Artificial transfer hydrogenases based on the biotin-(strept)avidin technology: Fine tuning the selectivity by saturation mutagenesis of the host protein

Article abstract of DOI:10.1021/ja061580o

Incorporation of biotinylated racemic three-legged d⁶-piano stool complexes in streptavidin yields enantioselective transfer hydrogenation artificial metalloenzymes for the reduction of ketones. Having identified the most promising organometallic catalyst precursors in the presence of wild-type streptavidin, fine-tuning of the selectivity is achieved by saturation mutagenesis at position S112. This choice for the genetic optimization site is suggested by docking studies which reveal that this position lies closest to the biotinylated metal upon incorporation into streptavidin. For aromatic ketones, the reaction proceeds smoothly to afford the corresponding enantioenriched alcohols in up to 97% ee (R) or 70% (S). On the basis of these results, we suggest that the enantioselection is mostly dictated by CH/π interactions between the substrate and the η⁶-bound arene. However, these enantiodiscriminating interactions can be outweighed in the presence of cationic residues at position S112 to afford the opposite enantiomers of the product.

Full text of DOI:10.1021/ja061580o

Published on Web 06/03/2006
Artificial Transfer Hydrogenases Based on the
Biotin-(Strept)avidin Technology: Fine Tuning the Selectivity
by Saturation Mutagenesis of the Host Protein
Christophe Letondor,‡ Anca Pordea,‡ Nicolas Humbert,‡ Anita Ivanova,‡
Sylwester Mazurek,§ Marjana Novic,§ and Thomas R. Ward*,‡
Contribution from the Institute of Chemistry, UniVersity of Neuchaˆtel, AV. BelleVaux 51,
CP 2, CH-2007 Neuchaˆtel, Switzerland, and Laboratory of Chemometrics, National Institute of
Chemistry, HajdrihoVa 19, SI-1001 Ljubljana, SloVenia
Received March 7, 2006; E-mail: thomas.ward@unine.ch
Abstract: Incorporation of biotinylated racemic three-legged d6-piano stool complexes in streptavidin yields
enantioselective transfer hydrogenation artificial metalloenzymes for the reduction of ketones. Having
identified the most promising organometallic catalyst precursors in the presence of wild-type streptavidin,
fine-tuning of the selectivity is achieved by saturation mutagenesis at position S112. This choice for the
genetic optimization site is suggested by docking studies which reveal that this position lies closest to the
biotinylated metal upon incorporation into streptavidin. For aromatic ketones, the reaction proceeds smoothly
to afford the corresponding enantioenriched alcohols in up to 97% ee (R) or 70% (S). On the basis of
these results, we suggest that the enantioselection is mostly dictated by CH/π interactions between the
substrate and the η6-bound arene. However, these enantiodiscriminating interactions can be outweighed
in the presence of cationic residues at position S112 to afford the opposite enantiomers of the product.
(ii) Initially proposed by Kaiser14 and Whitesides15 in the late
Introduction
1970s, artificial metalloenzymes based on either a covalent16-18
or a supramolecular anchoring19-23 of an organometallic catalyst
precursor in a macromolecular host (protein or DNA24), are
gaining attention in the context of enantioselective catalysis.25
In terms of catalyst performance, artificial metalloenzymes
offer the appealing prospect of a combined chemogenetic
optimization: the host protein can be optimized by genetic
means, and the organometallic catalyst precursor can be
chemically modified. In the spirit of enzymes, such hybrid
catalysts provide a well-defined chiral second coordination
sphere. By incorporating an achiral or racemic organometallic
The asymmetric reduction of CdO and CdN bonds is one
of the most fundamental transformations in organic chemistry.
Although efficient heterogeneous catalysts exist for such reduc-
tions,1 homogeneous and enzymatic catalysts dominate this
active field of investigation.2,3 Both latter approaches have
evolved independently and can achieve high levels of enan-
tioselection for selected substrate classes.
With the aim of alleviating some of the inherent limitations
of existing methodologies, two bio-inspired approaches have
recently witnessed a revival: (i) organocatalysis4,5 and (ii)
artificial metalloenzymes.6-8
(i) Originally developed in the late 1960s,9,10 metal-free
catalytic systems (organocatalysts), sometimes referred to as
minimal enzymes, have found numerous applications in the past
decade. In recent months, significant advances in the context
of enantioselective organocatalytic reduction of imines and
ketones have been published.11-13
(11) Akiyama, T.; Morita, H.; Itoh, J.; Fuchibe, K. Org. Lett 2005, 7, 2583-
2585.
(12) Hoffmann, S.; Seayad, A. M.; List, B. Angew. Chem., Int. Ed. 2005, 44,
7424-7427.
(13) Malkov, A. V.; Liddon, A. J. P. S.; Ramirez-Lopez, P.; Bendova, L.; Haigh,
D.; Kocovsky, P. Angew. Chem., Int. Ed. 2006, 45, 1432-1435.
(14) Kaiser, E. T.; Lawrence, D. S. Science 1984, 226, 505-511.
(15) Wilson, M. E.; Whitesides, G. M. J. Am. Chem. Soc. 1978, 100, 306-307.
(16) Panella, L.; Broos, J.; Jin, J.; Fraaije, M. W.; Janssen, D. B.; Jeronimus-
Stratingh, M.; Feringa, B. L.; Minnaard, A. J.; de Vries, J. G. Chem.
Commun. 2005, 5656-5658.
‡ University of Neuchaˆtel.
(17) Carey, J. R.; Ma, S. K.; Pfister, T. D.; Garner, D. K.; Kim, H. K.; Abramite,
J. A.; Wang, Z.; Guo, Z.; Lu, Y. J. Am. Chem. Soc. 2004, 126, 10812-
10813.
§ National Institute of Chemistry.
(1) Bu¨rgi, T.; Baiker, A. Acc. Chem. Res. 2004, 37, 909-917.
(2) Jacobsen, E. N.; Pfaltz, A.; Yamamoto, H. ComprehensiVe Asymmetric
Catalysis; Springer: Berlin, 1999.
(18) Kruithof, C. A.; Casado, M. A.; Guillena, G.; Egmond, M. R.; van der
Kerk-van Hoof, A.; Heck, A. J. R.; Klein Gebbink, R. J. M.; van Koten,
G. Chem. Eur. J. 2005, 11, 6869-6877.
(3) Faber, K. Biotransformations in Organic Chemistry. 5th ed.; Springer:
Berlin, 2004.
(4) Berkessel, R.; Gro¨ger, H. Metal-free organic catalysts in asymmetric
synthesis; Wiley-VCH: Weinheim, 2005.
(5) Dalko, P. I.; Moisan, L. Angew. Chem., Int. Ed. 2004, 43, 5138-5175.
(6) Qi, D.; Tann, C.-M.; Haring, D.; Distefano, M. D. Chem. ReV. 2001, 101,
3081-3111.
(7) Thomas, C. M.; Ward, T. R. Chem. Soc. ReV. 2005, 34, 337-346.
(8) Davis, B. G. Curr. Opin. Biotechnol. 2003, 14, 379-386.
(9) Pracejus, H.; Ma¨tje, H. J. Prakt. Chem. 1964, 24, 195-205.
(10) Eder, U.; Sauer, G.; Wiechert, R. Angew. Chem., Int. Ed. Engl. 1971, 10,
496-497.
(19) Okrasa, K.; Kazlauskas, R. J. Chem. Eur. J. 2006, 12, 1587-1596.
(20) Mahammed, A.; Gross, Z. J. Am. Chem. Soc. 2005, 127, 2883-2887.
(21) Lin, C.-C.; Lin, C.-W.; Chan, A. S. C. Tetrahedron Asymmetry 1999, 10,
1887-1893.
(22) Ohashi, M.; Koshiyama, T.; Ueno, T.; Yanase, M.; Fujii, H.; Watanabe,
Y. Angew. Chem., Int. Ed. 2003, 42, 1005-1008.
(23) Reetz, M. T.; Jiao, N. Angew. Chem., Int. Ed. 2006, 45, 2416-2419.
(24) (a) Roelfes, G.; Feringa, B. L. Angew. Chem., Int. Ed. 2005, 44, 3230-
3232. (b) Roelfes, G.; Boersma, A. J.; Feringa, B. L. Chem. Commun. 2006,
635-637.
(25) Kra¨mer, R. Angew. Chem., Int. Ed. 2005, 45, 858-860.
9
8320
J. AM. CHEM. SOC. 2006, 128, 8320-8328
10.1021/ja061580o CCC: $33.50 © 2006 American Chemical Society

Chemogenetics and Artificial Metalloenzymes
A R T I C L E S
Scheme 1. Artificial Metalloenzymes Based on
moiety within a chiral protein scaffold, the influence of the
second coordination sphere on enantioselectivity can be ad-
dressed.
Biotin-Streptavidin for Enantioselective Transfer Hydrogenation
Reactionsa
In the context of artificial metalloenzymes based on the
biotin-(strept)avidin technology, we recently demonstrated the
versatility of such a chemogenetic optimization for the enan-
tioselective hydrogenation of N-protected dehydroamino acids.26-28
As a second catalytic transformation to implement in this
context, we focus on the transfer hydrogenation of ketones
catalyzed by d6-piano stool complexes.29-36 This choice is
motivated by:
(i) the theoretical suggestion that this reaction proceeds
without coordination of the substrate to the metal.38-41 The chiral
recognition pattern for this organometallic transformation is thus
reminiscent of enzymatic catalysis. Indeed, the second coordina-
tion sphere provided by an enzyme is optimized to steer the
enantiodiscrimination step without necessarily requiring covalent
(or dative) binding of the substrate to the enzyme. The general
concept of this approach is depicted in Scheme 1.
(ii) the compatibility between d6-piano stool complexes and
enzymes at elevated temperatures, as established by Ba¨ckvall
in the context of dynamic kinetic resolution of alcohols via
acylation.37
We reasoned that, as the substrate does not bind to the metal
center during the transfer hydrogenation (Scheme 1), a well-
defined second coordination sphere provided by the host protein
around a piano stool complex offers an attractive means to
optimize the selectivity of transfer hydrogenation catalysts.
The starting point for the present work was the identification
of [η6-(arene)Ru(Biot-q-L)Cl] (q ) ortho, meta, para, see
Scheme 4) complexes as active and selective catalyst precursors
for the transfer hydrogenation of acetophenone derivatives. In
the presence of various streptavidin mutants, we recently
reported on enantioselectivities > 90% for the reduction of
p-methylacetophenone.32
a The host protein (streptavidin, red) displays a high affinity for the anchor
(biotin, violet); introduction of a spacer (green) and variation of the metal
and the ηn-bond arene (black) allows one to chemically optimize the activity
and the selectivity. Saturation mutagenesis (red star) allows a genetic
optimization of the host protein. In the transition state of the transfer
hydrogenation, both hydrogens are delivered to the prochiral substrate (blue)
without the substrate coordinating to the metal.
were performed using AutoDock 3.0.5.42,43 All docking simula-
tions presented below are based on a rigid host model and thus
should be considered as qualitative. Computational details can
be found in the Supporting Information.
Since tetrameric wild-type streptavidin (abbreviated hereafter
WT Sav) consists of a dimer of dimer with two proximal and
two distal binding sites, docking studies were carried out on
the dimeric structure (with two proximal binding sites: A and
C subunits) downloaded from the protein data bank (http://
www.rcsb.org/pdb/): 1stp for streptavidin.44 Following the
docking procedure, the B and D subunits of streptavidin were
included for the computation of distances.
The biotinylated complexes [η6-(arene)Ru(Biot-p-L)H] (are-
ne ) p-cymene, benzene) were built using Hyperchem 7.5 based
on the structurally characterized [η6-(p-cymene)Ru(Tos-DPEN)-
Cl] (Tos-DPEN: (R,R)-p-tolylsulfonamido-diphenyl-ethylene-
diamine, CSD code: TAXFON)45 as well as the (+)-biotin
anchor extracted from 1stp. For docking purposes, the (Z)-
configuration of the amide was enforced (Scheme 2 in red) and
the geometry around ruthenium was frozen. Both (R)- and (S)-
configurations at ruthenium were considered, and the dihedral
angle around amidic C-C bond was set to τ ) 0° and 180°
(Scheme 2, in violet), thus yielding a total of four different
isomers which were docked in streptavidin.46 The energy
minimization was performed by allowing rotation around the
bonds highlighted in green in Scheme 2.
With the aim of broadening the substrate scope and gaining
mechanistic insight, we screened 20 streptavidin isoforms in
combination with 21 biotinylated d6-piano stool complexes.
Results and Discussion
Docking Studies. To shed light on the localization of the
biotinylated three-legged piano stool catalyst, docking studies
(26) Skander, M.; Humbert, N.; Collot, J.; Gradinaru, J.; Klein, G.; Loosli, A.;
Sauser, J.; Zocchi, A.; Gilardoni, F.; Ward, T. R. J. Am. Chem. Soc. 2004,
126, 14411-14418.
(27) Klein, G.; Humbert, N.; Gradinaru, J.; Ivanova, A.; Gilardoni, F.; Rusbandi,
U. E.; Ward, T. R. Angew. Chem., Int. Ed. 2005, 44, 7764-7767.
(28) Collot, J.; Gradinaru, J.; Humbert, N.; Skander, M.; Zocchi, A.; Ward, T.
R. J. Am. Chem. Soc. 2003, 125, 9030-9031.
(29) Schlatter, A.; Kundu, M. K.; Woggon, W.-D. Angew. Chem., Int. Ed. 2004,
43, 6731-6734.
(30) Noyori, R.; Hashiguchi, S. Acc. Chem. Res. 1997, 30, 97-102.
(31) Clapham, S. E.; Hadzovic, A.; Morris, R. H. Coord. Chem. ReV. 2004,
248, 2201-2237.
To estimate the quality of the docking procedure, biotin and
the docked [η6-(arene)Ru(Biot-p-L)H] were superimposed in
streptavidin, and the root-mean-square error values (rms) were
(32) Letondor, C.; Humbert, N.; Ward, T. R. Proc. Natl. Acad. Sci. U.S.A. 2005,
102, 4683-4687.
(33) Reetz, M. T.; Li, X. J. Am. Chem. Soc. 2006, 128, 1044-1045.
(34) Palmer, M. J.; Wills, M. Tetrahedron Asymmetry 1999, 10, 2045-2061.
(35) Blaser, H.-U.; Malan, C.; Pugin, B.; Spindler, F.; Steiner, H.; Studer, M.
AdV. Synth. Catal. 2003, 345, 103-151.
(42) Morris, G. M.; Goodsell, D. S.; Halliday, R. S.; Huey, R.; Hart, W. E.;
Belew, R. K.; Olson, A. J. J. Comput. Chem. 1998, 19, 1639-1662.
(43) Goodsell, D. S.; Morris, G. M.; Olson, A. J. J. Mol. Recognit. 1996, 9,
1-5.
(36) Wu, X.; Vinci, D.; Ikariya, T.; Xiao, J. Chem. Commun. 2005, 4447-
4449.
(37) Pa`mies, O.; Ba¨ckvall, J.-E. Chem. ReV. 2003, 103, 3247-3261.
(38) Muniz, K. Angew. Chem., Int. Ed. 2005, 44, 6622-6627.
(39) Yamakawa, M.; Ito, H.; Noyori, R. J. Am. Chem. Soc. 2000, 122, 1466-
1478.
(44) Weber, P. C.; Ohlendorf, D. H.; Wendoloski, J. J.; Salemme, F. R. Science
1989, 243, 85-88.
(45) Uematsu, N.; Fujii, A.; Hashiguchi, S.; Ikariya, T.; Noyori, R. J. Am. Chem.
Soc. 1996, 118, 4916-4917.
(40) Yamakawa, M.; Yamada, I.; Noyori, R. Angew. Chem., Int. Ed. 2001, 40,
2818-2821.
(41) Brandt, P.; Roth, P.; Andersson, P. G. J. Org. Chem. 2004, 69, 4885-
4890.
(46) Cornell, W. D.; Cieplak, P.; Bayly, C. I.; Gould, I. R.; Merz, K. M., Jr;
Ferguson, D. M.; Spellmeyer, D. C.; Fox, T.; Caldwell, J. W.; Kollman, P.
A. J. Am. Chem. Soc. 1995, 117, 5179-5197.
9
J. AM. CHEM. SOC. VOL. 128, NO. 25, 2006 8321

A R T I C L E S
Letondor et al.
Figure 1. Computed average Ru-CR distances for [η6-(p-cymene)Ru(Biot-p-L)H] ⊂ WT Sav resulting from the 135 docking simulations with rms < 1
Å (a); zoom-in of the critical L7,8 Ru-CR loop distances including shortest and longest Ru-CR distances computed for the 135 docking simulations (blue
and green lines respectively) (b).
Scheme 2. Structure and Variable Parameters of
C-CR distances <8 Å are computed with residue S112. This
[η6-(p-cymene)Ru((Biot-p-L)H] Used for the Docking Studiesa
suggests that there may be significant interactions between the
side chain amino acid residue at this position and the η6-bound
arene.
The structure of the most stable [η6-(p-cymene)Ru(Biot-p-
L)H] ⊂ WT Sav is presented in Figure 2. As can be appreciated,
the ruthenium moiety (ball-and-stick representation) is embedded
within a cavity provided by monomers A, B, and C of
streptavidin (green, yellow, and blue solvent accessible surfaces).
In this supramolecular assembly [η6-(p-cymene)Ru(Biot-p-
a The dihedral angle (purple) was set to τ ) 0° and 180°, the absolute
configuration at ruthenium was set to (R) and (S) thus yielding a total of
L)H] ⊂ WT Sav, the catalytic moiety remains well accessible
four isomers used for the docking in dimeric streptavidin; rotation around
the bonds highlighted in green was allowed during the energy minimization
procedure.
to an incoming prochiral substrate.
The two closest lying amino acid residues, S112 and S122,
are highlighted (red and blue stick representation). The modeled
determined for biotin’s bicyclic scaffold. Only the structures
structure shows that the S112 side chain points toward the piano
with an rms < 1 Å were retained for further analysis.
stool moiety, whereas the S122 side chain points away. As a
Fifty runs were performed for each of the four [η6-(p-
consequence, the computed serine oxygens are repectively 5.61
cymene)Ru(Biot-p-L)H] isomers ((R)- and (S)- configurations
and 9.15 Å away from the ruthenium center. Very similar
at Ru and τ ) 0° and 180°), yielding a total of 200 docked
qualitative results are obtained for [η6-(benzene)Ru(Biot-p-
structures [η6-(p-cymene)Ru(Biot-p-L)H] ⊂ WT Sav. Follow-
L)H] ⊂ WT Sav.47
ing the docking in dimeric streptavidin, the complete tetrameric
In the absence of precise structural information, we assume
host was redisplayed. For the docked structures with rms < 1
that mutations close to the active site bring more diversity than
Å (135 in total), the Ru-CR distances (CR is the asymmetric
distant mutations.48 With the aim of steering the delivery of
carbon of each amino acid residue in the tetrameric streptavidin)
the incoming prochiral substrate, we favored position S112 over
were computed and are displayed in Figure 1. For the other B,
S122 for saturation mutagenesis as its side chain points toward
C, and D subunits, only one Ru-CR contact shorter than 10 Å
the ruthenium.
was computed (with Lys 121, subunit C, 9.42 Å).
From the Ru-CR distances, it appears that only two amino
acid residues display an average distance Ru-CR < 8 Å. These
are: S112, S122, both part to the L 7,8 loop of subunit A of
streptavidin, in which the [η6-(p-cymene)Ru(Biot-p-L)H] was
Screening Strategy. To generate diversity, we combine
chemical- with genetic optimization strategies. Combining m
streptavidin isoforms with n biotinylated catalyst precursors
yields a two-dimensional diversity matrix of n‚m experiments.
In our previous study on artificial metalloenzymes for the
docked. For the S112- and the S122 residues, the computed
hydrogenation of N-protected dehydroamino acids, we estab-
average Ru-CR distances are 7.41 and 7.51 Å with standard
deviations (SD) ) 0.91 Å and SD ) 1.43 Å respectively. The
(47) It is interesting to note that the amino acid P64 of subunit B (yellow stick
extreme distances displayed as green and blue lines in Figure
1b thus overemphasize the scattering of the calculated Ru-CR
distances for a particular set of docking simulations. A histogram
of the distances for both S112- and S122 residues is provided
in the Supporting Information. For the η6-bound arene, three
representation, average Ru-CR distance 15.15 Å) is within van der Waals
contact of the loop L7,8 (which contains residues 112-122) of subunit A
of streptavidin. This observation suggests that the P64G mutation may act
via a relay mechanism: influencing the L7,8 loop, which in turn, affects
the chiral discrimination event with [η6-(arene)Ru(Biot-p-L)H] ⊂ P64G
Sav.
(48) Morley, K. L.; Kazlauskas, R. J. Trends Biotechnol. 2005, 23, 231-237.
9
8322 J. AM. CHEM. SOC. VOL. 128, NO. 25, 2006

Chemogenetics and Artificial Metalloenzymes
A R T I C L E S
Figure 2. Result of the docking simulation between [η6-(p-cymene)Ru(Biot-p-L)H] (ball-and-stick representation) and WT streptavidin (transparent green,
blue, yellow, and violet for solvent accessibility in subunits A, B, C, and D respectively and schematic secondary structure for A and C subunits). Close
lying S112 and S122 residues are highlighted (red and blue stick representations respectively), residues P64 (subunit B, yellow stick representation), W120
(subunit C, white stick representation), the hydrophobic lid which shields the biotin binding site of subunit A as well as K121 (subunit C, white stick
representation).
lished that the chemical optimization brings more diversity than
the genetic counterpart.27 To reduce the number of screening
experiments, we opted for a representational search strategy49
to optimize the selectivity of the artificial metalloenzymes.
In a first screening round, all 21 catalyst precursors were
evaluated in combination with two streptavidin isoforms. For
the second round, the most promising biotinylated catalysts were
screened with the 20 streptavidin isoforms derived from
saturation mutagenesis at position S112. Finally, the best catalyst
⊂ protein combinations were tested with different ketones to
evaluate the substrate scope of the artificial metalloenzymes.
This optimization strategy is sketched in Figure 3. The substrates
and their corresponding reduction products used in this study
are depicted in Scheme 3.
Buffer-screening experiments revealed that the initial pH is
critical both for the activity and for the selectivity of the artificial
transfer hydrogenases.50-52 The pH-dependence profile for the
reduction of acetophenone is depicted in Figure 4. As the
conversion increases sharply between pH 6.0-6.5 at the cost
Figure 3. Representational search strategy49 allowing a rapid identification
of good catalyst ⊂ protein combinations without examination of all
possibilities.
of a modest erosion of enantioselectivity, we set the pHinitial
)
possesses no buffering capacity at this pH, the pH rises sharply
during catalysis, resulting from formate dehydrogenase activity
of the catalyst. Addition of MOPS (3-(N-morpholino)propane-
sulfonic acid sodium salt, 0.16 M) allows one to maintain the
pH < 7.3 (pHinitial ) 6.25) throughout catalysis.
6.25. Combining sodium formate (pKa ) 3.75) with boric acid
(pK′a ) 9.24) affords an acid-base mixture which, at its
isoelectric point, displays a pH ) 0.5‚(pKa + pK′a) ) 6.50.
Using a 0.48 M NaHCO2 combined with 0.41 M boric acid
yields a solution with pH ) 6.25. However, as this mixture
Genetic Diversity. Saturation mutagenesis was carried out
according to the quick-change mutagenesis protocol using the
degenerate NNS codons at position S112. Due to the abundance
of G‚C base pairs in the Sav gene, the polymerase chain reaction
(PCR) conditions were adapted to ensure proper amplification
of the gene. The optimized PCR conditions are: 5% DMSO,
(49) Shimizu, K. D.; Snapper, M. L.; Hoveyda, A. H. Chem. Eur. J. 1998, 4,
1885-1889.
(50) Wu, X.; Li, X.; Hems, W.; King, F.; Xiao, J. Org. Biomol. Chem. 2004, 2,
1818-1821.
(51) Ogo, S.; Abura, T.; Watanabe, Y. Organometallics 2002, 21, 2964-2969.
(52) Wu, X.; Li, X.; King, F.; Xiao, J. Angew. Chem., Int. Ed. 2005, 44, 3407-
3411.
9
J. AM. CHEM. SOC. VOL. 128, NO. 25, 2006 8323

A R T I C L E S
Letondor et al.
Scheme 3. Substrates and Reduction Products Used in This
Studya
Figure 4. pH-dependence profile of the activity (red circles) and of the
selectivity (blue triangles) for the reduction of acetophenone 1a with [η6-
(p-cymene)Ru(Biot-p-L)H] ⊂ WT Sav.
Complete experimental details for the ligand synthesis as well
as the catalytic runs are compiled in the Supporting Information.
For the first optimization step, the 21 catalyst precursors [ηn-
(CnRn)M(Biot-q-L)Cl] were screened in the presence of WT
Sav as well as P64G Sav. This mutant was selected as [η6-(p-
cymene)Ru(Biot-p-L)Cl] ⊂ P64G Sav had previously been
identified as the best mutant for the enantioselective reduction
of p-methylacetophenone 1c.32 To accelerate the optimization
process, a cocktail containing three aromatic substrates (ac-
etophenone 1a, p-bromoacetophenone 1b, and p-methylac-
etophenone 1c) were screened simultaneously using a 3 mol %
catalyst loading (i.e. 11 equiv of each substrate vs Ru). Control
experiments established that the selectivity and the conversion
obtained with this cocktail are similar, but not always identical,
to those obtained with a single substrate (i.e., small autoinduc-
tion).54
a Cocktail compositions: 1a, 1b and 1c; 5 and 7; 9 and 11. R-tetralone
3 was screened individually.
DNA denaturated at 95 °C for 5 min, followed by 16 cycles
(denaturation at 95 °C, 1 min; primer hybridization 65 °C, 1
min; DNA polymerization 68 °C, 15 min). A final elongation
(65 °C, 1 h) completed the PCR.
All 20 S112X Sav isoforms could be isolated and purified
by iminobiotin affinity chromatography, thus demonstrating the
biotin-binding capability of the mutants.53
Chemical Diversity. A total of 21 biotinylated d6-piano stool
complexes [ηn-(arene)M(Biot-q-L)Cl] (ηn-arene ) η6-p-
cymene, η6-benzene, η6-mesitylene, η6-durene, η6-hexameth-
ylbenzene, η5-pentamethylcyclopentadienyl; M ) Ru(II), Ir(III),
and Rh(III); q ) ortho-, meta-, and para-) were prepared in
situ by reacting [ηn-(arene)MCl2]2 with the appropriate ligand
in refluxing 2-propanol in the presence of NEt3 (Scheme 4).
The results of the screening experiments with all three
substrates are summarized in Figure 5 using a fingerprint display
for each substrate-protein-ligand combination.27,55 The selec-
tivity is color-coded: strawberry for (S)-selective and green for
(R)-selective ligand-protein combinations.56 The intensity of
the color reflects the conversion. This convenient display allows
Scheme 4. Chemical Diversity Generated by Combining [Cp*RhCl]-, [Cp*IrCl]- or [η6-(arene)RuCl] Fragments with (Biot-p-L) (q- ) ortho-,
meta- and para-)
9
8324 J. AM. CHEM. SOC. VOL. 128, NO. 25, 2006

Chemogenetics and Artificial Metalloenzymes
A R T I C L E S
selective, good conversions) > [η6-(durene)Ru(Biot-p-L)Cl]
((R)-selective, good conversions) > [η6-(benzene)Ru(Biot-p-
L)Cl] ((S)-selective, moderate conversions) > [η5-(C5Me5)Rh-
(Biot-p-L)Cl] ((S)-selective, moderate conversions) > [η5-
(C5Me5)Ir(Biot-p-L)Cl] ((R)-selective, moderate conversions).
These five complexes were combined with the 19 mutants
resulting from saturation mutagenesis at position S112X Sav.
The results of the cocktail screening are displayed in a
fingerprint format in Figure 6. For convenience, the column
vectors are arranged according to the amino acid properties at
position 112. To refine the trends observed in the fingerprint
display (Figure 6), enantioselectivity histograms are displayed
for five representative S112X mutants and for the five com-
plexes in Figure 7a and b, respectively. Table 1 summarizes
the results of the best catalytic runs, which were reproduced
with a single substrate and a 1 mol % catalyst loading (i.e. 100
equiv of substrate vs Ru).
The general trends that emerge from this chemogenetic
optimization can be summarized as follows:
(i) Both the [η6-(p-cymene)Ru(Biot-p-L)Cl] and the [η6-
(durene)Ru(Biot-p-L)Cl] complexes exert a strong prefence
in favor of the (R)-products (ee > 85% (R) in 9 cases, Table 1,
entries 1-5). This propensity can be partially overruled in the
presence of cationic residues at position S112, favoring (S)-
products with modest conversions and selectivity (up to 20(S)
% ee, Table 1 entry 6). Thus, using [η6-(p-cymene)Ru(Biot-
p-L)Cl], introduction of a single point mutation at position S112
can afford reduction product with differences in enantioselec-
tivity ∆ee < 110%. This corresponds to differences in transition-
state free energies δ∆Gq < 1.99 kcal‚mol-1 at room temperature.
(ii) Altough less pronounced, [η6-(benzene)Ru(Biot-p-L)-
Cl] favors mostly the (S)-enantiomers (Figure 7b). However,
certain mutations at position S112 (S112F Sav, S112Y Sav,
S112A Sav) can overrule this preference (Table 1, entry 7).
(iii) The selectivities of the [η5-(C5Me5)Rh(Biot-p-L)Cl]-
and [η5-(C5Me5)Ir(Biot-p-L)Cl]-containing artificial metal-
loenzymes depend very much on the host protein. For the
reduction of bromoketone 1b using [η5-(C5Me5)Rh(Biot-p-L)-
Cl], the S112F Sav and S112G Sav afford (R)-product (73%
conversion, 60% ee) and (S)-product (60% conversion, 52% ee),
respectively (Table 1, entries 8-9).
Figure 5. Fingerprint display of the results for the chemical optimization
of the transfer hydrogenation of a cocktail containing acetophenone 1a,
p-bromoacetophenone 1b and p-methylacetophenone 1c in the presence of
21 biotinylated d6-piano stool complexes [ηn-(arene)M(Biot-q-L)Cl] with
either WT Sav or P64G Sav as host proteins. The catalytic runs were
performed at 55 °C for 40 h using the mixed buffer HCO2Na (0.48 M) +
B(OH)3 (0.41 M) + MOPS (0.16 M) at pHinitial ) 6.25. Ru: substrates 1a-
c: formate ratio 1:33:4000 (i.e. 11 equiv of each substrate vs Ru). Please
refer to the Supporting Information for a tabulation of all catalytic runs.
rapid identification of interesting ligand-protein combinations.
Several interesting features are apparent from these data:
(i) Only catalyst precursors bearing a para-substituted ami-
nosulfonamide Biot-p-L afford any significant levels of
reduction products.
(ii) The nature of the ηn-(arene) plays a critical role in
determining both the activity and the selectivity of the corre-
sponding hybrid catalysts. Although the [η6-(mesitylene)Ru-
(Biot-p-L)Cl] is a poor catalyst, the [η6-(durene)Ru(Biot-p-
L)Cl] rivals with the best catalyst [η6-(p-cymene)Ru(Biot-p-
L)Cl].
(iv) The conversion decreases with increasing steric bulk at
position S112X Sav. For example: S112A > S112I ≈ S112L;
S112N > S112Q; S112F > S112W.
(v) Potentially coordinating amino acid residues at position
S112 (i.e. S112C, S112D, S112E, and S112H and to a lesser
extent S112M) shut off the catalytic activity. This observation
supports the docking simulations which suggest that the S112X
side chain indeed lies close to the [M(Biot-p-L)]-moiety.
(vi) Overall, the Ru catalysts outperform the Ir and Rh
analogues both in terms of activity and of selectivity.
Having identified the most efficient biotinylated d6-piano stool
⊂ S112X Sav mutant, we proceeded to test different substrates.
For this purpose, we selected the sterically constrained aryl
ketone R-tetralone 3, 2-acetylpyridine 5, and three dialkyl
ketones 7, 9, and 11. These five substrates were screened with
both [η6-(p-cymene)Ru(Biot-p-L)Cl] and [η6-(benzene)Ru-
(Biot-p-L)Cl] in the presence of representative Sav mutants:
S112G, S112A, S112L (aliphatic residues with increasing bulk);
S112F and S112W (aromatic residues with increasing bulk);
(iii) Although P64G Sav was previously identified as the best
host protein in terms of enantioselectivity, both WT Sav and
P64G Sav show similar trends.32
(iv) The most activated bromoketone 1b gives the highest
conversions. In general, the bulkier ketones 1b and 1c give the
best enantioselectivity.
On the basis of this initial screening, we selected five
biotinylated piano stool complexes for the genetic optimization
with S112X mutants: [η6-(p-cymene)Ru(Biot-p-L)Cl] ((R)-
(53) Humbert, N.; Zocchi, A.; Ward, T. R. Electrophoresis 2005, 26, 47-52.
(54) Satyanarayana, T.; Kagan, H. B. AdV. Synth. Catal. 2005, 347, 737-748.
(55) Wahler, D.; Badalassi, F.; Crotti, P.; Reymond, J.-L. Chem. Eur. J. 2002,
8, 3211-3228.
(56) To generate the RGB codes for a catalytic experiment yielding one of the
products in x% yield and y%(R) and z%(S) (100% - y% ) z% (S)) the
following formulas were implemented in an Excel Macro: RGB1 ) [100
- (y‚x)/100]‚2.55; RGB2 ) [100 - (z‚x)/100]‚2.55; RGB3 ) 0.5‚
(RGB1+RGB2).
9
J. AM. CHEM. SOC. VOL. 128, NO. 25, 2006 8325

A R T I C L E S
Letondor et al.
Figure 6. Fingerprint display of the results for the genetic optimization of the transfer hydrogenation of a cocktail containing acetophenone 1a,
p-bromoacetophenone 1b and p-methylacetophenone 1c in the presence of five biotinylated d6-piano stool complexes [ηn-(arene)M(Biot-p-L)Cl] combined
with the 20 streptavidin isoforms S112X Sav. The catalytic runs were performed at 55 °C for 40 h using the mixed buffer HCO2Na (0.48 M) + B(OH)3 (0.41
M) + MOPS (0.16 M) at pHinitial ) 6.25. Ru: substrates 1a-c: formate ratio 1:33:4000 (i.e. 11 equiv of each substrate vs Ru). Please refer to the Supporting
Information for a tabulation of all catalytic runs.
Figure 7. Enantioselectivity histograms as a function of the genetic component (S112X, X ) Y, F, A, R, K mutants displayed) (a); and the chemical
component (b).
WT Sav and S112T (polar residues); S112E (coordinating
residue); S112R (cationic residue) and S112N.
The results of the screening are summarized in a fingerprint
format in Figure 8. Numerical data are collected in Table 2.
The general features can be summarized as follows:
(i) The selectivity and conversion trends observed for these
substrates are similar to those observed with acetophenone
derivatives 1a-c: [η6-(p-cymene)Ru(Biot-p-L)Cl] and [η6-
(benzene)Ru(Biot-p-L)Cl] often yield opposite enantiomers,
with better conversions for the former catalyst precursor.
(ii) The dialkyl ketones give poor to modest enantioselec-
tivities (Table 2, entries 1-4).
(iii) For the reduction of 2-acetylpyridine 5 in the presence
of a cationic residue at position 112, (i.e., S112R and S112K),
both [η6-(p-cymene)Ru(Biot-p-L)Cl] and [η6-(benzene)Ru-
(Biot-p-L)Cl] afford (S)-products (51% ee and 70% ee,
9
8326 J. AM. CHEM. SOC. VOL. 128, NO. 25, 2006

Chemogenetics and Artificial Metalloenzymes
A R T I C L E S
Table 1. Selected Conversions and Enantioselectivities Obtained
for the Transfer Hydrogenation of Acetophenone Derivatives 1
a-c Catalyzed by [ηn-(CnRn)M(Biot-p-L)Cl] ⊂ S112Xa
Table 2. Selected Conversions and Enantioselectivities Obtained
for the Transfer Hydrogenation of Ketones 3, 5, 7, 9, and 11
Catalyzed by [η6-(arene)Ru(Biot-p-L)Cl] ⊂ S112Xa
conv.
conv.
entry
protein
complex
substrate
(%)
ee (%)
entry
protein
complex
substrate
(%)
ee (%)
1
2
3
4
5
6
7
8
9
S112A Sav [η6-(p-cymene)Ru(Biot-p-L)Cl]
WT Sav
[η6-(p-cymene)Ru(Biot-p-L)Cl]
1c
1c
1a
1b
1b
1a
1c
1b
1b
1b
1b
1b
1b
1b
98 91(R)
86 91(R)
95 90(R)
88 92(R)
89 83(R)
64 20(S)
74 41(R)
73 60(R)
60 52(S)
48 52(S)
90 55(S)
95 16(S)
96 80(R)
94 82(R)
1
2
3
4
5
6
7
8
9
S112A Sav [η6-(p-cymene)Ru(Biot-p-L)Cl]
S112A Sav [η6-(p-cymene)Ru(Biot-p-L)Cl]
S112A Sav [η6-(benzene)Ru(Biot-p-L)Cl]
S112A Sav [η6-(p-cymene)Ru(Biot-p-L)Cl]
S112R Sav [η6-(p-cymene)Ru(Biot-p-L)Cl]
S112R Sav [η6-(benzene)Ru(Biot-p-L)Cl]
S112F Sav [η6-(p-cymene)Ru(Biot-p-L)Cl]
S112F Sav [η6-(p-cymene)Ru(Biot-p-L)Cl]
S112Y Sav [η6-(p-cymene)Ru(Biot-p-L)Cl]
11
9
9
7
5
5
5
3
3
3
97 69(R)
98 48(R)
58 52(R)
71 30(R)
94 51(S)
95 70(S)
95 76(R)
70 96(R)
79 97(R)
44 51(S)
S112Y Sav [η6-(p-cymene)Ru(Biot-p-L)Cl]
S112Y Sav [η6-(durene)Ru(Biot-p-L)Cl]
S112F Sav [η6-(durene)Ru(Biot-p-L)Cl]
S112K Sav [η6-(p-cymene)Ru(Biot-p-L)Cl]
S112A Sav [η6-(benzene)Ru(Biot-p-L)Cl]
S112F Sav [η5-(C5Me5)Rh(Biot-p-L)Cl]
S112G Sav [η5-(C5Me5)Rh(Biot-p-L)Cl]
10 WT Sav
[η5-(C5Me5)Rh(Biot-p-L)Cl]
10 S112A Sav [η6-(benzene)Ru(Biot-p-L)Cl]
11 S112T Sav [η6-(benzene)Ru(Biot-p-L)Cl]
12 S112T Sav [η5-(C5Me5)Ir(Biot-p-L)Cl]
13 S112Y Sav [η5-(C5Me5)Ir(Biot-p-L)Cl]
14 S112N Sav [η6-(p-cymene)Ru(Biot-p-L)Cl]
a The catalytic runs were performed at 55 °C for 64 h using the mixed
buffer HCO2Na (0.48 M) + B(OH)3 (0.41 M) + MOPS (0.16 M) at pHinitial
) 6.25. Ru: substrates: formate ratio 1:100: 4000 (i.e. 100 equiv of substrate
vs Ru). The results presented here represent the average of at least two
catalytic runs.
a The catalytic runs were performed at 55 °C for 64 h using the mixed
buffer HCO2Na (0.48 M) + B(OH)3 (0.41 M) + MOPS (0.16 M) at pHinitial
) 6.25. Ru: substrates 1a-c: formate ratio 1:100: 4000 (i.e. 100 equiv of
substrate vs Ru). The results presented here represent the average of at
least two catalytic runs.
Scheme 5 . Computed Transition-State Structures Leading to Both
Enantiomers of the Product (adapted from ref 40)
(i) Both hydrogens required for the reduction of the ketone
are provided by the coordinatively saturated d6-piano stool
complex: one hydride and one acidic N-H proton from the
coordinated amine.
(ii) The computed difference in transition-state energies for
the diastereomeric complexes, which ultimately yield both
enantiomers of the product, is 2.1 kcal‚mol-1. This difference
in energy is by-and-large governed by attractive CH/π interac-
tions between the aromatic substrate and the η6-bound arene.
The theoretical study reveals that these CH groups can be either
aromatic or pendant aliphatic groups on the η6-bound arene
(Scheme 5).
To test the validity of the involvement of an acidic N-H
proton, the N-dimethyl analogue Biot-p-LMe2 was prepared
and tested. The catalytic performance of [η6-(p-cymene)Ru-
(Biot-p-LMe2)Cl] ⊂ WT Sav was evaluated: no conversion
(<3%) was observed either for the reduction of acetophenone
1a or 2-acetylpyridine 5. Introduction of a cationic residue at
position S112 (i.e., S112K) does not increase the yield, thus
suggesting that the protonation of the substrate indeed requires
an acidic proton in the first coordination sphere of the catalyst.
Although the enantioselectivity is mostly governed by the
nature of the biotinylated moiety (see Figure 7b), point mutations
can, in some cases, lead to an inversion of enantioselectivity.
In such cases, we speculate that protein-substrate interactions
partially outweigh the dominant CH/π interactions, which favor
one diastereomeric transition state over the other. This tendency
is particularly pronounced with cationic residues at position
S112, which yield preferentially (S)-reduction products, ir-
respective of the biotinylated catalyst (see Figure 7a). This trend
Figure 8. Fingerprint display of the results for the transfer hydrogenation
of the aromatic ketones 3 and 5 and the aliphatic ketones 7, 9, and 11 in
the presence of two biotinylated d6-piano stool complexes [η6-(arene)Ru-
(Biot-p-L)Cl] (arene ) p-cymene, benzene) combined with 10 representa-
tive S112X streptavidin isoforms. The catalytic runs were performed with
cocktail mixtures of the substrates at 55 °C for 64 h using the mixed buffer
HCO2Na (0.48 M) + B(OH)3 (0.41 M) + MOPS (0.16 M) at pHinitial
)
6.25. Ru: substrates: formate ratio 1:100: 4000. Please refer to the Supporting
Information for a tabulation of all catalytic runs.
respectively, Table 2, entries 5-6). This suggests that the
positive charge at this position by-and-large dictates which
prochiral face of the basic substrate is presented to the
biotinylated piano stool complex.
Mechanistic Considerations. The starting point for this study
was the theoretical prediction that, in the related homogeneous
catalytic system, both hydrogens are delivered to the prochiral
substrate without the substrate binding to the metal.39,40 In the
context of this work, two points are worthy of emphasis:
9
J. AM. CHEM. SOC. VOL. 128, NO. 25, 2006 8327

A R T I C L E S
Letondor et al.
is most prominent for 2-acetylpyridine, which may undergo
hydrogen-bonding interactions with the cationic residues at
position S112. For the other aromatic substrates, we suggest
that cation/π interactions57 favor TS b. In the presence of
nonaromatic ketones 7, 9, and 11, there is no clear-cut preference
between Re or Si face reduction as the CH/π attraction is less
pronounced or absent. The additional degrees of freedom
introduced between the aryl “steric recognition” element and
the prochiral ketone do not allow the second coordination sphere
interactions to impose an efficient discrimination. As a result,
the enantioselectivities for these substrates remain modest. On
the basis of these observations, we suggest that the enantiose-
lection mechanism is similar to that proposed by Noyori and
co-workers (see Schemes 1 and 5). Although this remains to
be demonstrated, we believe that, upon incorporation in the host
protein, the nature of the ηn-bound arene by-and-large dictates
the absolute configuration of the metal, which, in turn determines
the selectivity trends apparent in Figure 7b.
(iv) The presence of a coordinating amino acid residue at
position S112 totally shuts off catalytic actvity. This confirms
the docking simulations which suggest that this residue lies close
to the metal in [η6-(p-cymene)Ru(Biot-p-L)H] ⊂ WT Sav.
(v) Good enantioselectivities can be achieved for aromatic
ketones (up to 97% ee (R) for 4). Nonaromatic ketones are
reduced with modest enantioselectivity (up to 69% ee (R) for
12).
These results suggest that, saturation mutagenesis at a residue
close to the computed position of the biotinylated catalyst,
combined with chemical variations of the d6-piano stool moiety,
are a powerful tool for the optimization of artificial transfer
hydrogenases based on the biotin-avidin technology. Current
efforts include the structural characterization of the artificial
metalloenzyme as well as the implementation of a high-
throughput assay.58 Such an assay will allow us to fully exploit
the potential of directed-evolution methodologies combined with
chemical optimization (i.e., chemogenetic6) to produce artificial
metalloenzymes which unite the best of homogeneous and
enzymatic catalysis.
Summary and Outlook
Screening of 21 biotinylated d6-piano stool complexes [ηn-
(CnRn)M(Biot-q-L)] in the presence of 20 streptavidin isoforms
(S112X Sav) for the enantioselective transfer hydrogenation of
prochiral substrates has revealed several noteworthy features:
(i) The localization of the biotinylated catalyst, imposed by
the aromatic spacer, plays an essential role in determining the
activity of the resulting artificial metalloenzyme: only the para-
substituted ligand Biot-p-L affords significant conversions.
(ii) The Ru-based catalysts outperform both the Rh- and Ir-
based systems. This contrasts with a recent report that demon-
strated the versatility of a [Cp*Rh]-based catalyst for the
enantioselective aqueous transfer hydrogenation of acetophenone
derivatives.36
(iii) The nature of the η6-bound arene plays a critical role in
the enantioselectivity. Fine-tuning is achieved by saturation
mutagenesis at position S112. The highest enantioselectivities
are achieved with either aromatic residues (i.e., S112F and
S112Y) for (R)-products or with cationic residues (i.e., S112R
and S112K) for (S)-products.
Acknowledgment. We thank Prof. C. R. Cantor for the
streptavidin gene as well as Profs. J.-M. Neuhaus and P.
Schu¨rmann for their help with protein production. We thank
Dr. C. Malan and S. Guerra for the synthesis of Biot-p-LMe2.
This work was funded by the Swiss National Science Foundation
(FN 200021-105192/1), CERC3 (Grant FN20C321-101071), the
Roche Foundation, the Canton of Neuchaˆtel as well as FP6
Marie Curie Research Training Network (IBAAC network,
MRTN-CT-2003-505020). We thank Umicore Precious Metals
Chemistry for a loan of ruthenium.
Supporting Information Available: Listing of all catalytic
runs performed as cocktails, complete experimental procedures,
including characterization of intermediates and ligands, and
transfer hydrogenation protocol as well as HPLC and GC
methods used. This material is available free of charge via the
Internet at http://pubs.acs.org.
JA061580O
(57) Scha¨rer, K.; Morgenthaler, M.; Paulini, R.; Obst-Sander, U.; Banner, D.
W.; Schlatter, D.; Benz, J.; Stihle, M.; Diederich, F. Angew. Chem., Int.
Ed. 2005, 44, 4400-4404.
(58) Taran, F.; Gauchet, C.; Mohar, B.; Meunier, S.; Valleix, A.; Renard, P.
Y.; Cre´minon, C.; Grassi, J.; Wagner, A.; Mioskowski, C. Angew. Chem.,
Int. Ed. 2002, 41, 124-127.
9
8328 J. AM. CHEM. SOC. VOL. 128, NO. 25, 2006