Design, synthesis and evaluation of carbamate-linked uridyl-based inhibitors of human ST6Gal I

Article abstract of DOI:10.1016/j.bmc.2020.115561

Sialic acid at the terminus of cell surface glycoconjugates is a critical element in cell-cell recognition, receptor binding and immune responses. Sialyltransferases (ST), the enzymes responsible for the biosynthesis of sialylated glycans are highly upregulated in cancer and the resulting hypersialylation of the tumour cell surface correlates strongly with tumour growth, metastasis and drug resistance. Inhibitors of human STs, in particular human ST6Gal I, are thus expected to be valuable chemical tools for the discovery of novel anticancer drugs. Herein, we report on the computationally-guided design and development of uridine-based inhibitors that replace the charged phosphodiester linker of known ST inhibitors with a neutral carbamate to improve pharmacokinetic properties and synthetic accessibility. A series of 24 carbamate-linked uridyl-based compounds were synthesised by coupling aryl and hetaryl α-hydroxyphosphonates with a 5′-amino-5′-deoxyuridine fragment. The inhibitory activities of the newly synthesised compounds against recombinant human ST6Gal I were determined using a luminescent microplate assay, and five promising inhibitors with K_i’s ranging from 1 to 20 μM were identified. These results show that carbamate-linked uridyl-based compounds are a potential new class of readily accessible, non-cytotoxic ST inhibitors to be further explored.

Full text of DOI:10.1016/j.bmc.2020.115561

Journal Pre-proofs
Design, Synthesis and Evaluation of Carbamate–linked Uridyl-based Inhibi‐
tors of Human ST6Gal I
Andrew P. Montgomery, Christopher Dobie, Rémi Szabo, Laura Hallam,
Marie Ranson, Haibo Yu, Danielle Skropeta
PII:
DOI:
Reference:
S0968-0896(20)30391-6
https://doi.org/10.1016/j.bmc.2020.115561
BMC 115561
To appear in:
Bioorganic & Medicinal Chemistry
Received Date:
Revised Date:
Accepted Date:
20 February 2020
14 April 2020
16 May 2020
Please cite this article as: Montgomery, A.P., Dobie, C., Szabo, R., Hallam, L., Ranson, M., Yu, H., Skropeta, D.,
Design, Synthesis and Evaluation of Carbamate–linked Uridyl-based Inhibitors of Human ST6Gal I, Bioorganic
&
Medicinal Chemistry (2020), doi: https://doi.org/10.1016/j.bmc.2020.115561
This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover
page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version
will undergo additional copyediting, typesetting and review before it is published in its final form, but we are
providing this version to give early visibility of the article. 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.
©
2020 Elsevier Ltd. All rights reserved.

Design, Synthesis and Evaluation of
Carbamate-linked Uridyl-based Inhibitors of Human
ST6Gal I
1
1
1
1,2
1,2
Andrew P. Montgomery, Christopher Dobie, Rémi Szabo, Laura Hallam, Marie Ranson,
1
,2
1,2
Haibo Yu, Danielle Skropeta*
1
. Molecular Horizons and School of Chemistry & Molecular Bioscience, Faculty of Science, Medicine and Health,
University of Wollongong, Wollongong NSW 2522, Australia.
2
. Illawarra Health and Medical Research Institute, Wollongong, NSW 2522, Australia.
1

ABSTRACT
Sialic acid at the terminus of cell surface glycoconjugates is a critical element in cell-cell
recognition, receptor binding and immune responses. Sialyltransferases (ST), the enzymes
responsible for the biosynthesis of sialylated glycans are highly upregulated in cancer and the
resulting hypersialylation of the tumour cell surface correlates strongly with tumour growth,
metastasis and drug resistance. Inhibitors of human STs, in particular human ST6Gal I, are thus
expected to be valuable chemical tools for the discovery of novel anticancer drugs. Herein, we
report on the computationally-guided design and development of uridine-based inhibitors that
replace the charged phosphodiester linker of known ST inhibitors with a neutral carbamate to
improve pharmacokinetic properties and synthetic accessibility. A series of 24 carbamate-linked
uridyl-based compounds were synthesised by coupling aryl and hetaryl α-hydroxyphosphonates
with a 5′-amino-5′-deoxyuridine fragment. The inhibitory activities of the newly synthesised
compounds against recombinant human ST6Gal I were determined using a luminescent microplate
assay, and five promising inhibitors with K ’s ranging from 1-20 µM were identified. These results
i
show that carbamate-linked uridyl-based compounds are a potential new class of readily
accessible, non-cytotoxic ST inhibitors to be further explored.
2

INTRODUCTION
Sialic acid (N-acetylneuraminic acid, Neu5Ac) is one of the human body’s most important
1
sugars next to glucose. These negatively charged nine-carbon α-keto aldonic acids are located at
the terminal end of glycan chains on cell surface and secreted molecules, where they play essential
2
, 3
roles in cellular biology. These functions are mainly related to cellular and molecular recognition
events, such as activation or inhibition of intracellular and intramolecular interactions, cell-cell
recognition, receptor binding, protein-lectin interactions, protein targeting, cell adhesion, and
immune responses.4-7 The biosynthesis of sialic acid containing glycoconjugates in humans is
mediated by 20 sialyltransferases (STs) anchored within the Golgi apparatus’ membrane with the
catalytic domain present within the lumen.4-5 Sialylation catalysed by STs uses the sugar
nucleotide donor CMP-Neu5Ac and an oligosaccharide or glycoconjugate terminated by a
galactose (Gal), N-acetylgalactosamine (GalNAc), or another sialic acid residue as the acceptor.4,
8
Upregulation of ST activity and the resulting modified cell surface sialylation is strongly
9
associated with cancer, with hypersialylation of 30–50% observed in several cancers. This has
been directly correlated with an increased metastatic potential of tumours, facilitation of apoptotic
avoidance mechanisms and poor patient prognosis.10, 11 Hypersialylation as a result of ST
upregulation is also linked to chemo-resistance and thus reduced treatment efficacy in ovarian,12-
1
4
colorectal,15 and cervical cancer,16 hepatocellular carcinoma,17 pancreatic ductal
1
8
19
adenocarcinoma and myeloid leukaemia. Exposure to radiation induces an increased expression
2
0
of human ST6Gal I (hST6Gal I) and reduces radiation-induced cell death in colon cancer. Thus,
the critical role of STs in tumour growth, progression and resistance to both radio- and
3

chemotherapy demonstrates the importance of developing small molecule modulators of this
family of enzymes as tools to help understand a potential new anti-cancer drug target.
A number of ST inhibitors have been reported ranging from those isolated from natural sources
(e.g. lithocholic acid and soyasaponin) and high throughput screening to those specifically
9
designed to mimic ST substrates as reviewed recently. Of these, analogues mimicking the
proposed oxocarbenium ion-like transition state of the donor (Figure 1A) pioneered by RR
Schmidt, exhibit the highest affinity to STs.21, 22 The most potent are those incorporating a
2
-deoxy-2,3-didehydro-N-acetylneuraminic acid (Neu5Ac2en) moiety, where the C2–C3 double
bond mimics the planar anomeric carbon; and an extra carbon between the anomeric carbon and
CMP leaving group mimics the elongated distance in the transition state.21, 23 Of the various
modifications, substituting the glycerol side chain of the Neu5Ac2en moiety with a phenoxy group
gave derivative 1 with a K value of 29 nM against rat liver ST6Gal in a HPLC-based assay (Figure
i
2
4
1
B). More synthetically accessible derivatives that replace the Neu5Ac2en with an aryl moiety
have been developed.21, 22, 25 For example, the (R)-isomer of the 3-phenoxy derivative 2 (Figure
2
1
1
B) is a potent inhibitor of both rat liver ST6Gals (K = 70 nM) and recombinant human ST6Gal I
i
2
6
26
27
(
K = 19 nM), along with cyclopentyl (3) and amide (4) derivatives with K values of 28 nM
i
i
and 16 nM respectively against hST6Gal I, also in a HPLC-based assay. It has also been
demonstrated that the size of the aryl substituent does not impact binding to human ST6Gal I, with
fluorescein-labelled derivatives such as (R)-5 and (S)-5, exhibiting K values of 22 nM and 9 nM,
d
2
8
respectively, against human ST6Gal I. These fluorescein-labelled derivatives have since been
reported as important high-throughput in vivo screening tools to assist in the development of
inhibitors of human oligo- and polysialyltransferases.29
4

A)
NH2
N
NH2
N
O
O
O
O
CMP2-
HO
OH
OH
N
O
HO
OH
OH
-
N
O
P

P
O
ST
O
O
O
sialylated
glycan
O
CO2
+
O
O
CO2
R-OH
O
AcHN
AcHN
OH OH
OH OH
+
HO
HO
O H
R
CMP-Neu5Ac
-
N
NH
CMP-Neu5Ac
B)
NH2
N
NH2
N
O
O
O
O
N
O
N
O
P
O
P
O
O
O
O
O
O
O
O
H
H
O
OH OH
O
OH OH
AcHN
P
HO
P
O
O
HO
HO
1
, K (rat liver -2,6-ST) = 29 nM
2, K (rat liver -2,6-ST) = 70 nM
i
i
K (hST6Gal I) = 19 nM
i
NH2
N
NH2
N
O
O
O
O
N
O
N
O
P
P
O
O
BnO
O
O
O
O
O
H
O
N
AcHN
P
OH OH
P
OH OH
O
O
HO
, K (hST6Gal I) = 28 nM
O
HO
3
4, K (hST6Gal I) = 16 nM
i
i
NH2
N
O
O
OH
O
O
N
O
P
HO C
O
2
O
O
H
N
O
P
OH OH
O
O
HO
(
(
R)-5, K (hST6Gal I) = 22 nM
d
S)-5, K (hST6Gal I) = 9 nM
d
Figure 1. (A) General mechanism of sialylation catalysed by sialyltransferases. (B) Reported transition-state analogue
inhibitors 1–4 active against rat and human ST6Gal enzymes as determined using a HPLC-based assay21, 24, 26, 27 and
reported fluorescent cell-permeable chemical probe 5.28
5

The charged phosphodiester linkage present in the majority of known ST inhibitors (e.g. Figure
B) is considered essential for activity.21, 22, 25, 30 Yet it may also lead to low bioavailability in vivo
1
and loss of activity due to cleavage by phosphatases or instability in the ST active site.31, 32 The
charged nature of the linker may also lead to poor cellular permeability, although recent reports of
phosphodiester derivatives bearing a fluorescent probe (such as 5, Figure 1B) have suggested these
charged compounds may be able to cross cellular membrane via a vesicular uptake mechanism.28
In addition to the potential pharmacokinetic issues, the synthesis of the phosphodiester-linked
inhibitors utilises a capricious condensation of an α-hydroxyphosphonate and an highly
air-sensitive cytidine phosphitamide.21, 26, 27
To improve synthetic accessibility to ST inhibitors, we have replaced the phosphodiester linker
with an uncharged carbamate, which can be synthesised using a wide range of alkoxy
3
3
carbonylating agents. Structurally, the carbamate functionality is related to an amide-ester hybrid
and in general displays very good chemical and proteolytic stability. Their capability to permeate
3
3
cell membranes has resulted in carbamates being widely utilised as peptide bond isosteres, which
would also provide advantages in this case. More specifically a carbamate moiety would have
similar hydrogen bonding capabilities to a phosphodiester group and maintain the three-atom
distance between the nucleoside (Figure 2) and the Neu5Ac mimic observed in reported
phosphodiester-linked ST inhibitors such as 1–4 (Figure 1B).
We have demonstrated previously using docking and molecular dynamics (MD) simulations that
carbamate- and 1,2,3-triazole-linked derivatives have comparable interactions to their
phosphodiester-linked counterparts with the hST6Gal I active site (PDB ID: 4JS2).34, 35 Using free
energy perturbation (FEP) calculations we have shown that these compounds can successfully
mimic the charged phosphodiester linkage with a comparable binding affinity through an
6

3
6
enthalpy-entropy compensation. We have also explored replacing the cytidine moiety with
uridine, which has the advantage of requiring less protecting groups enabling inhibitors to be
produced more readily and in fewer steps. In addition, as CMP-Neu5Ac is the common natural
donor for all ST subtypes, the replacement of the cytidine moiety with uridine could be a hitherto
unexplored route to selectivity.
ionic interaction
and/or
strong HBA
O
O
P
O
O
N
H
O
O
3-atom
3-atom
distance
distance
Figure 2. The possible binding interactions of a carbamate and phosphodiester linker. White arrows represent
hydrogen bond acceptors and black arrows represent hydrogen bond donors.
Herein, the phosphodiester linker of classical ST inhibitors has been replaced with a carbamate,
along with uridine in place of cytidine and the effects investigated both computationally and
experimentally. Additionally, the difference of diastereomers and various substituents of the aryl
sialic acid mimic, particularly at the 3-position, were also explored. The target compounds were
prepared in 7 steps from two protected building blocks: the 5′-amino-5′-deoxynucleoside and an
α-hydroxyphosphonate. The inhibitory activity of the newly synthesised compounds was then
evaluated against recombinant hST6Gal I in a luminescence microplate assay, along with cellular
toxicity assessment in a pancreatic cancer cell line.
RESULTS AND DISCUSSION
1
. Free Energy Calculations. Building on previous computational studies demonstrating the
feasibility of using carbamates and 1,2,3-triazoles as phosphodiester isosteres,34-37 we have
7

performed FEP calculations, which are one of the most rigorous calculation methods available to
compare the relative binding affinity of cytidine- and uridine-based carbamate inhibitors with
hST6Gal I (PDB ID: 4JS2).38, 39 Alchemical transformation of the cytidine-based inhibitor (R)-6
to the uridine-based inhibitor (R)-7 in complex with hST6Gal I and in solution was performed
(
Figure 3).40, 41 Combining the data from all simulations, the ΔΔG value observed for this
b
transformation was −1.2 ± 0.3 kcal/mol (see Supporting Information, Table S1 and S2). As state-
of-the-art computational studies report a root-mean-square-deviation of ~1.0 kcal/mol for the
3
9
relative binding free energies when comparing to experimental data, our results suggests that
replacing the cytidine in carbamate-linked inhibitors with a uridine would not have a significant
impact on binding to hST6Gal I, with uridine potentially being slightly preferential.
NH2
N
O
O
O
NH
O
O
Gb = -1.2 ± 0.3 kcal/mol
O
O
HN
O
N
O
HN
O
N
O
O
O
P
P
HO O
HO O
OH OH
OH OH
(
R)-6
(R)-7
Figure 3. The perturbation of cytidine-based (R)-6 to uridine-based (R)-7 performed during FEP calculation of ΔΔG
b
to hST6Gal I (PDB ID: 4JS2).
To rationalise the result observed during the FEP calculation an additional 15 ns of MD
simulation was performed for uridine-based inhibitor (R)-7. From these trajectories the hydrogen
bonds and hydrophobic contacts of the uridine component and the binding pocket were analysed
3
6
and compared to those previously reported for the cytidine component of (R)-6 (Figure 4). This
comparison showed that a majority of interactions were maintained between the two nucleoside
components in support of the FEP calculations and suggesting that hST6Gal I can accommodate
uridine-based derivatives. Based on the outcomes of these calculations we prepared a series of
uridine-based compounds (see Table 1).
8

A)
B)
Val352
Val352
Cys364
Cys353
Thr365
Ala190
Tyr354
NH2
Thr365
O
N
Tyr354
N
NH
O
Lys376
Leu372
N
O
O
O
Leu372
Phe211
Ser188
OH OH
OH OH
Phe211
Phe343
Gly324
Ser189
Ser188
Ser322
Figure 4. Comparison of interactions with hST6Gal I between the (A) cytidine component of (R)-636 and the (B)
uridine component of (R)-7. Consistent interactions were observed between the remaining components of (R)-6 and
(R)-7. Interacting hST6Gal I residues (PDB ID: 4JS2) are represented as circles which are colour coded based on
interaction type as follows: Red = hydrogen bond donor; Green = hydrogen bond acceptor; Blue = water bridged
hydrogen bond; Orange = hydrogen bond donor and acceptor; Grey = hydrophobic contact. Dashed lines indicate
hydrophobic contacts and dotted lines represent the hydrogen bonds shown in Supplementary Table 3 and 4.
2
. Synthesis of Sialic Acid Mimic. The sialic acid mimics for the target compounds synthesised
in this work are α-hydroxyphosphonates that are derived from commercially available aldehydes.
The benzyl protecting group was selected for the phosphonate as it can be easily removed under
catalytic hydrogenation conditions. Based on reported methods,23, 42 the aldehydes were reacted
with dibenzyl phosphite and triethylamine in dichloromethane (CH Cl ) to give the requisite
2
2
α-hydroxyphosphonates (8–19) in good to excellent yields of 51–92% as racemic mixtures
(Scheme 1). As the diastereoisomeric final products were readily separable by RP-HPLC, racemic
mixtures of the α-hydroxyphosphonates were used in the following synthetic steps.21, 26, 27
9

Scheme 1. Synthesis of the α-hydroxyphosphonates 8–19.a
OH
O
a
O
R
P
R
H
OBn
OBn
R:
PhO
PhO
F
F
PrO
MeO
F
8
, 60%
9, 92%
10, 82%
11, 68%
12, 75%
13, 67%
cPeO
cPeO
S
S
MeO
MeO
14, 60%
15, 51%
16, 61%
17, 88%
18, 65%
19, 52%
aReagents and conditions: (a) Dibenzyl phosphite, triethylamine, CH
Cl , rt, overnight.
2 2
3
. Synthesis of the Nucleoside Fragment. The nucleoside fragments were prepared from
cytidine and uridine. Several protecting groups were trialled with the allyloxycarbonyl (Alloc)
protecting group favourable for uridine, and benzyloxycarbonyl (Cbz) preferred for cytidine to
facilitate global deprotection. Thus, Alloc protected uridine derivatives were prepared by the initial
4
3
5
′-selective installation of an azide via a Mitsunobu reaction to give 20. This was followed by
the reaction with allyl chloroformate to give the Alloc protected 5′-azidouridine 21 that was then
reduced to give the protected 5′-amino-5′-deoxyuridine 22 in 61% overall yield (Scheme 2).
Scheme 2. Synthesis of the Alloc protected 5′-amino-5′-deoxyuridine 22.a
O
N
O
N
O
N
O
N
NH
NH
NH
NH
a
b
c
HO
O
N3
O
N3
O
H N
2
O
O
O
O
O
OH OH
uridine
OH OH
AllocO
OAlloc
AllocO
OAlloc
21
22
20
aReagents and conditions: (a) Triphenylphosphine, DIAD, HN
pyridine, 20ºC–rt, 30 min, 90%; (c) (i) triphenylphosphine, THF, rt, 2 h; (ii) H
, THF, 0ºC–rt, 17 h, 97%; (b) allyl chloroformate,
O, reflux, 6 h, 70% over two steps.
3
2
1
0

The Cbz protected 5′-hydroxycytidine derivative 23 was synthesised from cytidine as reported.44
The N/O-protected compound 23 was subjected to azidonation under Mitsunobu conditions to
yield 24, then reduced to produce the protected 5′-amino-5′-deoxycytidine fragment 25 in 33%
overall yield in five steps from cytidine (Scheme 3).
Scheme 3. Synthesis of Cbz protected 5′-amino-5′-deoxycytidine 25.a
NHCbz
N
NHCbz
N
NHCbz
N
a
b
HO
N
O
N3
N
O
H2N
N
O
O
O
O
CbzO
OCbz
CbzO
OCbz
CbzO
OCbz
23
24
25
aReagents and conditions: (a) Triphenylphosphine, DIAD, HN
triphenylphosphine, THF, rt, 2 h; (ii) H O, reflux, 5 h, 50% over two steps.
, THF, 0ºC–rt, 18 h, 97%; (b) (i)
3
2
4
. Synthesis of Target Molecules. Coupling of the α-hydroxyphosphonate building blocks 8–
9 and the 5′-amino-derivatives 22 or 25 was performed using 4-nitrophenyl chloroformate
4-NPC) and catalytic DMAP (Schemes 4 and 5). The protected carbamate coupled derivatives
1
(
were obtained as a 1:1 mixture of diastereoisomers in 22–73% yield (Table 1). The fully protected
uridine-based derivatives 26a–l required two deprotection steps (Scheme 4), both performed in
excellent yields (Table 1). The first deprotection step involved removal of the Alloc protecting
groups using Pd(PPh ) , with dimedone as an allyl scavenger. The partially deprotected derivatives
3
4
2
7a–l were subjected to hydrogenation (10% Pd/C and one atmosphere of H ) to remove the benzyl
2
protecting groups to produce the crude mixture of target uridine-based compounds 28a–l.
1
1

Scheme 4. Synthesis of uridine-based target compounds 28a–l.a, b
O
O
N
O
NH
NH
OH
O
a
O
H N
HN
O
O
2
N
O
+
R
O
R
P
O
OBn
OBn
P
BnO OBn
AllocO
OAlloc
AllocO
OAlloc
8-19
22
26a-l
b
O
O
N
O
O
NH
O
NH
2
Na
O
O
c
HN
O
N
HN
O
O
R
O
R
O
P
P
O O
BnO OBn
OH OH
8a-l-s, 28a-l-l
OH OH
2
27a-l
aReagents and conditions: (a) (i) Compounds 8–19, 4-NPC, triethylamine, DMAP, CH
Compound 22, rt, 24 h; (b) Pd(PPh , dimedone, THF, rt, 30 min; (c) (i) 10% Pd/C, 1 atm. H , rt, 1 h; (ii) RP-HPLC;
iii) IR 120 Na . Refer to Table 1 for R group identity.
2
2
Cl , 0ºC–rt, 3 h; (ii)
3
)
4
2
+
b
(
The Cbz and benzyl protected cytidine-based derivative 29 underwent a global deprotection (10%
Pd/C/H ) to produce a crude mixture of the target cytidine-based compound 30 (Scheme 5). The
2
1
:1 diastereomeric mixture of compounds 28a–l and 30 were separated by preparative RP-HPLC,
converted to their corresponding sodium salt via ion exchange, and lyophilised to produce the
target compounds in excellent yields (Table 1). The separated diastereomers are designated as (s)
and (l), representing the shorter and longer HPLC retention times respectively, as the chemical
shift differences between the diastereomers were too small to use predictive computational
methods to assign the stereoisomers.45, 46 All final compounds were stable in their salt form at
room temperature in a sealed vial for over a year with negligible loss of activity, and stable for up
to 6 months in solution at -18°C.
1
2

Scheme 5. Synthesis of cytidine-based target compounds 30-(s) and 30-(l).a
NHCbz
N
NHCbz
N
O
OH
PhO
O
a
O
H N
N
O
HN
O
N
O
2
PhO
+
P
O
O
OBn
OBn
P
BnO OBn
CbzO
OCbz
CbzO
OCbz
8
25
29
b
NH2
N
O
2
Na
PhO
O
HN
O
N
O
O
P
O O
OH OH
30-s, 30-l
a
2 2
Reagents and conditions: (a) (i) Compound 8, 4-NPC, triethylamine, DMAP, CH Cl , 0ºC–rt, 3 h; (ii) Compound
+
2
2
5, rt, 24 h; (b) (i) 10% Pd/C, 1 atm. H , rt, 1 h; (ii) RP-HPLC; (iii) IR 120 Na .
5
. Enzymatic Assay. The majority of ST inhibition data reported to-date has been obtained using
a HPLC-based assay that relies on quantification of the transferred sialic acid onto a UV-labelled
2
5
acceptor and reported in terms of their K inhibitory constant values. In recent years, more rapid
i
4
7
and sensitive microplate assays have also been developed. Herein, the ST inhibitory activity was
determined using the CMP-Glo™ luminescence microplate assay from Promega as detailed by
4
8
Das et al. for the uridine-based compounds 28a–l and cytidine-based compound 30 against
recombinant hST6Gal I (aa 44–406) from R&D Systems. CMP-Neu5Ac was used as the sialic
acid donor and N-acetyllactosamine (LacNAc) used as the acceptor. CMP produced as a
by-product of the ST reaction was detected using a luciferase-based reaction with a linear response
to the concentration of CMP. The luminescence produced was detected using a POLARstar Omega
plate reader. With acceptor concentration set at 1 mM, the K for CMP-Neu5Ac was calculated to
m
be 37.2 ± 5.4 μM (see Supporting Information, Figure S15), which was comparable to reported
1
3

data.26, 49, 50 The inhibition rates of all compounds at 10 μM (with 1 mM of acceptor and 100 μM
CMP-Neu5Ac) was initially obtained (Table 1), and for those compounds with over 30%
inhibition, K values were also determined (Table 2). As shown in Table 1 all compounds tested
i
were active at 10 μM, with the Lineweaver–Burk double reciprocal plots suggesting a mixed mode
of inhibition as reported previously for ST inhibitors (see Supporting Information, Figure S16–
S22).9, 49 As the vast majority of ST inhibitors are reported in terms of their K inhibition constants,
i
the same measure has been used herein for comparison purposes. Future studies of the most
promising compounds using labelled substrates would also allow determination of the Kd
dissociation constants.
1
4

Table 1. Yields for the preparation of the carbamate-linked uridine-based compounds 28a–l and the cytidine-based
compound 30 and inhibition data against recombinant hST6Gal I. a
Coupling
Product
Inhibition at 10
R
Alloc Deprotection
Cbz/Bn Deprotectionb
μM
O
3
0-(s): 89%
0-(l): 92%
28a-(s): 90%
8a-(l): 92%
2.2%
2
9: 23%
-
3
6.3%
O
O
19.9%
2
6a: 49%
27a: 81%
27b: 84%
27c: 76%
2
28.2%
1.1%
28b-(s): 96%
28b-(l): 97%
2
6b: 38%
F
1.8%
2
8c-(s): 80%
8c-(l): 83%
34.6%
58.9%
2
6c: 40%
F
2
F
2
8d-(s): 72%
36.6%
18.0%
7.2%
2
6d: 22%
27d: 68%
27e: 91%
27f: 87%
27g: 85%
27h: 71%
27i: 92%
27j: 80%
27k: 97%
2
8d-(l): 70%
8e-(s): 83%
O
2
2
6e: 36%
2
8e-(l): 93%
8f-(s): 94%
30.9%
29.8%
23.6%
18.1%
7.1%
O
2
2
6f: 38%
2
8f-(l): 82%
8g-(s): 85%
8g-(l): 88%
8h-(s): 93%
8h-(l): 81%
8i-(s): 87%
8i-(l): 84%
8j-(s): 90%
8j-(l): 81%
8k-(s): 86%
8k-(l): 88%
8l-(s): 68%
8l-(l): 70%
2
2
6g: 62%
O
2
O
O
2
12.7%
7.9%
2
6h: 51%
O
2
2
19.0%
78.0%
7.1%
2
6i: 38%
2
2
2
6j: 44%
2
15.2%
66.4%
2
2
6k: 36%
S
2
13.2%
2
5.9%
6.9%
2
6l: 73%
27l: 70%
2
S
a
b
For details of the procedures, see Supporting Information. Yield of sodium salts displayed.
1
5

Comparing the activity of the cytidine-based (30) and uridine-based (28a) inhibitors
demonstrates that substituting the nucleoside is tolerated and even beneficial. This validates the
FEP calculations discussed above (Figure 3) and provides a rationale for exploring other
alternatives to cytidine.21, 22, 25, 26 The most active inhibitors in this series have K values that range
i
from 1–20 μM, comparable to many of the reported phosphodiester-based ST inhibitors.26, 27
Table 2. Affinity of CMP Neu5Ac (K
m
) to recombinant hST6Gal I and inhibition dataa
O
O
NH
2
Na
O
HN
O
N
O
R
O
P
O O
OH OH
Compound
R
Inhibition at 10 μM
K
m
or K
i
(μM)
m i
K /K
CMP-Neu5Ac
-
-
37.2 ± 5.4
306 ± 132
-
2
8c-(s)
8c-(l)
8d-(s)
34.6%
58.9%
0.1
F
2
1.1± 0.1
32.5
F
F
2
36.6%
30.9%
19.0%
19.2 ± 2.1
1.9
1.8
0.7
O
2
8e-(l)
8i-(s)
8i-(l)
8k-(s)
20.3 ± 2.0
O
O
2
55.5 ± 26.9
2
78.0%
66.4%
11.5 ± 5.6
8.5 ± 1.0
3.2
4.4
2
S
aFor details of the procedures, see Supporting Information.
The importance of the nature and position of the substituent on the α-hydroxyphosphonate
fragment is seen by comparing the activity of the phenyl parent compound 28j (s: 7.1%; l: 15.2%)
to the various substituted derivatives from which several key trends emerged. For 3-substituted
derivatives the alkyl ethers were more active, with the 3-cyclopentoxy derivative 28i-l being one
1
6

of the most active inhibitors overall (K = 11.5 ± 5.6 μM). The 3-phenoxy derivatives 28a, along
i
with the 3-propoxy (28e) and 3-methoxy (28f) derivatives showed moderate activity of up to 31%
inhibition. When 3-substitution was coupled with a substituent in the 4-position a decrease in
activity resulted as seen by comparing the 3-phenoxy derivative 28a (s: 19.9%; l: 28.2%) to the 4-
fluoro-3-phenoxy derivative 28b (s: 1.1%; l: 1.8%) or the 3-cyclopentoxy derivative 28i (s: 19.0%;
l: 78.0%) to the 3-cyclopentoxy-4-methoxy derivative 28h (s: 12.7%; l: 7.9%). However, when
there was only the single substituent at the 4-position activity was maintained as seen by comparing
the 4-methoxy derivative 28g (s: 18.1%; l: 7.7%) to the 3-cyclopentoxy-4-methoxy derivative 28h
(s: 12.7%; l: 7.9%) or improved in the case of the 4-fluoro derivative 28c (s: 34.6%; l: 58.9%)
compared to the 4-fluoro-3-phenoxy derivative 28b (s: 1.1%; l: 1.8%). Thus, substituents at either
the 3- or 4-position are beneficial for ST inhibition in this series of compounds, but not when
substituted at both positions presumably for steric reasons. Overall, the 4-fluoro derivative 28c-l
was the most active compound with a K value of 1.1 ± 0.1 μM.
i
The second most promising compound was the 2-benzothiophene derivative 28k-s (K = 8.5 ±
i
1
.0 μM). To our knowledge heterocycles such as 28k have not been explored in the context of ST
inhibition and therefore warrant further investigation. The difference in activity between the 2-
and 3-benzothiophene derivatives 28k and 28l highlights the importance of regiochemistry for
these compounds. Regarding stereochemistry, for the five most active compounds there were
significant differences in activity between stereoisomers. For the benzothiophenyl derivative 28k,
the s-isomer was more active with 66.4% inhibition compared to 13.2% for the l-isomer. For the
two most active aryl compounds, the l-isomer of the cyclopentyl and p-fluoro-derivative 28i and
2
8c exhibited 78.0 and 58.9% inhibition respectively, compared to 19.0 and 34.6% for their
diastereomers. The same trend was observed in their K values with a 5-fold difference in activity
i
1
7

for 28i s- and l-isomers and up to 300-fold difference for the 28c s- and l-isomers. The large
difference activity is observed for the most active compounds (28c-l, 28d-s, 28e-l, 28i-l and 28k-
s), whereas all other compounds showed comparable activity between their diastereomers. It has
previously been found that stereochemistry does not play a major role for cytidine-based ST
9
inhibitors, therefore this is a particularly interesting and warrants further in-depth calculations or
structural studies such as crystallisation of the inhibitors with hST6Gal 1 to rationalise the findings.
All final compounds were also evaluated in the MTS cell viability assay on a MiaPaCa-2 cancer
cell line at 200 μM and observed to be non-toxic. This is expected as inhibiting sialylation affects
cell adhesion and migration rather than cell viability. All compounds were assessed in silico as
being free from any pan-assay interference compounds (PAINS) as described by Baell et al., using
a publicly available filter (http://zinc15.docking.org/patterns/home).51, 52 The ADME parameters
and PK properties were computed for the uridyl final compounds using SwissADME
(www.swissadme.ch) and compared to their phosphodiester/cytidine-containing counterparts,
indicating improved properties including reduced size, H-bond acceptors, lipophilicity, molar
refractivity and polarity, along with greater synthetic accessibility. Along with screening against
other ST enzyme subtypes (ST3Gal I and ST8Sia II) for selectivity, efforts are currently underway
to further improve PK properties and cell permeability of our most promising compounds using
phosphonate ester prodrugs. Phosphonate dibenzyl esters (as present in the partially protected
compounds 27a-l), have been explored by others as prodrugs but drug release is less efficient than
from other prodrugs such as pivaloyloxymethyl (POM) and isopropyloxycarbonlyoxymethyl
(
POC) derivatives, to be investigated in future work.53, 54
1
8

CONCLUSIONS
Herein, we examined the replacement of the phosphodiester linker and cytidine of classical ST
inhibitors26, 27 with a carbamate linker and uridine respectively. Additionally, the difference of
diastereomers and various substituents of the aryl sialic acid mimic were also explored. FEP
calculations demonstrated that for binding to hST6Gal I, uridine was a beneficial alternative to
cytidine. Thus, 26 carbamate-linked compounds were synthesised by coupling various
α-hydroxyphosphonates with 5′-aminonucleosides and evaluated against recombinant hST6Gal I
in a luminescence-based microplate assay, giving five compounds (28c-l, 28d-s, 28e-l, 28i-l and
2
2
8k-s) with Ki values in the 1–20 μM range. The identification of the benzothiophene inhibitor
8k-s as one of the best compounds of this series suggests that various heterocycles should be
further explored in the future, including against a wider panel of ST enzymes as recently reported
by Moremen and Jarvis, as well for cross-reactivity towards UDP-sugar glycosyltransferases due
to the introduction of uridine.55, 56
Overall, these results show that both the carbamate and uridine modifications are viable changes
to the classical model of ST inhibitors. The promising activity of this first generation of
carbamate-linked, uridyl-based inhibitors provide a strong starting point for the development of
more drug-like and selective ST inhibitors as chemical tools to fully uncover role of STs in cancer
progression, metastasis and drug resistance. The facile 7-step synthesis developed is highly
amenable to scale-up to provide larger quantities of pure material for both crystallisation studies
and cell-based and animal studies, which is important for the field. Further studies are underway
to generate an asymmetric route to the most active fluorinated and heterocyclic diastereomers to
be evaluated in cell-based models of ST upregulation in pancreatic, ovarian and hepatic cancer.
1
9

EXPERIMENTAL SECTION
Free Energy Calculation. The FEP calculation describing the alchemical transformation40, 41 of the
cytidine-based inhibitor (R)-6 to the uridine based inhibitor (R)-7 was prepared using VMD version
5
7
58
1
.9.2. and carried out using the NAMD 2.10 package. The change in the binding free energy with the
perturbation from cytidine to uridine (Figure 5) can be expressed as:
ΔΔ퐺 = Δ퐺 ― Δ퐺 퐶푏 푦푡
푈푟푖
푏
푏
=
=
Δ퐺 ― Δ퐺1
2
ΔΔ퐺푒푥
=
∑푖ΔΔ퐺푒x,푖
Eq. 1
In the FEP method, one introduces a hybrid Hamiltonian, ℋ(λ)=(1 – λ) ℋ + λℋ ,
0
1
where H represents the Hamiltonian for the initial state and H for the final state. The interval between
0
1
0
and 1 is divided into n subintervals (λ , where i = 1,…, n – 1), and for each subinterval the free energy
i
difference is calculated from the ensemble average, ΔΔGex,i = –k T ln<exp[–(ℋ(λ ) – ℋ(λ ))/k T> .
B
i+1
i
B
λi
The free energy difference between initial and final states is obtained from the sum, ΔΔG = Σ ΔΔG .
i
ex,i
Alchemical transformations were carried out with 20 uniformly distributed λ values between 0 and 1
0.0, 0.05, 0.1, … , 0.9, 0.95, 1.0), with the transformation simulated in both the forward (0 ⟶ 1) and
(
backward (1 ⟶ 0) directions. To prevent the occurrence of singularities at values of λ where an atom
2
disappears/appears, a soft-core potential with α = 4.0 Å was used to reshape the Lennard-Jones potential
into a form devoid of singularity when λ approaches either 0 or 1.59, 60

GbCyt
Cyt
E
+
E-Cyt

G=0
G1
G2
E
+
Ur
E-Ur

GbUr
2
0

b
Figure 5. The thermodynamic cycle used in the cytidine to uridine FEP simulations to calculate ΔΔG (see Eq. 1). Cyt
represents the (R)-diastereoisomer of the cytidine-based inhibitor (R)-5, Uri represents the (R)-diastereoisomer of the uridine-
based inhibitor (R)-6 and E represents hST6Gal I.
The starting structure for each λ window was generated from the hST6Gal I-inhibitor MD simulations
3
6
of the cytidine-based inhibitor (R)-6, described in our previous work (PDB ID: 4JS2). Each λ window
simulation for both the complex and ligand only systems were solvated in an 82 Å × 82 Å × 82 Å cubic
6
1
+
-
TIP3P water box. Each box was neutralised with counter ions of Na and Cl , with the salt (NaCl)
concentration set to 0.15 mol/L. The protein was represented with the non-polarisable CHARMM
6
2
PARAM36 force field. The force field used for the ligands were those optimised in our previous work
using the GAAMP method.36, 63 All systems were simulated in periodic boundary conditions using the
Langevin algorithm for maintaining the temperature at 298.15 K, and the Langevin Piston Nose–Hoover
method64, 65 for maintaining a constant pressure of 1.0 bar. The electrostatic interactions were calculated
6
6
using the Particle Mesh Ewald method. The van der Waals forces were treated with a cut-off of 12 Å
and a smoothening function between 10 and 12 Å. All of the covalent bonds involving hydrogen were
6
7
kept rigid with the RATTLE algorithm. The integration time step was set to 1.0 fs, 2.0 fs and 4.0 fs for
bonded, non-bonded and long-range electrostatics, respectively.
Each λ window underwent an initial 5000 energy minimisation steps followed by 1 ns equilibration
simulation to provide starting coordinates and then simulated for 2 ns with a harmonic restraint placed
on the centre of mass of hST6Gal I for the complex simulations. A harmonic restraint was also applied
to one atom from the ligand in bulk simulations. Both restraints used a force constant of 32 kcal/mol/Å2.
6
8
ΔΔG and the associated errors were calculated using the ParseFEP plugin in VMD, which was used
b
to combine the results of the forward and backward simulations with the Bennett acceptance-ratio.69
To examine the H-bond interactions between the uridine-based inhibitors and the hST6Gal I active site
the λ window 1.0 from the backwards simulation was extended to 15 ns. H-bond interactions were
2
1

analysed according to a geometric criterion. A H-bond was defined by a minimum donor–hydrogen–
acceptor angle of 120° and a maximum heavy atom distance of 3 Å. Hydrophobic contacts were also
analysed according to a geometric criterion of a maximum distance of 4 Å between hydrophobic atom
5
7
pairs. To identify these interactions VMD version 1.9.2. was utilised.
General Chemistry. All solvents and reagents were purchased from Sigma Aldrich (USA),
Carbosynth (UK) or ChemSupply (SA) and used as supplied or purified by standard procedures. Unless
otherwise noted all reactions were performed under anhydrous conditions. Reactions were monitored by
TLC using Merck Silica Gel 60 F254 aluminium backed plates and components visualised by UV light or
staining. Solvents were removed under reduced pressure at a max. of 40ºC (water bath). Column
1
13
31
chromatography was performed under ‘flash’ conditions on silica gel 60 (40–63 μm mesh). H, C, P,
19F and 2D NMR spectra were acquired on a Varian VNMRS PS54 500 MHz and Inova 300 MHz, or
1
13
Bruker Advance Neo 500 MHz and 400 MHz NMR instruments. H and C NMR chemical shifts (in
3
1
ppm) were calibrated against residual solvent peaks or TMS; P NMR shifts against an external reference
1
9
of 85% H PO in D O; and F shifts against an external reference of 0.05% α,α,α-trifluorotoluene in
3
4
2
CDCl . High resolution electrospray ionisation mass spectra (HRMS) were performed on a Waters
3
XEVO G2 Q-TOF spectrometer with leucine enkephalin (LeuEnk) as an internal standard. Infrared
spectra were obtained on either a Shimadzu FTIR Affinity or Bruker Vortex 70 FTIR spectrophotometer.
Optical rotation was determined using a Jasco P-2000 polarimeter. Melting points were obtained using a
Büchi Melting Point M-560 apparatus and are uncorrected. Analytical RP-HPLC was performed with a
Shimadzu Prominence-I LC-2030C system with a PDA detector (190–800 nm), using a Luna C18 (2)
1
00Å (Phenomenex, 3 μm, 4.6 x 150 mm) or Shim-Pack GIS (Shimadzu, 5 μm, 4.6 x 150 mm) column.
Preparative RP-HPLC was performed on a Shimadzu Prominence LC-20AP system with a PDA detector
(190–800 nm) using a Prep C18 (Shimadzu, 5 μm, 20 x 150 mm) column. The purity of all final
1
compounds was determined to be ≥ 95% based on H NMR data and analytical HPLC (see Supporting
Information, Figures S1–S13).
2
2

Synthesis of α-Hydroxyphosphonates. The synthesis of all α-hydroxyphosphonates was based on the
4
2
method of Wong et al. To a mixture of dibenzyl phosphite (1.1 equiv.) and the desired benzaldehyde
(
1 equiv.) in CH Cl , was added triethylamine (2 equiv.). The reaction mixture was stirred at rt for 14 h
2
2
and monitored by TLC. Upon completion, the reaction mixture was diluted with CH Cl , washed with
2
2
saturated NaHCO solution and brine. The organic phase was separated, dried with anhydrous MgSO ,
3
4
filtered and the filtrate evaporated, with the resultant product purified by column chromatography.
Dibenzyl α-hydroxy(3-phenoxy)benzylphosphonate (8). From 3-phenoxybenzaldehyde (300 mg,
1
6
.51 mmol): purification by column chromatography (EtOAc : hexane, 1:1, R = 0.43) to give 8 (417 mg,
f
1
0%) as a white solid. mp 65.0–68.0ºC. H NMR (500 MHz, CDCl ) δ 7.24–7.31 (m, 12H), 7.19–7.22
3
(m, 3H), 7.13 (br s, 1H), 7.08 (t, J = 7.5 Hz, 1H), 6.95 (d, J = 8.5 Hz, 2H), 5.02 (d, JH,P = 10.0 Hz, 1H),
4
.89–5.00 (m, 4H). The spectral data match those reported.26
Dibenzyl α-hydroxy(4-fluoro-3-phenoxy)benzylphosphonate (9). From 4-fluoro-3-phenoxyben-zaldeh
yde (500 mg, 2.31 mmol): purification by column chromatography (EtOAc : pet. spirit, 2:3, R = 0.14)
f
1
to give 9 (1.11 g, 92%) as a white solid. mp 71.0–73.0ºC. H NMR (500 MHz, CDCl ) δ 7.04–7.30 (m,
3
1
3
1
6H), 6.90 (d, J = 8.0 Hz, 2H), 4.90–5.00 (m, 5H). C NMR (125 MHz, CDCl ) δ 157.3, 154.2 (dd, J
3 C,F
=
247.9 Hz, JC,P = 3.8 Hz), 143.7 (d, JC,F = 12.4 Hz), 136.0 (d, JC,P = 6.3 Hz), 133.05–133.16 (m), 129.8,
1
7
28.8, 128.72, 128.69, 128.12, 128.06, 123.5–123.6 (m), 123.4, 120.7–120.8 (m), 117.5, 117.1, 116.9,
3
1
0.5 (d, JC,P = 158.4 Hz), 68.8–68.9 (m), 68.7 (d, JC,P = 6.6 Hz). P NMR (162 MHz, CDCl ) δ 22.54 (d,
3
1
9
JP,F = 6.5 Hz). F NMR (235.3 MHz, CDCl ) δ −68.81 (d, J = 2.4 Hz). HRMS (ESI) calcd for
3
F,P
+
C H FO P [M + H] , 479.1424; Found 479.1425. IR (FTIR) ν 3307, 1505, 1489, 1208, 991, 740,
2
7
25
5
max
7
24, 698.
Dibenzyl α-hydroxy(4-fluoro)benzylphosphonate (10). From 4-fluorobenzaldehyde (1.00 g,
8
.06 mmol): purification by column chromatography (EtOAc : pet. spirit, 7:3, R = 0.55) to give 10
f
1
(
2.54 g, 82%) as a white solid. mp 102.0–104.0ºC. H NMR (500 MHz, CDCl ) δ 7.37–7.40 (m, 2H),
3
1
3
7
.18–7.28 (m, 10H), 6.94–6.97 (m, 2H), 5.04 (d, JH,P = 10.0 Hz, 1H), 4.88–5.00 (m, 4H). C NMR (125
2
3

MHz, CDCl ) δ 162.7 (dd, J = 245.3 Hz, JC,P = 3.8 Hz), 136.12 (d, JC,P = 3.4 Hz), 136.07 (d,
3
C,F
JC,P = 3.4 Hz), 132.29–132.33 (m), 129.1 (dd, JC,P = 8.4 Hz, JC,F = 5.5 Hz), 128.7, 128.63, 128.58, 128.5,
1
6
28.04, 127.99, 115.3 (dd, JC,F = 21.3 Hz, JC,P = 2.8 Hz), 70.5 (d, JC,P = 159.1 Hz), 68.8 (d, JC,P = 7.1 Hz),
3
1
19
8.5 (d, JC,P = 7.4 Hz). P NMR (162 MHz, CDCl ) δ 22.74 (d, J = 4.9 Hz). F NMR (235.3 MHz,
3
P,F
+
CDCl ) δ −50.94 (d, J = 2.4 Hz). HRMS (ESI) calcd for C H O PF [M + H] , 387.1161; Found
3
F,P
21 21
4
3
87.1160. IR (FTIR) νmax 3230, 1601, 1506, 1220, 101, 732, 695.
Dibenzyl α-hydroxy(3-fluoro)benzylphosphonate (11). From 3-fluorobenzaldehyde (500 mg, 4.03
mmol): purification by column chromatography (EtOAc : pet. spirit, 1:1 R = 0.44) to give 11 (1.06 g,
f
1
6
8%) as a white solid. mp 78.0–79.0ºC. H NMR (400 MHz, CDCl ) δ 7.16–7.33 (m, 13H), 6.96–7.01
3
1
3
(
m, 1H), 5.05 (d, JH,P = 12.0 Hz, 1H), 4.91–5.02 (m, 4H). C NMR (100 MHz, CDCl ) δ 163.3 (dd,
3
JC,F = 244.0 Hz, JC,P = 3.0 Hz), 139.1 (dd, JC,F = 8.0 Hz, JC,P = 3.0 Hz), 136.03 (d, JC,P = 3.0 Hz), 135.97
d, JC,P = 3.0 Hz), 129.9 (dd, JC,F = 8.0 Hz, JC,P = 3.0 Hz), 128.73, 128.71, 128.69, 128.67, 128.2, 128.1,
22.9 (dd, JC,P = 6.0 Hz, JC,F = 3.0 Hz), 115.2 (dd, JC,F = 21.0 Hz, JC,P = 3.0 Hz), 114.3 (dd,
JC,F = 23.0 Hz, JC,P = 6.0 Hz), 70.7 (dd, JC,P = 158.0 Hz, JC,F = 2.0 Hz), 68.9 (d, JC,P = 7.3 Hz), 68.7 (d,
(
1
3
1
19
JC,P = 7.3 Hz). P NMR (162 MHz, CDCl ) δ 22.43 (d, J = 8.1 Hz). F NMR (235.3 MHz, CDCl ) δ
3
P,F
3
+
−
49.90 (d, JF,P = 23.5 Hz). HRMS (ESI) calcd for C H O FP [M + H] , 387.1161; Found 387.1161. IR
21 21 4
(FTIR) νmax 3231, 1243, 1224, 1202, 734, 690.
Dibenzyl α-hydroxy(3-propoxy)benzylphosphonate (12). From 3-propoxybenzaldehyde (300 mg,
1
.82 mmol): purification by column chromatography (EtOAc : hexane, 1:1, R = 0.21) to give 12
f
1
(
587 mg, 75%) as a white solid. mp 56.0–57.5ºC. H NMR (500 MHz, CDCl ) δ 7.21–7.32 (m, 11H),
3
7
6
.01 (br s, 2H), 6.85 (d, J = 8.5 Hz, 1H), 5.03 (d, JH,P = 10.0 Hz, 1H), 4.87–4.98 (m, 4H), 3.81 (td, J =
1
3
.5 Hz, J = 1.5 Hz, 2H), 3.17 (br s, 1H), 1.76 (sxt, J = 7.1 Hz, 2H), 1.00 (t, J = 7.4 Hz, 3H). C NMR
(
75 MHz, CDCl ) δ 159.1 (d, JC,P = 3.8 Hz), 137.6 (d, JC,P = 2.3 Hz), 136.0–136.1 (m), 129.2 (d, JC,P =
3
2
1
.3 Hz), 128.5, 128.4, 128.33, 128.30, 127.90, 127.87, 119.3 (d, JC,P = 6.0 Hz), 114.9 (d, JC,P = 2.3 Hz),
12.7 (d, JC,P = 5.3 Hz), 69.3, 71.1 (d, JC,P = 158.3 Hz), 68.6 (d, JC,P = 8.3 Hz), 68.3 (d, JC,P = 8.3 Hz),
2
4

3
1
+
2
4
2.5, 10.5. P NMR (162 MHz, CDCl ) δ 23.00. HRMS (ESI) calcd for C H O PNa [M + Na] ,
3
24 27
5
49.1494; Found 449.1515. IR (FTIR) νmax 3430, 2978, 1232, 999, 729, 692.
Dibenzyl α-hydroxy(3-methoxy)benzylphosphonate (13). From 3-methoxybenzaldehyde (300 mg, 2.22
mmol): purification by column chromatography (EtOAc : hexane, 2:3, R = 0.45) to give 13 (584 mg,
f
6
1
7%) as a white solid. mp 77.5–80.0ºC (lit. mp 76–78ºC).70 1H NMR (500 MHz, CDCl ) δ 7.21–7.31 (m,
3
1H), 7.01–7.03 (m, 2H), 6.81 (d, J = 8.0 Hz, 1H), 5.04 (d, JH,P = 11.0 Hz, 1H), 4.97–4.84 (m, 4H), 3.64
(
s, 3H). The spectral data match those reported.70
Dibenzyl α-hydroxy(4-methoxy)benzylphosphonate (14). From 4-methoxybenzaldehyde (500 mg, 3.67
mmol): purification by column chromatography (EtOAc : hexane, 1:1, R = 0.34) to give 14 (884 mg,
f
6
0%) as a white solid. mp 103.0–105.0ºC (lit. mp 104–105ºC).70 1H NMR (300 MHz, CDCl ) δ 7.48 (d,
3
J = 7.3 Hz, 2H), 7.31 (br s, 10H), 6.86 (d, J = 8.5 Hz, 2H), 5.87 (br s, 1H), 5.16 (d, JH,P = 10.5 Hz, 1H),
4
.88–5.03 (m, 4H), 3.73 (s, 3H). The spectral data match those reported.70
Dibenzyl α-hydroxy(3-cyclopentoxy-4-methoxy)benzylphosphonate (15). From 3-cyclopentoxy-4-met
hoxybenzaldehyde (700 mg, 3.18 mmol): purification by column chromatography (EtOAc : pet. spirit,
1
1
:1, R = 0.23) to give 15 (789 mg, 51%) as a white solid. mp 116.5–119.2ºC. H NMR (400 MHz,
f
CDCl ) δ 7.27–7.32 (m, 8H), 7.20–7.22 (m, 2H), 7.01 (d, J = 8.0 Hz, 1H), 6.95 (d, J = 8.0 Hz, 1H), 6.80
3
(d, JH,P = 8.0 Hz, 1H), 4.85–4.99 (m, 5H), 4.61–4.62 (m, 1H), 3.83 (s, 3H), 1.75–1.81 (m, 6H), 1.54 (br
1
3
s, 2H). C NMR (100 MHz, CDCl ) δ 150.2 (d, J = 3.0 Hz), 147.8 (d, JC,P = 2.0 Hz), 136.2–136.3 (m),
3
C,P
1
28.68, 128.63, 128.52, 128.55, 128.11, 128.06, 119.8 (d, JC,P = 8.0 Hz), 113.8–113.9 (m), 111.8 (d,
3
1
JC,P = 2.0 Hz), 80.4, 71.1 (d, JC,P = 159.0 Hz), 68.5–68.6 (m), 56.2, 32.92, 32.87, 24.2. P NMR (162
+
MHz, CDCl ) δ 22.20. HRMS (ESI) calcd for C H O PNa [M + Na] , 505.1756; Found 505.1769. IR
3
27 31
6
(FTIR) νmax 3309, 2962, 1515, 1233, 989, 738, 698.
Dibenzyl α-hydroxy(3-cyclopentoxy)benzylphosphonate (16). From 3-cyclopentoxybenzaldeh-yde
(
973 mg, 5.11 mmol): purification by column chromatography (EtOAc : pet. spirit, 3:2, R = 0.30) to give
f
1
1
6 (1.42 g, 61%) as a white solid. mp 55.7–59.3ºC. H NMR (500 MHz, CDCl ) δ 7.15–7.27 (m, 11H),
3
2
5

6
1
.98–7.01 (m, 2H), 6.78 (d, J = 8.2 Hz, 1H), 4.85–5.07 (m, 5H), 4.59–4.62 (m, 1H), 1.68–1.81 (m, 6H),
.49–1.56 (m, 2H). 13C NMR (125 MHz, CDCl ) δ 158.1 (d, JC,P = 2.4 Hz), 137.9, 136.22 (d,
3
JC,P = 1.5 Hz), 136.18 (d, JC,P = 1.5 Hz), 129.3 (d, JC,P = 2.0 Hz), 128.51, 128.47, 128.4, 128.3, 128.0,
27.9, 119.2 (d, JC,P = 5.9 Hz), 115.8 (d, JC,P = 2.8 Hz), 114.1 (d, JC,P = 5.7 Hz), 79.1, 71.1 (d,
1
3
1
JC,P = 159.0 Hz), 68.7 (d, JC,P = 6.8 Hz), 68.4 (d, JC,P = 7.3 Hz), 32.83, 32.78, 24.1. P NMR (162 MHz,
+
CDCl ) δ 23.31. HRMS (ESI) calcd for C H O PNa [M + Na] , 475.1650; Found 475.1629. IR (FTIR)
3
26 29
5
νmax 3290, 2965, 1599, 1255, 989, 736, 694.
Dibenzyl α-hydroxybenzylphosphonate (17). From benzaldehyde (500 mg, 4.72 mmol): purification by
column chromatography (EtOAc : hexane, 1:1, R = 0.24) to give 17 (1.53 g, 88%) as a white solid. mp
f
1
1
15.0–116.0ºC (lit. mp 117–118ºC).70 1H NMR (300 MHz, CDCl ) δ 7.20–7.46 (m, 15H), 5.04 (d, JH,P
=
3
0.5 Hz, 1H), 4.84–4.96 (m, 4H). The spectral data match those reported.70
Dibenzyl α-hydroxy(benzo[b]thiophen-2-yl)methylphosphonate (18). From benzo[b]thiophene-2-carb
oxaldehyde (1.00 g, 6.16 mmol): recrystallisation from EtOAc gave 18 (1.67 g, 65%) as a white
1
crystalline solid. mp 155.3–156.6ºC. H NMR (400 MHz, CDCl ) δ 7.78–7.80 (m, 1H), 7.66–7.68 (m,
3
1
3
1
H), 7.22–7.36 (m, 13H), 5.33 (d, JH,P = 12.0 Hz, 1H), 4.99–5.10 (m, 4H). C NMR (100 MHz, CDCl )
3
δ 140.0 (2 x s), 139.8 (2 x s), 139.5 (2 x s), 135.95–136.06 (m), 128.72, 128.70, 128.67, 128.3, 128.2,
24.59–124.60 (2 x s), 124.5, 123.89–123.91 (2 x s), 123.1–123.2 (2 x s), 122.5, 69.0–69.1 (m), 68.1 (d,
1
3
1
+
JC,P = 164.0 Hz). P NMR (162 MHz, CDCl ) δ 21.02. HRMS (ESI) calcd for C H O PSNa [M + Na] ,
3
23 21
4
4
47.0796; Found 447.0804. IR (FTIR) νmax 3189, 1457, 1200, 1006, 727, 695.
Dibenzyl α-hydroxy(benzo[b]thiophen-3-yl)methylphosphonate (19). From benzo[b]thiophene-3-carb
oxaldehyde (250 mg, 1.54 mmol); recrystallisation from EtOAc gave 19 (339 mg, 52%), as a white
1
crystalline solid. mp 117.7–110.3ºC. H NMR (400 MHz, CDCl ) δ 7.82–7.85 (m, 2H), 7.67 (d, J = 3.4
3
1
3
Hz, 1H), 7.22–7.36 (m, 10H), 7.11–7.13 (m, 2H) 5.44 (d, JH,P = 10.6 Hz, 1H), 4.80–5.07 (m, 4H). C
NMR (100 MHz, CDCl ) δ 140.5, 137.4 (d, J = 6.1 Hz), 136.1 (d, JC,P = 5.8 Hz), 135.9 (d,
3
C,P
JC,P = 5.8 Hz), 131.4, 128.71, 128.65, 128.63, 128.59, 128.2, 128.1, 125.9 (d, JC,P = 7.5 Hz), 124.8,
2
6

3
1
1
24.4, 122.8, 122.7, 68.82 (d, JC,P = 3.3 Hz), 68.75 (d, JC,P = 3.3 Hz), 66.6 (d, JC,P = 164.1 Hz). P NMR
+
(
162 MHz, CDCl ) δ 22.66. HRMS (ESI) calcd for C H O PSNa [M + Na] , 447.0796; Found
3
23 21
4
4
47.0808. IR (FTIR) νmax 3274, 1497, 1254, 989, 764, 696.
Synthesis of Nucleoside Fragments.
5
′-Azido-5′-deoxyuridine (20). A solution of hydrazoic acid (HN , ~2.5 M) was prepared by the slow
3
addition of sulfuric acid (4.50 mL) to a suspension of NaN (10.6 g, 164 mmol) in H O (5.30 mL) and
3
2
toluene (64.8 mL) at 0ºC. The mixture was stirred for 30 min, at which time Na SO was observed as a
2
4
precipitate. The toluene layer was decanted, and used immediately in the following reaction. Uridine
(
2.00 g, 8.19 mmol) was suspended in dry THF (80 mL) under an atmosphere of N . Triphenylphosphine
2
(
6.44 g, 24.6 mmol) and the freshly prepared HN (49 mL, 123 mmol) were added to the suspension,
3
which had been cooled to 0ºC, followed by the dropwise addition of diisopropyl azodicarboxylate
DIAD, 5.20 mL, 24.6 mmol). Subsequently the reaction mixture was allowed to warm to rt and stir for
7 h. After this time the reaction mixture was concentrated under reduced pressure. The resulting residue
was dissolved in minimal amount of CH Cl and extracted into H O. After washing with CH Cl the
(
1
2
2
2
2
2
crude product was obtained by concentrating the aqueous phase, with the aid of a toluene azeotrope. The
crude product was purified via column chromatography (CH Cl :methanol (MeOH), 9:1, R = 0.44) to
2
2
f
afford 20 (2.11 g, 97%) as a white foam. mp 151.2–152.8ºC, (lit. mp 151.5–152.5ºC).71 1H NMR (500
MHz, MeOH-d ) δ 7.71 (d, J = 8.0 Hz, 1H), 5.84 (d, J = 4.0 Hz, 1H), 5.74 (d, J = 8.0 Hz, 1H), 4.22 (t, J
4
=
5.6 Hz, 1H), 4.10 (t, J = 5.6 Hz, 1H), 4.05–4.08 (m, 1H), 3.69 (dd, J = 13.5 Hz, J = 3.0 Hz, 1H), 3.60
(
dd, J = 13.5 Hz, J = 4.5 Hz, 1H). The spectral data match those reported.43
2
′,3′-O-Bis(allyloxycarbonyl)-5′-azido-5′-deoxyuridine (21). Allyl chloroformate (9.81 g, 82.0 mmol)
was added dropwise over a 10 min interval to a solution of 20 (2.20 g, 8.20 mmol) in pyridine (13.1 mL,
64 mmol), which had been cooled to −20ºC. The reaction mixture was allowed to warm to rt and was
1
stirred for 30 min. Volatiles were removed under reduced pressure and the residue was dissolved in
CH Cl . The product was partitioned between EtOAc and H O. The organic layer was washed with brine
2
2
2
2
7

and was then dried with MgSO and concentrated under reduced pressure. The resulting crude product
4
was purified by column chromatography (CH Cl :EtOAc, 3:2, R = 0.75) to give 21 (3.23 g, 90%) as an
2
2
f
off-white gum. mp 43.2–46.8ºC. [α]D25 +30.6 (c = 1.5, CHCl ). H NMR (500 MHz, CDCl ) δ 9.20 (br
1
3
3
s, 1H), 7.42 (d, J = 8.0 Hz, 1H), 5.97 (d, J = 5.0 Hz, 1H), 5.86–5.96 (m, 2H), 5.81 (d, J = 8.0 Hz, 1H),
5
3
1
.26–5.40 (m, 6H), 4.59–4.68 (m, 4H), 4.30 (d, J = 2.5 Hz, 1H), 3.78 (dd, J = 13.5 Hz, J = 2.5 Hz, 1H),
1
3
.67 (dd, J = 13.5 Hz, J = 4.0 Hz, 1H). C NMR (125 MHz, CDCl ) δ 162.9, 153.9, 153.8, 150.2, 140.4,
3
31.01, 130.96, 119.70, 119.68, 103.7, 88.6, 80.2, 75.6, 73.6, 69.6, 69.5, 51.7. HRMS (ESI) calcd for
+
C H N O Na [M + Na] , 460.1075; Found 460.1078. IR (FTIR) ν 2107, 1750, 1688, 1232, 1080,
1
7
19
5
9
max
9
08.
′,3′-O-Bis(allyloxycarbonyl)-5′-amino-5′-deoxyuridine (22). To a solution of 21 (3.23 g, 7.39 mmol)
2
in dry THF, triphenylphosphine(3.88 g, 14.8 mmol) was added with the resulting solution allowed to stir
at rt. After 2 h a few drops H O was added and the solution was heated at reflux for 6 h. Upon completion
2
volatiles were removed under reduced pressure and the subsequent crude purified by column
chromatography (CH Cl : MeOH, 9:1, R = 0.33) to yield 22 (2.13 g, 70%) as a white solid. mp 48.2–
2
2
f
9.5ºC. [α]D25 −4.0 (c = 1.0, MeOH). H NMR (500 MHz, DMSO-d ) δ 7.93 (d, J = 8.0 Hz, 1H), 5.86–
1
4
5
4
6
.98 (m, 3H), 5.69 (d, J = 8.0 Hz, 1H), 5.49 (t, J = 5.0 Hz, 1H), 5.24–5.37 (m, 5H), 4.56–4.69 (m, 4H),
.11–4.14 (m, 1H), 2.98 (dd, J = 15.0 Hz, J = 3.5 Hz, 1H), 2.92 (dd, J = 15.0 Hz, J = 3.5 Hz, 1H).
1
3
C NMR (125 MHz, DMSO-d ) δ 162.9, 153.3, 153.1, 150.5, 142.1, 131.75, 131.63, 118.7, 102.4, 87.4,
6
+
8
4
1.3, 75.1, 73.9, 68.7, 68.5, 41.8. HRMS (ESI) calcd for C H N O [M + H] , 412.1356; Found
1
7
21
3
9
12.1351. IR (FTIR) νmax 2107, 1750, 1688, 1232, 1080, 902.
4
2
′,3′-O-N -Tris(benzyloxycarbonyl)-5′-deoxycytidine (23). Cytidine (1.00 g, 4.12 mmol) and imidazole
(700 mg, 10.3 mmol) were dissolved in a minimum of DMF. Once dissolved tert-butyldimethylsilyl
(TBDMS) chloride (682 mg, 4.52 mmol) was slowly added. The resulting mixture was stirred at rt under
an atmosphere of N and judged to be complete by TLC after 16 h. Upon completion the reaction solution
2
was poured onto crushed ice with NaCl added, with the subsequent precipitate collected via vacuum
2
8

filtration. The crude 5′-O-TBDMS-protected cytidine and DMAP (1.38 g, 12.3 mmol) were dissolve in
CH Cl and the resulting suspension was cooled to 0ºC. Benzyl chloroformate (1.8 mL, 12.3 mmol) was
2
2
added dropwise and the suspension stirred at rt. After 16 h the solution was diluted with CH Cl and
2
2
washed with H O, followed by NaHCO and brine. The organic phase was collected, dried with MgSO
4
2
3
and concentrated under reduced pressure. The resulting crude product was purified by column
chromatography (CHCl :EtOAc, 4:1, R = 0.58) to produce the completely protected cytidine derivative
3
f
(2.18 g, 70%) as a clear solid. The clear solid (1.96 g, 2.58 mmol) was dissolved in THF (15 mL) and
glacial acetic acid (0.2 mL, 3.10 mmol) was added. The reaction mixture was cooled to 0ºC and 1.0 M
TBAF in THF (2.8 mL, 2.84 mmol) was added in a dropwise manner. The resulting reaction mixture was
allowed to gradually warm to rt over 2 h and left to stir at rt for an additional 16 h. After such time, the
volatile reactants were removed under reduced pressure with the resulting crude product was purified by
column chromatography (CHCl : EtOAc, 4:1, R = 0.26) to produce 23 (1.60 g, 96%) as a white solid.
3
f
1
mp 96.0–97.1ºC. H NMR (500 MHz, CDCl ) δ 8.28 (br s, 1H), 8.07 (d, J = 7.0 Hz, 1H), 7.26–7.37 (m,
3
1
5
6H), 5.94 (d, J = 4.6 Hz, 1H), 5.73 (t, J = 4.6 Hz, 1H), 5.52 (t, J = 5.0 Hz, 1H), 5.13–5.18 (m, 2H),
.04–5.11 (m, 4H), 4.28 (br s, 1H), 4.20 (br s, 1H), 3.97 (d, J = 12.5 Hz, 1H), 3.81 (d, J = 12.5 Hz, 1H).
Spectral data match those reported.44
4
5
′-Azido-2′,3′-O-N -tris(benzyloxycarbonyl)-5′-deoxycytidine (24). 23 (1.60 g, 2.48 mmol) was
suspended in dry THF (50 mL), with triphenylphosphine (1.95 g, 7.43 mmol) and the freshly prepared
HN (14.9 mL, 37.1 mmol; see 20 for details) were added to the suspension, which had been cooled to
3
0
ºC. To the cooled suspension, DIAD (1.57 mL, 7.43 mmol) was added slowly in a dropwise manner.
Subsequently the reaction mixture was allowed to warm to rt and stirred for 18 h. After this time, the
reaction mixture was concentrated under reduced pressure and the crude product was purified via column
chromatography (CH Cl : EtOAc, 4:1, R = 0.68) to afford 24 (1.62 g, 97%) as a white foam. mp 56.5–
2
2
f
9.2ºC. [α]D25 +13.2 (c = 1.0, CHCl ). H NMR (500 MHz, CDCl ) δ 8.00 (br s, 1H), 7.84 (d, J = 7.5 Hz,
1
5
1
3
3
H), 7.28–7.37 (m, 16H), 5.93 (d, J = 2.5 Hz, 1H), 5.51–5.52 (m, 1H), 5.34 (t, J = 5.5 Hz, 1H), 5.20 (s,
2
9