2003, 28:336-341.
8.
O’Brien PJ, Herschlag D: Catalytic promiscuity and the evolution of new enzymatic activities. Chem Biol 1999, 6:R91-R105.
9.
Jensen RA: Enzyme recruitment in evolution of new function. Annu Rev Microbiol 1976, 30:409-425.
14. Bohlmann J, Meyer-Gauen G, Croteau R: Plant terpenoid synthases: molecular biology and phylogenetic analysis. Proc Natl Acad Sci USA 1998, 95:4126-4133. 15. Trapp SC, Croteau RB: Genomic organization of plant terpene synthases and molecular evolutionary implications. Genetics 2001, 158:811-832. 16. Steele CL, Crock J, Bohlmann J, Croteau R: Sesquiterpene synthases from grand fir (Abies grandis) – comparison of constitutive and wound-induced activities, and cDNA isolation, characterization and bacterial expression of d-selinene synthase and g-humulene synthase. J Biol Chem 1998, 273:2078-2089. 17. Mathis JR, Back K, Starks C, Noel J, Poulter CD, Chappell J: Pre-steady-state study of recombinant sesquiterpene cyclases. Biochemistry 1997, 36:8340-8348. 18. Martin VJJ, Yoshikuni Y, Keasling JD: The in vivo synthesis of plant sesquiterpenes by Escherichia coli. Biotechnol Bioeng 2001, 75:497-503. 19. Tan XP, Liang WQ, Liu CJ, Luo P, Heinstein P, Chen XY: Expression pattern of (+)-d-cadinene synthase genes and biosynthesis of sesquiterpene aldehydes in plants of Gossypium arboreum L. Planta 2000, 210:644-651. 20. Back K, Chappell J: Cloning and bacterial expression of a sesquiterpene cyclase from Hyoscyamus muticus and its molecular comparison to related terpene cyclases. J Biol Chem 1995, 270:7375-7381. 21. Facchini PJ, Chappell J: Gene family for an elicitor-induced sesquiterpene cyclase in tobacco. Proc Natl Acad Sci USA 1992, 89:11088-11092. 22. Rademacher W: Growth retardants: effects on gibberellin biosynthesis and other metabolic pathways. Annu Rev Plant Physiol Plant Mol Biol 2000, 51:501-531. 23. Reiling KK, Yoshikuni Y, Martin VJJ, Newman J, Bohlmann J, Keasling JD: Mono and diterpene production in Escherichia coli. Biotechnol Bioeng 2004, 87:200-212. 24. Pylypenko O, Schlichting I: Structural aspects of ligand binding to and electron transfer in bacterial and fungal p450s. Annu Rev Biochem 2004, 73:991-1018. 25. Chau M, Croteau R: Molecular cloning and characterization of a cytochrome P450 taxoid 2 a-hydroxylase involved in Taxol biosynthesis. Arch Biochem Biophys 2004, 427:48-57. 26. DeJong JM, Liu YL, Bollon AP, Long RM, Jennewein S, Williams D, Croteau RB: Genetic engineering of Taxol biosynthetic genes in Saccharomyces cerevisiae. Biotechnol Bioeng 2006, 93:212-224. 27. Pal C, Papp B, Lercher MJ: An integrated view of protein evolution. Nat Rev Genet 2006, 7:337-348. 28. Motoo K: The neutral theory of molecular evolution. Cambridge: Cambridge University Press; 1983. 29. Jurgens C, Strom A, Wegener D, Hettwer S, Wilmanns M, Sterner R: Directed evolution of a (ba)8-barrel enzyme to catalyze related reactions in two different metabolic pathways. Proc Natl Acad Sci USA 2000, 97:9925-9930.
10. Glasner ME, Gerlt JA, Babbitt PC: Evolution of enzyme superfamilies. Curr Opin Chem Biol 2006, 10:492-497.
30. Schmidt DMZ, Mundorff EC, Dojka M, Bermudez E, Ness JE, Govindarajan S, Babbitt PC, Minshull J, Gerlt JA: Evolutionary potential of (b/a)8-barrels: functional promiscuity produced by single substitutions in the enolase superfamily. Biochemistry 2003, 42:8387-8393.
11. Gerlt JA, Babbitt PC: Divergent evolution of enzymatic function: mechanistically diverse superfamilies and functionally distinct suprafamilies. Annu Rev Biochem 2001, 70:209-246.
31. Aharoni A, Gaidukov L, Khersonsky O, Gould SM, Roodveldt C, Tawfik DS: The ‘evolvability’ of promiscuous protein functions. Nat Genet 2005, 37:73-76.
12. Segura MJR, Jackson BE, Matsuda SPT: Mutagenesis approaches to deduce structure–function relationships in terpene synthases. Nat Prod Rep 2003, 20:304-317.
32. Mildvan AS: Inverse thinking about double mutants of enzymes. Biochemistry 2004, 43:14517-14520.
13. Caruthers JM, Kang I, Rynkiewicz MJ, Cane DE, Christianson DW: Crystal structure determination of aristolochene synthase from the blue cheese mold, Penicillium roqueforti. J Biol Chem 2000, 275:25533-25539. Current Opinion in Chemical Biology 2007, 11:233–239
33. Wells JA: Additivity of mutational effects in proteins. Biochemistry 1990, 29:8509-8517. 34. Orr HA: The genetic theory of adaptation: a brief history. Nat Rev Genet 2005, 6:119-127. www.sciencedirect.com
Pathway engineering by designed divergent evolution Yoshikuni and Keasling 239
This extensive review discusses the rise and fall of various theories of adaptive evolution throughout history. 35. Lunzer M, Milter SP, Felsheim R, Dean AM: The biochemical architecture of an ancient adaptive landscape. Science 2005, 310:499-501. IMDH is an enzyme involved in leucine biosynthesis, and uses NAD as a coenzyme. Six amino acids (Asp236, Asp289, Ile290, Ala296, Gly337 and Arg341) are known to determine the co-enzyme specificity, and five amino acid substitutions (Asp236Arg, Asp289Lys, Ile290Tyr, Ala296Val, and Gly337Tyr) alter the co-factor specificity of IMDH to NADP by a factor of 20 000. The fitness landscape of co-factor use was successfully predicted
www.sciencedirect.com
on the basis of the quasi-additive assumption from a limited number of data points. These results indicated that the contributions of these residues were effectively additive to IMDH function and co-factor specificity. 36. Weinreich DM, Delaney NF, DePristo MA, Hartl DL: Darwinian evolution can follow only very few mutational paths to fitter proteins. Science 2006, 312:111-114. Five amino acid substitutions in a particular b-lactamase allele jointly increase bacterial resistance to a clinically important antibiotics by a factor of 100 000. In this study, 102 out of 120 mutational trajectories were shown to be inaccessible to Darwinian selection. The results imply that pleiotropy in b-lactamase is responsible.
Current Opinion in Chemical Biology 2007, 11:233–239