Artificial golgi apparatus: Globular protein-like dendrimer facilitates fully automated enzymatic glycan synthesis

Article abstract of DOI:10.1021/ja106955j

Despite the growing importance of synthetic glycans as tools for biological studies and drug discovery, a lack of common methods for the routine synthesis remains a major obstacle. We have developed a new method for automated glycan synthesis that employs

Full text of DOI:10.1021/ja106955j

Published on Web 10/29/2010
Artificial Golgi Apparatus: Globular Protein-like Dendrimer
Facilitates Fully Automated Enzymatic Glycan Synthesis
Takahiko Matsushita,† Izuru Nagashima,‡ Masataka Fumoto,§ Takashi Ohta,‡
Kuriko Yamada,| Hiroki Shimizu,‡ Hiroshi Hinou,† Kentaro Naruchi,† Takaomi Ito,§
Hirosato Kondo,§ and Shin-Ichiro Nishimura*,†,‡
Graduate School of Life Science, Hokkaido UniVersity, N-21, W-11, Kita-ku, Sapporo 001-0021, Japan,
Drug-Seeds DiscoVery Research Laboratory, National Institute of AdVanced Industrial Science
and Technology (AIST), 2-17-2-1, Tsukisamu-higashi, Toyohira-ku, Sapporo 062-8517, Japan,
Shionogi & Company, Ltd., 12-4, Sagisu 5-chome, Fukushima-ku, Osaka 541-0045, Japan, and
Hitachi High-Technologies Company, Minato-ku, Tokyo 105-8717, Japan
Received August 24, 2010; E-mail: shin@glyco.sci.hokudai.ac.jp
Abstract: Despite the growing importance of synthetic glycans as tools for biological studies and drug
discovery, a lack of common methods for the routine synthesis remains a major obstacle. We have developed
a new method for automated glycan synthesis that employs the enzymatic approach and a dendrimer as
an ideal support within the chemical process. Recovery tests using a hollow fiber ultrafiltration module
have revealed that monodisperse G6 (MW ) 58 kDa) and G7 (MW ) 116 kDa) poly(amidoamine)
dendrimers exhibit a similar profile to BSA (MW ) 66 kDa). Characteristics of the globular protein-like G7
dendrimer with high solubility and low viscosity in water greatly enhanced throughput and efficiency in
automated synthesis while random polyacrylamide-based supports entail significant loss during the repetitive
reaction/separation step. The present protocol allowed for the fully automated enzymatic synthesis of sialyl
Lewis X tetrasaccharide derivatives over a period of 4 days in 16% overall yield from a simple N-acetyl-
D-glucosamine linked to an aminooxy-functionalized G7 dendrimer.
Introduction
selective protections/deprotections of the multiple hydroxyl
groups present on individual sugars.6-8 Taking the ability to
It is clear that large amounts of structurally defined homo-
geneous samples of nucleic acids and peptides have greatly
promoted basic research and pharmaceutical/medical applica-
tions of these biomolecules. Solid-phase chemical synthesis for
peptides proposed by Merrifield1 was also utilized for oligo-
nucleotide (DNA/RNA) synthesis, and enabling the subsequent
use of automated synthesizers based on this methodology. As
is the case of glycans, which is the third major class of
biopolymers next to nucleic acids and proteins, solid-phase
chemical synthesis of oligosaccharides has also been studied2,3
since it was first reported by Frechet and Schuerch4 in 1971
and, Seeberger et al. developed the first automated carbohydrate
synthesizer based on a solid-phase chemical strategy in 2001.5
However, solid-phase chemical synthesis of oligosaccharides
still involves a longstanding characteristic problem. Although
extensive efforts have been made toward the development of
practical chemical synthetic protocols since the advent of these
powerful glycosylation techniques, stereoselective and regiose-
lective glycosylations definitely require multistep syntheses for
access the homogeneous glycans into account, these disadvan-
tages are extremely serious for the sequential chemical synthesis
on solid materials because purification of products is impossible
until the final cleavage from their solid supports. Furthermore,
tedious/time-consuming procedures are also needed for the
removal of multiple byproduct attached to solid supports such
as R/ꢀ stereoisomers, regioisomers, unreacted intermediates and
so on. On the contrary, enzyme-assisted synthesis is an attractive
alternative to chemical synthesis because there is the advantage
of accomplishing regio- and stereoselective glycosylations for
versatile glycoconjugate acceptors according to the substrate
specificity and/or tolerance inhered to enzymes.9-11 Although
the scope of glycan synthesis by mammalian glycosyltrans-
ferases is still restricted due to the limitation of available
recombinant glycosyltransferases with satisfactory activity and
stability required for synthetic purposes, it is generally believed
that enzymes from various microbial sources should provide a
† Hokkaido University.
‡ Drug-Seeds Discovery Research Laboratory.
§ Shionogi & Co., Ltd.
(6) Crich, D.; Dudkin, V. J. Am. Chem. Soc. 2001, 123, 6819–6825.
(7) Zhu, X. M.; Schmidt, R. R. Angew. Chem., Int. Ed. 2009, 48, 1900–
1934.
| Hitachi High-Technologies Co.
(1) Merrifield, R. B. J. Am. Chem. Soc. 1963, 85, 2149–2154.
(2) Sears, P.; Wong, C.-H. Science 2001, 291, 2344–2350.
(3) Seeberger, P. H. Chem. Soc. ReV. 2008, 37, 19–28.
(4) Frechet, J. M.; Schuerch, C. J. Am. Chem. Soc. 1971, 93, 492–496.
(5) Plante, O. J.; Palmacci, E. R.; Seeberger, P. H. Science 2001, 291,
1523–1527.
(8) Boltje, T. J.; Buskas, T.; Boons, G.-J. Nat. Chem. 2009, 1, 611–622.
(9) Unverzagt, C.; Kunz, H.; Paulson, J. C. J. Am. Chem. Soc. 1990, 112,
9308–9309.
(10) Koeller, K. M.; Wong, C.-H. Nature 2001, 409, 232–240.
(11) Hancock, S. M.; Vaughan, M. D.; Withers, S. G. Curr. Opin. Chem.
Biol. 2006, 10, 509–519.
9
10.1021/ja106955j  2010 American Chemical Society
J. AM. CHEM. SOC. 2010, 132, 16651–16656 16651

A R T I C L E S
Matsushita et al.
Figure 1. Polymers used for the recovery tests in the size-exclusion separation using a hollow fiber UF module.
Table 1. Summary of the Recovery of Various Water-Soluble
Polymers in UF Separation Test
promising alternative to mammalian glycosyltransferases for the
construction of a wide range of glycan analogues.12-16
polymer
MW (kDa)
recovery (%)
permeate/loss (%)
Toward the automated synthesis of glycoconjugates based
on chemical and enzymatic approaches, we have developed a
seamless protocol from chemical to enzymatic synthesis by
applying water-soluble polymers for supporting various glycosyl
acceptor substrates.17-24 Our recent strategy focusing on the
glycopeptide synthesis is as follows: (i) chemical synthesis of
BSA
PAA
1
2
3
4
66
80
26
82
93
92
7
14
29
58
116
85
10
52
0
82
0
1
3
12
85
86
15
90
48
100
18
100
99
97
87
15
14
G3 PAMAM dendrimer
G4 PAMAM dendrimer
G5 PAMAM dendrimer
G6 PAMAM dendrimer
G7 PAMAM dendrimer
(12) Weijers, C. A. G. M.; Franssen, M. C. R.; Visser, G. M. Biotechnol.
AdV. 2008, 26, 436–456.
(13) Naruchi, K.; Hamamoto, T.; Kurogochi, M.; Hinou, H.; Shimizu, H.;
Matsushita, T.; Fujitani, N.; Kondo, H.; Nishimura, S.-I. J. Org. Chem.
2006, 71, 9609–9621.
(14) Ma, B.; Audette, G. F.; Lin, S.; Palcic, M. M.; Hazes, B.; Taylor,
D. E. J. Biol. Chem. 2006, 281, 6385–6394.
designated glycosyl acceptors composed of small sugars such
as mono-, di-, or trisaccharide moiety and a heterobifunctional
linker that can shuttle between solid-phase and soluble polymer
platforms through specific catch and release reactions,21-24 (ii)
chemical ligation of the shuttles with water-soluble polymers
resulting in the primer polymers, (iii) sugar elongation reactions
by some recombinant glycosyltransferases, and (iv) releasing
(15) Yu, H.; Chen, X. Org. Biomol. Chem. 2007, 5, 865–872.
(16) Kushi, Y.; Kamimiya, H.; Hiratsuka, H.; Nozaki, H.; Fukui, H.;
Yanagida, M.; Hashimoto, M.; Nakamura, K.; Watarai, S.; Kasama,
T.; Kajiwara, H.; Yamamoto, T. Glycobiology 2010, 20, 187–198.
(17) Nishimura, S.-I.; Matsuoka, K.; Lee, Y. C. Tetrahedron Lett. 1994,
35, 5657–5660.
(18) Nishimura, S.-I.; Yamada, K. J. Am. Chem. Soc. 1997, 119, 10555–
10556.
(19) Yamada, K.; Fujita, E.; Nishimura, S.-I. Carbohydr. Res. 1998, 305,
443–461.
(22) Fumoto, M.; Hinou, H.; Ohta, T.; Ito, T.; Yamada, K.; Takimoto, A.;
Kondo, H.; Shimizu, H.; Inazu, T.; Nakahara, Y.; Nishimura, S.-I.
J. Am. Chem. Soc. 2005, 127, 11804–11818.
(20) Yamada, K.; Matsumoto, S.; Nishimura, S.-I. Chem. Commun. 1999,
50, 7–508.
(21) Fumoto, M.; Hinou, H.; Matsushita, T.; Kurogochi, M.; Ohta, T.; Ito,
T.; Yamada, K.; Takimoto, A.; Kondo, H.; Inazu, T.; Nishimura, S.-
I. Angew. Chem., Int. Ed. 2005, 44, 2534–2537.
(23) Nagashima, I.; Shimizu, H.; Matsushita, T.; Nishimura, S.-I. Tetra-
hedron Lett. 2008, 49, 3413–3418.
(24) Matsushita, T.; et al. Biochemistry 2009, 48, 11117–11133.
9
16652 J. AM. CHEM. SOC. VOL. 132, NO. 46, 2010

Artificial Golgi Apparatus
A R T I C L E S
Figure 2. Schematic procedure of automated glycan synthesis supported by aminooxy-functionalized G7 PAMAM dendrimer.
final products from polymer supports by photo-21 or protease-
selective cleavage22-24 of the linkers and purification. To
mechanize steps (iii) and (iv) in a repetitive manner, we have
developed a HPLC-based glycan synthesizer, namely, Golgi
(Supporting Information, Figure S1). Once a primer polymer
is set in the system, designed glycan derivatives or glycopeptides
can be synthesized automatically through four designated
functions under a computer control: (a) reaction, mixing samples
with reagents by an autosampler; (b) separation, ultrafiltration
(UF) of the reaction mixture by using a hollow fiber module to
separate polymers bearing products and surplus reagents; (c)
detection, quantitative monitoring of eluants from diafiltration
by UV or fluorescent detectors; and (d) repetition, repeating
the process for reactions and separations. Among these four
basic functions of this system, it was uncovered that the
efficiency of the separation is a crucial step because the loss
during these repetitive purification steps should directly influence
the outcome of automated synthesis. In fact, it was suggested
that solution property of polyacrylamide (PAA)-based polymer
supports strongly affects the overall yields of multistep enzy-
matic synthesis which is significantly dependent on the structures
of proximal moiety linking to glycans (data not shown). This
feature of PAA-based polymers has often been a serious hurdle
with regard to the achievement of the efficient and practical
9
J. AM. CHEM. SOC. VOL. 132, NO. 46, 2010 16653

A R T I C L E S
Matsushita et al.
automated synthesis of various glycoconjugates. Here we
hypothesized that the use of monodisperse and spherical
dendrimers25 having suitable molecular size and surface func-
tional groups might break through the above obstacle during
repetitive reaction/separation steps in automated synthesis.
Results and Discussion
Globular Protein-like Properties of Dendrimers. Our interest
was focused on the feasibility of a poly(amidoamine) (PAMAM)
dendrimer26 as an alternative candidate as a supporting polymer
due to its high monodispersity and spherical molecular shape
in addition to the feasibility of attaching surface ligand groups.
To assess the effect of molecular size, shape, and dispersity of
water-soluble polymers on the efficiency in the separation step
using a hollow fiber UF module (Microkros, 10 kDa MWCO),
the recovery of PAMAM dendrimers with different generations
(G3-G7) were evaluated in comparison with common PAA
and PAA-based polymers 1-419,21,23 (Figure 1) and bovine
serum albumin (BSA) as a control for the monodisperse
biopolymer. The results summarized in Table 1 indicate that
PAA and polymers 1-4 showed a wide range of recovery/
retentate from 0 to 82% while BSA (MW ) 66 kDa) as a
standard retained in 85%. It seems likely that permeability of
PAA-based polymers strongly depends on the solubility and
viscosity defined by molecular shape and dispersity rather than
the molecular weights of the core PAA, since PAA (MW ) 80
kDa) retained only 10% while polymer 1 (MW ) 26 kDa)
showed 52% recovery. Judging from the significant difference
of the recovery observed in the polymers bearing ꢀ-GlcNAc
residues through the similar hydrophobic linkers such as 1
(52%), 3 (82%), and 4 (0%), solution property of this class of
polydisperse random polymers appears to be highly sensitive
even to the minor changes in chemical structure of the linker
moiety. These characteristics of PAA-based polymers appear
to significantly affect the yields of enzymatic reactions as well
as the recovery in the repetitive separation steps. It should be
noted that there is no plausible theory or tendency to predict
the feasibility of the PAA-based polymers designed for indi-
vidual target glycoconjugates in automated glycan synthesis on
the Golgi apparatus. Although a few successful automated
syntheses were achieved for the construction of cancer-relevant
mucin glycopeptide libraries based on the functions of this
system,22,24 limitations due to unfavorable solution properties
of PAA-based primers still exist in enzyme-assisted automated
syntheses for a wide range of glycoconjugates. On the other
hand, PAMAM dendrimers showed excellent profiles in this
experiment, in which the G3-G7 dendrimers tested could be
divided into two subgroups by a clear threshold between G6
and G5 dendrimers. It was demonstrated that G6 (MW ) 58
kDa) and G7 (MW ) 116 kDa) retained 85 and 86% while G3
(MW ) 7 kDa), G4 (MW ) 14 kDa), and G5 (MW ) 29 kDa)
entirely permeated this UF module. These results indicate that
the profiles of G6 and G7 dendrimers are quite similar to those
of BSA (MW ) 66 kDa), a typical globular protein 2.5 nm in
diameter containing a single polypeptide chain 50 nm long that
is coiled and folded into a compact bundle. This clearly means
that the efficiency in the size-exclusion separation of water-
soluble polymers by means of the hollow fiber UF module is
greatly influenced by the molecular weight dispersity and
Figure 3. Capturing compound 6 by aminooxy-functionalized G7 PAMAM
dendrimer 5. (a) 0 h, (b) 8 h, and (c) 24 h. Elution condition: column,
YMC-Pack Dial-200 (φ 8.0 × 500 mm); eluent, 50 mM sodium phosphate
buffer, 0.3 M NaCl (pH 7.0); flow rate, 0.7 mL/min; column temperature,
25 °C; UV detector, 214 nm. Retention time for compound 6 is 30.1 min.
molecular shape rather than the monomer composition. These
results encouraged us to challenge the synthesis of a globular
protein-like polymer support by using G7 PAMAM as a starting
material.
Automated Glycan Synthesis Using a Functional Dendri-
mer. To demonstrate the feasibility of dendrimers as an ideal
polymer support in automated glycan synthesis, an aminooxy-
functionalized dendrimer 5 was derived from the G7 PAMAM
dendrimer by coupling with bis-Boc-aminooxyacetic acid suc-
cinimide ester27 and deprotection. We selected 3-[N-(5-oxo-
hexanoyl)-L-phenylanalyl-L-glutamyl-L-phenylalanyl-L-glycinyl]-
aminopropyl 2-acetamido-2-deoxy-ꢀ-D-glucopyranoside (6) as
a test substrate modified with a heterobifunctional linker having
a reactive ketone and the tetrapeptide moiety, Phe-Glu-Phe-
Gly, as a specific cleavage site by Bacillus licheniformis
BLase.23 In the present study, we examined the synthesis of a
sialyl Lewis X derivative 8 as a target compound due to its
various significant biological functions and its associated
difficulty in chemical synthesis in terms of the control of stereo-
and regioselective glycosylation accompanying with multistep
protection/deprotection procedures.6-8,28 Figure 2 indicates a
straightforward synthetic route to this target from a key starting
G7 poly(amidoamine) dendrimer. After capturing the shuttle 6
released from solid-phase synthetic platform by an aminooxy-
functionalized dendrimer 5, we designed a fully automated
multistep sugar elongation by three glycosyltransferases fol-
lowed by selective release of the final product by treating the
dendrimer with BLase.
As expected, HPLC monitoring revealed that shuttle molecule
6 (2.5 µmol) was quantitatively captured by employing 10 equiv
mol of dendrimer 5, in which 42 of 421 aminooxy-groups were
theoretically consumed for the immobilization to afford primer
7 (Figure 3). G7 PAMAM dendrimers bearing shuttle 6
exhibited an excellent solubility and low viscosity as compared
with a starting material, suggesting that an aminooxy-function-
alized G7 PAMAM dendrimer may become a versatile scaffold
for immobilizing relatively hydrophobic glycan derivatives and
glycosphingolipids as well as various glycans and glycopeptides.
Conditions and details of a programmed operation for the
fully automated synthesis of compound 8 from primer 7 on the
Golgi apparatus are summarized in Figure 4. Three separated
domains of the autosampler of Golgi were assigned as: (1) a
(25) Lee, C. C.; MacKay, J. A.; Frechet, J. M. J.; Szoka, F. C. Nat.
Biotechnol. 2005, 23, 1517–1526.
(27) Foillard, S.; Rasmussen, M. O.; Razkin, J.; Boturyn, D.; Dumy, P. J.
Org. Chem. 2008, 73, 983–991.
(26) Tomalia, D. A.; Baker, H.; Dewald, J.; Hall, M.; Kallos, G.; Martin,
S.; Roeck, J.; Ryder, J.; Smith, P. Polymer J. 1985, 17, 117–132.
(28) Boons, G. J.; Demchenko, A. V. Chem. ReV. 2000, 100, 4539–4566.
9
16654 J. AM. CHEM. SOC. VOL. 132, NO. 46, 2010

Artificial Golgi Apparatus
A R T I C L E S
Figure 4. A programmed operation of automated synthesizer Golgi for the synthesis of sialyl Lewis X derivative 8 from primer 7. (A) Details of a programmable
autosampler in Golgi composed of three areas such as sample rack at 4 °C, accessory rack, and injection rack. (B) Schematic diagrams for the separation
using UF module during sugar elongation reactions on the dendrimer and purification (a) as well as product release and separation from dendrimer (b),
respectively. Functions of the individual steps are as follows: (i) multiple glycosylations by mixing glycosyltransferases, sugar nucleotides, alkaline phosphatase,
and a primer solution; (ii) transferring a reaction mixture to the UF module; (iii) removing all small molecules and exchanging buffer solution from HEPES
to ammonium acetate in (a) or collecting product 8 as a low molecular weight fraction in (b) (<10 kDa); (iv) collecting a retained dendrimer-supported sugar
derivative (>10 kDa) in (a) or removing the retained dendrimer as a high molecular weight byproduct (>10 kDa) and cleaning up in (b). (C) Summary of
automated glycan synthesis operated by a programmable 27 steps procedure. (D) HPLC profiles of crude product 8 (a) and blank (b). Condition: column,
Shodex Asahipak NH2P-50 4E (φ 4.6 × 500 mm); eluent, A, 25 mM ammonium acetate (pH 5.8), eluent B, 25 mM acetic acid in acetonitrile (eluent A
gradient increased from 15% to 98% over 90 min and isocratic elution at 98% for another 30 min); column temperature, 25 °C; UV detector, 210 nm; flow
rate, 1.0 mL min-1. Retention time for compound 8 is 28.5 min.
sample rack for sugar nucleotides (D1-3), enzymes (E1-5),
and reservoirs to collect the dendrimer fraction (C1-2); (2) an
accessory rack for HEPES buffer (B1) and; (3) an injection rack
for primer 7 (R1) and reservoirs to collect the final product and
dendrimer (R2-3), respectively (Figure 4A). As indicated in
Figure 4B, a UF module was used for the purification of high-
molecular weight dendrimer fractions during enzymatic sugar
elongations (a, steps 1-18 in Figure 4C) while target 8 released
from the dendrimer should be recovered as a small molecular
fraction passed through the same hollow fiber UF module (b,
steps 19-27 in Figure 4C). The program is based on two cycles
of sequential functions, reaction and separation, for sugar
elongations of primer 7 by one-pot sequential reactions of three
glycosyltransferases in combination with designated sugar
nucleotides and alkaline phophatase (reaction steps 1-12) and
for purification of the dendrimer carrying sialyl Lewis X
9
J. AM. CHEM. SOC. VOL. 132, NO. 46, 2010 16655

A R T I C L E S
Matsushita et al.
tetrasaccharide by UF module (separation steps 13-18).
Furthermore, the following procedures are performed for
releasing 8 from dendrimer by BLase (reaction steps 19-22)
and for final purification of 8 by removing polymers such as
dendrimer and BLase using UF module (separation steps
23-27). The present synthetic platform allowed for fully
automated enzymatic synthesis via Golgi for sialyl Lewis X
tetrasaccharide derivative 8 (432 µg) within 4 days in 16%
overall yield from a simple N-acetyl-D-glucosamine derivative
(2.2 mg of shuttle 6). Surprisingly, the HPLC profile as well as
NMR and HRMS analyses of the crude product obtained
directly from automated synthesis showed a satisfactory purity
of 8 (Figure 4D). This means that no byproducts such as mono-,
di-, or trisaccharide derivatives were generated during the
sequential glycosylation reactions under a programmed operation.
also be emphasized that use of stable recombinant glycosyl-
transferases immobilized on the surface of some solid supports
greatly contribute to the establishment of totally ideal automated
glycan synthesis in a recyclable manner.31 Although several
limitations still exist in glycan synthesis and additional research
is required to further advance this field, the ease of acquiring
defined glycan structures with high purity from a machine will
definitely impact on our understanding of carbohydrates in
biological systems and on the development of carbohydrate-
based therapeutics. Unfortunately, no general methods are, at
present, available for the preparation of a wide range of complex
carbohydrates of biological importance. However, it is clear that
integrating dendrimer-supported enzymatic synthesis and chemi-
cal synthetic approaches such as solid-phase synthesis3,5 and
one-pot multistep glycosylations using unified monosaccharide
building blocks32 should expand the repertoire of synthetic
glycans and their non-natural analogues which can be made, as
well as allow for the creation of high-throughput screening
methods such as glycan microarrays.33,34
Conclusion
In summary, these results demonstrate that the use of globular
protein-like high molecular-weight dendrimer resulted in highly
efficient and practical synthesis using some different-types of
mammalian recombinant glycosyltransferases in as fully au-
tomatable manner as possible. The significance of the present
observations is both methodological and fundamental. On the
other hand, our findings should expand the scope of enzyme-
assisted automated synthesis to include a range of glycan
derivatives as important tools and probes for biological research
and leads for drug discovery. It is also considered that efforts
for the improvements and downsizing of the present HPLC-
based Golgi machine will enhance the potency of the dendrimer-
based strategy and versatility of automated glycan synthesis in
common biochemical and medical laboratories without the need
for specialized techniques.29,30 Considering a significant loss
of the product due to physical properties of hollow fiber-based
material, it seems likely that ultrafiltration of dendrimer-based
intermediates by some filter- or membrane-type modules might
be a potential alternative to improve overall yield. It should
Acknowledgment. This work was supported by a grant for the
“Development of Novel Diagnostic and Medical Applications
through Elucidation of Sugar Chain Functions” from the New
Energy and Industrial Technology Development Organization
(NEDO). We thank Ms. S. Oka, and T. Hirose at the Center for
Instrumental Analysis, Hokkaido University, for ESI-MS measure-
ment and amino acid analysis.
Supporting Information Available: Experimental section,
picture and profile of artificial Golgi, and supplemental data of
structural analysis of all new compounds. This material is
available free of charge via the Internet at http://pubs.acs.org.
JA106955J
(31) Ito, T.; Sadamoto, R.; Naruchi, K.; Togame, H.; Takemoto, H.; Kondo,
H.; Nishimura, S.-I. Biochemistry 2010, 49, 2604–2614.
(32) Zhang, Z.; Ollmann, I. R.; Ye, X.-S.; Wischnat, R.; Baasov, T.; Wong,
C.-H. J. Am. Chem. Soc. 1999, 121, 734–753.
(29) Seeberger, P. H.; Werz, D. B. Nature 2007, 446, 1046–1051.
(30) Ohyabu, N.; Hinou, H.; Matsushita, T.; Izumi, R.; Shimizu, H.;
Kawamoto, K.; Numata, Y.; Togame, H.; Takemoto, H.; Kondo, H.;
Nishimura, S.-I. J. Am. Chem. Soc. 2009, 131, 17102–17109.
(33) Fukui, S.; Feizi, T.; Galustian, C.; Lawson, A. M.; Chai, W. Nat.
Biotechnol. 2002, 20, 1011–1017.
(34) Blixt, O.; et al. Proc. Natl. Acad. Sci. U.S.A. 2004, 101, 17033–17038.
9
16656 J. AM. CHEM. SOC. VOL. 132, NO. 46, 2010