Biocatalytic Asymmetric Synthesis of Chiral Amines from Ketones

Applied to Sitagliptin Manufacture

Christopher K. Savile et al.
Science 329, 305 (2010);
DOI: 10.1126/science.1188934

This copy is for your personal, non-commercial use only.

If you wish to distribute this article to others, you can order high-quality copies for your
colleagues, clients, or customers by clicking here.

Downloaded from www.sciencemag.org on June 16, 2012

Permission to republish or repurpose articles or portions of articles can be obtained by
following the guidelines here.

The following resources related to this article are available online at

www.sciencemag.org (this information is current as of June 16, 2012 ):

Updated information and services, including high-resolution figures, can be found in the online
version of this article at:
http://www.sciencemag.org/content/329/5989/305.full.html
Supporting Online Material can be found at:
http://www.sciencemag.org/content/suppl/2010/06/17/science.1188934.DC1.html
A list of selected additional articles on the Science Web sites related to this article can be
found at:
http://www.sciencemag.org/content/329/5989/305.full.html#related
This article cites 24 articles, 2 of which can be accessed free:
http://www.sciencemag.org/content/329/5989/305.full.html#ref-list-1
This article has been cited by 4 articles hosted by HighWire Press; see:
http://www.sciencemag.org/content/329/5989/305.full.html#related-urls
This article appears in the following subject collections:
Chemistry
http://www.sciencemag.org/cgi/collection/chemistry

Science (print ISSN 0036-8075; online ISSN 1095-9203) is published weekly, except the last week in December, by the
American Association for the Advancement of Science, 1200 New York Avenue NW, Washington, DC 20005. Copyright
2010 by the American Association for the Advancement of Science; all rights reserved. The title Science is a
registered trademark of AAAS.
 REPORTS
counterintuitive. Based on the finding of a much occur, according to the existing data, and the order- 10. C. Stadler, S. Hansen, A. Schöll, C. Kumpf, E. Umbach,
stronger molecule-substrate interaction in the LT disorder transition can thus indeed be a real melting N. J. Phys. 9, 50 (2007).
11. J. Stöhr, NEXAFS Spectroscopy, R. Gomer, Ed. (Spinger
phase, we believe that the local atomic order at process or a transition into a glassy state. Series in Surface Science, Springer, Berlin, 1992),
the interface, although different from molecule to This scenario may bear some analogy to the vol. 25.
molecule (no long-range order), is considerably description of inverse melting in metallic alloys 12. A. Schöll et al., Phys. Rev. Lett. 93, 146406 (2004).
higher in the LT (compared to the RT) phase, (4). There, the phenomenon is possible because 13. Y. Zou et al., Surf. Sci. 600, 1240 (2006).
14. A. Bendounan et al., Surf. Sci. 601, 4013 (2007).
connected with a shorter distance between sub- the (low-temperature) amorphous phase gains 15. A. Schöll, Y. Zou, T. Schmidt, R. Fink, E. Umbach, J. Phys. Chem.
strate and adsorbate atoms. Thus, heating leads energy against the crystal because of increased B 108, 14741 (2004).
to a reduction of the short-range order (DS > 0), chemical short-range order (4). The very pro- 16. L. Kilian et al., Phys. Rev. Lett. 100, 136103 (2008).
which is not completely compensated for by the nounced anisotropy of the forces involved in the 17. F. S. Tautz, Prog. Surf. Sci. 82, 479 (2007).
18. J. Ziroff, P. Gold, A. Bendounan, F. Forster, F. Reinert,
appearance of long-range order (DS < 0). present strongly anisotropic (quasi-two-dimensional) Surf. Sci. 603, 354 (2009).
The reduction of the bonding strength is most case, where the ratio of the respective energy 19. N. Schupper, N. M. Shnerb, Phys. Rev. Lett. 93, 037202
probably accompanied by an increase of the scales is extremely crucial, requires a different (2004).
vertical bond lengths, perhaps induced by the theoretical description. 20. O. Skibbe et al., J. Chem. Phys. 131, 024701 (2009).
21. Fruitful discussions with M. Sokolowski and Th. Schmidt
excitation of vertical molecular vibrations (20).
are gratefully acknowledged. This work was funded by the
Consequently, the lateral interaction between the References and Notes German Bundesministerium für Bildung und Forschung
molecules gains more weight and induces the 1. N. Schupper, N. M. Shnerb, Phys. Rev. E Stat. Nonlin. Soft (contracts 05 KS1WWA-5 and 05 KS7WE1) and by the
Matter Phys. 72, 046107 (2005). Deutsche Forschungsgemeinschaft (contract DFG Re
lateral ordering (16).
2. A. L. Greer, Nature 404, 134 (2000). 1469/3-2). Experimental support by the BESSY (Berlin
At low temperature, the structural reorganiza-

Downloaded from www.sciencemag.org on June 16, 2012

3. P. M. Tedrow, D. M. Lee, Phys. Rev. 181, 399 (1969). Electron Storage Ring Society for Synchrotron Radiation)
tion may be explained by different, energetically 4. W. Sinkler, C. Michaelsen, R. Bormann, D. Spilsbury, staff during beam time is gratefully acknowledged.
favorable adsorption sites of the NTCDA mole- N. Cowlam, Phys. Rev. B 55, 2874 (1997).
cules if they are near the substrate, perhaps involv- 5. H. Y. Bai, C. Michaelsen, R. Bormann, Phys. Rev. B 56,
R11361 (1997). Supporting Online Material
ing a distortion of the adsorbed molecule. If the 6. M. R. Feeney, P. G. Debenedetti, F. H. Stillinger, J. Chem. Phys. www.sciencemag.org/cgi/content/full/329/5989/303/DC1
molecules thereby differ slightly, such as in their 119, 4582 (2003). Methods
SOM Text
rotational and/or translational alignment (the local 7. U. Stahl, D. Gador, A. Soukopp, R. Fink, E. Umbach,
Surf. Sci. 414, 423 (1998). Figs. S1 and S2
bonding geometry), a statistic population of the References
8. L. Kilian, thesis, Universität Würzburg, Würzburg,
respective sites would result in the observed loss Germany (2002).
of long-range order. The occurrence of both trans- 9. D. Gador et al., J. Electron Spectrosc. Relat. Phenom. 4 March 2010; accepted 8 June 2010
lational and rotational disorder upon cooling may 101–103, 523 (1999). 10.1126/science.1189106

at high pressure [250 pounds per square inch (psi)]

Biocatalytic Asymmetric Synthesis of using a rhodium-based chiral catalyst (Fig. 1A)
(4). Within the manufacturing scheme, the chem-
Chiral Amines from Ketones Applied to istry suffers from inadequate stereoselectivity and
a product stream contaminated with rhodium, ne-
Sitagliptin Manufacture cessitating additional purification steps at the ex-
pense of yield to upgrade both enantiomeric excess
(e.e.) and chemical purity. By using a transaminase
Christopher K. Savile,1* Jacob M. Janey,2* Emily C. Mundorff,1 Jeffrey C. Moore,2 Sarena Tam,1 (5–8) scaffold and various protein engineering
William R. Jarvis,1 Jeffrey C. Colbeck,1 Anke Krebber,1 Fred J. Fleitz,2 Jos Brands,2 technologies, we have developed a catalyst and
Paul N. Devine,2 Gjalt W. Huisman,1 Gregory J. Hughes2 process that substantially improve the efficiency
of sitagliptin manufacturing (Fig. 1B).
Pharmaceutical synthesis can benefit greatly from the selectivity gains associated with enzymatic Transaminases have a limited substrate range,
catalysis. Here, we report an efficient biocatalytic process to replace a recently implemented most accepting only substrates with a substituent
rhodium-catalyzed asymmetric enamine hydrogenation for the large-scale manufacture of the no larger than a methyl group at the position ad-
antidiabetic compound sitagliptin. Starting from an enzyme that had the catalytic machinery to jacent to the ketone (Fig. 2A) (9–14). Not sur-
perform the desired chemistry but lacked any activity toward the prositagliptin ketone, we applied prisingly, screening a variety of commercially
a substrate walking, modeling, and mutation approach to create a transaminase with marginal available transaminases provided no enzyme with
activity for the synthesis of the chiral amine; this variant was then further engineered via directed detectable activity for amination of the prosita-
evolution for practical application in a manufacturing setting. The resultant biocatalysts showed gliptin ketone (Fig. 2B). We therefore applied a
broad applicability toward the synthesis of chiral amines that previously were accessible only via combination of in silico design and directed evo-
resolution. This work underscores the maturation of biocatalysis to enable efficient, economical, lution in an effort to confer such activity.
and environmentally benign processes for the manufacture of pharmaceuticals. An (R)-selective transaminase [ATA-117 (15),
a homolog of an enzyme from Arthrobacter sp.]
symmetric hydrogenation technologies are insufficient stereoselectivity (requiring further up- (16) was previously used for R-specific transam-

A being increasingly used for commercial-

scale manufacture of fine chemicals and
pharmaceuticals (1). Solutions that address remain-
grading of the product), are still being sought. Even
though biocatalysis is rarely constrained by these
shortcomings, enzymes may suffer from other lim-
ination of methyl ketones and small cyclic ketones
1
Codexis, Incorporated, 200 Penobscot Drive, Redwood City, CA
94063, USA. 2Department of Process Research, Merck Research
ing shortcomings of such technologies, including itations, such as low turnover numbers, instability
Laboratories, Merck and Company, Incorporated, Rahway, NJ
the need for high-pressure hydrogen, the use and toward the demanding conditions of chemical 07065, USA.
subsequent removal of precious and toxic transi- processes, and postreaction processing issues. *To whom correspondence should be addressed. E-mail:
tion metals, the oftentimes lengthy process of lig- The current synthesis of sitagliptin (2, 3) in- [email protected] (C.K.S.); jacob_janey@
and screening and synthesis, and the generally volves asymmetric hydrogenation of an enamine merck.com (J.M.J.)

www.sciencemag.org SCIENCE VOL 329 16 JULY 2010 305

REPORTS
(11, 12). To assess the feasibility of developing an
enzyme for sitagliptin synthesis, we generated a
structural homology model of ATA-117 (17) to
develop hypotheses for initial library designs.
Docking studies using this model suggested that
the enzyme would be unable to bind prositaglip-
tin ketone (Fig. 2B) because of steric interference
in the small binding pocket and potentially un-
desired interactions in the large binding pocket
(Fig. 2C). By using a substrate walking approach
(18) with a truncated substrate (Fig. 2D), we first
engineered the large binding pocket of the en-
zyme and then evolved that enzyme for activity
toward prositagliptin ketone.
Consistent with the model, ATA-117 was poorly
active on the truncated methyl ketone analog
(Fig. 2D), giving 4% conversion at 2 g/l substrate
loading (table S2). Site saturation libraries of res-
idues lining the large pocket of the active site pro-

Downloaded from www.sciencemag.org on June 16, 2012

vided new variants with increased activity toward
the methyl ketone analog. The best variant con-
tained a S223 → P223 [S223P (19)] mutation and
showed an 11-fold activity improvement (Fig. 3
and table S1). On the basis of this improved var-
iant (ATA-117: S223P), we generated a small
library of enzyme variants for potential activity
on prositagliptin ketone. Analysis of the enzyme
model suggested four residues that could poten-
tially interact with the trifluorophenyl group [V69,
F122, T283, and A284 (19)]. Each of these po-
sitions was individually subjected to saturation Fig. 1. (A) The current synthesis of sitagliptin involves enamine formation followed by asymmetric hy-
mutagenesis and also included in a combinatorial drogenation at high pressure (250 psi) using a rhodium-based chiral catalyst, providing sitagliptin in 97% e.e.,
library that evaluated several residues at each po- with trace amounts of rhodium. Recrystallization to upgrade e.e. followed by phosphate salt formation
sition on the basis of structural considerations provides sitagliptin phosphate. (B) Our biocatalytic route features direct amination of prositagliptin ketone
[V69 → G69 and V69 → A69 (V69GA), F122AVLIG, to provide enantiopure sitagliptin, followed by phosphate salt formation to provide sitagliptin phosphate.
T283GAS, and A284GF; library size of 216 var-
iants]. A variant containing four mutations, three
in the small binding pocket and one in the large
pocket, provided the first detectable transaminase
activity on prositagliptin ketone (Fig. 3 and table
S3). No detectable activity was identified in any
of the variants from the single amino acid site sat-
uration libraries. Initial activity was accomplished
via an F122I, V, or L mutation in combination
with V69G or A284G. Docking studies indicated
that these mutations may relieve the steric inter-
ference in the small binding pocket (Fig. 2E). En-
zyme loading of 10 g/l provided 0.7% conversion
of 2 g/l of ketone over 24 hours, corresponding to
an estimated turnover of 0.1 per day. Screening
the same combinatorial library in the ATA-117
context without the S223P large binding pocket
mutation did not provide any variant with detect-
ible activity toward prositagliptin ketone. Having
attained activity through computer-aided catalyst
design, we started evolving an enzyme variant for
a practical, large-scale process.
The variant with the highest activity toward Fig. 2. Previous substrate range studies suggested that the active site of transaminase consists of large (L)
prositagliptin ketone from round 1b was chosen and small (S, typically limited to substituents about the size of a methyl group) binding pockets as mapped
as the parent for the second round of evolution, on the structure of acetophenone (A). Accordingly, the structure of prositagliptin ketone (B) can be mapped
and all the beneficial mutations from both the on these binding pockets and docked into the active site of the homology model (C). A prositagliptin ketone
small-pocket combinatorial library and the large- analog (D) was designed to fit the large pocket for initial optimization of this part of the active site. After
pocket saturation mutagenesis libraries were initial engineering of the large pocket, an enzyme variant was generated with activity on the desired
combined into a new library. Screening of this li- substrate (E) by excavating the small pocket (gray/blue, transaminase homology model; orange, large
brary resulted in a variant with 75-fold increased binding pocket; turquoise, small binding pocket; green, PLP and catalytic residues).

306 16 JULY 2010 VOL 329 SCIENCE www.sciencemag.org

 REPORTS
Biocatalyst libraries were generated by using
a variety of methods in subsequent iterative rounds
of directed evolution (22–25). Screening of these
libraries under processlike conditions provided var-
iants that were increasingly tolerant to the desired
process conditions. Specific reaction conditions
could not be devised at project inception because
no catalyst was available for evaluation of such
conditions. Consequently, chemical process devel-
opment and enzyme optimization were performed
in parallel, with process conditions iteratively op-
timized as improved enzymes became available.
Over the course of 11 rounds of evolution, high
throughput (HTP) screening conditions were ren-
dered more stringent with the rising activity and
tolerance of the biocatalyst. Specifically, we in-
creased the substrate concentration from 2 to 100 g/l,
the i-PrNH2 concentration from 0.5 to 1 M, the
cosolvent from 5 to 50% DMSO, the pH from

Downloaded from www.sciencemag.org on June 16, 2012

7.5 to 8.5, and the temperature from 22° to 45°C,
ultimately leading to a catalyst that met the re-
quired process targets (table S8). All active tran-
saminase variants tested met the >99.9% e.e.
target. Variants generated via this approach were
simultaneously optimized for several performance
criteria: activity; tolerance to DMSO, acetone, and
i-PrNH2; stability to the elevated temperature of
the reaction medium; and expression in an Esche-
richia coli manufacturing host.
The sequence changes accumulated through
this process gave rise to a compounded improve-
ment of four orders of magnitude in activity from
the initial prositagliptin transaminase (26). The
final catalyst contained 27 mutations; of the 17
noncatalytically essential amino acid residues pre-
dicted to be interacting with the substrate, 10 were
mutated in the final variant. The 27 mutations
were obtained by screening various libraries that
included diversity identified via structure-aided
design of the small binding pocket (4 mutations),
site saturation mutagenesis of the large binding
pocket (5 mutations), site saturation mutagenesis
of the small binding pocket (1 mutation), site
Fig. 3. Performance of evolvants in HTP screening over the course of evolution. *Mutations accumulated
saturation mutagenesis of positions outside the
in each round of evolution; reference amino acid refers to parent from previous round of evolution. †Fold
improvement over parent refers to the parent for that round of evolution. ‡Mutagenesis experiments binding pockets (2 mutations), homology libraries
showed that these mutations were not critical for activity toward prositagliptin ketone. The top variant (10 mutations), and random mutagenesis (5 muta-
from round 1b contained three mutations in the small pocket; however, two are sufficient for achieving tions). The progression of mutagenesis is depicted
detectable activity. Dash entry indicates starting enzyme; RT, room temperature. on the homology model (Fig. 4). After the active
site was engineered to accommodate the substrate
activity compared with that of the second-round optimal with regard to both enzyme performance in the initial rounds, further improvements to gen-
parent. This variant contained 12 mutations, 10 of and ketone solubility (table S7). Solubility can erate a productive catalyst came from substantial
which map to the binding pocket model (Fig. 3 also be enhanced by operating at higher tem- modification in the dimer interfacial region. The
and table S5). peratures, prompting our selection of higher boi- enzyme is presumably active only as a dimer (27);
Despite this substantial activity enhancement, ling DMSO as the preferred solvent. Because we can speculate that these mutations serve to
this catalyst was not yet of practical utility. Re- transaminase-catalyzed reactions are equilibrium strengthen dimer interactions in the face of in-
action conditions in a chemical plant are much controlled, product accumulation is favored by a large creasingly destabilizing reaction conditions (28).
different from the conditions that an enzyme en- excess of the amine donor isopropylamine (i-PrNH2) Under optimal conditions (17), the best var-
counters in nature, making the development of and/or removal of acetone coproduct. Thus, in iant converted 200 g/l prositagliptin ketone to
practical biocatalytic processes a multidimensional order to meet the required performance for practical sitagliptin of >99.95% e.e. (the undesired enan-
challenge (20, 21). The solubility of the prosita- application, the biochemical characteristics of the tiomer was never detected) by using 6 g/l enzyme
gliptin ketone in water is low (<1 g/l), necessitat- transaminase needed to be improved to withstand in 50% DMSO with a 92% assay yield at the end
ing a substantial amount of organic cosolvent to the harsh conditions of at least 100 g/l (250 mM) of reaction (table S11). In comparison with the
prevent precipitation during reaction. After an prositagliptin ketone, 1 M i-PrNH2, >25% DMSO, rhodium-catalyzed process (Fig. 1A), the bio-
evaluation of various cosolvents, methanol and and a temperature of >40°C for a period of 24 catalytic process provides sitagliptin with a 10 to
dimethylsulfoxide (DMSO) were identified as hours, and to give product with >99.9% e.e. 13% increase in overall yield, a 53% increase in

www.sciencemag.org SCIENCE VOL 329 16 JULY 2010 307

REPORTS

Fig. 4. Structural models of variants from sequential stages of the evolution mapped on the monomer with the active site of the enzyme exposed and
program. Accumulated mutations are highlighted in purple on the homology substrate bound. The bottom row shows the mutations mapped on the dimer.
model of the transaminase used in this study (size exclusion chromatography The active site, the subunit interface, and a few exposed surface areas can be
data show that the enzyme forms a dimer). The top row shows the mutations identified as structural hot spots for mutations.

Downloaded from www.sciencemag.org on June 16, 2012

Fig. 5. Transaminase
(TA)–catalyzed amination
reactions that were here-
tofore not feasible either
enzymaticallyor chemical-
ly but now proceed with
the enzymes described
in this work.

productivity (kg/l per day), a 19% reduction in by making substantial progress toward a general 11. M. D. Truppo, N. J. Turner, J. D. Rozzell, Chem. Commun.
total waste, the elimination of all heavy metals, approach for the safe, efficient, environmental- 2009, 2127 (2009).
12. D. Koszelewski, D. Clay, D. Rozzell, W. Kroutil, Eur. J. Org.
and a reduction in total manufacturing cost; the ly friendly production of chiral primary amines. Chem. 2009, 2289 (2009).
enzymatic reaction is run in multipurpose vessels, The manufacture of chiral alcohols is already 13. D. Koszelewski et al., Angew. Chem. Int. Ed. 47, 9337
avoiding the need for specialized high-pressure often accomplished with biocatalysts (30–32); (2008).
hydrogenation equipment. chiral amine synthesis via the enzymatic platform 14. D. Koszelewski, I. Lavandera, D. Clay, D. Rozzell,
W. Kroutil, Adv. Synth. Catal. 350, 2761 (2008).
The enzymes developed for sitagliptin syn- described here is expected to reach a similar level 15. The complete amino acid sequence of ATA-117 is as
thesis have a broad substrate range and increased of broad utility and robustness. This development follows: MAFSADTSEIVYTHDTGLDYITYSDYELDPANPLAGGAA-
tolerance to high concentrations of i-PrNH2 and will serve as a model for the implementation of WIEGAFVPPSEARISIFDQGYLHSDVTYTVFHVWNGNAFRLD-
organic solvent that enhances their practical util- other biocatalytic manufacturing processes in DHIERLFSNAESMRIIPPLTQDEVKEIALELVAKTELREAFVSVSITR-
GYSSTPGERDITKHRPQVYMYAVPYQWIVPFDRIRDGVHAM-
ity. For instance, various trifluoromethyl-substituted which enzymes can be evolved to meet desired VAQSVRRTPRSSIDPQVKNFQWGDLIRAVQETHDRG-
amines as well as phenylethylamines with electron- chiral process targets. FEAPLLLDGDGLLAEGSGFNVVVIKDGVVRSPGRAALPGITRKTV-
rich substituents (Fig. 5 and fig. S6), which cannot LEIAESLGHEAILADITLAELLDADEVLGCTTAGGVWPFVSVDGN-
be generated via traditional reductive amination, PISDGVPGPVTQSIIRRYWELNVESSSLLTPVQY.
References and Notes
were prepared with near-perfect stereopurity. The 1. N. B. Johnson, I. C. Lennon, P. H. Moran, J. A. Ramsden, 16. A. Iwasaki, Y. Yamada, N. Kizaki, Y. Ikenaka, J. Hasegawa,
Acc. Chem. Res. 40, 1291 (2007). Appl. Microbiol. Biotechnol. 69, 499 (2006).
engineered transaminases also enabled efficient 17. Materials and methods are detailed in supporting online
2. D. M. Kendall, R. M. Cuddihy, R. M. Bergenstal,
access to chiral pyrrolidines. As such, these en- Eur. J. Intern. Med. 20 (suppl.), S329 (2009). material on Science Online.
zymes provide a general approach for the practical 3. D. Williams-Herman et al., BMC Endocr. Disord. 8, 14 18. Z. Chen, H. Zhao, J. Mol. Biol. 348, 1273 (2005).
synthesis of chiral amines from prochiral ketones, (2008). 19. Single-letter abbreviations for the amino acid residues
4. K. B. Hansen et al., J. Am. Chem. Soc. 131, 8798 (2009). are as follows: A, Ala; C, Cys; D, Asp; E, Glu; F, Phe; G, Gly;
which is of general interest in pharmaceutical H, His; I, Ile; K, Lys; L, Leu; M, Met; N, Asn; P, Pro;
manufacturing (29). 5. a-Transaminases are pyridoxal 5′-phosphate
(PLP)-dependent enzymes widespread in nature for Q, Gln; R, Arg; S, Ser; T, Thr; V, Val; W, Trp; and Y, Tyr.
Although naturally occurring enzymes rarely the synthesis of a-amino acids from the corresponding 20. G. W. Huisman, J. J. Lalonde, in Biocatalysis in the
offer ideal manufacturing catalysts, directed evo- a-keto acids; comparatively few w-transaminases, for Pharmaceutical and Biotechnology Industries,
R. N. Patel, Ed. (CRC, Boca Raton, FL, 2006),
lution provides an effective means to overcome the synthesis of chiral amines that are not adjacent
to carboxylic acid functionalities, are known. pp. 717–742.
this limitation. We have demonstrated that com- 21. I. W. Davies, C. J. Welch, Science 325, 701 (2009).
6. S. P. Crump, J. D. Rozzell, in Biocatalytic Production of
bining modeling with directed evolution offers a Amino Acids and Derivatives, J. D. Rozzell, F. Wagner, 22. R. J. Fox et al., Nat. Biotechnol. 25, 338 (2007).
rapid means of creating an active enzyme that can Eds. (Wiley, New York, 1992), pp. 43–58. 23. J. H. Zhang, G. Dawes, W. P. Stemmer, Proc. Natl. Acad.
operate under the demanding conditions required 7. J.-S. Shin, B.-G. Kim, J. Org. Chem. 67, 2848 (2002). Sci. U.S.A. 94, 4504 (1997).
8. B.-K. Cho et al., Biotechnol. Bioeng. 99, 275 (2008). 24. K. Stutzman-Engwall et al., Metab. Eng. 7, 27 (2005).
for the manufacture of pharmaceuticals. In ad-
9. M. Höhne, S. Kühl, K. Robins, U. T. Bornscheuer, 25. J. Liao et al., BMC Biotechnol. 7, 16 (2007).
dition, we have removed the intrinsic limitation ChemBioChem 9, 363 (2008). 26. See “quantitation of catalyst improvement,” figs. S1 and
that transaminases accept only a methyl substi- 10. M. D. Truppo, J. D. Rozzell, J. C. Moore, N. J. Turner, S2, and tables S10 and S11 in supporting online material
tuent adjacent to the carbonyl functionality, there- Org. Biomol. Chem. 7, 395 (2009). (SOM) on Science Online.

308 16 JULY 2010 VOL 329 SCIENCE www.sciencemag.org

 REPORTS
27. Size exclusion chromatography data show the enzyme production system; R. Fox for bioinformatics support; process information should be directed to J.M.J.
forms a dimer (fig. S4). M. Jefferson, R. Trinidad, and M. Whitehorn for assay ([email protected]).
28. V. G. H. Eijsink et al., J. Biotechnol. 113, 105 support; V. Mitchell for molecular biology support;
(2004). J. Munger and O. Gooding for process development support; Supporting Online Material
29. D. J. C. Constable et al., Green Chem. 9, 411 (2007). P. Fernandez, S. Grosser, and M. Mohan for pilot scale www.sciencemag.org/cgi/content/full/science.1188934/DC1
30. S. M. A. De Wildeman, T. Sonke, H. E. Schoemaker, development; K. Morley and B. Grau for additional synthesis Materials and Methods
O. May, Acc. Chem. Res. 40, 1260 (2007). support; M. Foster and N. Wu for analytical support; and Figs. S1 to S6
31. J. C. Moore, D. J. Pollard, B. Kosjek, P. N. Devine, L. Moore, S. Lato, P. Seufer-Wasserthal, J. Grate, J. Liang, Tables S1 to S12
Acc. Chem. Res. 40, 1412 (2007). and J. Lalonde for helpful suggestions. Merck and Codexis References
32. G. W. Huisman, J. Liang, A. Krebber, Curr. Opin. have filed patent applications on the transaminases, the
Chem. Biol. 14, 122 (2010). genes encoding them as well as their use. Materials can be 1 March 2010; accepted 19 May 2010
33. We thank C. Ng, D. Standish, M. Mayhew, D. Gray, provided under a materials transfer agreement; requests for Published online 17 June 2010;
C. Fletcher, F. Mazzini, P. Lattik, T. Brandon, and enzymes should be directed to C.K.S. (christopher.savile@ 10.1126/science.1188934
J. Postlethwaite for developing the transaminase codexis.com), whereas requests for substrates, products, and Include this information when citing this paper.

dimethylacrylamide (Fig. 1, substrates 1 and 2,

Computational Design of an Enzyme respectively) (8).
The first step in de novo enzyme design is to
Catalyst for a Stereoselective decide on a catalytic mechanism and an asso-
ciated ideal active site. For normal-electron-demand

Downloaded from www.sciencemag.org on June 16, 2012

Bimolecular Diels-Alder Reaction Diels-Alder reactions, frontier molecular orbital
theory dictates that the interaction of the highest
occupied molecular orbital (HOMO) of the diene
Justin B. Siegel,1,2* Alexandre Zanghellini,1,2*† Helena M. Lovick,3 Gert Kiss,4 with the lowest unoccupied molecular orbital
Abigail R. Lambert,5 Jennifer L. St.Clair,1 Jasmine L. Gallaher,1 Donald Hilvert,6 (LUMO) of the dienophile is the dominant inter-
Michael H. Gelb,3 Barry L. Stoddard,5 Kendall N. Houk,4 Forrest E. Michael,3 David Baker1,2,7‡ action in the transition state (7). Narrowing the
energy gap between the HOMO and LUMO will
The Diels-Alder reaction is a cornerstone in organic synthesis, forming two carbon-carbon increase the rate of the Diels-Alder reaction. This
bonds and up to four new stereogenic centers in one step. No naturally occurring enzymes can be accomplished by positioning a hydrogen
have been shown to catalyze bimolecular Diels-Alder reactions. We describe the de novo bond acceptor to interact with the carbamate NH
computational design and experimental characterization of enzymes catalyzing a bimolecular of the diene (thus raising the energy of the HOMO
Diels-Alder reaction with high stereoselectivity and substrate specificity. X-ray crystallography energy and stabilizing the positive charge accumu-
confirms that the structure matches the design for the most active of the enzymes, and lating in the transition state), and a hydrogen bond
binding site substitutions reprogram the substrate specificity. Designed stereoselective catalysts donor to interact with the carbonyl of the dieno-
for carbon-carbon bond-forming reactions should be broadly useful in synthetic chemistry. phile (lowering the LUMO energy and stabilizing
the negative charge accumulating in the transition
ntermolecular Diels-Alder reactions are impor- relative orientation in order to accelerate the state) (9). Quantum mechanical (QM) calculations

I tant in organic synthesis (1–3), and enzyme

Diels-Alder catalysts could be invaluable in
increasing rates and stereoselectivity. No natu-
reaction and impart stereoselectivity. Also, pre-
vious successes with computational enzyme de-
sign have involved general acid-base catalysis
predict that these hydrogen bonds can stabilize the
transition state by up to 4.7 kcal mol–1. (fig. S1). In
addition to electronic stabilization, binding of the
rally occurring enzyme has been demonstrated and covalent catalysis, but the Diels-Alder two substrates in a relative orientation optimal for
(4) to catalyze an intermolecular Diels-Alder reaction provides the opportunity to alter the re- the reaction is expected to produce a large increase
reaction (1, 2), although catalytic antibodies have action rate by modulation of molecular orbital in rate through entropy reduction (10). Thus, a
been generated for several Diels-Alder reactions energies (7). To investigate the feasibility of de- protein with a binding pocket (Fig. 1) that po-
(3, 4). We have previously used the Rosetta signing intermolecular Diels-Alder enzyme cata- sitions the two substrates in the proper relative
computational design methodology to design lysts, we chose to focus on the well-studied model orientation and has appropriately placed hydro-
novel enzymes (5, 6) that catalyze bond-breaking Diels-Alder reaction between 4-carboxybenzyl gen bond donors and acceptors is expected to be
reactions. However, bimolecular bond-forming trans-1,3-butadiene-1-carbamate and N,N- an effective Diels-Alder catalyst.
reactions present a greater challenge, because
both substrates must be bound in the proper

1 O O
Department of Biochemistry, University of Washington,
Seattle, WA 98195, USA. 2Biomolecular Structure and Design N O N O NH N
Program, University of Washington, Seattle, WA 98195, USA. H Acceptor H N
3 Donor
Department of Chemistry, University of Washington, Seattle, O O O
WA 98195, USA. 4Department of Chemistry and Biochemistry, 2 O O
University of California, Los Angeles, CA 90095, USA. 5Division N
of Basic Sciences, Fred Hutchinson Cancer Research Center,
Seattle, WA 98109, USA. 6Laboratory of Organic Chemistry,
Eidgenössische Technische Hochschule (ETH) Zürich, 8093 CO 2 -
-
O2 C CO 2 H
Zürich, Switzerland. 7Howard Hughes Medical Institute (HHMI),
University of Washington, Seattle, WA 98195, USA. 1 3 4
*These authors contributed equally to this work.
Fig. 1. The Diels-Alder reaction. Diene (1) and dienophile (2) undergo a pericyclic [4 + 2] cycloaddition (3)
†Present address: Arzeda Corporation, 2722 Eastlake Avenue
East, Suite 150, Seattle, WA 98102, USA. to form a chiral cyclohexene ring (4). Also shown in (3) is a schematic of the design target active site, with
‡To whom correspondence should be addressed. E-mail: hydrogen bond acceptor and donor groups activating the diene and dienophile and a complementary
[email protected] binding pocket holding the two substrates in an orientation optimal for catalysis.

www.sciencemag.org SCIENCE VOL 329 16 JULY 2010 309