A Modular Approach to Phosphoglycosyltransferase Inhibitors Inspired by Nucleoside Antibiotics

Article abstract of DOI:10.1002/chem.201503986

Phosphoglycosyltransferases (PGTs) represent "gatekeeper" enzymes in complex glycan assembly pathways by catalyzing transfer of a phosphosugar from an activated nucleotide diphosphosugar to a membrane-resident polyprenol phosphate. The unique structures of selected nucleoside antibiotics, such as tunicamycin and mureidomycin A, which are known to inhibit comparable biochemical transformations, are exploited as the foundation for the development of modular synthetic inhibitors of PGTs. Herein we present the design, synthesis, and biochemical evaluation of two readily manipulatable modular scaffolds as inhibitors of monotopic bacterial PGTs. Selected compounds show IC₅₀ values down to the 40 μm range, thereby serving as lead compounds for future development of selective and effective inhibitors of diverse PGTs of biological and medicinal interest.

Full text of DOI:10.1002/chem.201503986

DOI: 10.1002/chem.201503986
Full Paper
&
Glycoconjugate Biosynthesis
A Modular Approach to Phosphoglycosyltransferase Inhibitors
Inspired by Nucleoside Antibiotics
[
a]
Marthe T. C. Walvoort, Vinita Lukose, and Barbara Imperiali*
Abstract: Phosphoglycosyltransferases (PGTs) represent
gatekeeper” enzymes in complex glycan assembly pathways
ment of modular synthetic inhibitors of PGTs. Herein we
present the design, synthesis, and biochemical evaluation of
two readily manipulatable modular scaffolds as inhibitors of
monotopic bacterial PGTs. Selected compounds show IC50
values down to the 40 mm range, thereby serving as lead
compounds for future development of selective and effec-
tive inhibitors of diverse PGTs of biological and medicinal in-
terest.
“
by catalyzing transfer of a phosphosugar from an activated
nucleotide diphosphosugar to a membrane-resident poly-
prenol phosphate. The unique structures of selected nucleo-
side antibiotics, such as tunicamycin and mureidomycin A,
which are known to inhibit comparable biochemical trans-
formations, are exploited as the foundation for the develop-
[
5]
Introduction
MraY and WecA are potently inhibited by mureidomycin A
[
6]
and tunicamycin, respectively (Figure 1). Recently, the impor-
tance of MraY in bacterial peptidoglycan synthesis has prompt-
ed major interest in the development of simplified nucleoside
antibiotic analogues in the search for novel antimicrobial strat-
Phosphoglycosyl transferases (PGTs), also designated as “pri-
ming glycosyltransferases”, act at a membrane interface and
catalyze the transfer of C1-phosphosugars to membrane-asso-
ciated polyprenol phosphates to afford polyprenol diphospho-
sugar (Pren-PP-CHO) products. These pivotal intermediates
then serve as acceptors for elaboration by a sequence of gly-
cosyl transferases, producing highly diversified polyprenol di-
phosphate-linked glycans for glycoprotein, proteoglycan, and
glycolipid biosynthesis. Since PGTs catalyze the first mem-
brane-committed step in many complex glycoconjugate bio-
synthesis pathways, they serve as “gatekeepers” of glycoconju-
[
7–11]
egies.
PGTs also feature in eukaryotic biology, most prominently at
the initiation of the dolichol-dependent glycosylation pathway,
where a GlcNAc-1-phosphate transferase (GPT), designated as
Alg7 in Saccharomyces cerevisiae and other eukaryotes, trans-
[
12]
fers phospho-GlcNAc to dolichol phosphate. Sequence anal-
ysis of WecA and Alg7 reveals that these enzymes are both in-
tegral membrane proteins with 11 predicted transmembrane
helices (TMHs) and that they share key predicted active site
residues, therefore it is unsurprising that Alg7 is also potently
[1]
gate production across all domains of life. For example, MraY
is a bacterial PGT that catalyzes transfer of phospho-MurNAc-
pentapeptide to undecaprenol phosphate, mediating a key
[
13]
inhibited by tunicamycin (Figure 1).
[2]
step in peptidoglycan biosynthesis, and WecA is a bacterial
phospho-GlcNAc transferase, which catalyzes the first step in
Unfortunately, despite the tantalizing inhibition properties of
the uridinyl nucleoside antibiotics, their biological activities are
hard to predict for other PGT targets, and must be empirically
determined. This is exemplified by the large differences in in-
[3]
the biosynthesis of lipopolysaccharide O-antigen. Nucleoside
[
4]
antibiotics that feature a common uridinyl motif have been
demonstrated to be potent inhibitors of selected PGTs, and
hibition properties of mureidomycin
Figure 1). Furthermore, the complexity of the natural product
structures makes it very challenging to repurpose the struc-
A and tunicamycin
(
[
a] Dr. M. T. C. Walvoort, Dr. V. Lukose, Prof. B. Imperiali
Departments of Chemistry and Biology
[
14–16]
tures of the natural products, by synthesis
or semisynthe-
[
17–22]
Massachusetts Institute of Technology
sis,
to target alternative PGTs with different substrate spe-
7
7 Massachusetts Avenue, Cambridge MA 02139 (USA)
cificities. This challenge is further exacerbated when working
with PGTs belonging to structural classes other than the well-
studied MraY and WecA-type integral membrane proteins,
which feature 10 and 11 predicted TMHs, respectively. For ex-
ample, recent bioinformatics and biochemical analysis has re-
vealed thousands of homologous small bacterial PGTs with
only a single TMH and a soluble globular domain within a 20–
E-mail: imper@mit.edu
Supporting information for this article is available on the WWW under
http://dx.doi.org/10.1002/chem.201503986. It includes experimental and
spectroscopic data for all compounds and overview of the click library. Ex-
perimental details of the chemo-enzymatic synthesis of radiolabeled UDP-
diNAcBac, the purification of PglC, the radioactivity-based activity assay
with PglC, the luminescence assay using UMP/CMP-Glo. IC50 curves and Mi-
chaelis–Menten plots for UDP-diNAcBac and Und-P.
[
23]
2
5 kDa protein. While these PGTs catalyze comparable bio-
Part of a Special Issue “Women in Chemistry” to celebrate International
Women’s Day 2016. To view the complete issue, visit: http://dx.doi.org/
chem.v22.11.
chemical processes and also play important roles at the initia-
tion of diverse glycoconjugate biosynthetic pathways, there
Chem. Eur. J. 2016, 22, 3856 – 3864
3856
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
Figure 1. Structures of mureidomycin A and tunicamycin, and their IC50 values against selected PGTs
are currently no small molecule inhibitors that can be used to
inform on the biology and essentiality of particular pathways
and which may ultimately represent novel targets for thera-
peutic intervention.
Currently, there is no structural information available to aid
in the design of inhibitors towards small PGTs such as PglC.
The only PGT that has been structurally characterized is
[
26]
MraY, and the complete absence of sequence homology to
the small PGTs reinforces that very little insight can be gleaned
from the MraY structure to address inhibitor development for
the small PGTs. Therefore, we have turned our attention to the
nucleoside antibiotics as inspiration for the blueprints of new
modular scaffolds for PGT inhibitor design. Examination of the
structures and reported biological activities of mureidomycin A
and tunicamycin suggests important design criteria. First, the
uridinyl moiety is a key “placeholder” for PGTs that act on
UDP-sugar substrates and therefore this moiety should feature
in projected inhibitors. Second, as illustrated in Figure 1, inter-
ception of a divalent cation-binding site, either through a stra-
tegic metal ion coordinating group(s) on the inhibitor (the tu-
A prototypic example of a small PGT is PglC from Campylo-
bacter jejuni, a Gram-negative food-borne pathogen that is the
major cause of gastroenteritis worldwide. PglC catalyzes the
first membrane-committed step in the biosynthesis of viru-
[
24]
lence-associated N-linked glycoproteins in the pathogen.
Specifically, PglC, acting at the cytoplasmic face of the inner
bacterial membrane, commences assembly of an undecaprenol
diphosphate-linked heptasaccharide, which is the ultimate sub-
strate for asparagine-linked glycosylation in the periplasm.
PglC catalyzes transfer of a modified phosphosugar, di-N-ace-
tylbacillosamine-phosphate, from the corresponding UDP-
sugar donor, onto undecaprenol phosphate with release of
[
25]
[1]
UMP (Scheme 1). The PGT reaction is isoenergetic; a diphos-
phate product is formed from a diphosphate starting material
nicamycin model ) or by metal ion displacement with a pro-
[22]
tonated primary amine (the mureidomycin A model ) repre-
sents an important design feature. In this context, despite the
(
UDP-sugar). However, it is likely that in nature flux through
2
+
the pathway is promoted by the action of the subsequent gly-
cosyltransferases, which elaborate the PGT product into the
polyprenol diphosphate-linked glycan.
diversity of PGT structures, Mg -dependence, presumably to
coordinate the uridine diphosphate to prime the substrate for
nucleophilic attack at the b-phosphate, is common to small
Scheme 1. PGT reaction catalyzed by PglC.
Chem. Eur. J. 2016, 22, 3856 – 3864
www.chemeurj.org
3857
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
PGTs, such as PglC, as well as the multi-TMH PGTs exemplified
by MraY, WecA, and Alg7. Finally, the design should be com-
patible with extended binding determinants, including glycan,
peptidic, and lipophilic components, to mimic the characteris-
tics of the native substrates. Based on these criteria, we have
devised two modular synthetic strategies, inspired by mureido-
mycin A and tunicamycin. Key features of both strategies in-
clude 1) a uridine moiety, 2) a functionality to either displace
or coordinate to the metal ion cofactor, 3) an aromatic moiety
that may intercept sugar binding, and 4) a hydrophobic acyl
moiety to either bind into the undecaprenol phosphate bind-
ing site, or potentially direct the inhibitor to the membrane in-
terface. Strategy 1 (mureidomimetic) is based on mureidomy-
extraction of the hydrophobic product followed by scintillation
3
counting. To test for PglC activity and inhibition directly, [ H]-
labeled UDP-di-N-acetylbacillosamine was prepared by chemo-
3
enzymatic synthesis using [ H]-labeled AcCoA (see Supporting
Information). Since the radioactivity-based assays require syn-
thesis of radiolabeled UDP-Bac and are labor intensive, we also
present the use of a prototype UMP/CMP-Glo assay from
Promega, which is a luminescent UMP detection assay that
greatly increased the reliability and throughput of activity and
inhibition assays.
cin A in that it includes a modification at the C5’-site to display Results and Discussion
three different lengths of alkylamines to explore placement of
Strategy 1: Mureidomimetics
a positive charge adjacent to the uridine, and strategy 2 (tuni-
camimetic) is based on the incorporation of amino acid build-
ing blocks, once again conjugated to the uridinyl moiety, to
explore an alternative, metal ion-coordinating functionality ad-
jacent to the uridine.
The first modular strategy takes its cue from mureidomycin A
and includes a primary amine adjacent to the uridinyl moiety
(see Figure 1). Starting from free uridine, the 2’,3’-syn diol of
the ribose was protected with an isopropylidene group, and
the 5’-hydroxyl was tosylated (Scheme 2). Substitution of the
tosylate with mono-Boc-protected alkyldiamines (C , C , and
The development of novel small PGT (e.g. PglC) inhibitors
has also demanded attention to practical details including the
establishment of reliable heterologous expression and purifica-
2
4
C6) was accomplished under basic conditions after prolonged
reaction times to yield intermediates 2–4 in good yield. Differ-
ent spacer lengths between the uridine core and amino group
were incorporated to investigate potential structure/function
dependencies in subsequent inhibition assays (vide infra). Re-
moval of the N-Boc and isopropylidene protecting groups was
accomplished simultaneously using aqueous TFA, and products
5–7 were obtained in quantitative yield after lyophilization. Ini-
tial inhibition assays revealed that these first generation com-
pounds 5–7 inhibited the PglC reaction completely at 5 mm,
while activity was almost completely unaffected at 1 mm (data
not shown). Qualitatively, small differences were observed with
the three congeners, so all series were forwarded for further
elaboration. In all cases, inhibition assays were carried out
[
27]
tion procedures for the target membrane-bound enzyme
and the application of robust assays to provide consistent
feedback on inhibitor efficiency. One of the standard ap-
proaches for assaying PGTs relies on radioactivity-based strat-
1
4
egies that use tritium or [ C]-labeled UDP-sugars and unde-
caprenol phosphate together with liquid/liquid extraction of
lipophilic undecaprenol diphosphosugar products. For exam-
ple, PglC may be assayed in a coupled assay, in which the PglC
product is further elaborated by PglA, an N-acetylgalactos-
3
amine transferase that uses UDP-[ H]GalNAc, from the pgl
[
28]
pathway of C. jejuni. In this way, tritium-labeled GalNAc is in-
corporated into the undecaprenol diphospho-disaccharide
(
Und-PP-Bac-GalNAc) and enzymatic activity is quantified by
Scheme 2. Synthesis of dyad inhibitors, varying alkyl length and alkyne. A) H
CO , THF (2: 78%, 3: 91%, 4: 96%); D) TFA, H
DCM; ii. NaN , DMF (toward 11: quant., toward 12: 94%, toward 13: quant.); G) alkyne, CuSO
dimethylamino)pyridine, TFA=trifluoroacetic acid, EDC = 3-(ethyliminomethylidenamino)-N,N-dimethyl-propan-1-amine, HOBt= hydroxybenzotriazole,
DCM=dichloromethane; DMF= dimethylformamide.
2
SO
4
, acetone (96%); B) p-TsCl, DMAP, DCM/pyridine (89%); C) N-Boc-alkylamine,
K
2
3
2
O (quant.); E) Ac-l-Ala-OH, EDC, HOBt, DMF (towards 8: 54%, 9: 37%, 10: 74%); F) i. chloroacetic anhydride,
, sodium ascorbate, tBuOH/H O (quant. by LC-MS). DMAP= 4-
3
4
2
(
Chem. Eur. J. 2016, 22, 3856 – 3864
www.chemeurj.org
3858
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
under conditions which minimized promiscuous binding ef-
sugar substrate, and therefore our attention was focused on
mono- and bicyclic aromatic compounds that could engage in
p-stacking interactions often identified in sugar-binding
[29]
fects resulting from inhibitor aggregation.
To reduce flexibility and potentially increase the opportunity
for productive binding interactions with the enzyme, the C5’-
secondary amine position was derivatized by coupling with N-
acetyl-alanine using standard coupling conditions (EDC, HOBt).
Although the use of the N-acetyl derivative of alanine under
these conditions resulted in epimerization at the C-a center,
the diasteromeric mixture was carried forward for inhibition
analysis. While epimerization could have been prevented using
the N-a-Fmoc-protected amino acid, deprotection under basic
conditions was known to result in decomposition of the result-
ing tertiary amide at the 5’-N in related compounds (vide
infra). Acidic treatment to remove the isopropylidine group af-
forded compounds 8–10, which showed improved inhibition
potency, with 78Æ14% inhibition at 1 mm for compound 10
[
38]
sites.
The synthesis of azide-functionalized alkylamine-uri-
dines 11–13 involved N-acylation with chloroacetic anhydride,
followed by chloride displacement with sodium azide in DMF
and global deprotection under acidic conditions. Next, a selec-
tion of commercially available terminal alkynes (a–i, Scheme 2)
II
was conjugated to 13 using standard Cu /ascorbate conditions
to catalyze the [3+2] cycloaddition click reaction. Reaction
progress was monitored using LC-MS and purification was per-
formed using SepPak C18 cartridges. For this study, the C6
compound 13 was used as the parent compound because the
alanine-conjugated equivalent, 10, had shown the best activity
in inhibition assays. Indeed, a broad range of activities was ob-
served when compounds 13a–i were assayed, with 13a result-
ing in complete inhibition of PglC activity at 3 mm (Figure 2B).
Interestingly, the parent compound 13 showed only modest
inhibition at 3 mm, reinforcing the importance of the naphthyl
moiety on binding. These results demonstrate the added value
of the dyad design, and might also suggest the importance of
a stereocenter appended to 5’-N which is lacking in 13 but
present in 10. With the aim of distinguishing between the dif-
ferent spacer lengths at this stage of inhibitor design, we con-
jugated the alkynes that had showed the greatest inhibition
potential (a, f–i) to the shorter parent compounds 11 and 12.
Interesting trends in inhibition properties were observed (see
Figure S2, Supporting Information), and although alkyne g also
yielded inhibitors with improved properties, the overall conclu-
sion was that inhibitor 13a, in which the naphthyl moiety was
(
the C dialkylamine, Figure 2A) at substrate concentrations of
6
2
0 mm UDP-diNAcBac (K =7.2Æ1.1 mm) and 20 mm Und-P (K
M
M,
=
15.6Æ5.1 mm) in the assay. At this point, it was possible to
app
discriminate between the different alkyl lengths, with com-
pound 10 being a superior inhibitor to 8 and 9 (see Figure S1,
Supporting Information). The activity of compound 10 repre-
sents a significant improvement relative to uridine, which
shows no inhibition of PglC at 1 mm, indicating that the uri-
dine core alone is not sufficient for binding at this concentra-
tion (Figure 2A). In contrast, uridine monophosphate (UMP),
which is also a side product of the PglC reaction, shows com-
parable inhibition relative to 10 at 1 mm. This is in line with
our observations that the PglC reaction saturates at 30% con-
version, supposedly due to UMP product inhibition. Although
uridyl-phosphates are successfully elaborated towards potent
combined with the C alkylamine, showed the highest activity.
6
[30–32]
glycosyltransferase inhibitors,
in the long term, the amine-
Because the dyad inhibitor strategy left little opportunity for
further functionalization, the inhibitor scaffold was modified to
allow further derivatization toward ‘triad inhibitors’. As illustrat-
ed in Scheme 3, N-a-Fmoc-protected azido-alanine (Aza) was
coupled to protected hexylamine-uridine 4, using standard
amide coupling conditions. In this way, the scaffold has the po-
tential to be modified first by click reactions at the azide
moiety with different alkynes and second by coupling different
carboxylic acid derivatives to the liberated amine. Ensuing
Fmoc deprotection of compound 14 using 20% piperidine in
DMF gave a high amount of the b-eliminated product, as well
substituted derivatives would be advantageous due to the
positive charge, which would enhance cell permeability rela-
tive to negatively charged analogues, as has been an inspira-
[33–37]
tion in several inhibitor design strategies.
Encouraged by these initial inhibition results, the mureido-
mimetic strategy was expanded to create ‘dyad inhibitors’ that
would display two binding modules. Instead of the alanine
that had principally been incorporated to increase rigidity, a tar-
geted screen was performed to identify a moiety that could
potentially bind in the carbohydrate-binding site of the UDP-
Figure 2. Biological evaluation of hexylamine-uridine inhibitors (A) and clicked product based on compound 13 (B), using the radioactivity-based activity
assay. Error bars indicate mean ÆSD.
Chem. Eur. J. 2016, 22, 3856 – 3864
www.chemeurj.org
3859
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
Scheme 3. Synthesis and biological evaluation of triad inhibitors based on uridine-hexylamine, using the UMP/CMP-Glo assay. A) Fmoc-l-Aza-OH, EDC, HOBt,
2
DMF (45%); B) immobilized piperazine, DCM; C) myristic acid, EDC, HOBt, DMF (48% 2 steps); D) TFA, H O (17: 96%, 18: 85%); E) 2-ethynyl-6-methoxynaph-
thalene, CuSO
4
, sodium ascorbate, DMF (77%). Error bars indicate mean ÆSD.
as decomposition back to starting compound 4 (as identified
using LC-MS), which was inseparable from the product. While
shorter reaction times and reduced levels of base suppressed
the formation of the elimination product, decomposition was
still detected. Ultimately, a systematic study of potential condi-
tions resulted in the use of piperazine immobilized on polystyr-
ene resin, which significantly reduced the amount of com-
pound 4 regenerated. Subsequently, myristic acid was coupled
to the amine in 15, followed by acidic hydrolysis to afford in-
hibitor 17. Additionally, compound 16 was subjected to click
chemistry conditions using 2-ethynyl-6-methoxynaphthyl to
give fully functionalized triad inhibitor 18 after acidic hydroly-
sis in good yield.
behavior was due to specific binding to the enzyme rather
than promiscuous binding activity resulting from inhibitor ag-
[
29]
gregation effects. In addition, the exact distance between
the uridine core and the naphthyl group is slightly longer in
18 than in the precursor 13a, which could indicate a slight
clash in the binding site, creating a suboptimal fit. However, it
is clear from these results that increasing complexity of the
modular scaffold results in increased inhibitory potency until
an unfavorable interaction is introduced. Most importantly, the
modular design of the mureidomycin mimetic strategy allows
rapid differentiation at three positions, which will be extremely
valuable in future optimization of this scaffold towards specific
PGT targets.
Inhibition assays using the UMP/CMP-Glo system revealed
that attaching the long chain (C ) fatty acid had a positive
1
4
Strategy 2 - Tunicamimetics
effect on the inhibition potency, as exemplified by 90Æ3% in-
hibition of PglC at 250 mm inhibitor with substrate concentra-
tions of 20 mm UDP-diNAcBac and 20 mm Und-P in the assay.
The IC50 of compound 17 was determined to be 85Æ28 mm.
This reinforces the importance of a hydrophobic acyl moiety in
the design of PGT inhibitors, to either compete with the unde-
caprenol phosphate acceptor for binding, or to direct the in-
hibitor to the membrane interface proximal to where catalysis
occurs. Interestingly, the inhibition potency decreased on
going from 17 to 18, which has the added methoxynaphthyl
moiety, to 75Æ1% inhibition at 250 mm inhibitor 18. This cor-
Having established that the combination of an alkylamine, a hy-
drophobic acyl group, and an aromatic moiety appended to
a uridine core produced inhibitors with three points of varia-
tion and micromolar IC50 values, we then investigated an alter-
native modular scaffold that would display similar functionali-
ties, but would facilitate easier differentiation at three modular
positions. As depicted in Scheme 4, this strategy started with
isopropylidene-protected 5’-amino-uridine (compound 19),
and involved amide coupling chemistry to introduce the vari-
ous modules. Instead of a primary amine that would displace
the metal in the active site, this strategy exploits a strategic
carboxylic acid functionality to investigate if such compounds
might offer advantages in binding by directly coordinating
responds to an IC5 value of about 207 mm for compound 18.
0
This decrease in inhibition potency with the added naphthyl
moiety is most likely explained by the poorer solubility of in-
hibitor 18 in the assay conditions. Importantly, inhibition
assays were carried out in the presence of 10% DMSO and
[
39,40]
a bound divalent cation at the active site.
Such a strategy
would mimic the proposed mechanism of tunicamycin-like nu-
cleosides, which are thought to intercept binding as illustrated
0
.1% Triton X-100 in order to ensure that observed inhibition
3860
Chem. Eur. J. 2016, 22, 3856 – 3864
www.chemeurj.org
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
Scheme 4. Synthesis and biological evaluation of triad inhibitors based on a peptidic backbone, using the UMP/CMP-Glo assay. A) Fmoc-l-Asp(tBu)-OH, EDC,
HOBt, DMF; B) 20% piperidine, DMF (62% 2 steps); C) Ac
quant., 29: 66%); E) Fmoc-l-Aza-OH, EDC, HOBt, DMF (71%); F) immobilized piperazine, DCM; G) myristic acid, EDC, HOBt, DMF (63% 2 steps); H) 2-ethynyl-6-
methoxynaphthalene, CuSO
, sodium ascorbate, DMF (toward 29: 93%). Error bars indicate mean ÆSD.
2 3 2
O, Et N, MeOH (24: 89%); D) TFA, H O (21: 86% 2 steps, 25: 74% 3 steps, 26: 61% 2 steps, 28:
4
in Figure 1. To this end, Fmoc-l-Asp(tBu)-OH was coupled to
compound 19, followed by Fmoc removal using 20% piperi-
dine. To generate the first generation inhibitor (compound 21),
the free amine in 20 was capped with an acetyl, and the tert-
butyl and isopropylidene groups were simultaneously removed
using aqueous TFA. Next, the Aza building block was conjugat-
ed to allow differentiation using click chemistry, and com-
pound 22 was obtained in good yield. Fmoc removal under
mild conditions proceeded smoothly to afford compound 23,
which was either capped with an acetyl group to give com-
pound 24, or coupled with myristic acid to generate com-
pound 27. The acetylated precursor 24 was either directly de-
protected under acidic conditions to give inhibitor 25, or sub-
jected to click conditions with 2-ethynyl-6-methoxynaphthyl to
generate a new class of tunicamimetic dyad inhibitor, 26, after
acidic hydrolysis of the protecting groups. To generate the
other dyad inhibitor displaying the hydrophobic acyl group,
myristic acid was coupled to compound 23 using standard
amide-coupling conditions. This resulted in compound 27,
which was either fully deprotected using aqueous TFA to give
dyad inhibitor 28, or subjected to click conditions with 2-eth-
ynyl-6-methoxynaphthyl to produce the triad inhibitor 29. Be-
cause this second synthetic strategy was more straightforward
than the first strategy, it allowed for thorough purification and
isolation of intermediate compounds throughout the sequence
of reactions. Advantageously, evaluating the inhibitory poten-
tial of the intermediates would provide information about the
additive effects of the different modules on the overall poten-
cy. The results of the comparative inhibition assays are dis-
played in Scheme 4. Using the UMP/CMP-Glo assay, it was
demonstrated that acid-functionalized uridine 21 inhibited
PglC activity by 57Æ6% at 250 mm. Notably, this inhibitory po-
tential is considerably higher than the inhibition observed with
first-generation inhibitors based on the alkylamine module
(Scheme 1, compound 7). The addition of the Aza building
block did not influence the inhibitory potency to a great
extent (25 in Scheme 4). In contrast, the addition of either a 2-
ethynyl-6-methoxynaphthyl through click chemistry (com-
pound 26), or a fatty acid moiety (compound 28) improved
the inhibition of PglC, resulting in 77Æ10% and 88Æ5% in-
hibition at 250 mm respectively. Using the UMP/CMP-Glo assay,
we were able to determine the IC50 values to be ~205 mm for
Chem. Eur. J. 2016, 22, 3856 – 3864
www.chemeurj.org
3861
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
2
6 and 43Æ5 mm for 28. This implies a modest improvement
evaluation of inhibitory potentials in a simple, 96-well format
assay.
over the analogous dyad inhibitor 17 from the first strategy.
Interestingly, a similar trend in inhibition potential was ob-
served with the triad inhibitor 29, which showed 81Æ3% in-
hibition at 250 mm, a slight decrease in potency compared to
Our efforts in both of these arenas have greatly contributed
to the generation of PglC inhibitors. In conclusion, the chemi-
cal and biochemical studies lay the foundation for a diversity-
oriented approach to explore the chemical space allowed for
inhibitors in the active sites of PGTs. Moreover, the results will
2
8. However, this corresponded to an IC50 of 64Æ10 mm, which
is an order of magnitude better than the corresponding alkyl-
[
43,44]
amine inhibitor 18.
direct diversification of existing uridine libraries.
Conclusion
Experimental Section
The development of inhibitors for PGTs is a relatively unex-
plored field, and it holds great potential to aid our understand-
ing of this important class of enzymes, and to generate target
molecules that inhibit crucial bacterial enzymes. Apart from di-
verse nucleosides that have been targeted to inhibit MraY,
there is little known about molecular scaffolds that inhibit
other PGTs specifically. Bioinformatics and structural investiga-
tions have revealed that PGTs display highly diverse topologies,
hinting at diverse active site architectures to catalyze a similar
reaction. Herein, we have presented the de novo design of
two modular scaffolds that resulted in compounds with low
micromolar inhibition of PglC, the monotopic PGT that is the
gatekeeper enzyme of N-linked glycoprotein biosynthesis in C.
jejuni. The modular approach allowed for evaluation of the
effect of adding different modules on the inhibitory potential
of PglC. Thus, it was clear that, starting from uridine, the at-
tachment of aspartate (21, tunicamimetic strategy) was superi-
or to the hexylamine moiety (7, mureidomimetic strategy), re-
sulting in 57% inhibition at 250 mm with the former. Attach-
ment of an aromatic moiety (i.e. naphthyl group in 26), and to
a greater extent a hydrophobic long-chain acyl group (17 and
General experimental procedures: All chemicals were used as re-
1
13
ceived unless stated otherwise. H and C NMR spectra were re-
corded on a Bruker DPX-400 (400/100 MHz), a Varian 500 (500/
125 MHz), and a Bruker AV-600 (600/150 MHz) spectrometer. Chem-
ical shifts (d) are given in ppm and coupling constants are given in
13
Hz. All given C spectra are proton-decoupled. Flash chromatogra-
phy was performed on Silicycle Siliflash P60 silica gel (40–60 mm).
TLC analysis was conducted on Agela Technologies TLC plates with
detection by UV absorption (254 nm) where applicable and by
spraying with 20% sulfuric acid in ethanol followed by charring at
about 1508C or by spraying with a solution of (NH ) Mo O ·H O
4
6
7
24
2
À1
À1
(25 gL ) and (NH ) Ce(SO ) ·2H O (10 gL ) in 10% sulfuric acid in
4
4
4 4
2
water followed by charring at about 1508C. LC-MS analysis was
performed on an HP 1000 series HPLC system and Finnegan
LCQDeca mass spectrometer. Standard eluents used were A: 0.1%
TFA in H O, B: 0.1% TFA in acetonitrile. The column used was
2
a YMC-Pack ODS-AQ column (3 mm, 100”3.0 mm I.D.). All analyses
À1
were carried out over 15 min, with a flow-rate of 0.1 mLmin . The
UMP/CMP-Glo assay (Promega) was used as received. Luminescent
H1
readings were performed on a Synergy platereader (Biotek) ac-
cording to the protocol from Promega. Scintillation counting was
performed on a Beckman Coulter LS6500 scintillation counting
system.
2
8) resulted in a further increase in inhibitory potential, while
the combination of both modules resulted in a slight decrease
18 and 29). This detrimental effect might be attributed to
General procedure A: Amino acid coupling: Fmoc-protected
amino acid (2 equiv) was pre-activated by mixing with EDC·HCl
(
(
2 equiv) and HOBt (2 equiv) in DMF (0.3m) for 10 min at room
a steric clash of the elaborated scaffold in the active site, or to
the poor solubility of the triad inhibitors. Interestingly, the in-
hibition potentials of these scaffolds are similar to that of tuni-
camycin, which was found to have an IC50 of 100Æ8 mm for
PglC (see Supporting Information). Together, these studies pro-
vide the proof-of-concept for our approach, and pave the way
for increasing the potency and specificity for diverse monotpic
PGTs and potentially the more complex polytopic PGTs by per-
forming structure optimization at the diverse point of contact
with the enzyme. Of paramount importance in these studies
was the establishment of a reproducible enzyme purification
protocol, and the development of a reliable activity assay,
which in contrast to existing fluorescence-based assays for
temperature, and this mixture was added to the amine (1 equiv).
The mixture was stirred overnight, diluted with DCM and washed
with H O (4”). The organic fraction was dried over Na SO , concen-
trated in vacuo, and re-dissolved in DMF (0.3m). Piperidine (20
vol%) was added, and the mixture was stirred until complete con-
sumption of starting material was observed using LC-MS. The mix-
ture was concentrated in vacuo and purified using flash column
chromatography (silica gel, DCM/MeOH) to obtain the desired
product.
2
2
4
General procedure B: Click reaction in tBuOH/H O: The azide
2
(
2 mmol, stock in DMSO) and alkyne (2.4 mmol, 100 mm stock in
DMSO) were together dissolved in tBuOH/H O (800 mL, 1/1, v/v).
2
CuSO4 (2 mmol, 125 mm stock in H O) and sodium ascorbate
2
[
41,42]
MraY,
represents a significant challenge. We found that
(4.8 mmol, 250 mm stock in H O) were added and the resulting so-
2
lution was incubated on a shaker at room temperature until com-
plete consumption of the azide was observed using LC-MS (24–
both assays described herein report accurately on PglC activity.
In general, the radiochemical assay is more robust with a varie-
ty of inhibitors, but necessitates preparation of radiolabeled
UDP-Bac and is also considerably more labor intensive due to
the need for manual liquid–liquid extractions for each
quenched reaction aliquot. In contrast, provided that appropri-
ate controls are carried out to check for interference with the
assay constituents, the UMP/CMP-Glo assay allows for rapid
4
8 h). The mixture was lyophilized, redissolved in H O/MeCN, and
2
purified using SepPak C18 cartridges. Product fractions were
lyophilized and redissolved in DMSO to give 50 mm stock solu-
tions.
General procedure C: Click reaction in DMF: A solution of the
azide (1 equiv) in DMF (0.025m) was treated with 2-ethynyl-6-me-
Chem. Eur. J. 2016, 22, 3856 – 3864
www.chemeurj.org
3862
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
thoxynaphthalene (4 equiv), CuSO (1 equiv, 1m stock in H O) and
which time the luminescence was read. The data was plotted as
percentage remaining activity compared to the positive control
(no inhibitor). The IC50 values were obtained by plotting residual
activity versus concentration (GraphPad Prism). To eliminate any
false negatives based on luminescence quenching and/or inhibi-
tion of the luminescence assay, a control experiment was included
in which the luminescence assay cocktail was incubated with and
without inhibitor (in the presence of 2 mm UMP). Background inhib-
ition of the UMP/CMP-Glo assay was established for each inhibitor,
and the luminescence values were adjusted accordingly.
4
2
sodium ascorbate (2 equiv, 1m stock in H O). Complete consump-
2
tion of the azide was confirmed using LC-MS (1–3 h) and the mix-
ture was diluted with DCM, washed with NH Cl (4”), dried over
4
Na SO , and concentrated in vacuo. The product was purified using
2
4
flash column chromatography (silica gel, DCM/MeOH).
General procedure D: Acidic hydrolysis: Protected compound
(
10 mg) was treated with TFA/H O (500 mL, 3/2, v/v) until complete
2
removal of the protecting groups was observed using LC-MS (3–
4 h). The mixture was lyophilized repeatedly to obtain the final
2
compound.
Comparison of radioactivity-based and UMP/CMP-Glo assays:
Using the exact same assay conditions, a similar turnover is detect-
ed using both the radiochemical and UMP/CMP-Glo assays. In the
radiochemical assay, approximately 600 DPM is generated in an un-
inhibited reaction, which corresponds to about 12% conversion of
UDP-diNAcBac. Using the UMP Glo assay, RLU values between
Expression and purification of PglC: PglC was expressed in E. coli
with a His -SUMO-tag for expression, solubility, and purification
6
purposes. The membrane fraction was isolated by centrifugation,
and the sample was homogenized in 1% n-dodecyl b-d-maltoside
2
000–2500 are typically generated, which correspond to a conver-
sion of about 10% UDP-diNAcBac to UMP and Und-PP-diNAcBac
calculated using a UMP standard curve).
(
DDM) to transfer PglC to detergent micelles. Buffers used for sub-
sequent purification steps included DDM at 0.03% (three times the
critical micelle concentration) to ensure that the protein was solu-
bilized in detergent micelles. PglC was purified using affinity chro-
matography with Ni-NTA resin, followed by gel filtration chroma-
tography. The specific activity of PglC was calculated to be
(
Acknowledgements
À1
À1
3
.43 mmol min mg . Purified enzyme was stored at À808C. A de-
This project was funded by the NIH (R21 AI101807) to B.I. We
thank Dr. Hicham Zegzouti (Promega) for supplying the UMP/
CMP-Glo assay. We also thank Dr. Debasis Das for helpful dis-
cussions.
tailed protocol is provided in the Supporting Information.
Radioactivity-based activity assay with PglC: Assays contained
2
1
0 mm Und-P, 10% DMSO, 0.1% Triton X-100, 50 mm HEPES pH 7.5,
00 mm NaCl, 5 mm MgCl , 20 mm [ H]-UDP-diNAcBac (5.4 mCi/
3
2
mmol), and 1 nm PglC in a final volume of 60 mL. Inhibitors were
added in DMSO, in a volume such that the total concentration of
DMSO in the reaction was equal to 10% (v/v). PglC was pre-incu-
Keywords: glycoconjugate biosynthesis · inhibitors · modular
approach · nucleoside antibiotic · phosphoglycosyl transferase
3
bated in the reaction mixture lacking [ H]-UDP-diNAcBac for five
[1] N. P. Price, F. A. Momany, Glycobiology 2005, 15, 29R–42R.
minutes at room temperature. After initiation of the reaction with
3
[2] A. Bouhss, D. Mengin-Lecreulx, D. Le Beller, J. Van Heijenoort, Mol. Mi-
crobiol. 1999, 34, 576–585.
[
H]-UDP-diNAcBac, aliquots (20 mL) were taken at twenty minute
time points and quenched in 1 mL CHCl :MeOH. The organic layer
3
[3] J. Lehrer, K. A. Vigeant, L. D. Tatar, M. A. Valvano, J. Bacteriol. 2007, 189,
was washed three times with 400 mL PSUP (Pure Solvent Upper
2618–2628.
Phase, composed of 15 mL CHCl , 240 mL MeOH, 1.83 g KCl in
3
[4] M. Winn, R. J. Goss, K. Kimura, T. D. H. Bugg, Nat. Prod. Rep. 2010, 27,
279–304.
[5] M. Inukai, F. Isono, A. Takatsuki, Antimicrob. Agents Chemother. 1993, 37,
2
5
35 mL H O). The resulting aqueous layers were combined with
mL EcoLite (MP Biomedicals) liquid scintillation cocktail. Organic
2
9
80–983.
6] B. Al-Dabbagh, D. Mengin-Lecreulx, A. Bouhss, J. Bacteriol. 2008, 190,
141–7146.
layers were combined with 5 mL OptiFluor (PerkinElmer). Both
layers were analyzed using scintillation counting. The data was
plotted as percentage remaining activity compared to the positive
control (no inhibitor). A typical uninhibited reaction yielded ap-
proximately 600 dpm (quench compensated), corresponding to
a turnover of roughly 12% of the total substrate in the reaction.
[
7
[
[
7] N. I. Howard, T. D. H. Bugg, Bioorg. Med. Chem. 2003, 11, 3083–3099.
8] S. Grüschow, E. J. Rackham, B. Elkins, P. L. Newill, L. M. Hill, R. J. Goss,
ChemBioChem 2009, 10, 355–360.
[
9] T. Tanino, B. Al-Dabbagh, D. Mengin-Lecreulx, A. Bouhss, H. Oyama, S.
Ichikawa, A. Matsuda, J. Med. Chem. 2011, 54, 8421–8439.
[
[
10] K. Okamoto, M. Sakagami, F. Feng, H. Togame, H. Takemoto, S. Ichikawa,
A. Matsuda, J. Org. Chem. 2012, 77, 1367–1377.
11] M. J. Fer, S. Olatunji, A. Bouhss, S. Calvet-Vitale, C. Gravier-Pelletier, J.
Org. Chem. 2013, 78, 10088–10105.
12] P. Burda, M. Aebi, Biochim. Biophys. Acta. 1999, 1426, 239–257.
13] R. K. Keller, D. Y. Boon, F. C. Crum, Biochemistry 1979, 18, 3946–3952.
14] T. Suami, H. Sasai, K. Matsuno, N. Suzuki, Carbohydr. Res. 1985, 143, 85–
96.
Luminescence assay using UMP/CMP-Glo: The quenching solu-
tion was prepared as described by Promega. Assays contained
2
1
0 mm Und-P, 10% DMSO, 0.1% Triton X-100, 50 mm HEPES pH 7.5,
00 mm NaCl, 5 mm MgCl , 20 mm UDP-diNAcBac, and 1 nm PglC in
2
[
[
[
a final volume of 25 mL. Inhibitors were added in DMSO, in
a volume such that the total concentration of DMSO in the reac-
tion was equal to 10% (v/v). PglC was pre-incubated in the reac-
tion mixture lacking UDP-diNAcBac for five minutes at room tem-
perature. The PglC reaction rate was dependent on the amount of
enzyme in the assay and the specific quantity of PglC used was
that which afforded a linear reaction rate over 20 min. After initia-
tion of the reaction with UDP-diNAcBac, the reaction was halted
after 20 min by the addition of 25 mL quenching buffer. The mix-
ture was transferred to a 96-well plate (white, half area, Corning)
and placed in the plate reader. The plate was shaken at low speed
for 16 min, and incubated at room temperature for 44 min, after
[15] A. G. Myers, D. Y. Gin, D. H. Rogers, J. Am. Chem. Soc. 1993, 115, 2036–
2038.
[
[
[
16] J. Li, B. Yu, Angew. Chem. Int. Ed. 2015, 54, 6618–6621; Angew. Chem.
015, 127, 6718–6721.
17] F. Sarabia, L. Martín-Ortiz, F. J. López-Herrera, Org. Biomol. Chem. 2003,
, 3716–3725.
2
1
18] S. J. Danishefsky, S. L. Deninno, S. Chen, L. Boisvert, M. Barbachyn, J.
Am. Chem. Soc. 1989, 111, 5810–5818.
[19] W. Karpiesiuk, A. Banaszek, Carbohydr. Res. 1997, 299, 245–252.
[20] J. Ramza, A. Zamojski, Tetrahedron 1992, 48, 6123–6134.
Chem. Eur. J. 2016, 22, 3856 – 3864
www.chemeurj.org
3863
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
[
[
[
[
[
[
[
21] A. Bozzoli, W. Kazmierski, G. Kennedy, A. Pasquarello, A. Pecunioso,
Bioorg. Med. Chem. Lett. 2000, 10, 2759–2763.
22] C. A. Gentle, S. A. Harrison, M. Inukai, T. D. H. Bugg, J. Chem. Soc. Perkin
Trans. 1 1999, 1287–1294.
23] V. Lukose, L. Luo, D. Kozakov, S. Vajfa, K. N. Allen, B. Imperiali, Biochem-
istry 2015, DOI: 10.1021/acs.biochem.5b01086.
24] J. Kelly, H. Jarrell, L. Millar, L. Tessier, L. M. Fiori, P. C. Lau, B. Allan, C. M.
Szymanski, J. Bacteriol. 2006, 188, 2427–2434.
[34] L. Ballell, R. J. Young, R. A. Field, Org. Biomol. Chem. 2005, 3, 1109–1115.
[35] N. Mitsuhashi, H. Yuasa, Eur. J. Org. Chem. 2009, 1598–1605.
[36] G. Mugunthan, D. Sriram, P. Yogeeswari, K. P. R. Kartha, J. Carbohydr.
Chem. 2012, 31, 553–570.
[37] S. Wang, J. A. Cuesta-Seijo, D. Lafont, M. M. Palcic, S. Vidal, Chem. Eur. J.
2013, 19, 15346–15357.
[38] S. Cecioni, A. Imberty, S. Vidal, Chem. Rev. 2015, 115, 525–561.
[39] A. Yamashita, E. Norton, P. J. Petersen, B. A. Rasmussen, G. Singh, Y.
Yang, T. S. Mansour, D. M. Ho, Bioorg. Med. Chem. Lett. 2003, 13, 3345–
3350.
[40] A. P. Spork, M. Buschleb, O. Ries, D. Wiegmann, S. Boettcher, A. Mihalyi,
T. D. H. Bugg, C. Ducho, Chem. Eur. J. 2014, 20, 15292–15297.
[41] A. Moberg, L. Z. Balderud, E. Hansson, H. Boyd, Assay Drug Dev. Technol.
2014, 12, 506–513.
25] K. J. Glover, E. Weerapana, M. M. Chen, B. Imperiali, Biochemistry 2006,
4
5, 5343–5350.
26] B. C. Chung, J. Zhao, R. A. Gillespie, D. Y. Kwon, Z. Guan, J. Hong, P.
Zhou, S. Y. Lee, Science 2013, 341, 1012–1016.
27] S. Wagner, L. Baars, A. J. Ytterberg, A. Klussmeier, C. S. Wagner, O. Nord,
P. A. Nygren, K. J. van Wijk, J. W. de Gier, Mol. Cell. Proteomics 2007, 6,
1
527–1550.
28] J. P. Morrison, J. M. Troutman, B. Imperiali, Bioorg. Med. Chem. 2010, 18,
167–8171.
[42] T. Stachyra, C. Dini, P. Ferrari, A. Bouhss, J. van Heijenoort, D. Mengin-Le-
creulx, D. Blanot, J. Biton, D. Le Beller, Antimicrob. Agents Chemother.
2004, 48, 897–902.
[
8
[
[
29] B. Y. Feng, B. K. Shoichet, Nat. Protoc. 2006, 1, 550–553.
30] L. V. Lee, M. L. Mitchell, S. J. Huang, V. V. Fokin, K. B. Sharpless, C. H.
Wong, J. Am. Chem. Soc. 2003, 125, 9588–9589.
[43] O. Moukha-Chafiq, R. C. Reynolds, ACS Comb. Sci. 2014, 16, 232–237.
[44] D. Sun, V. Jones, E. I. Carson, R. E. Lee, M. S. Scherman, M. R. McNeil,
R. E. Lee, Bioorg. Med. Chem. Lett. 2007, 17, 6899–6904.
[
31] N. Auberger, R. Frlan, B. Al-Dabbagh, A. Bouhss, M. Crouvoisier, C. Gravi-
er-Pelletier, Y. Le Merrer, Org. Biomol. Chem. 2011, 9, 8301–8312.
32] L. M. Tedaldi, M. Pierce, G. K. Wagner, Carbohydr. Res. 2012, 364, 22–27.
33] R. Wang, D. H. Steensma, Y. Takaoka, J. W. Yun, T. Kajimoto, C. H. Wong,
Bioorg. Med. Chem. 1997, 5, 661–672.
[
[
Received: October 5, 2015
Published online on December 10, 2015
Chem. Eur. J. 2016, 22, 3856 – 3864
www.chemeurj.org
3864
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim