In Silico Evaluation of the Binding Site of Fucosyltransferase 8 and First Attempts to Synthesize an Inhibitor with Drug-Like Properties

Article abstract of DOI:10.1002/cbic.201900289

Core fucosylation of N-glycans is catalyzed by fucosyltransferase 8 and is associated with various types of cancer. Most reported fucosyltransferase inhibitors contain non-drug-like features, such as charged groups. New starting points for the development of inhibitors of fucosyltransferase 8 using a fragment-based strategy are presented. Firstly, we discuss the potential of a new putative binding site of fucosyltransferase 8 that, according to a molecular dynamics (MD) simulation, is made accessible by a significant motion of the SH3 domain. This might enable the design of completely new inhibitor types for fucosyltransferase 8. Secondly, we have performed a docking study targeting the donor binding site of fucosyltransferase 8, and this yielded two fragments that were linked and trimmed in silico. The resulting ligand was synthesized. Saturation transfer difference (STD) NMR confirmed binding of the ligand featuring a pyrazole core that mimics the guanine moiety. This ligand represents the first low-molecular-weight compound for the development of inhibitors of fucosyltransferase 8 with drug-like properties.

Full text of DOI:10.1002/cbic.201900289

Accepted Article
Title: In silico evaluation of the binding site of fucosyltransferase 8 and
first attempts of synthesizing an inhibitor with drug-like properties
Authors: Claas Strecker, Melissa Baerenfaenger, Michaela Miehe,
Edzard Spillner, and Bernd Meyer
This manuscript has been accepted after peer review and appears as an
Accepted Article online prior to editing, proofing, and formal publication
of the final Version of Record (VoR). This work is currently citable by
using the Digital Object Identifier (DOI) given below. The VoR will be
published online in Early View as soon as possible and may be different
to this Accepted Article as a result of editing. Readers should obtain
the VoR from the journal website shown below when it is published
to ensure accuracy of information. The authors are responsible for the
content of this Accepted Article.
To be cited as: ChemBioChem 10.1002/cbic.201900289
Link to VoR: http://dx.doi.org/10.1002/cbic.201900289
A Journal of
www.chembiochem.org

10.1002/cbic.201900289
ChemBioChem
FULL PAPER
In silico evaluation of the binding site of fucosyltransferase 8 and
first attempts of synthesizing an inhibitor with drug-like
properties
Claas Strecker,[a] Melissa Baerenfaenger,[a,b] Michaela Miehe,[c] Edzard Spiller,[c] and Bernd Meyer*[a]
Abstract: Core fucosylation of N-glycans is catalyzed by
fucosyltransferase 8 and is associated with various types of cancer.
Most reported fucosyltransferase inhibitors carry non drug-like
features, like charged groups. New starting points for development
of inhibitors of fucosyltransferase 8 are presented using a fragment-
based strategy. First, we discuss the potential of a new putative
binding site of fucosyltransferase 8 that becomes accessible by a
significant motion of the SH3-domain according to an MD simulation.
This might enable design of completely new inhibitor types for
fucosyltransferase 8. Second, we performed a docking campaign
against the donor binding site of fucosyltransferase 8 yielding two
fragments that were linked and trimmed in silico. The resulting ligand
was synthesized. STD NMR confirmed binding of the ligand with a
Figure 1. Fucosyltransferase 8 (FUT8) catalyzes the transfer of a fucose
moiety from the donor substrate GDP-Fucose onto the 6-OH group of
pyrazole core that mimics the guanine moiety. This low affinity ligand
represents the first low-molecular weight molecule for development
of inhibitors of fucosyltransferase 8 with drug-like properties.
innermost N-acetylglucosamine of the acceptor N-glycan yielding a core
fucosylated N-glycan.
For therapeutic antibodies, it has been shown that depletion of
core fucosylation results in a 50-100 fold increase of Fcγ-
mediated cytotoxicity.2 As a result, methods for the production of
non-fucosylated antibodies are of significant commercial interest.
Functional disruption of FUT8 in CHO cell lines has previously
been achieved by the use of zinc-finger nucleases.3 Furthermore,
core fucosylation has been implicated to play a role in various
types of cancer. For example, upregulation of FUT8 is a driver of
melanoma metastasis,4 associated with an unfavourable clinical
outcome in patients with non-small cell lung cancer,5 and with
aggressive prostate cancer.6 The important physiological role of
FUT8 is underlined by the fact that 70% of FUT8-deficient
knock-out mice die within three days after birth.7 Inhibitors of
FUT8 would therefore constitute an efficient tool for more
detailed studies on the physiological role of FUT8 and might
demonstrate therapeutic potential as well.
Introduction
Fucosyltransferases catalyze the transfer of fucose from the
donor substrate GDP-Fucose onto an acceptor glycan. The
human genome encodes 13 fucosyltransferases (FUT1-11 and
POFUT1-2). Even though they show distinct substrate
preferences, a certain degree of redundancy exists. For example
FUT1 and FUT2 both yield α1,2-linked fucose and both are able
to act on O- as well as on N-glycans.1 In contrast,
fucosyltransferase 8 (FUT8) is the only fucosyltransferase that
aids the construction of α1,6-linked fucose and that acts
exclusively on N-glycans.1 The enzyme transfers fucose to the
innermost N-acetyl glucosamine of N-glycans, resulting in core
fucosylation (cf. Fig. 1).
Development of glycosyltransferase inhibitors has been held
back by a ubiquitous focus on substrate analogues.8 Analogues
of the donor substrate offer easily reachable binding affinity as
the phosphate group(s) of the donor substrate account for a
major portion of the binding affinity. Analogues of the acceptor
substrate offer a level of specificity that would be very tedious to
achieve otherwise. However, these close substrate analogues
have limited potential because they typically feature non drug-
like properties, e.g. a high polarity, that diminish their cell
viability as formulated in Lipinski’s rule of five.9 Specifically,
previous statements apply for fucosyltransferase inhibitors as
well (see the review by Tu et al.).10 To circumvent the problems
that are associated with inhibitors, which feature non drug-like
properties, we have embraced a fragment-based strategy for the
development of inhibitors of FUT8 that allows for more control
over molecular properties.
[a]
M. Sc. C. Strecker, M. Sc. M. Baerenfaenger, Prof. Dr. B. Meyer
Department of Chemistry
University of Hamburg
Martin-Luther-King-Platz 6, 20146 Hamburg, Germany
E-mail: bernd.meyer@chemie.uni-hamburg.de
Present address:
[b]
[c]
Department of Neurology
Radboud University Medical Center
Geert Grooteplein 10, Nijmegen, 6525 GA, The Netherlands
Dr. M. Miehe, Prof. Dr. E. Spillner
Department of Engineering
Aarhus University
Gustav Wieds Vej 10, 8000 Aarhus, Denmark
Supporting information for this article is given via a link at the end of
the document.
For internal use, please do not delete. Submitted_Manuscript
This article is protected by copyright. All rights reserved.

10.1002/cbic.201900289
ChemBioChem
FULL PAPER
So far, three publications have reported inhibitors of FUT8.11-13
Manabe et al.11 and Hosoguchi et al.12 reported analogues of
GDP-Fucose that inhibit FUT8. Their most potent inhibitors (as
shown in supplemental Fig. S1) exhibit affinities in the lower M
range. However, the corresponding clogP values of less than -
1.5 indicate that bioavailability is not present or extremely low.
Kamińska et al. reported a set of triazine dyes that inhibits
FUT8.13 Their most potent inhibitor reactive red 120 (see
supplemental Fig. S1) carries six ionic sulfonates that will
prevent membrane permeability. In addition, the molecular mass
of more than 1300 Da grossly violates Lipinski’s rule. While the
reported Ki of 2 M for reactive red 120 sounds promising on
first sight, the concept of ligand efficiency (LE = G/no. of heavy
atoms)14 proves that reactive red 120 is a very inefficient binder
with a ligand efficiency of only 0.09.
confusing, we used multidimensional scaling (a form of
dimensionality reduction)19 to point out this conformational
change (see the lower part of panel A in Fig. 2). Next, we
analyzed the origin of the conformational change. For this, for all
investigated structures the RMSD (of the C-atom) of each
residue is calculated in reference to the initial frame of this MD
simulation. The resulting data is plotted in panel B of Fig. 2. The
plot reveals that the conformational change is caused by
residues of the SH3 domain of FUT8 (colored in magenta in the
lower part of panel B in Fig. 2). The SH3 domain of FUT8 is
located at the C-terminus.16 SH3 domains are known to play a
role in protein-protein interactions, e.g. in signal-transduction
networks.20 However, for the SH3 domain of FUT8 no such
function has been described so far.16 Yet, the SH3 of FUT8 is
responsible for the accommodation of the 6-branch of the
acceptor N-glycan.18 It has been noted previously that FUT8 is
significantly more tolerant to modifications of the 6-branch of the
acceptor N-glycan than to modifications of the 3-branch.15 The
significant flexibility of the SH3 domain observed in this MD
simulations can explain this acceptor specificity well. Even more
interestingly, the observed movement of the SH3 domain of
FUT8 opens a channel (see panel C of Fig. 2) that connects to a
putative binding site (cf. Fig. 3 and the next paragraph).
The acceptor substrate preference of FUT8 has been
investigated: FUT8 shows its highest conversion rates on the
heptasaccharide displayed in Fig. 1.15 While FUT8 tolerates
modifications of the 6-branch, modifications of 3-branch are
usually not accepted.15 A X-ray crystal structure of apo FUT8
exists (PDB: 2DE0).16 On its basis, models of the binding mode
of GDP-Fucose and the acceptor N-glycan have been
developed.17-18
We assessed the druggability of this putative binding site in
silico. For this we used FTMap, an application for the
identification of binding „hot spots“.21 Prior work e.g. from Mattos
et al., who crystallized proteins in the presence of organic
solvent molecules to investigate binding „hot spots“, laid the
experimental foundation for in silico tools such as FTMap.22
FTMap places small molecular probes (such as acetamide,
ethanol, cyclohexane and urea) with varying functionality and
size onto the surface of the investigated protein and finds likely
binding sites for these probes by the use of energy functions.21
The MD-derived structure shown on the right side of panel C in
Fig. 2 was taken to FTMap. Interestingly, on the back side
(viewing from the acceptor site) of the channel that opens in the
described MD simulation four binding „hot spots“ were identified
by FTMap (as shown in Fig. 3). These four „hot
spots“ accommodate a total of 45 probe molecules (see
supplemental Tab. S1 for a list). In comparison, for the donor
binding site of FUT8 (of the same MD-derived structure) FTMap
identified four „hot spots“ as well. However, these accommodate
only a total of 23 probe molecules. This indicates that this
putative binding site might be druggable. A ligand binding to this
putative binding site and extending into the channel towards the
acceptor binding site might lock this conformation of the
acceptor site and therefore alter the acceptor preferences of
FUT8. A ligand that extends even further from the channel into
the acceptor binding site could possibly disrupt the enzymatic
activity of FUT8. Additionally, ligands of this putative binding site
might offer specificity over other fucosyltransferases. In contrast,
ligands that target the donor binding site of FUT8 will likely also
inhibit other fucosyltransferases as all of them use GDP-Fucose
as a donor substrate.
Results and Discussion
In the absence of potent and selective inhibitors of FUT8,
biochemical studies lack an efficient tool to clearly define the
physiological role of the product(s) of this enzyme. FUT8 has a
binding site that encompasses eight monosaccharide residues
(one fucosyl residue and seven monosaccharide building blocks
from the acceptor),
a purine based nucleoside and a
pyrophosphate group. It is very difficult to find molecules that
bind to such large binding sites and exert affinity and specificity.
Here, we present our approaches to discover inhibitors for FUT8.
In a first part, a putative new binding site and its potential is
discussed. In a second part, a fragment-based approach is
utilized to generate new chemical entities for binders of the
donor site.
A Putative Allosteric Binding Site:
An X-ray crystal structure of human FUT8 (PDB: 2DE0) served
as the starting point to our structure-based ligand discovery
process.16 This is an apo crystal structure and a key catalytic
residue (Arg-365) is collapsed into the donor binding site.
Therefore, we created a model for donor substrate binding that
has previously been developed in our working group.17 On the
basis of this model, we performed a MD simulation of a length of
20 ns to explore the flexibility of FUT8. The analysis of this MD
simulation showed evidence of a putative allosteric binding site
that will be discussed in the following. Panel A of Fig. 2 displays
a matrix of pairwise RMSD values (for all backbone atoms) of
structures taken from this MD simulation. The RMSD matrix
shows
a significant conformational change occurring after
approximately 14.4 ns. Because RMSD matrices can be rather
For internal use, please do not delete. Submitted_Manuscript
This article is protected by copyright. All rights reserved.

10.1002/cbic.201900289
ChemBioChem
FULL PAPER
Figure 2. An MD simulation of FUT8 with a length of 20 ns was performed. A) An RMSD matrix (for all backbone atoms) of this MD simulation reveals a
significant conformational change occurring after approximately 14.4 ns. Multidimensional scaling of this RMSD matrix highlights the conformational change in an
easily comprehensible fashion. B) RMSD (of the C-atom) of each residue (as referenced to the initial frame of the MD simulation) plotted against the simulation
time. The plot reveals that the conformational change is caused by residues of the SH3 domain (colored in magenta). C) Two structures illustrate the previous
findings: A significant snap back of the SH3 domain occurs. Interestingly, this movement opens a channel (colored in cyan) that connects to a putative allosteric
binding site (cf. Fig. 3 and the associated paragraph)
For internal use, please do not delete. Submitted_Manuscript
This article is protected by copyright. All rights reserved.

10.1002/cbic.201900289
ChemBioChem
FULL PAPER
(ZINC95830521) was identified as an interesting hit ranked
second place in the docking hit list (see panel A in Fig. 5 for a
binding pose, Tab. 1 for a short summary of docking results and
supplement Tab. S2 for a longer hit list). The NH of the pyrazole
moiety and a NH of the primary carboxamide function form
pincer-like hydrogen bonding to Asp-453. The carbonyl oxygen
of the second carboxamide function is able to hydrogen bond to
His-363. In the second docking campaign, a large number of
carboxylic acids hydrogen bonding to Arg-365 was identified
with high docking scores (see supplement Tab. S3 for a hit list).
In particular, 3,5-dihydroxyhydrocinnamic acid 2 (ZINC6091356)
that was ranked 42nd place in the docking hit list caught our
attention due to its structural simplicity and was subsequently
exempt from one of the phenolic OH functions to yield 3-(3-
hydroxyphenyl)propionic acid 3 (ZINC156346, see panel B in Fig.
5 for a binding pose). Both fragments, 1 and 3, were linked in
silico to yield ligand 4 (see panel C in Fig. 5 for a binding pose).
This ligand showed an improved calculated binding energy
compared GDP-Fucose in MM-GBSA calculations (see Tab. 1
for calculated binding energies). In order to reduce the synthetic
effort, we decided to remove the phenolic OH function of 4:
Otherwise, protection of the phenolic OH group would have
been inevitable and the synthesis of the corresponding building
block would have started from 5-hydroxyisophthalic acid. The
resulting ligand 5 (see panel D in Fig. 5 for a binding pose) still
showed an improved calculated binding energy compared to
GDP-Fuc (see Tab. 1). Importantly, ligand 5 exhibited stable
hydrogen bonding to the three previously defined key amino acid
Figure 3. FTMap identifies four binding „hot spots“ on the back side (viewing
from the acceptor site) of the channel (colored in cyan) that opens in the MD
simulation. The MD-derived structure shown on the right side of panel C in
Fig. 2 was used. This new putative binding site might be a target for inhibitor
design and might offer specificity over other fucosyltransferases.
Design of a Ligand for the Donor Site:
For FUT8, the purine base and the pyrophosphate represent the
structural proportions that dominantly contribute to the affinity of
substrate binding.17-18 The other parts, i.e. the fucosyl residue
and the heptasaccharide, exert specificity but little affinity.
Therefore, we primarily focused on the design and synthesis of a
donor site inhibitor. At a later stage, we hope to link this donor
site binder to structural proportions that are anchored in the new
binding site described above and induce specificity by this.
Therefore, we started a docking campaign against the donor site
of FUT8. For this, we used the MD-derived structure shown in
Fig. 4.
residues (Asp-453, His-363, and Arg-365) during
simulation of a length of 1.5 ns.
a MD
Table 1. Docking scores (from GlideScore SP5.0) and calculated binding
energies (from Prime MM-GBSA) for the donor substrate GDP-Fucose,
docked fragments and ligands after in silico linking.
Ligand
Docking Score
n.d.
MM-GBSA [kcal/mol]
GDP-Fuc
-48.5
-27.7
-33.4
-60.3
-53.8
1
3
4
5
-6.116
-6.413
-9.182
-8.363
Figure 4. Binding mode of GDP-Fucose. Hydrogen bonds are shown in yellow.
We hypothesized that three amino acid residues are of central significance for
donor substrate binding: Asp-453, His-363, and Arg-365 (shown as magenta
sticks).
We hypothesized that a potential ligand of the donor binding site
should exhibit hydrogen bonding to three key amino acid
residues: Asp-453, His-363, and Arg-365 (displayed as magenta
sticks in Fig. 4). We ran two separate docking campaigns
against the donor binding site of FUT8 employing a library of
700,000 fragments obtained from the ZINC12 database.23 Two
rather small grid boxes were used to ensure hits that cover the
entire binding site. The grid box used in the first docking
campaign was centered around the midpoint of the guanine
moiety, the second centered on the -phosphate (cf. supplement
Fig. S2). In the first campaign, 1H-pyrazole-3,5-dicarboxamide 1
For internal use, please do not delete. Submitted_Manuscript
This article is protected by copyright. All rights reserved.

10.1002/cbic.201900289
ChemBioChem
FULL PAPER
Figure 6. Retrosynthetic analysis of ligand 5.
The synthesis of building block 6 was started from 8 which was
accessed by employing a route similar to that described by
Skinner et al.24 The ester function of 8 was ammonolysed to the
primary amide 9 but in contrast to Skinner et al.,24 who
employed dry ammonia for this, we relied on a protocol by
Jagdmann et al.25 that releases ammonia by refluxing formamide
in the presence of sodium methanoate. Finally, the aromatic
methyl group of 9 was oxidized to give a carboxylic acid by the
action of potassium permanganate.
Figure 7. Synthesis of building block 6. Reaction conditions: a) 4.0 eq.
NaOMe, 6.0 eq. formamide, tetrahydrofuran, 5 h, 65 °C; b) 3.0 eq. KMnO4,
H2O, 3 h, 95 °C.
Building block
7 is currently commercially available. We
accessed 7 by a reaction sequence outlined in the supplement.
Both building blocks, 6 and 7, were subsequently linked via
amide coupling using propylphosphonic anhydride as coupling
reagent yielding 10. Finally, the ester function of 10 was
hydrolyzed by the action of potassium hydroxide to yield ligand 5.
Figure 5. Docking a fragment library yielded the two fragments, 1 and 3 (panel
A and B), that were subsequently linked in silico to yield ligand 4 (panel C).
After exemption of the phenolic OH function ligand 5 (panel D) was obtained.
Synthesis of a Ligand for the Donor Site:
We envisioned a synthesis of ligand 5 by amide coupling of
building blocks 6 and 7 and subsequent ester deprotection (see
Fig. 6).
Figure 8. Synthesis of ligand 5. Reaction conditions: a) 2.0 eq. DIPEA, 2.0 eq.
propylphosphonic anhydride, DMF, 24 h, 20 °C; b) 3.0 eq. KOH, H2O/MeOH
1:4, 24 h, 20 °C.
For internal use, please do not delete. Submitted_Manuscript
This article is protected by copyright. All rights reserved.

10.1002/cbic.201900289
ChemBioChem
FULL PAPER
Activity Assay:
We verified the functional activity of recombinantly expressed
FUT8 by monitoring the fucosylation of 1--N-acetylchitotriose
1
via H-NMR spectroscopy over a course of 3.5 h. The progress
of the reaction is visualized in Fig. 9. We derived initial rate
constants from this period that are summarized in Tab. 2.
Fucosylated 1--N-acetylchitotriose was produced with an
apparent initial rate constant of 0.010 s-1. This is in agreement
with results of Ihara et al. who found a 5% activity for the core-
fucosylation of chitotriose compared to the heptassacharide N-
glycan.26
Figure 10. Determination of the dissociation constant of ligand 5 via STD
NMR. The STD amplification factor of H-4’/H-6’ (phenyl ring) is plotted against
the concentration of 5. As a result, by employing a one-site binding model the
Kd of 5 was determined to be 1.65 ± 0.97 mM.
Results and Discussion
FUT8 has a very large and shallow binding site that is composed
of the donor and the acceptor binding site. The acceptor binding
site is known to stretch over five monosaccharide residues. It is
very difficult to find inhibitors for such large and open binding
sites. However, potent inhibitors of FUT8 are imperatively
needed to clearly define its physiological role and might open
new therapeutic avenues as well. We discussed a new putative
binding site of FUT8 that can become accessible by a significant
movement of the SH3 domain. An in silico analysis of this
putative new binding site showed that a number of small
fragments can bind to this area quite well. We expect a
significant impact on future development of FUT8 inhibitors if
this binding site can be verified experimentally: This putative
binding site offers the prospect of adding specificity over other
fucosyltransferases. Furthermore, we present a novel approach
to generate inhibitors for FUT8 in that a fragment-based
discovery process is utilized to discover new binders. For this,
we performed a docking campaign with more than 700,000
fragments. The campaign yielded two interesting fragments.
These two fragments were subsequently linked in silico and the
binding of the resulting low-molecular weight ligand 5 was
theoretically evaluated. Ligand 5 was synthesized and the
corresponding dissociation constant was determined to 1.65 mM
via STD NMR. Overall, the determined Kd is significantly higher
than anticipated from MM-GBSA calculations (see section
Ligand Design). This can be attributed to well-known
shortcomings of MM-GBSA calculations. These include
problems with entropic contributions and with charged ligands.27
Even though we certainly hoped for a more affine ligand, we
think that this molecule can be used as starting point for the
development of inhibitors of FUT8 with drug-like properties.
Taking the concept of ligand efficiency (LE = -G/no. of heavy
atoms) into consideration,14 the presented ligand 5 features a
ligand efficiency of LE = 0.17. In comparison, the most potent
inhibitor of FUT8 known to date, reactive red 120 with a Ki of
2 M,13 exhibits only a ligand efficiency of LE = 0.09 (due to its
high molecular weight of > 1.3 kDa). Surely, an efficient binder
should exhibit a LE > 0.3 but we take comfort in the fact that
Figure 9. Activity assay for recombinantly expressed FUT8. The transfer of
fucose from GDP-Fucose onto 1--N-acetylchitotriose was monitored via 1H-
NMR spectroscopy at 310 K.
Table 2. . Initial rate constants for the depletion of GDP-Fucose and 1--N-
acetylchitotriose and the production of GDP, fucosylated 1--N-
acetylchitotriose and fucose.
Fuc-Chito-
triose
GDP-Fuc
-0.026
GDP
Chitotriose
-0.011
Fucose
0.019
kini [s-1]
0.032
0.010
STD NMR:
Finally, we evaluated the dissociation constant for ligand 5. STD
NMR revealed a Kd of 1.65 ± 0.97 mM (evaluating H-4’/H-6’
(phenyl ring), see Fig. 10). A STD NMR spectrum acquired from
a sample without added enzyme revealed no STD artefacts.
Evaluation of the signal for the H-2’ (phenyl ring) and H-4’’
(pyrazole) yielded a higher Kd of 4.47 and 7.72 mM respectively.
The signals of H-2 and H-3 (right next to the carboxy function)
showed no STD effect. This is in accordance with the proposed
binding mode in that all of the latter protons should be more
solvent exposed than H-6’. Unfortunately, H-5’ (phenyl ring)
could not be evaluated because its signal is subsided by an
impurity originating from the protein solution. Similarly, the
benzyl protons could not be evaluated as they are strongly
affected by water suppression.
For internal use, please do not delete. Submitted_Manuscript
This article is protected by copyright. All rights reserved.

10.1002/cbic.201900289
ChemBioChem
FULL PAPER
potent inhibitors have emerged from low-affinity starting points
before. For example, Chessari et al. reported the development
of nanomolar inhibitors of cIAP1 starting from a fragment with an
IC50 > 5 mM and a ligand efficiency of LE < 0.21.28 We expect
that in order to obtain specific inhibitors of FUT8 an extension of
ligand 5 towards the acceptor binding site will be necessary.
Even though a considerable enhancement of binding potency
will be necessary in the ligand optimization process, we are
confident that the expansion of this process will yield potent
inhibitors of FUT8.
Synthesis:
5-Methyl-1H-pyrazole-3-carboxamide
9:
Synthesis
was
performed in accordance to a protocol by Jagdmann et al.25
8
(2.336 g, 15.15 mmol, 1.0 eq.) was dissolved in tetrahydrofuran
(40 mL). Formamide (3.62 mL, 4.09 g, 90.9 mmol, 6.0 eq.) was
added. Then, a 2.5 M solution of sodium methanoate (24.5 mL,
61.2 mmol, 4.0 eq.) in methanol was added. The reaction
mixture was heated to reflux for 5 h and neutralized with dilute
hydrochloric acid. Subsequently, the solvent was evaporated.
The crude product was desalted by filtration over silica gel
(eluent: dichloromethane/methanol 7:1). The residue was
purified by column chromatography on silica gel (eluent:
dichloromethane/methanol 7:1). 9 was obtained as a colorless
Experimental Section
solid
(841 mg,
6.72 mmol,
44%).
Rf:
0.23
(dichloromethane/methanol 7:1); 1H-NMR (500 MHz, DMSO-d6):
δ [ppm] = 12.83 (s, 1H, NH), 7.37 (s, 1H, CONH2), 7.10 (s, 1H,
CONH2), 6.37 (s, 1H, H-4), 2.22 (s, 3H, CH3); 13C-NMR
(126 MHz, DMSO-d6): δ [ppm] = 163.8 (CONH2), 146.9 (C-3),
139.7 (C-5), 104.2 (C-4), 10.4 (CH3); HRMS (ESI+): expt.
148.0466 ([M+Na]+), calc. 148.0487 ([M+Na]+).
Molecular Modeling:
We used Schrödinger’s Maestro for molecular modeling.
Schrödinger includes the software modules Prime, LigPrep,
Glide, Desmond, and QikProp. The default force field is OPLS
2005.29 On the basis of the only available X-ray crystal structure
of human FUT8 (PDB: 2DE0) we recreated a model for donor
substrate binding that has previously been developed in our
working group.17 The FUT8-GDP-Fucose complex was taken to
Desmond and fitted into an orthorhombic water box (SPC
model) expanding 10 Å in each direction from the complex. Then,
a 20 ns long MD simulation (Desmond v3) was performed using
the NPT ensemble at 310 K. For the RMSD matrix (as shown in
panel A of Fig. 2) every 100th structure (resulting in a total of 402
structures) of this MD simulation was exported to PyMol. All non-
backbone atoms were deleted. Then, pairwise RMSD values for
all structures were calculated by iterating PyMOL’s “align”
command with the cycles option set to zero employing a basic
Python script. Multidimensional scaling of the resulting RMSD
matrix was performed using Matlab’s “mdscale” command using
“sammon” as criterion and two dimensions. For the plot shown in
panel B of Fig. 2, all atoms except C-atoms were deleted. Then,
structures were aligned to the initial frame of the MD simulation
(employing PyMOL’s “align” command with the cycles option set
to zero). Finally, for the C of each residue the RMSD value was
calculated by employing PyMOL’s “rms_cur” command. For the
druggability assessment, the specified structure was uploaded to
the FTMap server (http://ftmap.bu.edu/serverhelp.php).
3-Carbamoyl-1H-pyrazole-5-carboxylic acid 6.
9
(100 mg,
800 μmol, 1.0 eq.) was suspended in water (10 mL). Potassium
permanganate (379 mg, 2.40 mmol, 3.0 eq.) was added. The
reaction mixture was stirred at 95 °C for 3 h and then filtrated
while still hot. The filter cake was washed with hot water twice.
The filtrate was treated with sodium sulfite until discoloration and
then the pH value was adjusted to 1 with concentrated sulfuric
acid. The solution was stored overnight at 4 °C during which
product precipitated. The product was filtered off.
6 was
1
obtained as a colorless solid (89 mg, 570 μmol, 72%). H-NMR
(500 MHz, DMSO-d6): δ [ppm] = 10.11 (bs, 2H, COOH, NH),
7.84 (s, 1H, CONH2), 7.46 (s, 1H, CONH2), 7.17 (s, 1H, H-4);
13C-NMR (126 MHz, DMSO-d6): δ [ppm] = 161.6, 161.3 (CONH2,
COOH), 142.9, 140.0 (C-3, C-5), 108.3 (C-4); MS (ESI-): expt.
154.0174 ([M-H]-), calc. 154.0258 ([M-H]-).
Methyl 3-(3-((5-carbamoyl-1H-pyrazole-3-carboxamido)methyl)-
phenyl)propanoate 10. 6 (69 mg, 447 μmol, 1.0 eq.) and 7
(95 mg, 492 μmol, 1.1 eq.) were dissolved in N,N-
dimethylformamide (2 mL). Then N,N-diisopropylethylamine
(152 μL, 116 mg, 894 μmol, 2.0 eq.) was added and
subsequently a solution of propylphosphonic anhydride (284 mg,
894 μmol, 2.0 eq.) in N,N-dimethylformamide (50wt%, 522 μL).
The reaction mixture was stirred at room temperature for 24 h
and then freed of the solvent in vacuo. The residue was taken
up in 5% hydrochloric acid and extracted with ethyl acetate. The
combined organic phases were dried over sodium sulfate,
filtered and freed of the solvent in vacuo. The residue was
purified via RP-HPLC (column: Nucleodur C18 Isis; solvent A:
95% H2O + 5% MeCN; solvent B: 95% MeCN + 5% H2O;
gradient: 0-5 min, 35% B; 5-13 min, 100% B; 13-15 min, 100%
B; 15-18 min, 100% B; 18-20 min, 35% B; flow rate: 20 mL/min;
Rt = 5.2 min). 10 was obtained as a colorless solid (51.5 mg,
156 μmol, 32%). Rf: 0.37 (dichloromethane/methanol 9:1); 1H-
NMR (500 MHz, DMSO-d6): δ [ppm] = 13.92 (bs, 1H, NH
(pyrazole)), 8.88 (bs, 1H, R1CONHCH2-R2), 7.86 (bs, 1H,
CONH2), 7.44 (bs, 1H, CONH2), 7.25 (s, 1H, H-4’’), 7.23 (dd, 1H,
3J = 7.7 Hz, 3J = 7.7 Hz, H-5’), 7.16-7.14 (m, 1H, H-2’), 7.14-7.11
(m, 1H, 3J = 7.7 Hz, H-4’/H-6’), 7.10-7.07 (m, 1H, 3J = 7.5 Hz, H-
We used the “FragsNow” subset of the ZINC12 databank for our
docking campaigns. This library contains fragments with a
molecular weight of less than 250 Da, a clogP of less than 3.5
and less than 5 rotatable bonds. This library was prepared at pH
7 ± 0.2 using Maestro’s LigPrep (with default options except pH)
totaling about 700,000 ligands. Two independent docking
campaigns were performed: The first one using a grid box
centered around guanine (expanding 15, 17 and 17 Å in X, Y
and Z direction) and the second one centered on the -
phosphate of GDP-Fuc (expanding 17 Å in X, Y, and Z direction,
cf. supplement Fig. S2). Docking was performed using Glide
(using default options). Initially the “HTVS” scoring function was
used. Top hits were then redocked using the “Standard
Precision” scoring function (GlideScore SP5.0). For MM-GBSA
calculations, we used Prime (v3.0, OPLS3 force field). QikProp
was used for calculations of log P values (QPlogPo/w is
reported).
3
4’/H-6’), 4.40 (d, 2H, J = 5.9 Hz, R1CONHCH2R2), 3.56 (s, 3H,
For internal use, please do not delete. Submitted_Manuscript
This article is protected by copyright. All rights reserved.

10.1002/cbic.201900289
ChemBioChem
FULL PAPER
COOCH3), 2.83 (t, 2H, 3J = 7.6 Hz, H-3), 2.60 (t, 2H, 3J = 7.7 Hz,
300 mM imidazole. Purification was confirmed by SDS-PAGE
and immunoblotting.
H-2); 13C-NMR (126 MHz, DMSO-d6):
δ [ppm] = 172.6
(COOCH3), 140.5, 139.6 (C-1’, C-3’), 128.3 (C-5’), 127.1 (C-2’),
126.6, 125.1 (C-4’, C-6’), 105.7 (C-4’’), 51.3 (COOCH3), 41.9
(R1CONHCH2R2), 34.8 (C-2), 30.2 (C-3); HRMS (ESI+): expt.
331.1360 ([M+H]+), calc. 331.1401 ([M+H]+).
NMR experiments:
Activity assay and STD-NMR were measured on a Bruker
Avance 700 MHz spectrometer in 3 mm NMR tubes. Samples
were prepared in D2O containing MES-d13 (50 mM) and TMSP-
d4 (1 mM) at pD 7.0. The FUT8 solution was rebuffered to the
buffer specified above by using Amicon Ultra-4 cellulose filter
(molecular weight cut-off 5 kDa) and the protein concentration
was determined by using a nanodrop at 280 nm.
3-(3-((5-Carbamoyl-1H-pyrazole-3-carboxamido)methyl)phenyl)-
propanoic acid 5. 10 (41 mg, 124 μmol, 1.0 eq.) was dissolved in
methanol (8 mL). Then, potassium hydroxide (21 mg, 372 μmol,
3.0 eq.) was dissolved in water (2 mL) and added to the solution
described above. The reaction mixture was stirred at room
temperature for 24
h
and then neutralized with dilute
Activity assay: The fucosylation of 1--N-acetylchitotriose was
1
hydrochloric acid. Subsequently, the solvent was evaporated.
The residue was desalted by filtration over silica gel (eluent:
dichloromethane/methanol 9:1 + 0.5% formic acid). The crude
product was purified via RP-HPLC (column: Nucleodur C18 Isis;
solvent A: H2O + 0.01% NH3; solvent B: MeCN; gradient: 0-10
min, 10% B; 10-20 min, 40% B; 20-22 min, 90% B; 22-25 min,
90% B; 25-27 min, 10% B; 27-30 min, 10% B; flow rate: 1
mL/min; Rt = 5.0-7.0 min). 5 was obtained as a colorless solid
(22.7 mg, 71.9 μmol, 58%). Rf: 0.17 (dichloromethane/methanol
monitored via H-NMR over a course of 3.5 h. Every 15 min a
1H-NMR spectra with excitation sculpting was acquired at 310 K
using a pseudo 2D pulse program. The first data point was
acquired after 11.5 min. The sample contained
concentration of 2 μM, a GDP-Fucose concentration of 2.6 mM
and 1--N-acetylchitotriose concentration of 2.3 mM.
a FUT8
a
Additionally, the sample contained 10 U of alkaline phosphatase
(EC 3.1.3.1) and 1 mg/mL bovine serum albumin. Spectra were
acquired with 32,768 data points and a total of 64 scans. FIDs
were multiplied with an exponential function (line broadening
1
9:1 + 0.5% formic acid); H-NMR (500 MHz, DMSO-d6): δ [ppm]
= 8.95 (bs, 1H, R1CONHCH2R2), 7.88 (bs, 1H, CONH2), 7.42 (bs, 0.2) before Fourier transformation. The concentration of assay
3
3
1H, CONH2), 7.23 (s, 1H, H-4’’), 7.21 (dd, 1H, J =7.6 Hz, J =
7.6 Hz, H-5’), 7.17-7.15 (m, 1H, H-2’), 7.12-7.07 (m, 2H, H-4’, H-
components was determined from H-8 of GDP-Fuc and GDP,
H-1 of (fucosylated) 1--N-acetylchitotriose and the H-6 methyl
group of fucose.
3
3
6’), 4.40 (d, 2H, J = 6.1 Hz, R1CONHCH2R2), 2.78 (t, 2H, J =
7.6 Hz, H-3), 2.43 (t, 2H, 3J = 7.6 Hz, H-2); 13C-NMR (126 MHz,
DMSO-d6): δ [ppm] = 174.5 (COOH), 161.3 (CONH2), 160.1
(R1CONHCH2R2), 141.5 (C-1’), 139.4 (C-3’), 128.2 (C-5’), 127.4
(C-2’), 126.6, 124.8 (C-4’, C-6’), 105.8 (C-4’’), 42.0
(R1CONHCH2R2), 36.2 (C-2), 30.8 (C-3); HRMS (ESI-): expt.
315.1027 ([M-H]-), calc. 315.1093 ([M-H]-). Purity of 5 was
determined from HPLC-MS (cf. supplement Fig. S12) to 96%.
STD NMR: The standard pulse program “stddiffesgp2d” was
used. On resonance irradiation was applied at 0 ppm. Saturation
was achieved by a cascade of 40 Gaussian pulses with a
duration of 50 ms. To reduce the protein background a spinlock
pulse of 15 ms length was used. Experiments were performed at
300 K. All samples contained a FUT8 concentration of 5 μM and
ligand excesses of 5 ranging from 41 to 199. Spectra were
acquired with 24,576 data points and a total of 512 scans. FIDs
were multiplied with an exponential function (line broadening 2)
before Fourier transformation. For the determination of
dissociation constants STD amplifications factors were plotted
against the ligand concentration. The data points were then
fitted using a one-site binding model in Origin2016G.
Protein Expression:
The FUT8 cDNA was inserted into the baculovirus transfer
vector pAcGP67B (BD Pharmingen, Heidelberg, Germany) and
modified by addition of an N-terminal 10-fold His-tag, a V5
epitope, and a factor Xa cleavage site. Spodoptera frugiperda
insect cells (Sf9) (Invitrogen) were grown at 27 °C in serum-free
medium (SFX-Insect cell culture medium HyClone (GE
Healthcare) containing 10 μg/ml gentamycin; Invitrogen).
Recombinant baculovirus was generated by cotransfection of
Sf9 cells with BaculoGold bright DNA (BD Pharmingen,
Heidelberg, Germany) and the baculovirus transfer vector pAC-
GP67-B containing FUT8. High titer stocks were produced by
three rounds of virus amplification and optimal MOI for protein
expression was determined empirically by infection of Sf9 cells
in 6 well plates (1.0×106 cells/well) with serial dilutions of high
titer virus stock. The high titer stock of recombinant baculovirus
was used to infect 500 mL suspension cultures of Sf9 cells
(1.0×106 cells per mL) in a roller bottle (850 cm2 growth area,
Greiner). For protein production, the cells were incubated at
27 °C and 110 rpm for 5 days. The supernatant of baculovirus-
infected cells was collected, and applied to a nickel-chelating
affinity matrix (HisTrap Excel, GE-Healthcare). The column was
rinsed with binding buffer (50 mM sodium phosphate, pH 8,
500 mM NaCl) and pre-eluted with NTA binding buffer
containing 20 mM imidazole. The recombinant protein was
eluted from the matrix using NTA-binding buffer containing
Keywords: Fragment-based drug discovery •
Fucosyltransferase 8 • Glycosylation • Glycosyltransferase
inhibitor • Molecular modeling
[1]
[2]
[3]
B. Ma, J. L. Simala-Grant, D. E. Taylor, Glycobiology, 2006, 16, 158-
184.
R. L. Shields, J. Lai, R. Keck, L. Y. O'Connell, K. Hong, G. Meng, S. H.
A. Weikert, L. G. Presta, L. G. J. Biol. Chem. 2002, 277, 26733-26740.
L. Malphettes, Y. Freyvert, J. Chang, P. Q. Liu, E. Chan, J. C. Miller, Z.
Zhou, T. Nguiyen, C. Tsai, A. W. Snowden, T. N. Collingwod, P. D.
Gregory, G. J. Cost, Biotechnol. Bioeng. 2010 106, 774-783.
P. Agrawal, B. Fontanals-Cirera, E. Sokolova, S. Jacob, C. A. Vaiana,
D. Argibay, V. Davalos, M. McDermott, S. Nayak, F. Darvishian, M.
Castillo, B. Ueberheide, I. Osman, D. Fenyö, L. K. Mahal, E. Hernando,
Cancer Cell 2017 31, 804-819.
[4]
[5]
R. Honma, I. Kinoshita, E. Miyoshi, U. Tomaru, Y. Matsuno, Y. Shimizu,
S. Takeuchi, Y. Kobayashi, K. Kaga, N. Taniguchi, H. Dosaka-Akita,
Oncology 2015, 88, 298-308.
For internal use, please do not delete. Submitted_Manuscript
This article is protected by copyright. All rights reserved.

10.1002/cbic.201900289
ChemBioChem
FULL PAPER
[6]
X. Wang, J. Chen, Q. K. Li, S. B. Peskoe, B. Zhang, C. Choi, E. A. Platz,
H. Zhang, Glycobiology 2014, 24, 935-944.
[7]
X. Wang, S. Inoue, J. Gu, E. Miyoshi, K. Noda, W. Li, Y. Mizuno-
Horikawa, M. Nakano, M. Asahi, M. Takahashi, N. Uozumi, S. Ihara, S.
H. Lee, Y. Ikeda, Y. Yamaguchi, Y. Aze, Y. Tomiyama, J. Fujii, K.
Suzuki, A. Kondo, S. D. Shapiro, C. Lopez-Otin, T. Kuwaki, M. Okabe,
K. Honke, N. Taniguchi, Proc. Natl. Acad. Sci. USA 2005, 102, 15791-
15796.
[8]
[9]
L. Tedaldi, G. K. Wagner, Med. Chem. Commun. 2014, 5, 1106-1125.
C. A. Lipinski, F. Lombardo, B. W. Dominy, P. J. Feeney, Adv. Drug
Deliv. Rev. 2001, 46, 3-26.
[10] Z. Tu, Y.-N. Lin, C.-H. Lin, Chem. Soc. Rev. 2013, 42, 4459-4475.
[11] Y. Manabe, S. Kasahara, Y. Takakura, X. Yang, S. Takamatsu, Y.
Kamada, E. Miyoshi, D. Yoshidome, K. Fukase, Bioorg. Med. Chem.
2017, 25, 2844-2850.
[12] K. Hosoguchi, T. Maeda, J.-i. Furukawa, Y. Shinohara, H. Hinou, M.
Sekiguchi, H. Togame, H. Takemoto, H. Kondo, S.-I. Nishimura, J.
Med. Chem. 2010, 53, 5607-5619.
[13] J. Kamińska, J. Dzięciol, J. Kościelak, Glycoconj. J. 1999, 16, 719-723.
[14] A. L. Hopkins, G. M. Keserü, P. D. Leeson, D. C. Rees, C. H. Reynolds,
Nat. Rev. Drug Discov. 2014, 13, 105-121.
[15] A. D. Calderon, Y. Liu, X. Li, X. Wang, X. Chen, L. Li, P. G. Wang, Org.
Biomol. Chem. 2016, 14, 4027-4031.
[16] H. Ihara, Y. Ikeda, S. Toma, X. Wang, T. Suzuki, J. Gu, E. Miyoshi, T.
Tsukihara, K. Honke, A. Matsumoto, A. Nakagawa, N. Taniguchi,
Glycobiology 2007, 17, 455-466.
[17] M. P. Kötzler, S. Blank, F. I. Bantleon, E. Spillner, B. Meyer, Biochim.
Biophys. Acta. 2012, 1820, 1915-1925.
[18] M. P. Kötzler, S. Blank, F. I. Bantleon, M. Wienke, E. Spillner, B. Meyer,
ACS Chem. Biol. 2013, 8, 1830-1840.
[19] J. B. Kruskal, Psychometrika 1964, 29, 1-27.
[20] B. J. Mayer, J. Cell Sci. 2001, 114, 1253-1263.
[21] D. Kozakov, D. R. Hall, R. L. Napoleon, C. Yueh, A. Whitty, S. Vajda, J.
Med. Chem. 2015, 58, 9063-9088.
[22] C. Mattos, D. Ringe, Nature Biotechnol. 1996, 14, 595–599.
[23] J. J. Irwin, T. Sterling, M. M. Mysinger, E. S. Bolstad, R. G. Coleman, J.
Chem. Inf. Model. 2012, 52, 1757-1768.
[24] P. J. Skinner, M. C. Cherrier, P. J. Webb, Y.-J. Shin, T. Gharbaoui, A.
Lindstrom, V. Hong, S. Y. Tamura, H. T. Dang, C. C. Pride, R. Chen, J.
G. Richman, D. T. Connolly, G. Semple, Bioorg. Med. Chem. Lett.
2007, 17, 5620-5623.
[25] G. E. Jagdmann, H. R. Munson, T. W. Gero, Synth. Commun. 1990 20,
1203-1208.
[26] H. Ihara, S. Hanashima, T. Okada, R. Ito, Y. Yamaguchi, N. Taniguchi,
Y. Ikeda, Glycobiology 2010, 20, 1021-1033.
[27] S. Genheden, U. Ryde, Expert Opin. Drug Discov. 2015, 10, 449-461.
[28] G. Chessari, I. M. Buck, J. E. H. Day, P. J. Day, A. Iqbal, C. N. Johnson,
E. J. Lewis, V. Martins, D. Miller, M. Reader, D. C. Rees, S. J. Rich, E.
Tamanini, M. Vitorino, G. A. Ward, P. A. Williams, G. Williams, N. E.
Wilsher, A. J.-A. Woolford, J. Med. Chem. 2015, 58, 6574-6588.
[29] J. L. Banks, H. S. Beard, Y. Cao, A. E. Cho, W, Damm, R. Farid, A. K.
Felts, T. A. Halgren, D. T. Mainz, J. R. Maple, R. Murphy, D. M. Philipp,
M. P. Repasky, L. Y. Zhang, B. J. Berne, R. A. Friesner, E. Gallicchio, R.
M. Levy, J. Comp. Chem. 2005, 26, 1752-1780.
For internal use, please do not delete. Submitted_Manuscript
This article is protected by copyright. All rights reserved.

10.1002/cbic.201900289
ChemBioChem
FULL PAPER
Entry for the Table of Contents
FULL PAPER
Fucosyltransferase 8 catalyzes the
core fucosylation of N-glycans. Here,
we perform an in silico analysis of a
new putative binding site of
Claas Strecker, Melissa Baerenfaenger,
Michaela Miehe, Edzard Spiller, and
Bernd Meyer*
Page No. – Page No.
fucosyltransferase 8. Secondly, we
describe the design and synthesis of a
donor-site binder to start the
development of inhibitors of
fucosyltransferase 8 with drug-like
properties.
In silico evaluation of the binding site
of fucosyltransferase 8 and first
attempts of synthesizing an inhibitor
with drug-like properties
For internal use, please do not delete. Submitted_Manuscript
This article is protected by copyright. All rights reserved.