Covalent inhibitors of LgtC: A blueprint for the discovery of non-substrate-like inhibitors for bacterial glycosyltransferases

Article abstract of DOI:10.1016/j.bmc.2017.04.006

Non-substrate-like inhibitors of glycosyltransferases are sought after as chemical tools and potential lead compounds for medicinal chemistry, chemical biology and drug discovery. Here, we describe the discovery of a novel small molecular inhibitor chemot

Full text of DOI:10.1016/j.bmc.2017.04.006

Accepted Manuscript
Covalent inhibitors of LgtC: a blueprint for the discovery of non-substrate-like
inhibitors for bacterial glycosyltransferases
Yong Xu, Ruth Smith, Mirella Vivoli, Masaki Ema, Niina Goos, Sebastian
Gehrke, Nicholas J. Harmer, Gerd K. Wagner
PII:
S0968-0896(17)30179-7
DOI:
http://dx.doi.org/10.1016/j.bmc.2017.04.006
BMC 13675
Reference:
To appear in:
Bioorganic & Medicinal Chemistry
Received Date:
Revised Date:
Accepted Date:
2 February 2017
1 April 2017
4 April 2017
Please cite this article as: Xu, Y., Smith, R., Vivoli, M., Ema, M., Goos, N., Gehrke, S., Harmer, N.J., Wagner, G.K.,
Covalent inhibitors of LgtC: a blueprint for the discovery of non-substrate-like inhibitors for bacterial
glycosyltransferases, Bioorganic & Medicinal Chemistry (2017), doi: http://dx.doi.org/10.1016/j.bmc.2017.04.006
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers
we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and
review of the resulting proof before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Covalent inhibitors of LgtC: a blueprint for the discovery of non-substrate-like
♯
inhibitors for bacterial glycosyltransferases
Yong Xu [a], Ruth Smith [b], Mirella Vivoli [c], Masaki Ema [a], Niina Goos [b], Sebastian
Gehrke [b,d], Nicholas J. Harmer [c] & Gerd K. Wagner [a]*
[
a] King’s College London, Department of Chemistry, Faculty of Natural &
Mathematical Sciences, Britannia House, 7 Trinity Street, London, SE1 1DB, UK.
Phone: +44 (0)20 7848 1926; e-mail: gerd.wagner@kcl.ac.uk
[
[
[
b] King’s College London, Institute of Pharmaceutical Science, 150 Stamford Street,
London, SE1 9NH, UK
c] University of Exeter, Henry Wellcome Building for Biocatalysis, Stocker Road,
Exeter, EX4 4QD, UK
d] University of East Anglia, School of Pharmacy, Earlham Road, Norwich, NR4 7TJ,
UK
*
Corresponding author
1
April 2017
Keywords: covalent inhibitor, enzyme, glycosyltransferase, chemical tool, virulence
factor
♯
th
GW dedicates this work to his mother, on the occasion of her 65 birthday.
-
1 -

Abstract
Non-substrate-like inhibitors of glycosyltransferases are sought after as chemical tools
and potential lead compounds for medicinal chemistry, chemical biology and drug
discovery. Here, we describe the discovery of a novel small molecular inhibitor
chemotype for LgtC, a retaining α-1,4-galactosyltransferase involved in bacterial
lipooligosaccharide biosynthesis. The new inhibitors, which are structurally unrelated to
both the donor and acceptor of LgtC, have low micromolar inhibitory activity, comparable
to the best substrate-based inhibitors. We provide experimental evidence that these
inhibitors react covalently with LgtC. Results from detailed enzymological experiments
with wild-type and mutant LgtC suggest the non-catalytic active site residue Cys246 as a
likely target residue for these inhibitors. Analysis of available sequence and structural
data reveals that non-catalytic cysteines are a common motif in the active site of many
bacterial glycosyltransferases. Our results can therefore serve as a blueprint for the
rational design of non-substrate-like, covalent inhibitors against a broad range of other
bacterial glycosyltransferases.
-
2 -

1
. Introduction
Glycosyltransferases (GTs) are key enzymes for the biosynthesis of complex glycans
and glycoconjugates in all domains of life [1]. In bacteria, GT activity is required for
virulence and viability, and individual GTs are emerging as novel targets for anti-
microbial and anti-virulence drug discovery [2,3]. Small molecular inhibitors of GTs are
therefore sought after as chemical tools for the interrogation, manipulation and
disruption of cellular glycosylation pathways [4], and as potential lead compounds for
drug discovery. Most existing GT inhibitors are substrate analogues that are structurally
derived from the respective GT donor and/or acceptor [5]. Such substrate-based GT
inhibitors are usually not drug-like and often suffer from modest bioactivity and/or
intrinsic physicochemical liabilities, such as limited stability and poor cell penetration.
While some of these drawbacks can be circumnavigated by using metabolic precursors
[
6-8], the identification of alternative, non-substrate-like GT inhibitors remains an
important goal in medicinal chemistry and chemical biology [9]. Despite recent progress
10,11], very few such non-substrate-like GT inhibitors are currently available, in stark
[
contrast to other enzyme classes of similar size and biological importance (e.g. kinases,
proteases).
Covalent enzyme inhibitors are currently undergoing a renaissance in chemical
biology [12] and drug discovery [13]. Covalent inhibitors generally display a range of
attractive features, including high potency, prolonged duration of action, and amenability
to rational design. A systematic study has recently shown that even reactive
electrophiles, such as the Michael acceptor acrylamide, do not react indiscriminately
with any biological nucleophile [14]. This differential reactivity can be harnessed for the
development of covalent inhibitors, especially those that target non-catalytic residues
[
15]. This approach has been used successfully for inhibitor development against
several challenging enzyme targets, including drug-resistant kinases [16] and proteases
-
3 -

[
17]. The covalent targeting of non-catalytic residues, in particular cysteines, therefore
represents a promising strategy for inhibitor development against intractable targets
such as GTs. To date, no inhibitors with this mode of action have been reported for any
bacterial GT. This is particularly striking, as non-catalytic cysteines are a common motif
in the active site of bacterial GTs. Even beyond bacterial enzymes, there is only a single
mechanistically related example in the entire GT family [11]. Inhibitors of the mammalian
O-linked N-acetylglucosamine transferase (OGT) act via an unusual double-
displacement mechanism, which draws on the unique architecture of this enzyme – not
found in other GTs – and requires a very specific arrangement of two separate target
residues [11].
Herein, we describe a new class of non-substrate-like, covalent inhibitors of the
retaining -1,4-galactosyltransferase LgtC from Neisseria meningitidis. LgtC catalyzes
the transfer of D-galactose (D-Gal) from a UDP-Gal donor to a lactose acceptor (Fig. 1A)
[
18]. In pathogenic Gram-negative bacteria such as Neisseria and Haemophilus, LgtC is
required for the biosynthesis of the terminal digalactoside epitope -D-Galp-(1,4)--D-
Galp in the lipooligosaccharide (LOS) structures of the outer core of the cell envelope
[
19]. LgtC expression has been associated with high-level serum resistance in non-
typeable Haemophilus influenzae (NTHi) [19], and inhibition of LgtC and related GTs
that are involved in LOS biosynthesis has been suggested as a strategy for anti-
microbial drug discovery [18]. LgtC is an excellent starting point for such an approach
not only because of its role in bacterial virulence, but also because of its structural and
mechanistic communalities with other bacterial GTs [18], which have made it a widely
used model system for mechanistic and structural studies in this enzyme family.
We have identified small molecular inhibitors of LgtC that are structurally
unrelated to both the LgtC donor and acceptor (Fig. 1B). We demonstrate that these
inhibitors act via a covalent mechanism of inhibition, targeting a non-catalytic cysteine
-
4 -

residue in the LgtC active site. Analysis of sequence and structural data shows that
active-site cysteines are a common feature in bacterial GTs. Our results may therefore
serve as a blueprint for the rational development of non-substrate-like, covalent
inhibitors of other bacterial enzymes in this family.
a
Figure 1
(A) The LgtC reaction, and its inhibition by pyrazol-3-one 1. (B) Chemical
structures of pyrazol-3-ones 1-3.
a
Assay conditions: LgtC (activated with 5 mM DTT), lactose (2 mM), UDP-Gal (28 µM),
MnCl (5 mM), CIP (10 U/ml), CEL (1 mg/ml), Triton (0.01%) and inhibitor (0.1-100 µM)
2
were incubated in HEPES buffer (pH 7.0) for 20 min at 30 °C. The progress of the
reaction was determined with malachite green as previously described [20].
2
. Results and discussion
2
.1 Rationale and chemical synthesis of inhibitors. We have recently adapted a
biochemical GT assay for inhibitor studies with non-substrate-like chemotypes [20]. We
have subsequently used this assay to evaluate a collection of 130 small molecules as
potential non-substrate-like inhibitors of LgtC. The collection was designed around
-
5 -

structural scaffolds that are known from other GT inhibitors, including nucleosides,
steroids, and pyrazol-3-ones (Fig. 1B). Pyrazol-3-ones have previously been reported as
screening hits against other GTs [21,22], but their molecular mode of action has not
been investigated. From our own screen against LgtC, pyrazol-3-one 1 emerged as the
most promising hit from this series and is the primary target molecule of the present
study. 1 was synthesized in two steps from phenylhydrazine via a ring-closing
condensation with ethyl acetoacetate, followed by a Knoevenagel condensation with 2-
trifluoromethyl benzaldehyde (Scheme 1). This synthesis was readily adapted to also
provide access to the congeners 2 and 3 (Scheme 1).
a
Scheme 1 Synthesis of target molecules 1-3.
R1
R1
H
N
N
N
R2
R4
NH2
(i)
(ii) or (iii)
N
N
R3
O
O
6
7
(R = CH3)
(R = CF3)
1
2
3
4
1 (R = CH , R = CF , R = R = H)
3
3
1
2
4
3
2 (R = CF , R = R = H, R = OBn)
3
1
2
3
4
3
(R = CF , R = H, R = Br, R = OMe)
3
a
Reagents & conditions: (i) ethyl 3-oxobutanoate, acetic acid, 110 ºC, 97% (6) or ethyl
,4,4-trifluoroacetoacetate, acetic acid, 110 ºC, 72% (7); (ii) 2-(trifluoromethyl)
benzaldehyde, acetic acid, reflux, 48% (for 1); (iii) 3-(benzyloxy)benzaldehyde (for 2) or
4
o
3
-bromo-4-methoxybenzaldehyde (for 3), 160 C/microwave, 15 min, 53% (2) or 75% (3)
2
.2 Biochemical evaluation of inhibitors. In initial experiments, pyrazol-3-one 1 inhibited
LgtC with an IC value of 27 μM (Fig. 1A). All inhibition assays were carried out in the
50
presence of surfactant to suppress non-specific aggregation and avoid assay artefacts
23]. This assay protocol provided confidence that the inhibitory activity of 1 was
genuine. Equally importantly, 1 showed no inhibitory activity in control experiments
[
-
6 -

against the calf intestinal phosphatase (CIP) used as a secondary enzyme in the
biochemical assay (SI).
We hypothesised that 1 may act as a covalent inhibitor towards LgtC, due to the
presence of the Michael acceptor system, which is a known cysteine-reactive
electrophile in covalent inhibitors and probes [24]. As no reports on pyrazol-3-ones as
covalent enzyme inhibitors have previously been published, we set out to experimentally
test this hypothesis. First, we compared enzyme activity in the assay mixture of LgtC
and 1 before and after diafiltration. The inhibitory activity of 1 was preserved after
diafiltration, which strongly suggested a covalent mode of inhibition (Fig. 2).
a
Figure 2
LgtC activity in the presence of inhibitor 1, before and after diafiltration.
a
Conditions: LgtC (activated with 5 mM DTT) was pre-incubated with inhibitor (100 M)
or DMSO, in the presence of UDP-Gal (28 µM), for 10 min at 30 °C. Lactose (2 mM) was
added, and the reactions were incubated for 20 min at 30 °C. Samples were drawn, and
enzyme activity before diafiltration was determined under standard assay conditions.
Enzyme reactions (3 mL) were subjected to diafiltration in Vivaspin concentrator tubes
(centrifugation at 4 °C, 4000 rpm). The initial residual volume (300 µL) was washed once
with HEPES buffer (pH 7.0, total volume 3 mL). The residual volume (300 µL) from the
wash step was diluted to 450 µL with HEPES buffer (pH 7.0), and enzyme activity after
diafiltration was determined under standard assay conditions (2 mM lactose, 28 µM
UDP-Gal, 20 min incubation at 30 °C). Each experiment was carried out in triplicate
(100% activity = complete conversion of UDP-Gal). Error bars represent standard
deviation.
-
7 -

Next, we directly analysed the reaction of LgtC with 1 by mass spectrometry.
These experiments showed the formation of a covalent adduct, providing direct evidence
for the covalent reaction of 1 and LgtC (Fig. 3A). Importantly, only a single adduct was
observed in these experiments, at a 1:1 ratio of inhibitor and enzyme. This suggests that
1
reacts only with a single residue in LgtC.
Figure 3
Mass spectrometry reveals the formation of covalent adducts of LgtC with
a
pyrazol-3-ones 1 (panel A), 2 (panel B) and 3 (panel C).
A
B
-
8 -

C
a
Conditions: LgtC in HEPES buffer (pH 7.0) was incubated with inhibitor (1 and 2: 100
µM in DMSO; 3: 50 µM in DMSO) for 20 min at 30 °C (10% final DMSO concentration).
Samples were directly analysed by electrospray mass spectrometry.
1
Figure 4
cysteine.
H-NMR spectra of the time-dependent, covalent reaction of 1 with
a
H3N
CO2
CH3
S
3
CH (diastereomers)
N
N
CF3
H
CH
cysteine
OH
CH3
O
H
N
N
CF3
CH
3
CH (E-/Z-isomers)
a
Conditions: Solutions of 1 (19.8 mg, 0.06 mmol, in d6-DMSO) and L-cysteine (7.26 mg,
.06 mmol, in d6-DMSO) were mixed in a glass NMR tube. H-NMR spectra were
1
0
recorded at various time points from 0 min to 190 min ( 3.3 ppm: H O).
2
-
9 -

In order to identify potential target residues for the covalent reaction, we
incubated 1 separately with three amino acids containing different side chain
nucleophiles (L-cysteine, L-serine and L-lysine), and followed the individual reactions by
1
H-NMR. The time-dependent formation of a covalent adduct at the exocyclic double
bond was observed upon incubation of 1 with L-cysteine (Fig. 4), but not with L-serine or
L-lysine (SI). These results show that under otherwise identical conditions, the Michael
acceptor system in 1 reacts preferentially with a sulfur nucleophile.
This preference suggested that the target residue in LgtC may be a cysteine,
although the reactivity of a specific residue in the environment of the protein will
necessarily depend on neighboring groups. LgtC contains five cysteines, of which three
are solvent-accessible [18]. One of these is Cys246, which sits at the interface of the
LgtC acceptor and donor binding sites, in close proximity to the lactose acceptor (SI, Fig.
S1). Molecular simulations using a covalent docking protocol [25] suggested that 1 is
binding at the substrate binding site of LgtC in an ideal orientation for the covalent
reaction of the Michael acceptor system with Cys246 (SI, Fig. S1). In order to
experimentally test the hypothesis that 1 may indeed bind at the substrate binding site,
we investigated the effect the concentration of the LgtC acceptor, lactose, may have on
inhibition. Under standard assay conditions (no pre-incubation, turnover 20-50%),
inhibitory activity decreased markedly at higher acceptor concentrations (Fig. 5A).
Qualitatively, this is consistent with an acceptor-competitive binding mode for 1. In order
to obtain quantitative data and enable detailed enzyme kinetics analysis, we repeated
this experiment at turnover rates of <10%, at a fixed concentration of UDP-Gal donor
(
100 M), and variable concentrations of lactose acceptor (5-25 mM) and inhibitor (1-5
M). Dixon and Cornish-Bowden analyses of the resulting data show that under these
conditions, 1 acts as a competitive inhibitor of LgtC relative to the lactose acceptor (Fig.

-
10 -

5
B). These results show that 1 does indeed bind at the substrate binding site, as
required for a potential covalent reaction with Cys246.
Figure 5
Effect of different acceptor concentrations on the inhibitory activity of
pyrazol-3-one 1 towards LgtC under standard assay conditions (A) and enzyme kinetics
conditions (B).
a
A
Standard assay conditions (donor turnover 20-50%)
a
Conditions: LgtC was incubated with 1 (0.1-100 µM), UDP-Gal (28 µM), MnCl (5 mM),
2
CIP (10 U/ml), CEL (1 mg/ml), Triton (0.01%) and lactose acceptor (2-50 mM) for 20 min
at 30 °C in 13 mM HEPES buffer (pH 7.0). Each experiment was carried out in triplicate.
a
B
Enzyme kinetics conditions (donor turnover <10%)
Dixon plot
Cornish-Bowden plot
5
4
3
2
1
0
100
80
60
40
20
0
5
mM
5mM
Lactose
15mM
Lactose
25mM
Lactose
Lactose
15mM
Lactose
25mM
Lactose
-14 -12 -10 -8
-6
-4
-2
0
2
4
6
-10
-8
-6
-4
-2
0
2
4
6
[
1
6j] uM
(M)
[6j] uM
1 (M)
-
11 -

a
Conditions: LgtC was incubated with 1 (1-5 µM), UDP-Gal (100 µM), MnCl (5 mM), CIP
2
(
3
10 U/ml), CEL (1 mg/ml), Triton (0.01%) and lactose acceptor (5-25 mM) for 20 min at
0 °C in 13 mM HEPES buffer (pH 7.0). Data points were fitted to a Dixon or Cornish-
Bowden plot with GraFit 7.
In order to investigate the role of Cys246 for the inhibitory activity of 1, we created
two LgtC mutants, in which Cys246 is replaced with a non-nucleophilic (C246A) or only
weakly nucleophilic (C246S) residue. Both mutants were catalytically active under our
standard conditions, albeit less so than wild-type LgtC. At the highest enzyme
concentration tested, the observed turnover of donor was approximately 36% (C246A)
or 22% (C246S) of that of the wild-type enzyme (SI). This level of enzymatic activity was
sufficient for inhibition experiments. Under the standard assay conditions, pyrazol-3-one
1
at the highest concentration (100 μM) showed only weak inhibition against C246S
LgtC, and almost none against C246A LgtC (Fig. 6). These results provide direct
evidence that Cys246 is a critical residue for the inhibition of LgtC by pyrazol-3-one 1,
possibly as the target for a covalent interaction with the inhibitor. They also suggest that
1
is only weakly effective as a non-covalent inhibitor of LgtC, at least at the
concentrations tested.
Figure 6
C246S.
Compound 1 shows only weak inhibition of the LgtC mutants C246A and
a
-
12 -

a
Conditions: LgtC was incubated with 1 (0.1-100 µM), UDP-Gal (28 µM), MnCl (5 mM),
2
CIP (10 U/ml), CEL (1 mg/ml), Triton (0.01%) and lactose acceptor (2 mM) for 20 min at
0 °C in 13 mM HEPES buffer (pH 7.0). Control experiments without inhibitor were
3
carried out in parallel, showing a donor turnover of, respectively, 23% (wild-type), 21%
or 39% (C246A), and 26% (C246S). Each experiment was carried out in triplicate.
To further dissect the structural requirements for the interaction between 1 and
LgtC, we next carried out inhibition experiments with structural analogues of 1 (SI). As
expected, compound 4, a hydrogenated variant of 1 lacking the exocyclic double bond
and hence the capacity for covalent interaction, showed no inhibitory activity against
LgtC (SI, Table S1). Compound 5, which contains a different heterocyclic scaffold, but
the same Michael acceptor system as pyrazol-3-one 1, was also inactive (SI, Table S1).
The lack of activity of compounds 4 and 5 shows that the Michael acceptor system in 1
is necessary, but not sufficient, for the inhibition of LgtC. It suggests that non-covalent
interactions also contribute to the LgtC inhibitory activity of 1, probably by orienting the
inhibitor correctly in the substrate binding site prior to a covalent interaction with Cys246.
The substrate binding sites of GTs with the same donor and/or acceptor are
highly conserved. We therefore decided to explore the behaviour of 1 against three other
GTs (Table 1), which use either the same UDP-Gal donor (bovine -1,4-GalT) or the
same lactose acceptor (Pasteurella multocida α-2,3-SiaT, human FUT-2) as LgtC, but
lack an active-site cysteine. Pyrazol-3-one 1 was inactive against all three of these GTs
(Table 1 & SI). These results demonstrate that 1 does not act simply as a non-covalent
GT substrate mimic. Taken together, our results suggest strongly that it is the capacity to
bind non-covalently at the substrate binding site of LgtC, in combination with a covalent
interaction, most likely with Cys246, which forms the basis for the inhibitory activity of 1
towards LgtC.
-
13 -

1
Table 1
Substrate specificities and inhibition of four glycosyltransferases
Donor
substrate
Acceptor
substrate
Inhibition
(cmpd 1)
GT
Organism
2
LgtC
Neisseria meningitidis
Bos taurus
UDP-Gal
UDP-Gal
β-lactose
27 ± 5 μM
3
4
4
β-1,4-GalT
α-2,3-SiaT
FUT-2
GlcNAc
no inhibition
no inhibition
Pasteurella multocida CMP-Neu5Ac β-lactose
H. sapiens
GDP-Fuc α-lactose
no inhibition
1
2
3
For experimental details of the inhibition experiments see SI; IC ; at highest
50
4
concentration of inhibitor (100 μM); at highest concentration of inhibitor (150 μM)
Having established that prototype pyrazol-3-one 1 does indeed act via a covalent
mode of action, we set out to improve the potency of this new LgtC inhibitor class. To
increase the reactivity of the Michael acceptor system, we replaced the 5-CH group in 1
3
with a strongly electron-withdrawing 5-CF group (Fig. 1B). The exact 5-CF congener of
3
3
1
, bearing a (2-trifluoromethyl)benzylidene substituent in position 4, was not synthetically
accessible. We therefore introduced alternative benzylidene substituents in position 4
Fig. 1B). While the new inhibitors 2 and 3 showed no, or only modest, activity under our
(
standard assay conditions, their inhibitory activity increased markedly upon pre-
incubation of LgtC with inhibitor, prior to starting the enzyme reaction (Fig. 7 & 8). Such
a time-dependent inhibition is characteristic for a covalent mode of action. Interestingly,
the pronounced effect of pre-incubation on inhibitory activity was largely independent of
the presence of donor or acceptor substrate (Fig. 7, methods B-D).
-
14 -

Figure 7
Pre-incubation with LgtC significantly increases the inhibitory activity of 2
a
and 3, both in the presence and absence of LgtC substrates.
a
Conditions: Following pre-incubation (Methods A-D, see Table 2), LgtC was incubated
with inhibitor (0.1-100 µM), UDP-Gal (28 µM), MnCl (5 mM), CIP (10 U/ml), CEL (1
2
mg/ml), Triton (0.01%) and lactose acceptor (2 mM) for 20 min at 30 °C in 13 mM
HEPES buffer (pH 7.0). Each experiment was carried out in triplicate.
Table 2
Pre-incubation protocols and enzyme turnovers.
1
Turnover (%)
Pre-incubation protocol
No pre-incubation
2
3
Method A
Method B
29
29
30
30
Pre-incubation (10 min) of LgtC, inhibitor and donor,
prior to addition of acceptor
Method C
Pre-incubation (10 min) of LgtC and inhibitor,
prior to addition of donor and acceptor
25
23
25
23
Method D
Pre-incubation (10 min) of LgtC, inhibitor and acceptor,
prior to addition of donor
1
Turnover is defined as the percentage consumption of donor UDP-Gal in the
biochemical assay, relative to the total amount of UDP-Gal used.
For the two inhibitors 2 and 3, inhibitory activity increased consistently with
increasing pre-incubation time, albeit at different rates, before reaching a plateau after
6
0 mins (2) and 20 mins (3), respectively (Fig. 8 & Table 3). The covalent mode of action
for 2 and 3 was further confirmed with several key experiments, as for the parent
compound 1. Thus, for both 2 and 3, covalent adducts with LgtC were detected by mass
spectrometry (Fig. 3B & C), and the time-dependent formation of a covalent adduct with
-
15 -

1
cysteine was observed by H NMR (SI). In control experiments, neither 2 nor 3 showed
activity against CIP or -1,4-GalT (SI).
Figure 8
LgtC inhibition by pyrazolones 2 and 3 at different pre-incubation times.
For conditions and IC values see Table 3.
50
Table 3
LgtC inhibition by pyrazolones 2 and 3 at different pre-incubation times.
a
IC values extracted from inhibition curves in Figure 8.
50
Pre-incubation time (min)
20 30
0
5
10
60
180
2
>100
>50
>50
38 ± 13.0 17.6 ± 1.1
9.0 ± 0.2
6.0 ± 0.2
IC50 (µM) ± SD
LgtC
turnover (%)
2
9
30
24 ± 7.1
32
41
15.8 ± 0.8
46
27
6.5 ± 1.6
32
24
3.1 ± 0.5
32
29
3.1 ± 0.3
39
19
5.5 ± 1.2
26
3
>50
IC50 (µM) ± SD
LgtC
turnover (%)
30
a
Conditions: LgtC was pre-incubated with inhibitor (2: 0.1-100 µM; 3: 0.1-50 µM) or
DMSO for the given time (0-180 min) at 30 °C. Lactose (2 mM) and UDP-Gal (28 µM)
were added, and the reactions were incubated for 20 min at 30 °C. Enzyme activity was
determined under standard assay conditions. Each experiment was carried out in
triplicate.
The activity of a covalent inhibitor is governed by two parameters: the strength of
the initial, non-covalent binding interaction with the target enzyme, and the reactivity of
the electrophilic warhead (Fig. 9). In order to dissect the contribution of these two steps
towards the overall activity of 2 and 3, we determined kobs at different inhibitor
concentrations. By plotting kobs over inhibitor concentration, the kinetic parameters for
-
16 -

the first (K) and second (kinact) interaction step can be extracted [13]. This analysis
i
shows that kinact for both 2 and 3 is of a similar order of magnitude, whereas K is about
i
1
0-times greater for compound 2 than for 3. These results suggest that due to the
different substitution pattern of the 4-substituent, 2 and 3 differ significantly in their non-
covalent binding affinity for LgtC, but not so much in the reactivity of the Michael
acceptor system. This profile corresponds to the results from experiments with different
pre-incubation times, which show that 2 requires a longer pre-incubation period for
maximal activity than 3 (Table 3). The different non-covalent affinities of 2 and 3 are also
consistent with the slightly different binding modes that were obtained for both inhibitors
from covalent docking simulations. These docking results suggest that both 2 and 3 can
bind at the substrate binding site of LgtC in an orientation that allows a covalent reaction
with residue Cys246, but that the smaller 3-bromo-4-methoxyphenyl substituent in 3 can
be more readily accommodated than the bulky 3-benzyloxyphenyl substituent in 2 (SI,
Fig. S1).
Figure 9
Covalent inhibition kinetics for 2 and 3.
(a) Two-step model of covalent inhibition.
-
17 -

a
(b)
Determination of kinetic parameters k , K and kinact.
obs i
A Semi-logarithmic plots for the time-dependent inactivation of LgtC at different inhibitor concentrations
B Observed rate constants kobs at different inhibitor concentrations, extracted from the plots in A
a
Conditions: A LgtC activity was determined at various concentrations of 2 (0.5-200 µM)
or 3 (0.5-50 µM) and in function of different pre-incubation times (0, 5, 10, 20 and 30
min). Enzyme activity is expressed as percentage of control (DMSO only) and plotted on
a semi-logarithmic scale over pre-incubation time. Values for kobs were extracted from
-kobs x t
exponential regression using the equation enzyme activity [%] = Ae
. B Observed
rate constants kobs were extracted from the plots in A, re-plotted over inhibitor
concentrations, and fitted to the hyperbolic equation kobs = kinact × [I]/(K+[I]). From these
i
plots, K and k
were obtained as previously described [13]. All experiments were
i
inact
performed in triplicate.
3
. Conclusions
Existing inhibitors of the bacterial α-1,4-galactosyltransferase LgtC are substrate
analogues based on a sugar-nucleotide scaffold [18,26]. Our results establish suitably
substituted pyrazol-3-ones as the first non-substrate-like, covalent inhibitor chemotype
-
18 -

for this enzyme. The activity of inhibitors 2 and 3 upon preincubation is comparable to
the most potent substrate-based LgtC inhibitors reported to date [26]. Our data suggest
that pyrazol-3-ones 1-3 bind at the substrate binding site of LgtC prior to the covalent
reaction, which most likely occurs at the non-catalytic residue Cys246 in the enzyme
active site. Importantly, the new inhibitors are inactive against three related GTs that use
either the same donor or acceptor substrate as LgtC, but lack a non-catalytic cysteine in
the active site. This behaviour is notable, as other pyrazol-3-ones have previously been
classified as non-selective, pan-assay interfering compounds (PAINS) [27]. It suggests
that, while it is likely that the unoptimised inhibitors 1-3 may also react with other
proteins that contain a carbohydrate- or nucleotide-binding site with a non-catalytic
cysteine, target selectivity could be engineered into these inhibitors through further
structural modification.
Non-catalytic cysteine residues are a common feature of bacterial GTs. In
principle, the direct targeting of non-catalytic cysteines can therefore serve as a general
inhibitor discovery strategy against many other bacterial GTs. In the sequence-based
CAZy database of carbohydrate-active enzymes [28], LgtC is a member of family GT-8.
Neisseria meningitidis LgtC is the only bacterial member of this CAZy family for which
structural information is currently available. Sequence alignment with other members of
CAZy family GT-8 shows that Cys246 in LgtC is part of a highly conserved 6-aa motif,
which is present in several other GTs from different bacteria (SI, Fig. S2). Interestingly,
these enzymes are not limited to galactosyltransferases, but also include GTs with other
substrate specificities as well as GTs with unknown function.
Beyond family GT-8, structural information is currently available for a total of 48
GTs from bacterial organisms that have a human host. 23 of these structures, including
Neisseria meningitidis LgtC, have been solved at a resolution of <3Å and contain a
ligand in the active site (SI, Table S3). Analysis of these 23 structures reveals that all but
-
19 -

three, across 14 different CAZy families, have a non-catalytic cysteine at or near the
active site. In 11 of these structures, this cysteine is within 8Å of the active site ligand
(see Fig. 10 for selected examples, and SI, Table S3). This analysis demonstrates that
the presence of non-catalytic, active site cysteines is widespread in bacterial GTs from
both Gram-negative and Gram-positive organisms, independent of enzyme substrate
specificity, CAZy family, overall fold type and biological function (Fig. 10).
a
Figure 10
A
Examples for bacterial GTs with non-catalytic active-site cysteines.
B
C
D
F
E
a
Proteins are shown in cartoon representation (wheat), active site residues are shown as
lines (wheat), and the active-site cysteine as sticks (yellow). Substrates or substrate
analogues are shown as sticks (green). Broken red lines indicate the distance from the
α-carbon of the non-catalytic cysteine to the nearest atom of the respective ligand. (A) E.
coli chondroitin polymerase (PDB 2Z87), donor: UDP-GalNAc, CAZy family: GT-2,
general fold type: A; (B) E. coli glycogen synthase (PDB 2QZS) [40], donor: ADP-Glc,
CAZy family: GT-5, general fold type: B; (C) Bacteroides ovatus BoGT6a (PDB 4CJC)
[
41], donor: UDP-GalNAc, CAZy family: GT-6, general fold type: A; (D) Bradyrhizobium
sp. WM9 fucosyltransferase NodZ (PDB 3SIX) [42], donor: GDP-Fuc, CAZy family: GT-
3, general fold type: B; (E) Campylobacter jejuni sialyltransferase-II (PDB 2X61) [29],
2
donor: CMP-Neu5Ac, CAZy family: GT-42, general fold type: A variant; (F) Rhodobacter
sphaeroides cellulose synthase (PDB 4HG6) [43], donor: UDP-Glc, CAZy family: GT-2,
general fold type: A.
-
20 -

A
representative example is the bifunctional CMP-Neu5Ac:α-2,3/-2,8-
sialyltransferase Cst-II from Campylobacter jejuni [29]. In Cst-II, the non-catalytic
cysteine residue Cys30 is located to the rear of active site, at a distance of 6.0Å from the
CMP ligand (SI, Fig. S3). Cst-II is a member of CAZy family GT-42. Sequence
alignments within this family suggest that Cys30 in Cst-II is part of a conserved 10-aa
sequence, which is also present in several other bacterial sialyltransferases in this family
(SI, Fig. S4).
This analysis, in conjunction with our experimental results, suggests that the
presence of non-catalytic active site cysteines can be systematically exploited for the
discovery of new inhibitor classes against a broad range of bacterial GTs, including
enzymes with very different, or indeed unknown, substrate specificities. In the latter
case, such inhibitors would be of considerable use to annotate the biochemical and
biological function of these GTs. Our results provide a blueprint for such an inhibitor
discovery strategy. Such a strategy will be facilitated further by the recent publication of
freely available, web-based protocols for covalent docking [25] and virtual screening
[
30].
Many of the bacterial GTs that are amenable to this general inhibitor discovery
strategy are involved in bacterial virulence processes and complex pathogen/host
interactions [31]. Covalent inhibitors would therefore be eminently useful as tool
compounds for anti-virulence drug discovery [32], not only for inhibition studies, but also
as affinity probes for the labelling and imaging of bacterial GTs in vivo. In preliminary
experiments, we have already successfully used a fluorescent derivative of 1 for the
labelling of LgtC (unpublished results). Results from this ongoing work will be reported in
due course.
-
21 -

4
. Experimental section
4
.1 Chemistry
All chemical reagents were obtained commercially and used as received. Target
compounds and synthetic intermediates were purified by flash chromatography column
1
13
and characterized by TLC, H-NMR, C-NMR, and ESI-MS. Flash chromatography
columns were packed wet. Thin layer chromatography (TLC) was performed on
precoated aluminium plates (Silica Gel 60 F254, Merck). Compounds were visualized by
exposure to UV light (254/365 nm). NMR spectra were recorded on a Bruker BioSpin at
1
13
4
00 MHz ( H) or 100 MHz ( C). Mass spectra were recorded at the EPSRC National
Mass Spectrometry Service Centre, Swansea. Analytical HPLC was carried out on
Agilent 1260 Infinity machine equipped with an Agilent Poroshell 120 EC-C18 column
(2.7 µm, 4.6 x 50 mm) under the following conditions: 0.1% formic acid in
water/methanol, flow rate: 0.5 mL/min, detection wavelength: 254 nm. The purity of all
target compounds was >95%.
4
.1.1 5-Methyl-2-phenyl-2,4-dihydro-3H-pyrazol-3-one (6). Phenylhydrazine (219 mg,
2
.03 mmol) and ethyl 3-oxobutanoate (264 mg, 2.03 mmol) were dissolved in glacial
acetic acid. The reaction mixture was stirred at 110 ºC until TLC (hexane/EtOAc 1:1)
showed complete consumption of the starting material. Upon cooling a white solid
precipitated from the solution and was filtered and washed with ice-cold ethanol.
Purification by flash column chromatography afforded the title compound [33] as a light
1
yellow solid (343 mg, 1.97 mmol, 97 %). H-NMR (400 MHz, CDCl ) δ = 2.26 (s, 3H),
3
13
3
.49 (s, 2H), 7.27-7.32 (m, 1H), 7.48-7.54 (m, 2H), 7.96-8.01 (m, 2H) ppm; C-NMR
(
100 MHz, CDCl ) δ = 16.9, 43.0, 118.8, 125.0, 128.8, 138.0, 156.4, 170.6 ppm.
3
-
22 -

4
.1.2 1-Phenyl-3-(trifluoromethyl)-1H-pyrazol-5-ol (7). The title compound [34] was
obtained as a white solid (652 mg, 2.86 mmol, 72 %) from phenylhydrazine (433 mg, 4
mmol) and ethyl 4,4,4-trifluoroacetoacetate (736 mg, 4 mmol) via the method described
1
for 6. H-NMR (400 MHz, DMSO-d6) δ = 5.94 (s, 1H), 7.39 (t, 1H, J = 8.0 Hz), 7.52 (t,
13
2
H, J = 8.0 Hz), 7.71 (d, 2H, J = 8.0 Hz), 12.49 (s, 1H) ppm; C-NMR (100 MHz,
1
2
DMSO-d6) δ = 85.6, 121.3 (q, JCF = 267 Hz), 122.3, 127.2, 129.1, 137.7, 140.4 (d, J
CF
=
37 Hz), 153.7 ppm.
4
.1.3 (Z)-5-Methyl-2-phenyl-4-(2-(trifluoromethyl)benzylidene)-2,4-dihydro-3H-pyrazol-3-
one (1). 6 (220 mg, 2.03 mmol) and 2-(trifluoromethyl)benzaldehyde (706 mg, 4.06
mmol) were dissolved in glacial acetic acid. The reaction was heated to reflux until TLC
(hexane/EA 3:1) showed complete consumption of starting material. The reaction
mixture was cooled and the product was isolated by column chromatography. The title
1
compound was obtained as an orange solid (320 mg, 0.97 mmol, 48%). H-NMR (400
MHz, CDCl ) δ = 2.36 (s, 3H), 7.18 (t, 1H, J = 8.0 Hz), 7.39 (t, 2H, J = 8.0 Hz), 7.59 (t,
3
1
H, J = 8.0 Hz), 7.66 (t, 1H, J = 8.0 Hz), 7.77-7.78 (m, 2H), 7.90 (d, 2H, J = 8.0 Hz), 8.49
13
1
(
d, 1H, J = 8.0 Hz) ppm; C-NMR (100 MHz, CDCl ) δ = 13.1, 118.9, 123.9 (q, J
=
CF
3
3
2
3
2
72 Hz), 125.1, 126.1 (q, J = 6 Hz), 128.8, 129.4 (d, J = 30 Hz), 129.7 (d, J = 2
CF CF CF
Hz), 129.8, 131.3, 131.5, 132.7, 138.0, 141.3, 150.4, 161.0 ppm. ESI-MS: m/z 331.1
+
+
+
+
(
100 %) [M+H] , 348.1 (8 %) [M+NH ] ; HR-MS: m/z 331.1058 [M+H] , [C H F N O]
4 18 14 3 2
calcd for 331.1053. HPLC: 98%.
4
.1.4
(Z)-4-(3-(Benzyloxy)benzylidene)-2-phenyl-5-(trifluoromethyl)-2,4-dihydro-3H-
pyrazol-3-one (2). 7 (57 mg, 0.25 mmol) and 3-(benzyloxy)benzaldehyde (64 mg, 0.30
o
mmol) were placed in a microwave-proof glass tube and heated for 15 min at 160 C in a
commercial microwave apparatus. The reaction was cooled to room temperature. The
-
23 -

reaction product was precipitated by addition of ethylacetate and hexane, collected by
filtration, and recrystallized from hexane and ethylacetate. The title compound was
1
obtained as orange solid (56 mg, 0.13 mmol, 53%). H-NMR (400 MHz, CDCl ) δ = 5.23
3
(s, 2H), 7.27-7.33 (m, 2H), 7.35-7.39 (m, 1H), 7.41-7.45 (m, 2H), 7.47-7.53 (m, 5H), 7.76
13
(
s, 1H), 7.84 (d, 1H, J = 8.0 Hz), 7.92-7.95 (m, 2H), 8.68 (t, 1H, J = 2.0 Hz) ppm; C-
1
NMR (100 MHz, CDCl ) δ = 70.3, 118.1, 119.8 (q, J = 270 Hz), 120.1, 121.6, 122.9,
3
CF
2
1
26.3, 127.9, 128.2, 128.5, 128.6, 129.0, 129.8, 133.8, 136.4, 137.5, 140.7 (d, JCF = 37
+
Hz), 150.6, 159.0, 161.1 ppm. ESI-MS: m/z 423.1 (91 %) [M+H] , 455.2 (100 %)
+
+
+
[
[
M+MeOH+H] , 477.1 (65 %) [M+MeOH+Na] ; HR-MS: m/z 423.1314 [M+H] ,
C H F N O ] calcd for 423.1315. HPLC: 97%.
24
18
3
2
2
4
3
.1.5
(Z)-4-(3-Bromo-4-methoxybenzylidene)-2-phenyl-5-(trifluoromethyl)-2,4-dihydro-
H-pyrazol-3-one (3). The title compound was obtained as a white solid (80 mg, 0.19
mmol, 75%) from 7 (54 mg, 0.25 mmol) and 3-bromo-4-methoxybenzaldehyde (65 mg,
1
0
3
1
.30 mmol) under the conditions described for 2. H-NMR (400 MHz, CDCl ) δ = 4.05 (s,
3
H), 7.06 (d, 1H, J = 12.0 Hz), 7.30 (t, 1H, J = 8.0 Hz), 7.48 (t, 2H, J = 8.0 Hz), 7.63 (s,
1
3
H), 7.93 (d, 2H, J = 8.0 Hz), 8.71-8.74 (m, 1H), 8.88 (d, 1H, J = 4.0 Hz); C-NMR (100
1
MHz, CDCl ) δ = 57.8, 111.6, 112.3, 119.8, 119.9 (q, J = 270 Hz), 120.1, 126.3,
3
CF
2
1
27.0, 129.0, 136.7, 137.6, 140.0, 140.6 (d, J = 37 Hz), 148.2, 160.9, 161.4 ppm.
CF
+
+
ESI-MS: m/z 425.0 (100 %) [M+H] , 479.0 (60 %) [M+MeOH+Na] ; HR-MS: m/z
+
+
4
25.0107 [M+H] , [C H BrF N O ] calcd for 425.0107. HPLC: 98%.
18 13 3 2 2
4
.2 Biochemistry. LgtC was expressed and purified as previously described [26].
Recombinant LgtC was activated with DTT (10 mM, in HEPES buffer) in a 1:1 ratio for
0 min at 30 °C prior to each experiment, unless otherwise stated. For enzyme activity
and inhibition experiments, we used a recently reported modification [20] of a
3
-
24 -

colorimetric glycosyltransferase assay originally developed by Wu and co-workers [35].
All assays were carried out in Nunc clear, flat-bottom 96-well microplates, unless
otherwise stated.
4
.3 Inhibition assays. For IC50 experiments, LgtC activity was adjusted to 20-50%
turnover of UDP-Gal donor. We have previously shown that within this turnover range
IC values are obtained reproducibly [20]. Assay components were added in the order
50
required for the respective experiment (Methods A-D). All concentrations for the assay
components are final concentrations.
4
.3.1 Method A (standard assay protocol): no pre-incubation with inhibitor. Aliquots (15
µL each) of activated LgtC, MnCl (5 mM), chicken egg-white lysozyme (CEL, 1 mg/mL),
2
calf-intestinal alkaline phosphatase (CIP, 10 U/mL), Triton (0.01%) and HEPES buffer
(
13 mM, pH 7.0) were combined with lactose acceptor (30 µL, 2 mM) or HEPES buffer
30 µL, control for non-specific hydrolysis) in the requisite microplate wells. Inhibitor at
(
various concentrations in DMSO (15 µL, 10% final DMSO concentration) or DMSO only
15 µL, control) was added. Reactions were started by addition of UDP-Gal donor (15
(
µL, 28 µM) and incubated for 20 min at 30 °C. Reactions were stopped by addition of
Malachite Green reagent A (30 µL) [20]. The microplate was shaken carefully, and
Malachite Green Reagent B (30 µL) [20] was added. The colour was allowed to develop
over 20 min, and the absorbance in each well was recorded at 620 nm on a Polarstar
Optima plate reader (BMG Labtech). The absorbance measurements were used to
calculate enzyme activity.
4
.3.2 Method B: pre-incubation of LgtC with inhibitor, in the presence of donor. Aliquots
(
15 µL each) of activated LgtC, MnCl (5 mM), CEL (1 mg/mL), CIP (10 U/mL), Triton
2
-
25 -

(0.01%) and HEPES buffer (13 mM, pH 7.0) were combined with inhibitor at various
concentrations in DMSO (15 µL, 10% final DMSO concentration) or DMSO only (15 µL,
control) in the requisite microplate wells. UDP-Gal donor (15 µL, 28 µM) was added, and
the mixtures were pre-incubated for the requisite time at 30 °C. Lactose acceptor (30 µL,
2
mM) or HEPES buffer (30 µL, control) were added, and the reactions were incubated
for 20 min at 30 °C. The reactions were stopped and analysed as described.
4
.3.3 Method C: pre-incubation of LgtC with inhibitor. Aliquots (15 µL each) of activated
LgtC, MnCl (5 mM), CEL (1 mg/mL), CIP (10 U/mL), Triton (0.01%) and HEPES buffer
2
(13 mM, pH 7.0) were combined with inhibitor at various concentrations in DMSO (15
µL, 10% final DMSO concentration) or DMSO only (15 µL, control) in the requisite
microplate wells. The mixtures were pre-incubated for the requisite time at 30 °C. UDP-
Gal donor (15 µL, 28 µM) and lactose acceptor (30 µL, 2 mM) or HEPES buffer (30 µL,
control) were added, and the reactions were incubated for 20 min at 30 °C. The
reactions were stopped and analysed as described.
4
.3.4 Method D: pre-incubation of LgtC with inhibitor, in the presence of acceptor.
Aliquots (15 µL each) of activated LgtC, MnCl (5 mM), CEL (1 mg/mL), CIP (10 U/mL),
2
Triton (0.01%) and HEPES buffer (13 mM, pH 7.0) were combined with inhibitor at
various concentrations in DMSO (15 µL, 10% final DMSO concentration) or DMSO only
(15 µL, control) in the requisite microplate wells. Lactose acceptor (30 µL, 2 mM) or
HEPES buffer (30 µL, control) were added, and the mixtures were pre-incubated for the
requisite time at 30 °C. UDP-Gal donor (15 µL, 28 µM) was added, and the reactions
were incubated for 20 min at 30 °C. The reactions were stopped and analysed as
described.
-
26 -

4
.4. Data analysis. On each microplate, a UDP calibration curve (0-12.5 µM) was
included. Wells for the calibration curve comprised of all components of the standard
reaction except for acceptor (replaced with buffer), inhibitor (replaced with DMSO), and
UDP-Gal (replaced with UDP). For each experiment, absorbance at 620 nm (AU) was
converted to [UDP] (µM) using linear regression of the respective calibration curve. A
negative control (0 µM inhibitor) and a blank value (0 µM inhibitor, 0 µM acceptor) were
included on each plate in triplicate. After linear regression, subtracting the blank from the
negative control afforded the assay window. The background value for each inhibitor
concentration (no acceptor, but otherwise identical components) was subtracted from
each inhibitor concentration data point. Once corrected for the background, the
absorbance in the presence of each inhibitor concentration was divided by the assay
window and represented as a percentage. These percentage values were plotted
against log [inhibitor] and analysed using GraphPad Prism (version 6) to afford relative
IC values, if the data represented a sigmoidal curve. Each experiment was carried out
50
at least in triplicate.
4
.5. Inhibition kinetics. The kinetic parameters k , K and k
for covalent inhibition by
obs
i
inact
2
and 3 were determined with the assay protocols described above, following the
general approach described by Singh and co-workers [13]. LgtC activity was determined
at various concentrations of 2 (0.5-100 µM) or 3 (0.5-50 µM) and in function of different
pre-incubation times (0, 5, 10, 20 and 30 min). Enzyme activity was expressed as
percentage of control (with DMSO) and plotted on a semi-logarithmic scale over pre-
incubation time. From these plots, the observed rate constant kobs at each inhibitor
concentration was extracted. kobs values were re-plotted over inhibitor concentrations
and fitted to a hyperbolic equation (kobs = kinact × [I]/(K+[I])) with GraphPad Prism (version
i
-
27 -

6
) to obtain the values for K and k
as previously described [13]. All experiments were
inact
i
performed in triplicate.
4
.6. Acceptor kinetics. The inhibition kinetics of 1 relative to acceptor were determined
with the standard assay protocol (Method A) at a fixed concentration of UDP-Gal (100
M) and variable concentrations of 1 (1, 2, 3, 4 and 5 M) and lactose acceptor (5, 15

and 25 mM). LgtC activity was adjusted to <10% turnover of UDP-Gal donor. The type of
enzyme inhibition was determined graphically by fitting the relevant set of data points to
Dixon or Cornish-Bowden plots with GraFit 7 software.
4
.7. LgtC mutants. C246A and C246S mutants of LgtC were prepared in the pCWori+
vector. Primers (Life Technologies) were designed using the Agilent primer design tool
available online [36], and mutagenesis performed using the Quikchange Lightning II site-
directed mutagenesis kit (Agilent). Mutations were confirmed by sequencing. The mutant
plasmids were transformed into the BL21 (DE3) strain of E. coli, and incubated in 1 L of
LB broth supplemented with 100 μg/mL ampicillin. Cells were incubated at 37 °C until
the optical density at 600 nm reached 0.6, induced with 0.5 mM IPTG, and incubated
overnight at 20 °C. Cells were harvested, resuspended in 25 mL of buffer A (20 mM Tris-
HCl, 0.5 M NaCl, 20 mM imidazole pH 8.0) supplemented with one tablet of Complete
EDTA free protease inhibitor cocktail (Roche), and lysed using a sonicator. The sample
was clarified by centrifugation at 20,000g, and the protein purified using an ÄKTAxpress
automated purification system. The sample was first purified over a 1 mL HisTrap crude
FF column (equilibrated with buffer A) and eluted with buffer A supplemented with 250
mM imidazole. It was then purified over a Superdex 200 16/600 hr column and eluted
isocratically with 10 mM HEPES pH 7.0, 0.5 M NaCl. Glycerol was added to 20% (v/v)
and the samples stored at -80 °C until use. The activity of both mutants and inhibition by
-
28 -

compound 1 (method A) were determined in our colorimetric glycosyltransferase assay
20].
[
4
.8. Diafiltration. Aliquots (15 µL each) of activated LgtC, MnCl (5 mM), CEL (1 mg/mL),
2
CIP (10 U/mL), Triton (0.01%), HEPES buffer (13 mM, pH 7.0), and inhibitor in DMSO
100 µM) or DMSO only (control) were combined in individual wells of a 96-well
(
microplate. UDP-Gal donor (15 µL, 28 µM) was added, and the reactions were
incubated for 10 min at 30 °C. Lactose acceptor (30 µL, 2 mM) or HEPES buffer (30 µL,
control for non-specific hydrolysis) was added to each well, and the reactions were
incubated for another 20 min at 30 °C. Samples for pre-diafiltration activity
measurements were diluted with HEPES buffer at a ratio of 1:1 and analysed
colorimetrically as described. Samples for diafiltration were collected in Vivaspin
concentrators (MW cut-off: 10 kDa, total volume: 2-6 mL) and kept on ice prior to
centrifugation. Samples (3 mL) were centrifuged at 4 °C (4000 rpm) for 50 min. The
residual volume (300 µL) was diluted to 3 mL with HEPES buffer (pH 7.0) and
centrifuged again for 50 min. The residual volume (300 µL) from the wash step was
diluted to 450 µL with HEPES buffer (pH 7.0) and combined with aliquots (450 µL each)
of MnCl (5 mM), CEL (1 mg/mL), CIP (10 U/mL), Triton (0.01%) and HEPES buffer (13
2
mM, pH 7.0). To this master mix (90 µL), lactose acceptor (30 µL, 2 mM) or HEPES
buffer (30 µL, control for non-specific hydrolysis) were added, followed by DMSO (15
µL). Reactions were started by addition of UDP-Gal donor (15 µL, 28 µM), incubated for
2
0 min at 30 °C, and analysed colorimetrically as described, to determine enzyme
activity after diafiltration.
4
.9. Time-dependent NMR experiments. A solution of the requisite amino acid (0.06
mmol; cysteine in d -DMSO, lysine and serine in D O) was added to a solution of the
6
2
-
29 -