Structures of alkaloid biosynthetic glucosidases decode substrate specificity

Article abstract of DOI:10.1021/cb200267w

Two similar enzymes with different biosynthetic function in one species have evolved to catalyze two distinct reactions. X-ray structures of both enzymes help reveal their most important differences. The Rauvolfia alkaloid biosynthetic network harbors two O-glucosidases: raucaffricine glucosidase (RG), which hydrolyses raucaffricine to an intermediate downstream in the ajmaline pathway, and strictosidine glucosidase (SG), which operates upstream. RG converts strictosidine, the substrate of SG, but SG does not accept raucaffricine. Now elucidation of crystal structures of RG, inactive RG-E186Q mutant, and its complexes with ligands dihydro-raucaffricine and secologanin reveals that it is the "wider gate" of RG that allows strictosidine to enter the catalytic site, whereas the "slot-like" entrance of SG prohibits access by raucaffricine. Trp392 in RG and Trp388 in SG control the gate shape and acceptance of substrates. Ser390 directs the conformation of Trp392. 3D structures, supported by site-directed mutations and kinetic data of RG and SG, provide a structural and catalytic explanation of substrate specificity and deeper insights into O-glucosidase chemistry.

Full text of DOI:10.1021/cb200267w

Articles
pubs.acs.org/acschemicalbiology
Structures of Alkaloid Biosynthetic Glucosidases Decode Substrate
Specificity
Liqun Xia,†,# Martin Ruppert,‡,# Meitian Wang,*,§ Santosh Panjikar,*,∥ Haili Lin,† Chitra Rajendran,§
Leif Barleben,⊥ and Joachim Stockigt*,†
̈
†Institute of Materia Medica, College of Pharmaceutical Sciences, Zhejiang University, 866 Yu Hang Tang Road, Hangzhou 310058,
P.R. China
‡Lehrstuhl fur Pharmazeutische Biologie, Institut fur Pharmazie und Biochemie, Johannes Gutenberg-Universitat, Staudinger Weg 5,
̈
̈
̈
D-55099 Mainz, Germany
§Swiss Light Source, Paul Scherrer Institute, CH-5232 Villigen, Switzerland
∥Australian Synchrotron, 800 Blackburn Road, Clayton VIC, Australia 3168
⊥Department of Medical Biochemistry and Biophysics, Karolinska Institute, SE-17177 Stockholm, Sweden
S
* Supporting Information
ABSTRACT: Two similar enzymes with different biosynthetic
function in one species have evolved to catalyze two distinct
reactions. X-ray structures of both enzymes help reveal their most
important differences. The Rauvolfia alkaloid biosynthetic network
harbors two O-glucosidases: raucaffricine glucosidase (RG), which
hydrolyses raucaffricine to an intermediate downstream in the
ajmaline pathway, and strictosidine glucosidase (SG), which
operates upstream. RG converts strictosidine, the substrate of
SG, but SG does not accept raucaffricine. Now elucidation of
crystal structures of RG, inactive RG-E186Q mutant, and its
complexes with ligands dihydro-raucaffricine and secologanin
reveals that it is the “wider gate” of RG that allows strictosidine to enter the catalytic site, whereas the “slot-like” entrance of SG
prohibits access by raucaffricine. Trp392 in RG and Trp388 in SG control the gate shape and acceptance of substrates. Ser390
directs the conformation of Trp392. 3D structures, supported by site-directed mutations and kinetic data of RG and SG, provide
a structural and catalytic explanation of substrate specificity and deeper insights into O-glucosidase chemistry.
EC 4.3.3.2) initiates ajmaline biosynthesis by catalyzing the
Pictet−Spengler reaction between the biosynthetic precursors,
tryptamine and the monoterpene secologanin. It is this step
that is central to the biosynthesis of about 2000 plant-derived
monoterpenoid indole alkaloids.6 Because of its fundamental
importance, STR1 has become the most extensively charac-
terized enzyme of the network, with much being known about
its 3-dimensional structure and mechanism.7−9
A second enzyme, polyneuridine aldehyde esterase (PNAE,
EC 3.1.1.78), shows exceptionally high substrate specificity.
The structural and mechanistic characteristics of PNAE have
also been recently determined.10 While this enzyme belongs to
the well-known α/β hydrolase superfamily, it also functions as
the “enzymatic gate” to the biosynthesis of C9 monoterpenoid
indole alkaloids from their C10 progenitor, polyneuridine
aldehyde (PNA).
Alkaloids represent one of the most interesting groups of plant-
derived natural products due, in part, to the enormous diversity
they display in carbon frameworks. They also possess
therapeutic value for the treatment of human diseases.1 Many
efforts have been made over the past decades to unravel the
complex biosynthetic formation of single alkaloids at the
enzyme level in order to fully comprehend their multistep
pathways and their biosynthetic networks. To date, this
objective has been fully satisfied for only a very few alkaloids:
the isoquinoline morphine;2 the diterpene alkaloid taxol;3,4 and
the monoterpenoid indole alkaloids, ajmalicine and ajmaline.5
To date, characterization of the largest alkaloidal networks
based on isolated enzymes has been achieved in Papaver, Taxus,
and Rauvolfia.
The biosynthetic enzymes of the alkaloid ajmaline belong to
a variety of enzyme families including synthases, esterases,
glucosidases, reductases, oxidases, and transferases. It is only
since 2004, through structural biology techniques leading to the
successful crystallization and 3D X-ray analysis, that a much
more thorough understanding of the major enzymes participat-
ing in the Rauvolfia ajmaline biosynthetic network has been
gained (Figure 1). The enzyme strictosidine synthase (STR1,
The PNAE product epi-vellosimine serves as the substrate for
the second synthase of the metabolic network, vinorine
Received: July 28, 2011
Accepted: October 17, 2011
Published: October 17, 2011
© 2011 American Chemical Society
226
dx.doi.org/10.1021/cb200267w|ACS Chem. Biol. 2012, 7, 226−234

ACS Chemical Biology
Articles
Figure 1. Enzymatic biosynthesis of monoterpenoid indole alkaloids in cell suspension cultures of the Indian medicinal plant Rauvolfia. Single steps
were elucidated by isolation, characterization and partial sequencing of individual enzymes, cloning and expression of corresponding cDNAs by
“reverse genetics” followed by crystallization and 3D X-ray analysis of the major enzymes. (STR1, strictosidine synthase; SG, strictosidine
glucosidase; PNAE, polyneuridine aldehyde esterase; VS, vinorine synthase; VH, vinorine hydroxylase; CPR, cytochrome P 450 reductase; RG,
raucaffricine glucosidase.).
synthase (VS, EC 2.3.1.160), whose reaction product, vinorine,
already exhibits the complete carbon skeleton of the target
alkaloid, ajmaline. Elucidation of the crystal structure of VS has
not only provided significant insight into its binding pocket and
catalytic center but also represented the first 3D example of the
small BAHD enzyme family.11,12 This superfamily is particularly
important for the biosynthesis of several naturally derived
therapeutics, such as pain killers (morphine), anticancer drugs
(taxol and vinblastine/vincristine), and antiarrhythmia drugs
(ajmaline), as well as a number of nonalkaloidal plant products.
Two additional O-β-D-glucosidases are present in the
Rauvolfia alkaloid proteome, but they exhibit quite different
functions and substrate specificities. The first, strictosidine O-β-
D-glucosidase (SG, EC 3.2.1.105),13 acts at the beginning of the
ajmaline route by deglucosylation of the glucoalkaloid
strictosidine. The second glucosidase, raucaffricine O-β-D-
glucosidase (RG, EC 3.2.1.125),14−16 hydrolyzes the glucoalka-
loid raucaffricine forming the aglycone vomilenine, an
intermediate that appears in the middle of the ajmaline
pathway. RG can also hydrolyze strictosidine, the substrate of
SG, but in sharp contrast, SG does not accept raucaffricine. The
structural data presented here provide an explanation of the
different functional behavior of each of these enzymes.
The X-ray structure of SG and its complex with the substrate
strictosidine has recently been elucidated.17 The completion of
the 3D structure of wild-type RG and its ligand complexes
together with site-directed mutagenesis studies allow a much
clearer understanding of the catalyzed reaction. These
structures enable comparisons of the structural and functional
relationships of both glucosidases. The extensive, high quality
3D X-ray data reported here provide rare structural examples of
two different but closely related enzymes of plant secondary
metabolism that in principle catalyze the same reaction type in
the same biosynthetic network and originate in the same plant
but show remarkable differences in their substrate recognition
and function (Figure 1).
RESULTS AND DISCUSSION
■
Metabolic Function and Significance of RG. The β-
glucosidase RG likely performs a special metabolic function
within the Rauvolfia alkaloid metabolic network. The hydrolysis
of its substrate, raucaffricine, is only one reaction step away
from the major biosynthetic route to ajmaline (Figure 1).
Soluble RG directly delivers the key ajmalicine pathway
intermediate, vomilenine. This aglycone is also the substrate
for the raucaffricine-synthesizing vomilenine glucosyltransfer-
ase, which is responsible for the accumulation of raucaffricine.
Concentrations of this glucoalkaloid can reach 1.6 g L−1 of
cultured Rauvolfia cells, which is the highest amount of a
monoterpenoid indole alkaloid ever observed in plant cell
suspension cultures. This accumulation exceeds the highest
concentration found in plants by >60-fold.15
RG-catalyzed reutilization of raucaffricine could be a crucial
metabolic step in ajmaline biosynthesis. Raucaffricine accumu-
lation could be important for future production in cell culture
systems by blocking the raucaffricine-forming enzyme or by
improving the RG-catalyzed reaction in vivo. Furthermore, RG
is a promising candidate for the generation of libraries of novel
structurally related alkaloids through chemo-enzymatic ap-
proaches starting with raucaffricine (Supplementary Figure S1).
Both objectives require a better insight into the structure of the
enzyme and the mechanism of the reaction that it catalyzes.
Hence comparison of RG and SG at the structural level
presents an interesting challenge from a mechanistic point of
view and may offer significantly more detailed insight into
higher plant β-glucosidases.
227
dx.doi.org/10.1021/cb200267w|ACS Chem. Biol. 2012, 7, 226−234

ACS Chemical Biology
Articles
Figure 2. Structural comparison of the overall architecture of RG substrate complex and SG structure. ( A) Ligand complex of RG-E186Q mutant
with dihydro-raucaffricine (DHR) in stick presentation is illustrated in green and the N- and C-termini are marked. The figure resembles the (β/α)8
barrel fold, the binding site, and the secondary structure elements. The α-helices, β-sheets of the barrel, and loops are in cyan, magenta, and salmon,
respectively. Loops with missing density are marked as “- - - - -”. (Aa) Stereoview of raucaffricine binding site of RG. The 2Fo − Fc SIGMAA-
weighted electron density of DHR contoured at 1.0 σ is shown in yellow and DHR in green. In complex with the inactive mutant (E186Q), residues
within 4.0 Å distance from DHR are shown in magenta, and Gln186 is in cyan. Supplementary Figure S3 better displays the aglycone part of DHR
following rotation of Aa by about 90°. (B) RG-DHR complex superimposed with the overall structure of SG (in red) indicating the high degree of
structural similarity of both glucosidases. Both structures superimposed with a rms deviation of <0.52 Å.
a
Table 1. Structural Alignment of Plant β-Glucosidases of the GH-1 Family
protein name
EC no.
plant
PDB[chain D]
rmsd
Cα
raucaffricine β-glucosidase
3.2.1.125
3.2.1.105
3.2.1.21
3.2.1.21
3.2.1.21
3.2.1.21
3.2.1.21
3.2.1.21
3.2.1.21
3.2.1.21
3.2.1.21
Rauvolfia serpentina
Rauvolfia serpentina
Trifolium repens
4A3Y[A]
0
472
423
426
425
417
408
409
406
408
408
396
strictosidine β-glucosidase
2JF7[A,B]
1CBG[A]
0.513
0.711
0.600
0.701
0.673
0.850
0.737
0.713
0.764
0.760
β-glucosidase 2 (cyanogenic)
β-glucosidase (Os04g0474800;Os4bglu12)
β-glucosidase (Bglu1;Os03g0703000)
β-glucosidase (Os03g0212800;Os3bglu6)
zeatin β-glucosidase (p60.1;Zm-p60.1)
β-glucosidase 1 (Glu1)
Oryza sativa Japonica Group
Oryza sativa Japonica Group
Oryza sativa Japonica Group
Zea mays
3PTK[A,B]
3F5L[A,B]
3GNR[A]
1HXJ[A,B]
1V08[A,B]
1V02[A,B,C,D,F]
2DGA[A]
3AIU[A]
Zea mays
cyanogenic β-glucosidase (dhurrinase 1)
β-glucosidase (TaGlu1b)
Sorghum bicolor P721N
Triticum aestivum
β-glucosidase (ScGlu)
Secale cereale
a
3D structures of all of these O-β-glucosidases were elucidated. They are from plant sources and belong to the GH-1 family; myrosinases (S-β-
glucosidase) are excluded. Total compared backbone atoms are 472. The rmsd values are obtained by PYMOL program (July 6, 2011 in CAZy).
Overall Structure of RG. Extensive investigations into
plant-derived glucosidases of the GH-1 family, to which RG and
SG belong, have been published over the past decade. The
occurrence of both of these enzymes in plant cells of Rauvolfia
and their closely related properties make their detailed
comparison interesting from both functional and structural
perspectives. Their shared physical, biochemical, and molecular
properties (Supplementary Table S1) are proof of a tight link
between SG and RG. This similarity is now also supported by
the newly obtained 3-dimensional structure of RG. The refined
crystal structure of RG consists of 13 α-helices and 13 β-
strands. The overall structure of the wild-type RG enzyme
possesses the expected TIM (β/α)8 barrel fold belonging to
GH1 (http://www.cazy.org18,19), as seen earlier for SG.17 RG
now represents the 11th (excluding myrosinase) structural
example from a total of ∼380 identified GH1 members of plant
origin and the 32nd example from all structurally elucidated
GH1 proteins to date. The (β/α)8 barrel of RG hosts binding
sites for the natural substrate, raucaffricine, and its derivative,
dihydro-raucaffricine (DHR). The groove leading to the
228
dx.doi.org/10.1021/cb200267w|ACS Chem. Biol. 2012, 7, 226−234

ACS Chemical Biology
Articles
catalytic center is formed by irregular loops between the
secondary structures located on the surface of the enzyme.
Similarly to SG, RG contains the catalytic residues Glu186 and
Glu420. The most striking difference is at the catalytic center of
RG and SG. The identical positions of RG-Trp392 and SG-
Trp388 but different conformations of the tryptophan side
chains in the two glucosidases most likely help each protein to
recognize its natural substrate (see next subtitle, paragraph 4).
Figure 2A and B illustrate the overall fold of the inactive mutant
RG-E186Q in complex with its substrate dihydro-raucaffricine
(see Figure 2Aa and Supplementary Figure S3 for the electron
density) and its superimposition with the fold of the SG-E207Q
mutant. These figures illustrate the high degree of structural
similarity between the enzymes.
Sequence-based alignment and a phylogenetic tree of the 11
structurally characterized GH1 enzymes (Supplementary Figure
S2A and B) displays their phylogenetic distribution as well as
the conserved residues during the evolution process. Further
comparison of the structure of RG together with SG and other
β-glucosidases by structural alignment demonstrates, on the
one hand, the excellent correspondence of RG with SG by
indicating smaller rmsd values (0.52 Å) for the 423 super-
imposed backbone carbons and, on the other hand, nearly
identical but higher rmsd values (from 0.6−0.9 Å) for the
backbone carbons ranging from 396 to 426 for other plant
glucosidases (Table 1).
Substrate Recognition and Active Site Relationship of
RG and SG. Although RG and SG are structurally closely
related, surprisingly their substrate specificity differs consid-
erably. Whereas in addition to its natural substrate raucaffricine
RG also hydrolyses the glucoalkaloid strictosidine (rel activity
∼1.2%), SG does not exhibit any measurable conversion of
raucaffricine (Table 2). The reason for this high specificity has
not been known but can now be explained by rigorous
structural comparison of the overall binding pockets, their 3D
ligand complexes, and especially the shape of the entrance to
the active site of both enzymes.
Table 2. Relative Enzyme Activity of RG Wild-Type and RG
Mutants
a
raucaffricine (%)
(dihydro-
strictosidine
(%)
secologanin
(%)
arbutin
(%)
substrates
raucaffricine %)
b
c
RG-WT
100 (98 )
1.18
2.92
0.24
0.04
nd
R G -
1.21
43.66
4.60
63.94
24.24
30.91
0.62
0.57
0.1
W392A
R G -
1.72
0.02
0.77
0.08
5.37
nd
H193A
R G -
Y200A
R G -
T189A
R G -
F485Y
R G -
S390G
R G -
E476L
R G -
nd
E476A
R G -
E186D
R G -
E186Q
0.1
R G -
E420Q
0.5
R G -
nd
E186Q/
E420Q
SG-WT
nd
53.01
a
Comparison of the relative enzyme activity of RG-WT (0.27 mM
min−1 μg−1, decrease measured of substrate concentrations under the
conditions described in Supplemental Experimental Procedures,
activity set at 100%), and mutants for putative substrates raucaffricine,
1,2-dihydro-raucaffricine (DHR), strictosidine, secologanin, and
arbutin; detection limit 0.02%; nd = not detectable; blank = not
b
determined. WT, wild-type enzymes with N-terminal His6-tag.
c
Measured with RG isolated from R. serpentina cell suspension culture.
As is the case for SG and for other β-glucosidases,17,20 the
gate to the active site should host the binding region of the
aglycone part of the substrate. This part is always solvent-
exposed, whereas the glucose moiety extends into and is mainly
localized deep within the active center. As illustrated in Figure
3A this assumption is clearly supported by the 3D structure of
the complex of RG with its ligand dihydro-raucaffricine (DHR).
The aglycone binding site is believed to be responsible for the
substrate specificity of β-glucosidases (and also for catalysis) as
shown by structural and computational analysis.20−23 The
shape and space of the entrance of RG and SG differ
remarkably, providing insight into their substrate recognition.
Whereas RG leaves significant space for the ligand DHR
(Figure 3A), there is much less space in the slot-like entrance of
SG (Figure 3B) in which the ligand strictosidine shows a
“tighter” fit. Moreover, the aglycones of both ligands occupy
different positions, particularly when compared to the
appropriate tryptophans 392 in RG and 388 in SG. The
parallel arrangement of Trp388 and the aglycone of
strictosidine in SG will fix the substrate strictosidine more
tightly by stronger hydrophobic, sandwich-like interactions
compared to RG, its Trp392, and the substrate raucaffricine,
which are present in more of a perpendicular orientation. The
above findings help to explain the variability in substrate affinity
of RG and SG, whereby wild-type RG has significantly weaker
binding of its natural substrate raucaffricine, with a Km of ∼1.3
mM, compared to the approximately 10-fold higher affinity of
SG for its substrate strictosidine (Km = 0.12 mM) and ∼15-fold
lower for RG toward strictosidine (Km = 1.8 mM) compared to
SG (Km = 0.12 mM). A similar result (∼7-fold lower RG
affinity) was obtained with the His6-tagged RG and SG
(Supplementary Table S2).
If Trp392 is important in allowing RG to distinguish between
the substrates raucaffricine and strictosidine, mutation of this
residue to a smaller amino acid should result in a mutant with
increased SG activity. Indeed, the RG-W392A mutant exhibited
a large decrease in relative activity against the substrate
raucaffricine from 100% for the wild-type to about 1.2% for the
mutant, probably due to loss of hydrophobic interaction by
Trp392. However, when activity for the substrate strictosidine
was measured, a 2.5-fold increase from ∼1.2% to ∼2.9% was
observed for the mutant, indicating that creation of additional
space for strictosidine in RG by the substitution described
above changes substrate acceptance in favor of the natural
substrate of SG (Table 2). This finding is also supported by the
catalytic activity kcat and kcat/Km of the RG-W392A mutant,
which was about 3-times higher for strictosidine compared to
raucaffricine (Supplementary Table S2), a result emphasizing
even more so the prominent structural but also functional role
of this particular Trp392 residue.
Except in SG, the conformation of Trp392, which we called
plant glucosidase-typical conformation,17 is the same in all
229
dx.doi.org/10.1021/cb200267w|ACS Chem. Biol. 2012, 7, 226−234

ACS Chemical Biology
Articles
Figure 3. Surface of the cavity representing the binding pockets of RG-E186Q (blue) and SG-E207Q (red). Ligand complexes with their substrates
DHR and strictosidine (STR), respectively, in stick presentation. ( A) The whole binding pocket of RG-E186Q with particular emphasis on the
shape of the entrance to the binding site and the aglycone part of DHR (green), which is somewhat sandwiched with Trp392. (B) The different
shape of the slot-like entrance of SG-E207Q mutant compared to panel A and the ligand STR in sandwich-like π−π interaction with Trp388. (C)
Superimposed meshes of entrances of the RG-DHR ligand complex (blue) with that of SG (substrate STR not shown), which illustrates that the RG
substrate can move easily through the RG entrance but is hindered with regards to the SG entrance (red). In addition, different conformations of
both Trp392 of RG and Trp388 of SG, respectively, are shown. (D) The whole binding pocket of the RG-DHR complex after rotation of A around
the X axis by 90°. Trp392 of RG is also illustrated (in blue, stick representation) at the bottom of the figure.
structurally characterized plant-derived β-glucosidases. This
conformation is obviously due to interactions with second
sphere residues Ser390 in RG and Asn or Thr in the other
enzymes (Supplementary Figure S2). An early publication put
forward the idea that the conformation of the conserved Trp
may vary depending on the nature of neighboring amino acid,
e.g., of Tyr473 in the enzyme Glu1 (Figure 4G and Table 1)
because it will form a hydrogen bond with the nitrogen of that
Trp.20 Tyr is replaced by Phe in RG (Phe485) and 7 other
enzymes, while SG with its variant Trp conformation still
features Tyr (Tyr481). This observation therefore indicates that
at least for 8 of the 11 GH-1 enzymes Tyr is not essential for
plant glucosidase-typical conformation of Trp. In contrast, in
SG the Ser residue is replaced by Gly386, which is unable to
form a hydrogen bond with Trp388, resulting in the SG-typical
conformation of Trp388.17 We thus suggest that the crucial
conformation of this particular Trp residue is influenced more
by the neighboring Ser (Asn or Thr) than by Tyr. The
corresponding mutations in RG support this suggestion, for
which the relative activity of RG-S390G (SG-Gly386) for
strictosidine increased ∼4.5 times compared to that of RG-WT,
while the relative activity of RG-F485Y (SG-Tyr481) for
strictosidine decreased. Replacement of Ser390 by Gly leads to
a more flexible conformation of Trp392, whereas Phe485
mutated to Tyr results in a more fixed Trp392 conformation, in
turn resulting in a less productive accommodation of
strictosidine.
obvious that for steric reasons alone DHR can relatively easily
pass through the entrance into the RG binding pocket but
cannot enter the SG pocket. This structural result is in
agreement with both observations that (i) raucaffricine is not
converted by SG and (ii) SG activity is not inhibited by
raucaffricine (data not shown), since raucaffricine cannot access
the active site of SG. These results provide strong evidence that
the wider entrance of the binding pocket of RG contributes to
the lower substrate specificity compared to that of SG. This
structural arrangement is important in the modulation of
glucosidase specificity.24 This result also points to the
significant structural role played by residues Trp392 and
Trp388. The shape of the entrance to the catalytic pocket
determines which glycoalkaloid will enter and be converted.
Aglycone Binding Site. The RG aglycone binding site is
basically very similar to that of other β-glucosidases, since
interactions are nearly all of a hydrophobic nature. Surprisingly,
the amino acids surrounding the aglycone are not identical to
those in SG, although both enzymes are specific for the
conversion of indole alkaloid glucosides. The conformation of
Trp392 in relation to Trp388 has already been discussed above.
When complexes of RG and its substrate and SG and its
substrate, respectively, are superimposed, only a few residues
are found to be identical. Tyr347 of RG occupies the same
position as Tyr345 in SG. The same situation is observed for
Thr189 and Thr210, but their substrate binding is different.
Whereas Thr189 interacts with C-14 of DHR, the second
Thr210 of SG interacts with the exocyclic double bond of the
secologanin part of strictosidine. It is evident that both
When the mesh of the RG complex with DHR was
superimposed with the mesh for SG (Figure 3C), it became
230
dx.doi.org/10.1021/cb200267w|ACS Chem. Biol. 2012, 7, 226−234

ACS Chemical Biology
Articles
Figure 4. Conformation of conserved Trp and its neighboring residues in GH1 enzymes of higher plants. ( A) The superimposed residues of RG
(purple, PDB 4A3Y) and SG (red, PDB 2JF7). The side chains of the two Trp point to the opposite direction. Panels B−J represent other plant
GH1 enzymes with plant glucosidase-typical Trp conformation (see Table 1 and Supplementary Figure S2A); their PDB codes are 1CBG, 3PTK,
3F5L, 3GNR, 1HXJ, 1V08, 1V02, 2DGA, and 3AIU, respectively. The residues were marked according to their position in each enzyme. In panel A−
E and H−J, Phe is >4.1 Å from Trp to form strong interaction. It is the hydrogen binding of Trp to the neighboring residues (≤3.9 Å distance) that
illustrates the essential role of Ser/Asn/Thr for directing the conformation of Trp.
threonines help to fix the aglycone and indirectly keep the
glycone moiety in a favorable position for hydrolysis. In
agreement with this observation, following site-directed
mutation of Thr189 into Ala, the mutant shows only 64%
activity toward raucaffricine compared to that of RG-WT
(Table 2). Leu199 in RG and Gly386 of SG are completely
different residues; however, it appears that they have the same
function of shielding the 10 and 11 positions of the indole
moiety of the corresponding substrate by the Leu199 side chain
and by the loop region in which Gly386 is located, respectively.
A similar case of two different residues is observed for
Tyr200 in RG and its counterpart Phe220 in SG. While they
overlap perfectly in both complexes, binding of the former is
likely to be more efficient due to the presence of two
interactions: one sandwich-like hydrophobic interaction with
the indole part, and the second being the H-bonding to the
indole nitrogen (Figure 5A). Further assays showed that the
hydrolytic activity of the RG-Y200A mutant decreases
dramatically for both substrates, proving the significant role
of the two interactions of Tyr200.
The biggest difference between both ligand complexes is
represented by His193 in RG, a residue that, in fact, is replaced
by Asn in SG and exceeds 4 Å distance to the substrate
strictosidine. It has three binding opportunities represented by
fixing the acetyl group of DHR and keeping the indole part
fixed by steric, sandwich-like interaction and a hydrogen bridge
with the indole nitrogen as illustrated in Figure 5A (see also
Figure 2Aa and Supplementary Figure S5). Consequently, the
RG-H193A mutant shows only low transformation of
raucaffricine (44%). The relative activity toward strictosidine
increases slightly, probably because of the absence of the
interactions. It is clear that the indole nitrogen of the two
substrates have different orientations when overlapping with
the raucaffricine complex of RG and the strictosidine complex
of SG (Figure 5Bb). As a result, strictosidine can interact more
freely with this RG mutant.
In conclusion, although both glucosidases are strongly
related, their aglycone binding sites are very different,
particularly when compared with published structures of
other β-glucosidases, indicating that the aglycone must have a
significant influence on the enzyme activity.
Insight into the Catalytic Site of RG. Mechanistic
insights into glycosidase chemistry have been reviewed several
times in the past.25,26 In order to illustrate the hydrolytic
mechanism of RG, the structure in Figure 3A was rotated,
allowing much improved visualization of the pocket for the
glucosidic part (Figure 3D). Interactions of amino acid residues
<4.0 Å in distance are two-dimensionally illustrated in Figure
5A. Two glutamic acids constitute the catalytic residues as in all
structures of β-glucosidases of the GH-1 family. In RG, these
residues are represented by Glu186 and Glu420, as verified by
the corresponding mutants E186D, E186Q, E420Q, and the
double mutant E186Q/E420Q, in which hydrolase activity is
≤0.5% of RG-WT (Table 2). Glu186 and Glu420 are separated
by a distance of ∼5.1 Å between their carboxylate carbons,
which provides the space for reacting with the glucosidic bond.
This situation is typical for enzymes of the GH1 retaining the
β-configuration of the anomeric carbon of glucose during
catalysis. Whereas in SG a total of 11 hydrophilic residues
interact with the glucose part of its substrate strictosidine,17 in
RG there appear to be only seven residues responsible for
keeping the glucose in the catalytically desired position, all
being identical with those in the SG complex. Glu476 is one of
the seven residues and participates in an important hydrophilic
interaction with both O6 and O4 of the glycone part of the
substrate (Figure 5). The activity of mutants RG-E476L and
231
dx.doi.org/10.1021/cb200267w|ACS Chem. Biol. 2012, 7, 226−234

ACS Chemical Biology
Articles
DHR and secologanin ligand complexes each Glu476 side chain
points away from the oxygen of its backbone peptide bond (C-
terminal). The same observation was made for the SG wild-
type structure. Different conformations of Glu476 seem to be
rare in other β-glucosidase structures, with the only exception
being the zeatin β-glucosidase from Zea mays (PDB 1HXJ) and
a rice glucosidase (PDB 3GNO28). In the SG complex with
strictosidine the Glu472 points to the oxygen of that peptide
bond, which is the N-terminal direction (Supplementary Figure
S4). Independently from these observations, the binding
distance of Glu476 and Glu472 residues to oxygens O4 and
O6 of the glucose moiety is in all cases very similar. The results,
therefore, suggest that the positions of the appropriate Glu side
chain in β-glucosidases differ; however, this difference appears
not to influence the interactions with the glucose part very
much, since distances to O4 and O6 are nearly the same.
When all three ligand complexes, RG with DHR and
secologanin and SG with strictosidine, were compared by
superimposition (Figure 5B), it was found that the binding site
for the glucose unit is identical in the first two cases. This
observation is in contrast to the rate of catalytic conversion,
which is the highest for the substrate raucaffricine and its
derivative DHR (100% and 98% rel activity, respectively) but
much lower for the glucoside secologanin, showing only 0.24%
relative activity (Table 2). The finding again supports the
influence of the structure of the aglycone on enzyme
conversion rates.
Our results indicate the importance of the nature of the
aglycone for glycosidase action. The catalytic site of each
enzyme has to adapt not only to the structure of the glycosyl
residue but also to that of the aglycone, which can occur in
highly diverse and unexpected ways. Using structural differ-
ences in order to understand differences in substrate acceptance
by closely related members of the GH1 is exceptionally
complexed, as has been recently discussed.28
In sharp contrast, comparison of RG and SG substrate
complexes has now provided a clear structural explanation of
their distinct substrate specificities and the basis for molecular
recognition of their intrinsic ligands.
Figure 5. Detail of the binding mode of different substrates for RG
and SG. ( A) Hydrogen binding network between 1,2-(S)-dihydro-
raucaffricine and residues in a distance of ≤4.0 Å in the ligand
structure of His6-RG inactive mutant E186Q, illustrating the
interactions of amino acids with both the aglycone and the glucosidic
part of the substrate. (Ba) Superimposed glucose part of RG
complexes with DHR (green), SG with strictosidine (red), and RG
with secologanin (black), displaying the identical position of the
glucose moiety in the RG complexes and slightly different binding in
the SG complex. (Bb) Panel Ba is rotated in order to better display the
different direction of indole nitrogen (blue) and the slight shift in the
glucose part.
RG-E476A dropped to less than 1% compared to the wild-type,
clearly demonstrating the importance of Glu476’s hydrophilic
interaction with the substrate. The remaining four residues,
though conserved in SG (except Tyr481, which is replaced in
RG by Phe485, which cannot form a hydrogen bridge) and in
other β-glucosidases (Supplementary Figure S2A), seem not to
be essential for binding (>4.0 Å) the glucose unit of the
substrate of RG. This might be due to the fact that the glucose
unit has slightly shifted (∼1.5 Å) in the RG complex
(compared to the SG complex), indicating a higher structural
flexibility within the active site. This shift has been observed
several times in different crystals and data sets of RG
complexes, which might be based on the bigger pocket size
and/or the different aglycone structures of RG and SG
substrates. As previously discussed, the aglycone may indeed
have an influence on the binding properties of the glucose in
the catalytic center.27 The superimposed ligand complexes of
RG and SG (Figure 5Ba and Bb) disclose the structural
differences in the catalytic pocket of both enzymes.
METHODS
■
RG Crystallization and Preparation of Its Complexes with Its
Substrates. The crystallization of wild-type RG was described by
Ruppert, et al.16 Crystallization of the inactive mutant RG-E186Q and
preparation of its complex with the substrate 1,2-(S)-dihydro-
raucaffricine and secologanin were carried out with the optimal
crystallization conditions found to be in the range of 0.01−0.3 M
ammonium sulfate, 0.1 M sodium acetate, pH 4.5−5.0, 9−12% (w/v)
PEG 4000 as precipitant buffer, and 20 °C for incubation temperature.
RG complexes with raucaffricine showed always insufficient density of
the ligand in contrast to 1,2-dihydro-raucaffricine, which exhibits
nearly identical conformation. After stable crystals were formed (≥10
d), freeze-dried ligands 1,2-(S)-dihydro-raucaffricine and secologanin
were added directly to the drops that contained appropriate crystals.
These soaking experiments were carried out for 10 min to 3 h.
Thereafter crystals were measured at RT.
Structure Determination of Wild-Type RG, Model Building
and Refinement. Prior to X-ray data collection, the crystals were
treated with cryoprotectant [10 mM calcium acetate, 10 mM Tris pH
7.5, 7% DMSO, 20% (v/v) glycerol] and flash-cooled to100 K.
Diffraction data were then collected on the X11 beamline of EMBL-
Hamburg, Germany; 720 images were measured with 0.5° rotation per
frame using aMAR CCD detector at wavelength 0.8051 Å. The data
set statistics are summarized in Supplementary Table S3.
A more detailed inspection and comparison of the binding
sites of RG and SG wild-type and their substrate complexes has
also revealed several additional structural differences. In the RG
and SG structures, the side chains of Glu476 and Glu472,
respectively, can occupy different conformations (Supplemen-
tary Figure S4). The distance between the carboxy carbons in
the two positions is about 2.4 Å. In RG wild-type and in RG/
232
dx.doi.org/10.1021/cb200267w|ACS Chem. Biol. 2012, 7, 226−234

ACS Chemical Biology
Articles
The 3D-structure of wild-type RG was determined by molecular
replacement. The applied software pipeline was AUTO-RICK-
SHAW.29 Packing considerations30 suggested that the content of
one asymmetric unit of RG crystals was two monomers. The initial
models were solved using the program MOLREP31 with the structure
of homologous SG (PDB 2JF7) as a target model. The output model
ACKNOWLEDGMENTS
■
Staff members at the EMBL beamline at the DORIS storage
ring (DESY, Hamburg, Germany), beamline X06SA at Paul
Scherrer Institute (Villigen, Switzerland) and beamline BL17U1
at Shanghai Synchrotron Radiation Facility (China) are kindly
appreciated for their help. We thank Deutsche Forschungsge-
meinschaft (Bonn-Bad Godesberg, Germany), Fonds der
Chemischen Industrie (Frankfurt/Main, Germany) and
K.P.Chao Foundation (Zhejiang University, Hangzhou,
China) for support. We also acknowledge H. Zou (Zhejiang
University, Hangzhou, China) for samples of secologanin and
D. E. Cane (Brown University, USA) for critical advices.
was subjected to iterative refinement using CNS32 and REFMAC5.33
A
random set containing 2.4% of the total data was excluded from the
refinement, and the agreement between calculated and observed
structure factors corresponding to these reflections (Rfree) was taken to
follow the progress of the refinement.34 The model phase was
subjected to bias reduction and density modification using the
program PIRATE and continued with automated model building
program ARP/wARP.35
At this stage, ARP/wARP generated 2mfo-Dfc maps36 were used for
careful examination of the maps and allowed corrections and
incorporation of missing residues into the model. The graphics
program COOT37 was used for rebuilding of the model. Manual
building alternated with additional REFMAC cycles, which included
bulk solvent correction, anisotropic scaling, and with each molecule
defined as a TLS group in the modeling of anisotropy in the program
REFMAC5.38 Overall geometric quality of the model was assessed
using PROCHECK.39
Structure Determination of RG-E186Q Complexes, Model
Building and Refinement. All RT crystal mounting and data
collection were carried out at Swiss Light Source (SLS) macro-
molecular crystallography beamline X06SA.
REFERENCES
■
(1) Nicolaou, K. C. Montagnon, T. (2008) Molecules That Changed
the World, Chapter 10: Morphine, pp 67−78, Wiley-VCH, Weinheim.
(2) Zenk, M. H., and Juenger, M. (2007) Evolution and current
status of the phytochemistry of nitrogenous compounds. Phytochem-
istry 68, 2757−2772.
(3) Chau, M., Jennewein, S., Walker, K., and Croteau, R. (2004)
Taxol biosynthesis. Chem. Biol. 11, 663−672.
(4) Kaspera, R., and Croteau, R. (2006) Cytochrome P450
oxygenases of Taxol biosynthesis. Phytochem. Rev. 5, 433−444.
(5) Ruppert, M., Ma, X, and Stockigt, J. (2005) Alkaloid biosynthesis
̈
The MiTegen RT system40 was used for the crystal mounting.
Crystals were measured in continuous data acquisition mode with a
noise-free and high dynamic range pixel-array detector - PILATUS 6
M.4142 All room-temperature data were collected in continuous mode
with 5 Hz frame rate. After mounting, crystals were quickly transferred
for diffraction data collection in order to reduce possible dehydration.
X-ray beam was defocused to about 100 μm × 100 μm to match the
size of RG crystals, which are about 100 μm × 200 μm in general.
Collection protocol for RG crystal is 0.5° oscillation per 0.2 s, 1 Å
radiation, and attenuated flux of about 60 × 109 photons/s. All
diffraction data were processed with automated data processing script
GO.COM, which combines XDS robust indexing, integration, scaling,
and merging processes43,44 with high-performance computing facility
at X06SA beamline. Typical data processing time for a data set of 270
images is about 1−2 min, which is proven to be very valuable for real-
time data quality assessment. Structures were determined with
molecular replacement package MOLREP using refined RG model
from cryo-data as search model. Atomic coordinates and atomic
displacement parameters were refined with REFMAC package. Manual
and semiautomated model building of protein, solvent, and substrate
were carried out with COOT. PROCHECK and MolProbity were
used for structure validation. All figures were made with PyMOL.45
in Rauvolfia−cDNA cloning of major enzymes of the ajmaline
pathway. Curr. Org. Chem. 9, 1431−1444.
(6) Stockigt, J., and Zenk, M. H. (1977) Strictosidine (isovincoside):
̈
Key intermediate in biosynthesis of monoterpenoid indole alkaloids. J.
Chem. Soc., Chem. Commun., 646−648.
(7) Ma, X., Panjikar, S., Koepke, J., Loris, E., and Stockigt, J. (2006)
The structure of Rauvolfia serpentina strictosidine synthase is a novel
six-bladed beta-propeller fold in plant proteins. Plant Cell 18, 907−
920.
̈
(8) Maresh, J., Giddings, L., Friedrich, A., Loris, E., Panjikar, S.,
̀
Trout, B., Stockigt, J., Peters, B., and OConnor, S. (2008) Strictosidine
̈
synthase: Mechanism of a Pictet-Spengler catalyzing enzyme. J. Am.
Chem. Soc. 130, 710−723.
(9) Stockigt, J., Barleben, L., Panjikar, S., and Loris, E. (2008) 3D-
̈
Structure and function of strictosidine synthase−the key enzyme of
monoterpenoid indole alkaloid biosynthesis. Plant Physiol. Biochem. 46,
340−355.
(10) Yang, L., Hill, M., Wang, M., Panjikar, S., and Stockigt, J. (2009)
̈
Structural basis and enzymatic mechanism of the biosynthesis of C(9)-
from C(10)-monoterpenoid indole alkaloids. Angew. Chem., Int. Ed. 48,
5211−5213.
(11) Ma, X., Koepke, J., Panjikar, S., Fritzsch, G., and Stockigt, J.
̈
(2005) Crystal structure of vinorine synthase, the first representative
ASSOCIATED CONTENT
* Supporting Information
This material is available free of charge via the Internet at
of the BAHD superfamily. J. Biol. Chem. 280, 13576−13583.
■
̀
(12) DAuria, J. C. (2006) Acetyltransferases in plants: a good time to
S
be BAHD. Curr. Opin. Plant Biol. 9, 331−340.
(13) Gerasimenko, I., Sheludko, Y., Ma, X., and Stockigt, J. (2002)
̈
http://pubs.acs.org.
Heterologous expression of a Rauvolfia cDNA encoding strictosidine
glucosidase, a biosynthetic key to over 2000 monoterpenoid indole
alkaloids. Eur. J. Biochem. 269, 2204−2213.
Accession Codes
Coordinates and structure factors have been deposited in
Protein Data Bank with the codes 4A3Y for RG-WT-glycerol
complex, 3U5U for RG-E186Q, 3U57 for RG-E186Q-
dihydroraucaffricine complex, and 3U5Y for RG-E186Q-
secologanin complex.
(14) Warzecha, H., Obitz, P., and Stockigt, J. (1999) Purification,
̈
partial amino acid sequence and structure of the product of
raucaffricine-O-beta-D-glucosidase from plant cell cultures of Rauvolfia
serpentina. Phytochemistry 50, 1099−1109.
(15) Warzecha, H., Gerasimenko, I., Kutchan, T. M., and Stockigt, J.
̈
(2000) Molecular cloning and functional bacterial expression of a plant
glucosidase specifically involved in alkaloid biosynthesis. Phytochemis-
try 54, 657−666.
AUTHOR INFORMATION
Corresponding Author
*E-mail: meitian.wang@psi.ch; panjikar.santosh@embl-
hamburg.com; joesto2000@yahoo.com.
■
(16) Ruppert, M., Panjikar, S., Barleben, L., and Stockigt, J. (2006)
̈
Heterologous expression, purification, crystallization and preliminary
X-ray analysis of raucaffricine glucosidase, a plant enzyme specifically
involved in Rauvolfia alkaloid biosynthesis. Acta Crystallogr., Sect. F:
Struct. Biol. Cryst. Commun. 62, 257−260.
Author Contributions
#These authors contributed equally to this work.
233
dx.doi.org/10.1021/cb200267w|ACS Chem. Biol. 2012, 7, 226−234

ACS Chemical Biology
Articles
(17) Barleben, L., Panjikar, S., Ruppert, M., Koepke, J., and Stoeckigt,
J. (2007) Molecular architecture of strictosidine glucosidase: the
gateway to the biosynthesis of the monoterpenoid indole alkaloid
family. Plant Cell 19, 2886−2897.
(37) Emsley, P., and Cowtan, K. (2004) Coot: model-building tools
for molecular graphics. Acta Crystallogr., Sect. D: Biol. Crystallogr. 60,
2126−2132.
(38) Winn, M. D., Isupov, M. N., and Murshudov, G. N. (2001) Use
of TLS parameters to model anisotropic displacements in macro-
molecular refinement. Acta Crystallogr., Sect. D: Biol. Crystallogr. 57,
122−133.
(39) Laskowski, R. A., MacArthur, M. W., Moss, D. S., and
Thornton, J. M. (1993) PROCHECK: A program to check the
stereochemical quality of protein structures. J. Appl. Crystallogr. 26,
283−291.
(40) Kalinin, Y., Kmetko, J., Bartnik, A., Stewart, A., Gillilan, R.,
Lobkovsky, E., and Thorne, R. (2005) A new sample mounting
technique for room-temperature macromolecular crystallography. J.
Appl. Crystallogr. 38, 333−339.
(41) Henrich, B., Bergamaschi, A., Broennimann, C., Dinapoli, R.,
Eikenberry, E. F., Johnson, I., Kobas, M., Kraft, P., Mozzanica, A., and
Schmitt, B. (2009) PILATUS: A single photon counting pixel detector
for X-ray applications. Nucl. Instrum. Methods Phys. Res., Sect. A 607,
247−249.
(42) Broennimann, C., Eikenberry, E. F., Henrich, B., Horisberger,
R., Huelsen, G., Pohl, E., Schmitt, B., Schulze-Briese, C., Suzuki, M.,
Tomizaki, T., Toyokawa, H., and Wagner, A. (2006) The PILATUS
1M detector. J. Synchrotron Rad. 13, 120−130.
(43) Kabsch, W. (2010) XDS. Acta Crystallogr., Sect. D: Biol.
Crystallogr. 66, 125−132.
(18) Henrissat, B., and Davies, G. (1997) Structural and sequence-
based classification of glycoside hydrolases. Curr. Opin. Struct. Biol. 7,
637−644.
(19) Ketudat Cairns, J. R., and Esen, A. (2010) β-Glucosidases. Cell.
Mol. Life Sci. 67, 3389−3405.
(20) Czjzek, M., Cicek, M., Zamboni, V., Bevan, D. R., Henrissat, B.,
and Esen, A. (2000) The mechanism of substrate (aglycone)
specificity in β-glucosidases is revealed by crystal structures of mutant
maize beta-glucosidases-DIMBOA, -DIMBOAGlc, and -dhurrin
complexes. Proc. Natl. Acad. Sci. U.S.A. 97, 13555−13560.
(21) Dopitova, R., Mazura, P., Janda, L., Chaloupkova, R., Jerabek, P.,
Damborsky, J., Filipi, T., Kiran, N. S., and Brzobohaty, B. (2008)
Functional analysis of the aglycone -binding site of the maize β-
glucosidase Zm-p60.1. FEBS J. 275, 6123−6135.
(22) Hill, A. D., and Reilly, P. J. (2008) Computational analysis
glycoside hydrolase family 1 specificities. Biopolymers 89, 1021−1031.
(23) Mendonca, L. M., and Marana, S. R. (2008) The role in the
substrate specificity and catalysis of residues forming the substrate
aglycone-binding site of a β-glucosidase. FEBS J. 275, 2536−2547.
(24) Sue, M., Nakamura, C., Miyamoto, T., and Yajima, S. (2011)
Active-site architecture of benzoxazinone-glucoside β-D-glucosidases
in Triticeae. Plant Sci. 180, 268−275.
(25) Vocadlo, D. J., and Davies, G. J. (2008) Mechanistic insights
into glycosidase chemistry. Curr. Opin. Chem. Biol. 12, 539−555.
(26) Vasella, A., Davies, G. J., and Bohm, M. (2002) Glycosidase
mechanisms. Curr. Opin. Chem. Biol. 6, 619−629.
(27) Verdoucq, L., Moriniere, J., Bevan, D. R., Esen, A., Vasella, A.,
Henrissat, B., and Czjzek, M. (2004) Structural determinants of
substrate specificity in family 1 beta-glucosidases−Novel insights from
the crystal structure of sorghum dhurrinase-1, a plant beta-glucosidase
with strict specificity, in complex with its natural substrate. J. Biol.
Chem. 279, 31796−31803.
(44) Kabsch, W. (2010) Integration, scaling, space-group assignment
and post-refinement. Acta Crystallogr., Sect. D: Biol. Crystallogr. 66,
133−144.
(45) DeLano, W. L. (2002) The PyMOL Molecular Graphics System,
DeLano Scientific, San Carlos, CA.
(28) Seshadri, S., Akiyama, T., Opassiri, R., Kuaprasert, B., and
Cairns, J. K. (2009) Structural and enzymatic characterization of
Os3BGlu6, a rice beta-glucosidase hydrolyzing hydrophobic glycosides
and (1→3)- and (1→2)-linked disaccharides. Plant Physiol. 151, 47−
58.
(29) Panjikar, S., Parthasarthy, V., Lamzin, V. S., Weiss, M. S., and
Tucker, P. A. (2005) Auto-Rickshaw: an automated crystal structure
determination platform as an efficient tool for the validation of an X-
ray diffraction experiment. Acta Crystallogr., Sect. D: Biol. Crystallogr.
61, 449−457.
(30) Matthews, B. W. (1968) Solvent content of protein crystals. J.
Mol. Biol. 33, 491−497.
(31) Vagin, A., and Teplyakov, A. (1997) MOLREP: an automated
program for molecular replacement. J. Appl. Crystallogr. 30, 1022−
1025.
(32) Brunger, A. T., Adams, P. D., Clore, G. M., DeLano, W. L.,
̈
Gros, P., Grosse-Kunstleve, R. W., Jiang, J. S., Kuszewski, J., Nilges, M.,
Pannu, N. S., Read, R. J., Rice, L. M., Simonson, T., and Warren, G. L.
(1998) Crystallography & NMR system: A new software suite for
macromolecular structure determination. Acta Crystallogr., Sect. D: Biol.
Crystallogr. 54, 905−921.
(33) Murshudov, G. N., Vagin, A. A., and Dodson, E. J. (1997)
Refinement of macromolecular structures by the maximum-likelihood
method. Acta Crystallogr., Sect. D: Biol. Crystallogr. 53, 240−255.
(34) Brunger, A. T. (1993) Assessment of phase accuracy by cross
̈
validation: the free R value. Methods and applications. Acta
Crystallogr., Sect. D: Biol. Crystallogr. 49, 24−36.
(35) Perrakis, A., Morris, R., and Lamzin, V. S. (1999) Automated
protein model building combined with iterative structure refinement.
Nat. Struct. Biol. 6, 458−463.
(36) Read, R. J. (1986) Improved Fourier coefficients for maps using
phases from partial structures with errors. Acta Crystallogr., Sect. A:
Found. Crystallogr. 42, 140−149.
234
dx.doi.org/10.1021/cb200267w|ACS Chem. Biol. 2012, 7, 226−234