COCHBI-1035; NO. OF PAGES 8
Available online at www.sciencedirect.com
De novo enzymes by computational design
Hajo Kries1,3, Rebecca Blomberg2,3 and Donald Hilvert1
Computational enzyme design has emerged as a promising
tool for generating made-to-order biocatalysts. In addition to
improving the reliability of the design cycle, current efforts in
this area are focusing on expanding the set of catalyzed
reactions and investigating the structure and mechanism of
individual designs. Although the activities of de novo enzymes
are typically low, they can be significantly increased by directed
evolution. Analysis of their evolutionary trajectories provides
valuable feedback for the design algorithms and can enhance
our understanding of natural protein evolution.
Addresses
1
Laboratory of Organic Chemistry, ETH Zurich, Zurich CH-8093,
Switzerland
2
Corporate R&D Division, Firmenich SA, Geneva CH-1211, Switzerland
Corresponding author: Hilvert, Donald ([email protected])
H.K. and R.B. contributed equally to this work.
3
Current Opinion in Chemical Biology 2013, 17:XXX–XXX
This review comes from a themed issue on Biocatalysis and
biotransformation
Edited by Nicholas Turner and Matthew Truppo
Available online XXX
S1367-5931/$ – see front matter, # 2013 Elsevier Ltd. All rights
reserved.
http://dx.doi.org/10.1016/j.cbpa.2013.02.012
Introduction
Tailoring enzymatic function, a long-standing goal in
chemical biology, could provide significant benefits to
both biomedicine and industrial biocatalysis. Although
Pauling proposed the conceptual requirements for
enzyme design more than 60 years ago [1], selective
stabilization of transition states through design is difficult
to realize in practice. The vastness of protein sequence
space and incomplete understanding of enzyme structure–function relationships have hampered the development of general design strategies.
Catalytic antibody technology is one source of artificial
biocatalysts [2]. Rather than crafting protein structure
from scratch, clonal selection in the immune system is
harnessed to create tailored protein binding pockets in
response to stable transition state analogs or mechanismbased inhibitors. A wide range of chemical transformations has been catalyzed in this way, often with high
selectivity, but the modest rate accelerations achieved by
www.sciencedirect.com
antibodies make them generally poor substitutes for
natural enzymes.
Redesign of existing proteins, for example through semisynthesis, is another approach that has been explored
toward novel enzymatic activities [3,4]. In this strategy,
organic or organometallic cofactors are embedded in a
protein scaffold to afford complexes in which the intrinsic
reactivity of the cofactor is modulated by the steric and
electronic properties of the protein. In a topical example,
a biotinylated Rh(III) metal center bound to an engineered streptavidin was shown to catalyze C–H bond
activation with enantiomeric ratios as high as 93:7 for
the coupling of benzamides and alkenes (Figure 1a) [5].
Due to participation of an engineered glutamate side
chain in catalysis, reaction rates also increased by ca.
100 fold over the free complex.
Computational enzyme design represents a potentially
wider-ranging approach to new enzymes [6]. Powerful
computer programs such as the Rosetta software suite [7–
9], developed in the laboratory of David Baker (University of Washington, Seattle, USA), and ORBIT [10,11]
from Stephen Mayo (California Institute of Technology,
Pasadena, USA) can facilitate the search of protein
sequence space. Although applied only recently to
enzyme design [10,12–14], in silico methods promise to
address many of the limitations of other design technologies. Computer-based screening of protein scaffolds
provides access to structures beyond the antibody fold,
for instance, and active sites can now be sculpted around
realistic transition state models for diverse reactions,
including those lacking biological counterparts.
Initial successes
Computational enzyme design comprises several steps
(Figure 2). First, a target reaction is chosen and an
appropriate catalytic mechanism, including all requisite
functional groups, defined. Second, an idealized active
site that positions the catalytic residues so as to maximize transition state stabilization is modeled quantum
mechanically. These structures, termed ‘theozymes’
[15], are docked into diverse protein scaffolds, and
the resulting complexes are optimized in silico by a
combinatorial search for amino acid substitutions that
improve transition state binding [7,16–18]. Finally,
the designs are filtered on the basis of their catalytic
geometry, transition state binding energy, and shape
complementarity between the designed pocket and
the transition state. Synthesis, production, and characterization of the top ranking designs complete the
procedure.
Current Opinion in Chemical Biology 2013, 17:1–8
Please cite this article in press as: Kries H, et al.: De novo enzymes by computational design, Curr Opin Chem Biol (2013), http://dx.doi.org/10.1016/j.cbpa.2013.02.012
COCHBI-1035; NO. OF PAGES 8
2 Biocatalysis and biotransformation
The first steps toward computationally designed enzymes
were taken a little over a decade ago when the E. coli
protein thioredoxin was successfully converted into a
primitive esterase [10]. Introducing a histidine nucleophile
into an engineered binding site on the protein surface
yielded a catalyst that hydrolyzed p-nitrophenyl acetate
(Figure 1b) with a 180-fold rate acceleration (kcat/kuncat).
Subsequently, artificial biocatalysts were generated for
several additional reactions, including carbon–carbon bond
cleavage via retroaldolization (kcat/kuncat = 102 to 104;
Figure 1c) [14], proton transfer from carbon (kcat/kun2
5
cat = 10 to 10 ; Figure 1d) [12], and a Diels-Alder cycloaddition (kcat/kuncat = 89 M; Figure 1e) [13].
New challenges
Encouraged by these initial successes, computational
designers are turning toward increasingly sophisticated
design challenges. Mimicking the active sites of natural
Figure 1
Semisynthetic metalloenzymes
Ref.
O
O
(a)
N
H
OPiv
[5]
NH
O
O
Rh(III)
O
O
Computationally designed enzymes
R
O
(b)
O
O2N
His-Cys
or His
OH
O
OH
[10,19]
R
O2N
O
OH
O
O
[14,31]
(c)
Lys
O
O
(d)
N
O2N
O2N
N
[12,33]
His-Asp,
Asp, or Glu
O
OH
O
(e)
Ar
N
NH
O
O
Ar
O
O
N
O
O
O
O
NH
[13,32]
(f)
O
O
P
O
O
O
OH
–O
Zn
2+
O
O
P
O
[21]
Challenges for computational design
(g)
–
O
OH
O
O
P
O–
O
CO2–
–O
O
O
P
O–
O
OH
[8]
CO2–
–
O2C
O
(h)
O
CO2–
OH
OH
Current Opinion in Chemical Biology
Representative target reactions for designer enzymes.
Current Opinion in Chemical Biology 2013, 17:1–8
www.sciencedirect.com
Please cite this article in press as: Kries H, et al.: De novo enzymes by computational design, Curr Opin Chem Biol (2013), http://dx.doi.org/10.1016/j.cbpa.2013.02.012
COCHBI-1035; NO. OF PAGES 8
De novo enzymes by computational design Kries, Blomberg and Hilvert 3
Figure 2
transition
state
‡
theozyme
‡
EΨ=HΨ
(1) define
catalytic groups
(2) optimize
geometry
dock
into scaffold
design
improve
packing
(1) characterize
(2) optimize
tailor-made
enzyme
Current Opinion in Chemical Biology
Computational enzyme design methodology. Once a target reaction is chosen, the structure of the rate-limiting transition state is calculated at a high
level of quantum mechanical theory and catalytic groups are provided in silico to stabilize it. The resulting theozyme, which has all side-chains
optimally oriented for catalysis but lacks a protein backbone to hold them in place, is then computationally docked into a protein scaffold from the
Protein Data Bank (PDB). In order to improve packing, side-chains surrounding the active site are subsequently mutated using a computational search
algorithm. Finally, the designs are ranked by a scoring function, and the most promising are produced and kinetically characterized in vitro. Although
first-generation designs typically exhibit low catalytic efficiency, they can be optimized experimentally, for example by directed evolution.
hydrolytic enzymes is one example. In an attempt to
extend the thioredoxin-based esterase designs [10], a
more complex catalytic apparatus consisting of a nucleophilic cysteine–histidine dyad and backbone amides for
oxyanion stabilization was targeted [19]. The resulting
proteins were found to cleave activated ester substrates
by a two-step acylation/deacylation mechanism as programmed. Cysteine acylation was rapid, with rate accelerations up to 4000-fold over background. However,
turnover was limited by slow deacylation of the nucleophilic cysteine. Moreover, crystal structures of several
designs revealed that the histidine had been misplaced,
highlighting difficulties associated with correctly positioning multiple, mutually dependent residues with present-day design software.
range as many de novo designs, showcasing the unique
potential of metal ions. To capitalize on metal ion catalysis
in deliberate fashion, a zinc-based adenosine deaminase
was recently computationally redesigned to hydrolyze
diethyl 7-hydroxycoumarinyl phosphate (Figure 1f)
[21]. Eight mutations engendered the new substrate
specificity, enabling stabilization of a trigonal bipyramidal,
instead of tetrahedral, transition state geometry. Though
the phosphotriesterase activity of the redesigned catalyst is
low (kcat/KM = 4 M 1 s 1), it could be improved some
2500-fold by only three rounds of directed evolution. A
range of other metalloenzymes has been designed, among
them ‘true’ de novo designs of the entire polypeptide
backbone, auguring a bright future for this catalyst class
[22].
Ironically, misplacement of a histidine in a computationally designed Zn(His)4 structural motif opened up a
metal coordination site, leading serendipitously to significant hydrolytic activity [20]. The artificial metalloprotein cleaves p-nitrophenyl acetate and p-nitrophenyl
phosphate with kcat/KM values of 630 M 1 s 1 and
14 M 1 s 1, respectively. This activity is in the same
Limitations
www.sciencedirect.com
Despite evident progress, the palette of reactions catalyzed by computationally designed enzymes remains
limited. Efforts to accelerate a number of relatively
simple transformations have been unsuccessful. For
example, despite much effort, de novo design of a triosephosphate isomerase (TIM; Figure 1g) is still an unsolved
Current Opinion in Chemical Biology 2013, 17:1–8
Please cite this article in press as: Kries H, et al.: De novo enzymes by computational design, Curr Opin Chem Biol (2013), http://dx.doi.org/10.1016/j.cbpa.2013.02.012
COCHBI-1035; NO. OF PAGES 8
4 Biocatalysis and biotransformation
problem [8,23]. Also, attempts to mimic chorismate
mutase (CM; Figure 1h), which promotes a biologically
important Claisen rearrangement, have not yet yielded
proteins with detectable activity (F Richter et al., unpublished data).
Though it is not entirely clear why CMs or TIMs are
difficult to design, it is notable that many designable
reactions, such as benzisoxazole deprotonation and retroaldolization, are susceptible to general acid/base or
nucleophilic catalysis. Almost any base (e.g. triethylamine) will promote the Kemp elimination in an appropriate solvent [24], whereas aldol reactions are accelerated
by small amine catalysts such as proline [25]. Not surprisingly, single basic or nucleophilic protein side chains,
placed in a hydrophobic pocket and thus activated by
medium effects, are also good catalysts. In fact, rudimentary rational design of such constructs can be remarkably
successful [26–28]. An amine-containing b-peptidic foldamer, for example, was designed from simple chemical
principles and shown to accelerate a retroaldol reaction by
more than three orders of magnitude [27]. Symmetry too
has been exploited to generate retroaldol catalysts with
activities comparable to the first generation in silico
designs [28]. In this context, combining rational design
with minimal computation can be effective. An allosterically regulatable Kemp eliminase was engineered
by computationally placing an aspartate or glutamate
residue in the peptide binding cleft of the calcium binding protein calmodulin. The best variant had activity (kcat/
KM = 6 M 1 s 1) comparable to many Rosetta designs
(kcat/KM = 10–160 M 1 s 1), but required far less effort
to generate [29].
What, then, are the benefits of computational design?
Kipnis and Baker addressed this question by comparing
computationally designed Kemp eliminases and aldolases
in a TIM barrel scaffold with randomly generated
libraries containing mutations at the rim of the barrel
[30]. They found that the random libraries yielded a
much smaller fraction of active clones (<1%), most of
which exploited preexisting scaffold residues as functional groups. Nonetheless, the catalytic efficiency of
the hits from the random libraries was similar to that of
Rosetta designs in the same scaffold.
Computational algorithms evidently improve the likelihood of finding active enzymes, but the design protocols
are clearly not robust. Poor predictability originates in the
trade-off between speed and accuracy. Current algorithm
force fields are relatively primitive; for instance, they fail
to consider long-range electrostatic interactions and make
the simplifying assumption that protein scaffolds are rigid
and undergo no large conformational changes upon
mutation. It is unsurprising that these programs are consequently unable to distinguish between active and inactive variants in silico, and that ‘shotgun’ approaches must
Current Opinion in Chemical Biology 2013, 17:1–8
be adopted wherein many designs are screened experimentally to identify a few catalytically competent variants.
Learning from experience
In an attempt to improve the efficacy of current design
methods, the aldol reaction was recently revisited. In the
original study [14], theozymes employing a water molecule as a proton shuttle performed better than designs
based on more complex arrays of functional groups.
Focusing subsequent design efforts on that configuration,
success rates could be increased from 44 to 75% by
introducing an aspartate or glutamate for water positioning and by finer sampling of catalytic and binding pocket
residues [31].
Weak initial activity can sometimes be increased by
further computational design of a lead structure. For
example, the low efficiency of a first-generation DielsAlderase catalyst [13] was improved by subsequent computational remodeling of the protein backbone to minimize active site exposure to bulk solvent and thereby
enhance interactions with the substrates. Because unrestrained backbone remodeling presents automated design
algorithms with an almost intractable number of combinatorial possibilities, human expertise was sought from
the foldit community, an online network of computer
players who fold and design proteins as a part of a
computer game. ‘Crowdsourcing’ in this case yielded a
novel helix-loop-helix motif that enhanced Diels-Alderase activity 18-fold by increasing catalyst affinity for the
diene and dienophile [32].
Failed designs offer the chance to begin again more
intelligently. Rather than discard the large fraction of
inactive proteins that emerge from a typical design campaign, detailed biophysical and structural analysis can
help rationalize failures in mechanistic terms and suggest
strategies for improvement. Such an approach was used
to rescue an unsuccessful Kemp eliminase, which was
reengineered to give one of the fastest de novo enzymes
for this reaction [33]. The redesigned catalyst exhibited
higher kinetic parameters (kcat = 1.7 s–1; kcat/KM =
710 M 1 s 1) than the best Rosetta design selected from
a screen of 59 mostly inactive proteins (kcat = 0.29 s–1; kcat/
KM = 160 M 1 s 1) [12].
How to proceed
While rescuing or improving initial designs may represent
a thrifty rebound process, the ultimate goal remains to
engineer more efficient enzymes from the start. Assuming
that ab initio theozyme geometries are reasonably accurate [34,35], problems often arise from imperfect realization of the idealized active site in available protein
scaffolds. The use of implicit solvent models and neglect
of long-range electrostatic interactions are particularly
crude approximations made by the design software.
www.sciencedirect.com
Please cite this article in press as: Kries H, et al.: De novo enzymes by computational design, Curr Opin Chem Biol (2013), http://dx.doi.org/10.1016/j.cbpa.2013.02.012
COCHBI-1035; NO. OF PAGES 8
De novo enzymes by computational design Kries, Blomberg and Hilvert 5
Improved force fields [36,37], coupled with increased
computing power, can be expected to minimize these
problems. Backbone flexibility [38] and the highly
dynamic nature of protein structures [39] must also be
considered in the design process. Finally, since many
reactions proceed via multiple transition states, as in the
instances of aldol and ester cleavage, multistate design
[40] could also be implemented.
In principle, classical molecular dynamics (MD) and
combined quantum mechanics/molecular mechanics
(QM/MM) simulations should facilitate the search for
suitable active enzymes. Although MD simulations are
too time-consuming at present to be brought into the
design routine, they have been successfully utilized in
scoring and analysis at a late stage of the process
[33,39,41–43]. Indeed, active and inactive Kemp eliminases are distinguishable by differences determined in
MD simulations in key angle and distance relationships in
the respective enzyme–substrate complexes [43]. In the
future, more reliable scoring functions may reduce the
number of proteins that have to be experimentally characterized. As computational power improves, it is probable
that slow but physically more accurate energy functions in
MD programs will also complement faster knowledgebased, empirically oriented design algorithms used today.
Coupling design and experiment
While improved algorithms have the potential to move the
enzyme design field forward, true enzyme-like activities
are likely to remain out of reach for in silico design, at least
for the near future. To generate practically useful catalysts,
experimental optimization methods such as directed evolution must therefore be applied. In principle, any assayable
property of a protein is optimizable by multiple rounds of
mutagenesis and screening [44]. First-generation computationally designed enzymes have proven to be particularly
amenable to this approach as demonstrated by 102-103-fold
increases in activity for several retroaldolases [31], Kemp
eliminases [45–47], and a phosphotriesterase [21]. Generation of an improved Kemp eliminase with a kcat/KM of
570 000 M 1 s 1 is notable in this context (Figure 3) [45].
Figure 3
(a)
Kemp elimination theozyme
(b) Experiment vs. design (R1)
–
Asp/Glu
O
‡
O
H
O
+
N
N
O
–O
H
O
Phe/Trp
log ( (kcat/KM)/kAcO-)
(c)
Ser/Thr/Tyr
(d) Evolved variant (R13)
Directed Evolution
9
8
7
6
Design
R4
R8
R13
Current Opinion in Chemical Biology
Computational design and directed evolution of a Kemp eliminase. (a) A theozyme containing a carboxylate base, an aromatic side-chain for substrate
binding, and a hydrogen bond donor to stabilize developing negative charge in the transition state was used as the template for design. (b) The crystal
structure of the first-generation enzyme (blue) complexed with the transition state analog 5,7-dichlorobenzotriazole shows good qualitative agreement
with the computational design model (green). However, the ligand in the crystal structure (orange) is flipped relative to the designed orientation (pink).
(c) The catalytic efficiency of the starting enzyme (R1) relative to the nonenzymic acetate-promoted reaction was increased by more than two orders of
magnitude over thirteen rounds of mutagenesis and screening (R2–R13). (d) The structure of the best evolved variant R13 (blue) cocrystallized with
benzotriazole (orange) shows conformational changes of active site residues, enhanced desolvation of the catalytic glutamate, and a water molecule in
hydrogen bonding distance to the transition state analog.
www.sciencedirect.com
Current Opinion in Chemical Biology 2013, 17:1–8
Please cite this article in press as: Kries H, et al.: De novo enzymes by computational design, Curr Opin Chem Biol (2013), http://dx.doi.org/10.1016/j.cbpa.2013.02.012
COCHBI-1035; NO. OF PAGES 8
6 Biocatalysis and biotransformation
Detailed characterization of computationally designed
enzymes and their evolutionary descendants can help to
identify the limitations of current design models. For
instance, mechanistic investigations of the retroaldolase
RA60 suggested that binding of substrate in a hydrophobic
pocket proximal to a reactive lysine largely accounted for
the observed 105-fold rate acceleration [48], in accord with
design. However, residues supposed to position a water
molecule for proton shuttling did not show the expected
sensitivity to mutation, suggesting that this design feature
did not work as intended. Illuminating how favorable
mutations increase the activity of these catalysts might
show how such problems can be corrected.
Structural analysis promises to play an increasingly
important role in this context. For example, the crystal
structure of another retroaldolase, RA34, in complex with
a 1,3-diketone inhibitor confirmed the general features of
the design model, while highlighting a few important
differences [49]. The mechanism-based inhibitor covalently modifies the enzyme by forming a Schiff base
intermediate with the catalytic lysine as intended, but
the density assigned to the ligand exhibited considerable
positional heterogeneity within the pocket. Different
binding modes arise — at least in part — because the
lysine adopts more than one conformation. Additionally,
although the substrate naphthyl group binds in the appropriate hydrophobic pocket, it is rotated and translated
relative to its position in the model. Water-mediated
interactions at the active site are also more extensive
than predicted. Together, these observations strongly
suggest that obtaining more active catalysts will require
improved control over substrate binding and better preorganization of the active site.
Several Kemp eliminases and their evolutionary descendents were similarly subjected to structural and biochemical analysis (Figure 3). Cocrystallization with
substituted benzotriazoles, which are close structural
analogs of benzisoxazoles, revealed that the designed
active site residues mediate catalysis and binding but,
surprisingly, not in the predicted mode. The small ligand
was found to adopt unanticipated, and sometimes even
unproductive, binding orientations [33,45,47]. Loose
positioning and ineffective desolvation of the catalytic
base were identified as additional problems. During evolutionary optimization, the floppy starting pockets were
often reshaped, resulting in tighter substrate binding and
improved alignment with respect to the base. In addition,
depending on the design, the efficacy of the base was
enhanced either by fine-tuning its electrostatic microenvironment [47] or minimizing its interactions with water
molecules from bulk solvent [45]. Creating active sites
with tighter, unambiguous substrate binding modes
directly by design is therefore desirable, particularly as
the best evolved variants generally derive from the most
active starting points.
Current Opinion in Chemical Biology 2013, 17:1–8
Conclusion
Recent years have seen continuous evolution of computational design procedures, yet the activities of the
resulting artificial enzymes are more typical of catalytic
antibodies than natural biocatalysts. As computational
algorithms are further fine-tuned and more powerful
computer hardware becomes available, the number
and activity of protein catalysts designed de novo will
certainly increase [50]. In contrast to catalytic antibodies,
computationally designed enzymes have proven to be
highly evolvable [45]. In the future, the powerful
combination of theory and experiment may afford deeper
insights into the tenets underlying the remarkable catalytic efficiencies of natural enzymes as well as practical
catalysts for streamlined reaction sequences [51] and
other applications.
Acknowledgements
Work in the authors’ lab was supported by the Defense Advanced Research
Projects Agency (DARPA), the Schweizer Nationalfonds, and the ETH
Zürich. Fellowships from the Fonds des Verbandes der chemischen
Industrie (to R.B.), the Stipendienfonds der Schweizer Chemischen
Industrie (to H.K.), and the Studienstiftung des deutschen Volkes (to R.B.
and H.K.) are gratefully acknowledged.
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.
Pauling L: Chemical achievement and hope for the future. Am
Sci 1948, 36:51-58.
2.
Hilvert D: Critical analysis of antibody catalysis. Annu Rev
Biochem 2000, 69:751-793.
3.
Kaiser ET, Lawrence DS: Chemical mutation of enzyme activesites. Science 1984, 226:505-511.
4.
Ward TR: Artificial metalloenzymes based on the biotin-avidin
technology: enantioselective catalysis and beyond. Acc Chem
Res 2011, 44:47-57.
5.
Hyster TK, Knörr L, Ward TR, Rovis T: Biotinylated Rh(III)
complexes in engineered streptavidin for accelerated
asymmetric C–H activation. Science 2012, 338:500-503.
A semisynthetic enzyme was created by docking a biotinylated Rh(III)
metal catalyst to a variant of streptavidin containing an engineered
carboxylate residue. The hybrid catalyst promotes asymmetric C–H
activation, forming dihydroisoquinolones from benzamides and alkenes,
with a 100-fold rate acceleration.
6.
Nanda V, Koder RL: Designing artificial enzymes by intuition
and computation. Nat Chem 2010, 2:15-24.
7.
Zanghellini A, Jiang L, Wollacott AM, Cheng G, Meiler J, Althoff EA,
Röthlisberger D, Baker D: New algorithms and an in silico
benchmark for computational enzyme design. Protein Sci
2006, 15:2785-2794.
8.
Richter F, Leaver-Fay A, Khare SD, Bjelic S, Baker D: De novo
enzyme design using Rosetta3. PLoS ONE 2011, 6:e19230.
9.
Meiler J, Baker D: Rosettaligand: protein-small molecule
docking with full side-chain flexibility. Proteins 2006, 65:538-548.
10. Bolon DN, Mayo SL: Enzyme-like proteins by computational
design. Proc Natl Acad Sci U S A 2001, 98:14274-14279.
11. Lassila JK, Privett HK, Allen BD, Mayo SL: Combinatorial
methods for small-molecule placement in computational
enzyme design. Proc Natl Acad Sci U S A 2006, 103:
16710-16715.
www.sciencedirect.com
Please cite this article in press as: Kries H, et al.: De novo enzymes by computational design, Curr Opin Chem Biol (2013), http://dx.doi.org/10.1016/j.cbpa.2013.02.012
COCHBI-1035; NO. OF PAGES 8
De novo enzymes by computational design Kries, Blomberg and Hilvert 7
12. Röthlisberger D, Khersonsky O, Wollacott AM, Jiang L,
DeChancie J, Betker J, Gallaher JL, Althoff EA, Zanghellini A,
Dym O et al.: Kemp elimination catalysts by computational
enzyme design. Nature 2008, 453:190-195.
13. Siegel JB, Zanghellini A, Lovick HM, Kiss G, Lambert AR, Clair JLS,
Gallaher JL, Hilvert D, Gelb MH, Stoddard BL et al.:
Computational design of an enzyme catalyst for a
stereoselective bimolecular Diels-Alder reaction. Science
2010, 329:309-313.
14. Jiang L, Althoff EA, Clemente FR, Doyle L, Röthlisberger D,
Zanghellini A, Gallaher JL, Betker JL, Tanaka F, Barbas CF et al.:
De novo computational design of retro-aldol enzymes. Science
2008, 319:1387-1391.
15. Tantillo DJ, Chen JG, Houk KN: Theozymes and compuzymes.
Theoretical models for biological catalysis. Curr Opin Chem
Biol 1998, 2:743-750.
16. Malisi C, Kohlbacher O, Höcker B: Automated scaffold selection
for enzyme design. Proteins 2009, 77:74-83.
17. Dahiyat BI, Mayo SL: Protein design automation. Protein Sci
1996, 5:895-903.
18. Hellinga HW, Richards FM: Construction of new ligand-binding
sites in proteins of known structure. 1. Computer-aided
modeling of sites with predefined geometry. J Mol Biol 1991,
222:763-785.
19. Richter F, Blomberg R, Khare SD, Kiss G, Kuzin AP, Smith AJT,
Gallaher JL, Pianowski Z, Helgeson RC, Grjasnow A et al.:
Computational design of catalytic dyads and oxyanion
holes for ester hydrolysis. J Am Chem Soc 2012, 134:
16197-16206.
This paper explores the computational design of artificial esterases using
catalytic Cys-His dyads for nucleophilic catalysis and backbone amides
for oxyanion stabilization. Although efficient acylation of the active site
cysteine was achieved, slow hydrolysis of the acyl-enzyme intermediate
limited overall catalytic efficiency, most likely due to misplacement of the
catalytic histidine.
20. Der BS, Edwards DR, Kuhlman B: Catalysis by a de novo zincmediated protein interface: implications for natural enzyme
evolution and rational enzyme engineering. Biochemistry 2012,
51:3933-3940.
29. Korendovych IV, Kulp DW, Wu Y, Cheng H, Roder H, DeGrado WF:
Design of a switchable eliminase. Proc Natl Acad Sci U S A
2011, 108:6823-6827.
30. Kipnis Y, Baker D: Comparison of designed and randomly
generated catalysts for simple chemical reactions. Protein Sci
2012, 21:1388-1395.
The efficacy of computational enzyme design was compared to that of
enzymes from random libraries constructed in the same TIM barrel
scaffold. Although the random libraries yielded fewer clones with Kemp
eliminase and retroaldolase activities, the most active variants compete
well with analogous computational designs in terms of efficiency.
31. Althoff EA, Wang L, Jiang L, Giger L, Lassila JK, Wang Z, Smith M,
Hari S, Kast P, Herschlag D et al.: Robust design and
optimization of retroaldol enzymes. Protein Sci 2012,
21:717-726.
Modifications to the computational design procedure are shown to
increase the fraction of catalytically competent retro-aldolases from
44% to 75% compared to an earlier study. Successful designs were
further optimized by site-directed mutagenesis and directed evolution.
32. Eiben CB, Siegel JB, Bale JB, Cooper S, Khatib F, Shen BW,
Players F, Stoddard BL, Popovic Z, Baker D: Increased DielsAlderase activity through backbone remodeling guided by
foldit players. Nat Biotechnol 2012, 30:190-192.
Players of the online protein folding game foldit were given the task to
improve the poor activity of a Diels-Alderase. A helix–turn–helix motif,
designed as an active site lid, increased activity 18-fold by lowering the
KM for both substrates. A crystal structure of the apo enzyme confirmed
that the remodeled backbone adopts the predicted conformation.
33. Privett HK, Kiss G, Lee TM, Blomberg R, Chica RA, Thomas LM,
Hilvert D, Houk KN, Mayo SL: Iterative approach to
computational enzyme design. Proc Natl Acad Sci U S A 2012,
109:3790-3795.
This study shows that MD simulations and crystal structure analysis of a
failed design can guide the generation of highly active enzymes. After two
iterations of redesign and characterization, a Kemp eliminase was
obtained with a kcat value of 1.7 s–1. A cocrystal structure with a transition
state analog showed that the ligand is flipped relative to the design model
but still able to interact productively with the correctly positioned catalytic
base.
34. DeChancie J, Clemente FR, Smith AJT, Gunaydin H, Zhao Y-L,
Zhang X, Houk KN: How similar are enzyme active site
geometries derived from quantum mechanical theozymes to
crystal structures of enzyme–inhibitor complexes?
Implications for enzyme design. Protein Sci 2007, 16:
1851-1866.
21. Khare SD, Kipnis Y, Greisen P, Takeuchi R, Ashani Y, Goldsmith M,
Song Y, Gallaher JL, Silman I, Leader H et al.: Computational
redesign of a mononuclear zinc metalloenzyme for
organophosphate hydrolysis. Nat Chem Biol 2012, 8:294-300.
A zinc dependent adenosine deaminase was converted into an organophosphate hydrolase by introducing mutations predicted by Rosetta. A
combination of computational design and evolutionary optimization
increased the very low phosphotriesterase activity of the scaffold protein
by more than seven orders of magnitude.
35. Zhang X, DeChancie J, Gunaydin H, Chowdry AB, Clemente FR,
Smith AJT, Handel TM, Houk KN: Quantum mechanical design
of enzyme active sites. J Org Chem 2008, 73:889-899.
22. Lu Y, Yeung N, Sieracki N, Marshall NM: Design of functional
metalloproteins. Nature 2009, 460:855-862.
37. Acevedo O, Jorgensen WL: Advances in quantum and
molecular mechanical (QM/MM) simulations for organic and
enzymatic reactions. Acc Chem Res 2010, 43:142-151.
23. Check Hayden E: Chemistry: designer debacle. Nature 2008,
453:275-278.
24. Kemp DS, Cox DD, Paul KG: The physical organic chemistry
of benzisoxazoles. IV. The origins and catalytic nature
of the solvent rate acceleration for the decarboxylation
of 3-carboxybenzisoxazoles. J Am Chem Soc 1975, 97:
7312-7318.
36. Lazaridis T, Karplus M: Effective energy function for proteins in
solution. Proteins 1999, 35:133-152.
38. Murphy PM, Bolduc JM, Gallaher JL, Stoddard BL, Baker D:
Alteration of enzyme specificity by computational loop
remodeling and design. Proc Natl Acad Sci U S A 2009,
106:9215-9220.
39. Ruscio JZ, Kohn JE, Ball KA, Head-Gordon T: The influence of
protein dynamics on the success of computational enzyme
design. J Am Chem Soc 2009, 131:14111-14115.
25. List B, Lerner RA, Barbas CF: Proline-catalyzed direct
asymmetric aldol reactions. J Am Chem Soc 2000, 122:
2395-2396.
40. Davey JA, Chica RA: Multistate approaches in computational
protein design. Protein Sci 2012, 21:1241-1252.
26. Merski M, Shoichet BK: Engineering a model protein cavity to
catalyze the Kemp elimination. Proc Natl Acad Sci U S A 2012,
109:16179-16183.
41. Alexandrova AN, Röthlisberger D, Baker D, Jorgensen WL:
Catalytic mechanism and performance of computationally
designed enzymes for Kemp elimination. J Am Chem Soc 2008,
130:15907-15915.
27. Müller MM, Windsor MA, Pomerantz WC, Gellman SH, Hilvert D: A
rationally designed aldolase foldamer. Angew Chem Int Ed
2009, 48:922-925.
28. Wörsdörfer B, Henning LM, Obexer R, Hilvert D: Harnessing
protein symmetry for enzyme design. ACS Catalysis 2012,
2:982-985.
www.sciencedirect.com
42. Frushicheva MP, Cao J, Chu ZT, Warshel A: Exploring challenges
in rational enzyme design by simulating the catalysis in
artificial Kemp eliminase. Proc Natl Acad Sci U S A 2010,
107:16869-16874.
43. Kiss G, Röthlisberger D, Baker D, Houk KN: Evaluation and
ranking of enzyme designs. Protein Sci 2010, 19:1760-1773.
Current Opinion in Chemical Biology 2013, 17:1–8
Please cite this article in press as: Kries H, et al.: De novo enzymes by computational design, Curr Opin Chem Biol (2013), http://dx.doi.org/10.1016/j.cbpa.2013.02.012
COCHBI-1035; NO. OF PAGES 8
8 Biocatalysis and biotransformation
44. Jäckel C, Kast P, Hilvert D: Protein design by directed evolution.
Ann Rev Biophys 2008, 37:153-173.
45. Khersonsky O, Kiss G, Röthlisberger D, Dym O, Albeck S,
Houk KN, Baker D, Tawfik DS: Bridging the gaps in design
methodologies by evolutionary optimization of the stability
and proficiency of designed Kemp eliminase KE59. Proc Natl
Acad Sci U S A 2012, 109:10358-10363.
The catalytic efficiency of a designed Kemp eliminase was enhanced 2000fold over 13 evolutionary cycles. Extensive biochemical, structural and
molecular dynamics simulations of selected variants along the evolutionary
trajectory helped rationalize this improvement in molecular terms.
46. Khersonsky O, Röthlisberger D, Dym O, Albeck S, Jackson CJ,
Baker D, Tawfik DS: Evolutionary optimization of
computationally designed enzymes: Kemp eliminases of the
KE07 series. J Mol Biol 2010, 396:1025-1042.
47. Khersonsky O, Röthlisberger D, Wollacott AM, Murphy P, Dym O,
Albeck S, Kiss G, Houk KN, Baker D, Tawfik DS: Optimization of
the in-silico-designed Kemp eliminase KE70 by computational
design and directed evolution. J Mol Biol 2011, 407:391-412.
Current Opinion in Chemical Biology 2013, 17:1–8
48. Lassila JK, Baker D, Herschlag D: Origins of catalysis by
computationally designed retroaldolase enzymes. Proc Natl
Acad Sci U S A 2010, 107:4937-4942.
49. Wang L, Althoff EA, Bolduc J, Jiang L, Moody J, Lassila JK,
Giger L, Hilvert D, Stoddard B, Baker D: Structural analyses of
covalent enzyme-substrate analog complexes reveal
strengths and limitations of de novo enzyme design. J Mol Biol
2012, 415:615-625.
The crystal structure of a computationally designed retro-aldolase in
complex with a covalently bound inhibitor shows qualitative agreement
with the design model. However, inaccuracies in placing the ligand and
active site residues highlight limitations of current design algorithms.
50. Baker D: An exciting but challenging road ahead for
computational enzyme design. Protein Sci 2010, 19:1817-1819.
51. Savile CK, Janey JM, Mundorff EC, Moore JC, Tam S, Jarvis WR,
Colbeck JC, Krebber A, Fleitz FJ, Brands J et al.: Biocatalytic
asymmetric synthesis of chiral amines from ketones
applied to sitagliptin manufacture. Science 2010, 329:
305-309.
www.sciencedirect.com
Please cite this article in press as: Kries H, et al.: De novo enzymes by computational design, Curr Opin Chem Biol (2013), http://dx.doi.org/10.1016/j.cbpa.2013.02.012