2-Substituted α,β-Methylene-ADP Derivatives: Potent Competitive Ecto-5′-nucleotidase (CD73) Inhibitors with Variable Binding Modes

Article abstract of DOI:10.1021/acs.jmedchem.9b01611

CD73 inhibitors are promising drugs for the (immuno)therapy of cancer. Here, we present the synthesis, structure-activity relationships, and cocrystal structures of novel derivatives of the competitive CD73 inhibitor α,β-methylene-ADP (AOPCP) substituted in the 2-position. Small polar or lipophilic residues increased potency, 2-iodo- and 2-chloro-adenosine-5′-O-[(phosphonomethyl)phosphonic acid] (15, 16) being the most potent inhibitors with K_i values toward human CD73 of 3-6 nM. Subject to the size and nature of the 2-substituent, variable binding modes were observed by X-ray crystallography. Depending on the binding mode, large species differences were found, e.g., 2-piperazinyl-AOPCP (21) was >12-fold less potent against rat CD73 compared to human CD73. This study shows that high CD73 inhibitory potency can be achieved by simply introducing a small substituent into the 2-position of AOPCP without the necessity of additional bulky N⁶-substituents. Moreover, it provides valuable insights into the binding modes of competitive CD73 inhibitors, representing an excellent basis for drug development.

Full text of DOI:10.1021/acs.jmedchem.9b01611

pubs.acs.org/jmc
Article
2‑Substituted α,β-Methylene-ADP Derivatives: Potent Competitive
Ecto-5′-nucleotidase (CD73) Inhibitors with Variable Binding Modes
Sanjay Bhattarai, Jan Pippel, Emma Scaletti, Riham Idris, Marianne Freundlieb, Georg Rolshoven,
Christian Renn, Sang-Yong Lee, Aliaa Abdelrahman, Herbert Zimmermann, Ali El-Tayeb,
̈
̈
Christa E. Muller,* and Norbert Strater*
Cite This: J. Med. Chem. 2020, 63, 2941−2957
Read Online
sı
*
Metrics & More
Article Recommendations
Supporting Information
ACCESS
ABSTRACT: CD73 inhibitors are promising drugs for the (immuno)therapy of cancer. Here, we present the synthesis, structure−
activity relationships, and cocrystal structures of novel derivatives of the competitive CD73 inhibitor α,β-methylene-ADP (AOPCP)
substituted in the 2-position. Small polar or lipophilic residues increased potency, 2-iodo- and 2-chloro-adenosine-5′-O-
[(phosphonomethyl)phosphonic acid] (15, 16) being the most potent inhibitors with Ki values toward human CD73 of 3−6 nM.
Subject to the size and nature of the 2-substituent, variable binding modes were observed by X-ray crystallography. Depending on
the binding mode, large species differences were found, e.g., 2-piperazinyl-AOPCP (21) was >12-fold less potent against rat CD73
compared to human CD73. This study shows that high CD73 inhibitory potency can be achieved by simply introducing a small
substituent into the 2-position of AOPCP without the necessity of additional bulky N6-substituents. Moreover, it provides valuable
insights into the binding modes of competitive CD73 inhibitors, representing an excellent basis for drug development.
ions activate the enzyme. The C- and N-terminal domains are
connected by a hinge region, which allows the enzyme to
undergo large domain reorientations and to switch between an
open and a closed state.6 In the closed conformation, the
substrate is buried between the N-terminal and the C-terminal
domains with the terminal phosphate group in close proximity
to the catalytically active zinc ions.8 It is likely that hydrolysis
only occurs in the closed conformation.
Human CD73 is expressed on lymphocytes, endothelial, and
epithelial cells, and its expression is elevated during stress
responses, e.g., under hypoxic or inflammatory conditions.9−11
Upregulation of human CD73 has been further reported for
INTRODUCTION
■
The main players in purinergic signaling are the nucleotides
ATP and ADP and the nucleoside adenosine. Nucleotides can
activate two subfamilies of membrane receptors: G protein-
coupled P2Y receptors and ATP-gated P2X ion channels.1
Adenosine acts as a potent anti-inflammatory and immuno-
suppressive agent by activating the G protein-coupled
adenosine receptor subtypes A2A and A2B that are expressed
on immune cells.2 Moreover, many cancer cells upregulate the
expression of adenosine A2B receptors, whose activation leads
to increased proliferation, angiogenesis, and metastasis.3 The
extracellular concentration of adenosine is controlled by ecto-
5′-nucleotidase (CD73), which hydrolyzes AMP to adenosine
(see Figure 1).4,5 Human CD73 is a glycosylphosphatidylino-
sitol (GPI)-anchored cell surface protein, which can also be
cleaved off and act as a soluble enzyme in the extracellular
fluid.5−7 Mammalian CD73 functions as a homodimer. Each
monomer contains a dimetal active center, and the natural
enzyme probably contains Zn2+ but also other divalent metal
Received: September 27, 2019
Published: February 11, 2020
https://dx.doi.org/10.1021/acs.jmedchem.9b01611
© 2020 American Chemical Society
J. Med. Chem. 2020, 63, 2941−2957
2941

Journal of Medicinal Chemistry
pubs.acs.org/jmc
Article
Figure 1. Catalysis of the natural substrate AMP by CD73.
Figure 2. Structures of selected nucleotide-derived CD73 inhibitors.
various types of cancer, where it impairs antitumor responses
and enhances tumor growth and metastasis by producing a
microenvironment that contains high extracellular adenosine
levels.12,13 In agreement with these observations, a number of
studies have demonstrated that targeted inhibition or
disruption of human CD73, e.g., by knockout, siRNAs, small
molecules, or antibodies, provides immune defense against
cancers and inhibits tumor growth.14−16
Recently, monoclonal antibodies for CD73 have been
moved forward into clinical trials, e.g., for treating pancreatic
cancer,17 and most recently, the development of the first small
molecule clinical candidate (AB680, I) for blocking CD73 was
announced and its characterization was described.18
Small molecule inhibitors may have significant advantages
compared to antibodies, which typically show only partial
enzyme inhibition and cannot be expected to penetrate well
into solid tumors. Despite their great therapeutic potential,
only a few CD73 inhibitors have been reported to date.
Currently, the most promising class of CD73 inhibitors are
analogues of nucleoside-5′-diphosphates, such as the ADP
analogue α,β-methylene-ADP (AOPCP, II), its derivatives
III19 and IV,20 and the uracil nucleotide analogue V21 (see
Figure 2), which act as competitive inhibitors blocking the
substrate binding site.19−22 All non-nucleotide-derived CD73
inhibitors published so far are only weakly potent, e.g.,
anthraquinone, sulfonamide, and polyphenol derivatives.22−28
Recently, we published the structure-guided development of
a subnanomolar CD73 inhibitor displaying high selectivity and
metabolic stability.20 In our previously published crystal
structures of human CD73 in complex with II (AOPCP,
PDB 4H2I)8 and III (PSB-12379, PDB 6S7F),20 a pocket with
polar as well as hydrophobic features close to the 2-position of
the adenine moiety was observed. We found that a chloro
substitution in this position led to a large increase in potency
(see nucleotide analogue IV, PSB12489). In the present study,
we synthesized a series of 2-substituted AOPCP derivatives to
probe the size of the pocket and to explore the flexibility of the
orthosteric CD73 binding site. The selection of 2-substituents
was not only based on size but also on polarity/lipophilicity
and structural diversity. For selected inhibitors, we determined
cocrystal structures with human CD73 to support structure−
activity relationship (SAR) analysis and to assess the quality of
earlier molecular modeling and docking approaches.19,21,23,24
Surprisingly, different binding modes were observed depending
on the nature and size of the 2-substituent. These results
provide an excellent basis for further drug design approaches
aimed at the development of potent CD73 inhibitors.
2942
https://dx.doi.org/10.1021/acs.jmedchem.9b01611
J. Med. Chem. 2020, 63, 2941−2957

Journal of Medicinal Chemistry
pubs.acs.org/jmc
Article
Scheme 1. Synthesis of 2-Amino- and 2-Oxo-Substituted Purine Riboside-5′-O-[(phosphonomethyl)phosphonic acid]
a
Derivatives
a
Reagents and conditions: (a) acetic anhydride, DMAP (4-dimethylaminopyridine), EDMA (ethyl dimethylamine), 40 °C, 1 h, yield 97%; (b)
POCl3, N,N-dimethylaniline, tetraethylammonium chloride, reflux, 110 °C, 15 min, yield 85%; (c) 2% NaOMe in methanol, rt, 24 h, yield 80%; (d)
ammonia in ethanol, rt, 24 h, yield 75%; (e) acetic acid, sodium nitrite, H2O, 2 h, rt, yield 95%; (f) two steps: (i) methylenebis(phosphonic
dichloride), trimethyl phosphate, 4 °C, 40 min, (ii) TEAC (triethylammonium hydrogencarbonate) buffer pH 7.4−7.6, rt, 15 min, yield 35−60%.
For details see Supporting Information.
AOPCP derivatives but lower than those for 6-substituted
AOPCP derivatives.19 To introduce an iodo-substitution, 2-
amino-6-chloro-2′,3′,5′-O-acetyl-guanosine (4) was reacted
with copper(I) iodide and diiodomethane yielding 6-chloro-
2-iodo-2′,3′,5′-O-acetylpurine riboside (11), followed by
deprotection with ethanolic ammonia providing 2-iodoadeno-
sine (9) in a yield of 75%.32 This iodination method
substantially decreases the reaction time as compared to
standard procedures and obviates the need for tedious HPLC
purification. Similarly, 4 was diazotized with benzyltriethylam-
monium nitrate in the presence of acetyl chloride to give 2,6-
dichloro-2′,3′,5′-O-acetylpurine riboside 13, followed by
deprotection to provide 2-chloroadenosine (14) in a yield of
80%.20 2-Iodoadenosine (13) and 2-chloroadenosine (14)
were subsequently phosphonylated, providing the final
nucleotide analogues 15 and 16 in overall yield of 38−42%
(Scheme 2). 2-Hydrazinyladenosine (17) was prepared from
2-chloroadenosine (14) by reaction with hydrazine hydrate
according to a published procedure in a high yield of 97%,
simply by separating the product from impurities by
precipitation through the addition of water.33 Similarly, 14
was also reacted with N-tert-butyloxycarbonyl- (Boc-)
protected piperazine followed by deprotection, yielding 2-(N-
piperazinyl)adenosine (19) in a yield of 95%.34 2-Hydraziny-
ladenosine (17) and 2-(N-piperazinyl)adenosine (19) were
subsequently phosphonylated, providing the final nucleotide
RESULTS AND DISCUSSION
■
Chemistry. For the preparation of the target compounds, a
convergent synthetic strategy was applied which involved the
synthesis of the intermediate nucleosides in a multistep
procedure, followed by phosphonylation according to a
previously described procedure with some modifications (see
Schemes 1−3; for details, also see Supporting Information).19
The reaction started from commercially available guanosine
(1), which was acetylated by acetic anhydride to obtain
2′,3′,5′-O-acetylguanosine (2) in a high yield of 97%.
Subsequent chlorination of the 6-position gave 2-amino-6-
chloro-2′,3′,5′-O-acetylpurine riboside (3), which was depro-
tected to 2-amino-6-chloro-purine riboside (4) in a good yield
of 80%, comparable to reported methods.29,30 Treatment of 4
with ammonia in ethanol resulted in nucleophilic substitution,
affording 2,6-diaminopurine riboside (5). Nucleoside 5 was
diazotized with sodium nitrite and acetic acid in water to give
isoguanosine (6) in a high yield of 95%.31 Finally, guanosine
(1), 2-amino-6-chloropurine riboside (4), 2,6-diaminopurine
riboside (5), and isoguanosine (6) were phosphonylated by
using methylenebis(phosphonic dichloride) followed by
hydrolysis with aqueous triethylammonium bicarbonate
(TEAC) solution to give the final nucleotides 7−10 in overall
yields of 35−60%. (Scheme 1). The yields of the
phosphonylation of 2-substituted-AOPCP derivatives were
generally higher compared to that observed with 8-substituted
2943
https://dx.doi.org/10.1021/acs.jmedchem.9b01611
J. Med. Chem. 2020, 63, 2941−2957

Journal of Medicinal Chemistry
pubs.acs.org/jmc
Article
Scheme 2. Synthesis of 2-Chloro-, 2-Iodo-, 2-Hydrazinyl-, and 2-Piperazinyl-Substituted Adenosine-5′-O-
a
[(phosphonomethyl)phosphonic acid] Derivatives
a
Reagents and conditions: (a) CuI, I2, CH2I2, isoamyl nitrite in THF, 80 °C, 2 h, yield 75%; (b) benzyltriethylammonium nitrite, acetyl chloride,
DCM, 2 h, yield 80%; (c) 2.0 M ammonia in ethanol, rt, 24 h, yield 75%; (d) two steps: (i) methylenebis(phosphonic dichloride), trimethyl
phosphate, 4 °C, 30 min, (ii) TEAC buffer pH 7.6, rt, 15 min, yield 40−65%; (e) 1-Boc-piperazine, triethylamine, ethanol, 140 °C, 24 h, yield 85%;
(f) 8% TFA, CH2Cl2, H2O, 4 h, rt, yield 95%; (g) hydrazine hydrate, 4 h, rt, yield 97%. For details, see Supporting Information.
Scheme 3. Synthesis of 2-Allylthio- and 2-Cyclohexylethylthio-Substituted Adenosine-5′-O-[(phosphonomethyl)phosphonic
a
acid] Derivatives
a
Reagents and conditions: (a) acetic acid, H2O2, rt, 4 h, yield 86%; (b) NaOH, 2 h, rt; (c) carbon disulfide, MeOH, H2O, autoclave, 980 kPa, 120
°C, 5 h, rt, yield 65%; (d) NaOH, alkyl bromide, 2 h, rt, yield 56−60%; (e) two steps: (i) methylenebis(phosphonic dichloride), trimethyl
phosphate, 4 °C, 30 min, (ii) TEAC buffer pH 7.6, rt, 15 min, yield 35−45%. For details, see Supporting Information.
2944
https://dx.doi.org/10.1021/acs.jmedchem.9b01611
J. Med. Chem. 2020, 63, 2941−2957

Journal of Medicinal Chemistry
pubs.acs.org/jmc
Article
Table 1. Inhibitory Potencies of 2-Substituted Purine Riboside-5′-O-[(phosphonomethyl)phosphonic acid] Derivatives against
a
Rat and Human CD73 in Comparison to Standard Inhibitor
rat soluble CD73 Ki
human soluble CD73 Ki
human CD73 in MDA-MB-231 cancer
a
a
a
compd
R1
R2
H
SEM (nM)
SEM (nM)
cells Ki SEM (nM)
II AOPCP
amino
197
5
88.4 4.019
207 6.020
7 (guanosine derivative)
see structure
above
see structure
above
1110 350
410 90
255 83
8
chloro
amino
amino
amino
see structure
above
iodo
chloro
hydrazinyl
N-piperazinyl
allylthio
268 21
90.6 7.3
326 42
23.6 1.3
38.8 4.5
32.2 2.7
35.7 9.4
9 (PSB-12676)
10 (isoguanosine derivative) see structure
b
b
nd
nd
(PSB-12690)
above
amino
amino
amino
amino
amino
amino
15
16
15.1 1.2
18.6 3.8
116 18
2290 240
65.7 5.6
56.6 8.3
3.59 0.82
5.93 1.64
15.5 1.2
184 27
61.9 13.4
46.9 10.2
3.33 1.12
4.57 1.20
15.4 0.9
238 26
65.4 9.1
49.1 15.1
20 (PSB-12646)
21 (PSB-12604)
28
29
cyclohexyl-
ethylthio
a
[3H]AMP (5 μM) was used a substrate; Km values 59, 17, and 14.8 μM, respectively, for purified recombinant soluble rat CD73, purified
b
recombinant soluble human CD73, and native membrane-anchored human CD73 (in MDA-MB-231 cell membrane preparations). nd, not
determined
analogues (20, 21) in overall yields of 33−40% (Scheme 2).
The preparation of 2-thioadenosine (25) started from
commercial adenosine (22), which was oxidized by hydrogen
peroxide in the presence of acetic acid yielding adenosine-N1-
oxide (13). Subsequent ring opening with sodium hydroxide
afforded carboximidamide intermediate 24, which was
thionated with carbon disulfide in an autoclave (pressure of
980 kPa) to give 25. Subsequent alkylation with allyl bromide,
or 2-cyclohexylethyl bromide, respectively, yielded the desired
2-alkylthio-substituted derivatives 26 and 27 in yields of 56−
60%,35 which were further phosphonylated to provide the final
nucleotide analogues 28, 29 in overall yields of 25−32%
(Scheme 3).
cells),39 which shows a high expression level of CD73.21
Results are summarized in Table 1 and selected concen-
tration−inhibition curves are shown in Figure 3.
The phosphonylation of nucleosides generally yielded
mixtures of products and side products. Therefore, careful
purification was required to obtain the target nucleotide
analogues as pure products. All 2-substituted nucleotide
analogues were purified by anion exchange chromatography,
followed by reverse-phase high performance liquid chromatog-
raphy (RP-HPLC).
Figure 3. Concentration−inhibition curves of selected potent
compounds in comparison to the lead structure AOPCP (II) at rat
CD73. Rat enzyme Km, 59 μM; AMP concentration, 5 μM. Data
points represent means
performed in triplicate.
SEM from three separate experiments
Pharmacological Evaluation. The inhibitory potency of
the nucleotide analogues was determined in a previously
described sensitive enzyme inhibition assay using [3H]AMP as
a substrate.36 Recombinant soluble rat CD73 was obtained by
expression in Spodoptera frugiperda (Sf9) insect cells as
previously described.37 For all compounds, full concentra-
tion−response curves were determined and Ki values were
calculated from the obtained IC50 using the Cheng−Prusoff
equation.38 The most potent compounds were also tested
against recombinant soluble human CD73 expressed in the
same system21 in order to study potential species differences.
In addition, we investigated the compounds with native
membrane-bound CD73 from membrane preparations of a
triple-negative human breast cancer cell line (MDA-MB-231
Structure−Activity Relationships. The lead structure
AOPCP (II), tested under the same conditions, showed a Ki
value of 197 nM against rat CD73. Substitution in the 2-
position with a small, polar amino group (2-amino-AOPCP, 9)
increased potency by 2-fold (Ki 90.6 nM). Replacing the 6-
amino group in 2-amino-AOPCP by an oxo-function
(guanosine analogue GOPCP, 7) led to a strong reduction
in potency (12-fold, Ki 1110 nM) compared to 9. This
indicated that a keto group at the purine C6-position was not
tolerated, possibly due to the N1-H,C6O lactam structure of
7, which abolishes the aromaticity of the pyrimidine ring. This
result correlates with the previously observed low potency of
2945
https://dx.doi.org/10.1021/acs.jmedchem.9b01611
J. Med. Chem. 2020, 63, 2941−2957

Journal of Medicinal Chemistry
pubs.acs.org/jmc
Article
Figure 4. Concentration−inhibition curves of selected inhibitors against (A) recombinant human CD73 and (B) native CD73 expressed in triple-
negative breast cancer cells (membrane preparations of MDA-MB-231 cells). AMP concentration, 5 μM. Results are plotted as the percentage of
maximal activity measured in the absence of inhibitors. Ki values are shown in Table 1.
the inosine derivative IOPCP, an analogue of 7 lacking the 2-
amino group (Ki 2830 nM),19 which further confirms that the
2-amino group (in 7) is beneficial. A 6-chloro substituent (8)
preserving the aromatic ring structure, in contrast to the oxo
substitution in 7, was better tolerated (Ki 268 nM, only a 3-fold
decrease compared to the analogous 2-amino-AOPCP, 9).
For the following step, the 6-amino group in AOPCP was
kept, whereas the 2-substituent was varied. Introduction of a 2-
oxo function (in iso-GOPCP, 10, Ki 326 nM) led to a slight
reduction in potency compared to the lead structure AOPCP
(II, 197 nM). The lactam tautomerism resulting in the loss of
aromaticity could be the reason for the moderate potency of
10, which contains a polar 2-substituent that was expected to
increase the potency of the compound. In contrast, all further
investigated 2-substituents (in 15, 16, 20, 28, 29) led to
increased potency of the AOPCP derivatives, with one
exception (21). Small polar and small lipophilic residues in
the 2-position of AOPCP were well tolerated and led to
significant increases in potency. The order of potency by rank
was as follows: 2-I (15, 15.1 nM) ≈ 2-Cl (16, 18.6 nM) > 2-
(cyclohexylethylthio) (29, 56.6 nM) ≈ 2-(allylthio) (28, 65.6
nM) > 2-NH-NH2 (20, 116 nM) > H (I, 197 nM) (Figure 3
and Table 1). Only the 2-(1-piperazinyl)-AOPCP (21)
showed significantly reduced potency (>12-fold, Ki 2290
nM). It can therefore be concluded that small polar or
lipophilic substituents in the 2-position of the adenine moiety
of AOPCP can drastically improve potency. For larger
substituents, the situation appears to depend on the flexibility
and size of the substituent as well as on its hydrogen-bonding
properties.
Rat and human CD73 are highly homologous sharing 89%
protein sequence identity.8 The substrate binding site is highly
conserved, having only three amino acid differences near the
active site (F500Y, N499S, and L389V, Supporting Informa-
tion, Figure S5). F500 in human CD73 is involved in
nucleobase stacking of AOPCP8 and is replaced by tyrosine
in rat CD73.
In previous studies, we observed that AOPCP and other
nucleotide analogues were similarly potent against human and
rat CD73; in some cases, they were somewhat more potent
against the human enzyme, however the differences were
always moderate.19,20 To investigate whether this was also the
case for various 2-substituted AOPCP derivatives, we further
tested the inhibitors against recombinant soluble human
CD73. In addition, we studied the same set of compounds
using human membrane-bound CD73, which was natively
expressed in a triple-negative breast cancer cell line (see Table
1). There were virtually no differences between the potencies
of the inhibitors toward human soluble and human membrane-
bound CD73 (Figure 4). However, several compounds were
more potent against human compared to the rat enzyme (up to
>12-fold). The most potent CD73 inhibitors of the present
series were 2-iodo-AOPCP (15, human membrane-bound
CD73, Ki 3.33 nM) and 2-chloro-AOPCP (16, Ki 4.57 nM).
This indicates that a bulky N6-substituent which is present in
the previously developed inhibitors III19 and IV20 is not
required for high potency. Low nanomolar inhibitors can be
obtained by simple modification such as 2-halogen-substitution
of N6-unsubstituted AOPCP, resulting in CD73 inhibitors with
molecular weights of less than 500 g/mol, while the previously
published potent CD73 inhibitors (e.g., III−V) display
significantly higher molecular weights of well above 500 g/
mol. The largest species differences were observed for 2-
amino-6-chloro-AOPCP (8, 6.3-fold), 2-hydrazinyl-AOPCP
(20, 7.5-fold), and 2-(N-piperazinyl)-AOPCP (21, 12.5-fold)
indicating that significant species differences may be found for
orthosteric CD73 inhibitors depending on the substitution
pattern. In contrast, 2-allylthio- and 2-cyclohexylthio-AOPCP
were almost equally potent at rat and human CD73.
Crystal Structures of 2-Substituted Derivatives
Bound to Human CD73. To explain the observed SARs
for 2-substituted AOPCP derivatives and to investigate the
flexibility of the binding pocket (in the following also
designated as “C2-pocket”) in regard to its diverse
accommodation of polar, lipophilic, small, and large 2-
substituents, we obtained cocrystal structures of human
CD73 with the following selected inhibitors: 9 (PSB-12676),
10 (PSB-12690), 20 (PSB-12646), and 21 (PSB-12604).
These compounds were cocrystallized in the closed state of the
enzyme.
In addition, we analyzed X-ray structures of unliganded
(PDB 6TVE) and AOPCP-bound (PDB 6TVG) CD73 in the
open state of the enzyme at resolutions of 1.05 Å and 1.53 Å,
respectively (Table 2). In contrast to the crystals of the closed
form, which usually diffract to around 2.5 Å resolution, these
high-resolution crystal structures allow for a detailed analysis of
water structure within the active site. The rearrangement and
displacement of water molecules upon ligand binding can play
an important role in understanding ligand affinity.40
A
comparison of the binding modes of AOPCP in the open
and closed enzyme forms shows that the ligands adopt very
similar binding modes (Supporting Information, Figure S7D)
2946
https://dx.doi.org/10.1021/acs.jmedchem.9b01611
J. Med. Chem. 2020, 63, 2941−2957

Journal of Medicinal Chemistry
pubs.acs.org/jmc
Article
2947
https://dx.doi.org/10.1021/acs.jmedchem.9b01611
J. Med. Chem. 2020, 63, 2941−2957

Journal of Medicinal Chemistry
pubs.acs.org/jmc
Article
2948
https://dx.doi.org/10.1021/acs.jmedchem.9b01611
J. Med. Chem. 2020, 63, 2941−2957

Journal of Medicinal Chemistry
pubs.acs.org/jmc
Article
Figure 5. Water structure of the C2-pocket in unliganded CD73 and in the CD73 × AOPCP complex (both in the open state). (A) Water
molecules in the pocket next to F500 and F417 as identified in the 1.05 Å structure of unliganded human CD73 in the open conformation. Only
fully occupied water binding sites or predominantly occupied water molecules of alternative binding sites are shown. The view is from the
nucleobase clamp into the pocket. Refer to Supporting Information, Figure S6 for further information. The carbon atoms of the protein residues are
depicted in different colors for the four regions lining the C2-pocket. (B) Change in the water structure upon binding of AOPCP to the open state.
Water molecules from the unliganded structure are shown in yellow, whereas waters from the ligand-bound complex are shown in red. The
unliganded structure (PDB 6TVE, gray carbons, partial transparency) is superimposed onto AOPCP-bound CD73 (PDB 6TVG).
Figure 6. Crystal structures of human CD73 in the closed conformation in complex with different 2-substituted AOPCP derivatives 10 (A), 20 (B),
9 (C), and 21 (D). The color scheme for the protein residues of (A−C) is the same as in Figure 5. (D) Residues of the N-terminal and C-terminal
domain are shown in blue and green, respectively. The inhibitors are colored cyan. For comparison, the binding mode of AOPCP (I, PDB 4H2I)8
is depicted in gray together with the water structure. Water molecules in the C2-pocket are not visible for cocrystal structures of 9, 10, and 21 due
to crystal quality. For the CD73 × 20 structure, fully occupied waters H and I are shown in red. Waters D and N (pink) are alternatively occupied
with water F (purple) due to steric overlap.
and water structure in the C2-pocket. Therefore, high-
resolution X-ray structures of CD73 in the open form are
also suitable for the characterization of the C2-pocket structure
in the inhibited closed-state structures bound with AOPCP
derivatives.
The unliganded C2-pocket and the neighboring nucleobase
clamp (F417 and F500) are filled by nine water molecules
(designated A−I, Figure 5A, and Supporting Information,
Figure S6). Upon binding of AOPCP, water binding sites B
and C are not occupied due to steric overlap with the adenine
ring (Figure 5B, and Supporting Information, Figure S7).
Waters A, D, I, F, and H stay close to their positions in the
unliganded structure and maintain their interactions with other
water molecules and protein residues. Water molecule A forms
a hydrogen bond to the N1 nitrogen of AOPCP. An interesting
structural feature upon ligand binding is observed for water E,
which moves ∼1.1 Å closer to the center of the C2-pocket
together with its coordinating protein residue N390. The
2949
https://dx.doi.org/10.1021/acs.jmedchem.9b01611
J. Med. Chem. 2020, 63, 2941−2957

Journal of Medicinal Chemistry
pubs.acs.org/jmc
Article
movement of this side chain is probably caused by the
formation of hydrogen bonds between the side chain nitrogen
and N3 of the adenine ring (3.3 Å) as well as O2′ of the ribose
ring (3.0 Å). In the ligand-bound structures, water E is H-
bonded to Oδ1 of N390 and water molecule D. A remarkable
feature is its close distance to the C2-atom of the adenine base
at 3.10 Å (PDB 4H2I, closed state) and 3.04 Å (PDB 6TVG,
open state). The C−H···O angle in the two structures
corresponds to 154°, in agreement with the existence of a
C−H···X weak hydrogen bond. As a result of this movement,
water binding site G is not occupied. Therefore, in the
presence of the adenine base, binding of water E is more
favorable than that of water G.
On the basis of the CD73 × AOPCP cocrystal structure
(PDB 4H2I), we assumed that polar substituents at the C2-
position of AOPCP might improve the inhibitory potency as
they were expected to be accommodated in the water-filled
C2-pocket formed by residues of the C-terminal domain of
CD73.8,20 The upper part of the pocket, in close proximity to
the nucleobase, is predominantly hydrophilic but also contains
some apolar patches formed by F417, F500, and G393. The
base of the pocket has an apolar character (Figure 5A, and
Supporting Information, Figure S6B). We now were interested
in investigating, how various 2-substituents, 2-amino (9), 2-oxo
(10), 2-hydrazinyl (20), and 2-(1-piperazinyl) (21), could be
accommodated in the small binding pocketadjacent to C2 in
the AOPCP-bound enzyme.
indicates that water binding sites H and I are fully occupied,
whereas alternate occupancy is observed for waters F and D.
Either water site F or water site D together with a new water
position (labeled “N” in Figure 6B) are occupied. Both
alternative water structures have unmatched H-bond donors or
acceptors; when water F is not present, D504 lacks a hydrogen-
bonding partner. When site N is not occupied, the terminal
nitrogen of the hydrazinyl group has only one H-bond partner.
As noted above, a relatively large difference is observed for the
Ki values of 20 toward human vs rat CD73. We assume that
the influence of the N499S substitution in rat CD73 on the
water structure of the C2-pocket is responsible for the different
inhibition parameters (Supporting Information, Figure S5B).
In contrast, for 9, the situation for the amino-substituent at
the C2-position is more similar to that of the oxo-substituent
in 10; the distance to the side chain of N390 is too long for a
hydrogen-bonding interaction (Figure 6C). Formation of a H-
bond with water D is likely (the water structure in the C2-
pocket could not be characterized due to weak diffraction of
the crystal). A remarkable characteristic of the nucleobase
binding mode is that the base is shifted partly toward F500 and
partly toward the C2-pocket. This shift is caused by F417,
which interacts with the adenine ring of 9 in a side-on T-
shaped orientation via two C−H···π interactions instead of the
usual sandwich binding mode observed so far for all other
AOPCP derivatives. We assume that the additional amino-
substituent is responsible for this change of the π−π
interaction mode.
In the case of the 2-oxo-substituent in 10 (Figure 5B), the
inhibitory potency is slightly lower compared to AOPCP. The
cocrystal structure shows that the substituent is 4.0 Å away
from the side chain nitrogen of N390 and also forms no
hydrogen-bonding interactions with other protein residues
(Figure 6A). The limited resolution obtained for the cocrystal
structure with 10 in combination with overall high B factors
does not allow for an analysis of the water structure of the C2-
pocket. However, it is obvious that water binding site E cannot
be occupied in the presence of any substituent at the C2 of
AOPCP due to steric overlap. A superposition of the CD73 ×
10 complex with the CD73 × AOPCP structure indicates that
the 2-oxo substituent could interact with water D (Figure 6A).
Thus, it is likely that the oxo-group can only form one
favorable hydrogen-bonding interaction with water molecule
D. The other lone electron pair of the oxygen either has no
well-positioned donor partner, or a water molecule is
positioned in a previously unobserved water binding site in
order to interact with the oxo-group, however, this water
molecule lacks a favorable low-energy hydrogen-bonding
environment. Interestingly, a shift of the nucleobase is
observed in the binding mode of 10 compared to AOPCP.
This may be caused by the altered electrostatic potential due to
the oxo-substituent in the C2-position, resulting in a loss of
aromaticity41 and a less favored π-stacking interaction with
F417 and F500. Together with the suboptimal hydrogen-
bonding interactions of the oxo group in the bound state, this
in turn results in an increased Ki value.
The piperazinyl-substituted derivative 21 features a bulky
substituent with limited flexibility. The crystal structure
analysis shows that the inhibitor adopts a new binding mode,
in which the ribose group and the adenine ring are flipped by
about 180°, while the terminal phosphate group only shifts
slightly (Figure 6D). The N6-amino group points into the C2-
pocket and the 2-piperazinyl substituent is positioned in the
interdomain cleft, where N6-substituents bind in the canonical
AOPCP binding mode. The 2-piperazinyl substituent of 21
most likely fails to adopt a low-energy conformation in the
polar environment of N390 at the top rim of the C2-pocket. A
consequence of the reorientation is the loss of the well-
matched hydrogen bond network of the CD73 × AOPCP
complex, which most likely is the major contributor to the loss
in binding affinity. The AMP group of AOPCP forms eight
hydrogen bonds with protein residues directly, whereas only
four such interactions for this group are present in the CD73 ×
21 complex.
We observed a remarkable species difference for binding of
21 to rat and human CD73. The 2-piperazinyl substitution of
AOPCP results in an affinity loss by a factor of more than 10 in
rat CD73, whereas the affinity is only reduced by a factor of 2
in human CD73. We attribute this difference in affinity change
to the F500Y substitution in the rat enzyme (residue Y502 in
rat CD73). F500 of the nucleobase clamp adopts a different
side chain orientation in CD73 × 21 compared to CD73 ×
AOPCP, and it forms a favorable hydrophobic contact to L184
(Supporting Information, Figure S8A). The F → Y substitution
in the rat enzyme leads to a close contact between the tyrosine
OH group and the side chain of L184 (Supporting
Information, Figure S8B), which is unfavorable compared to
the unbound state, where the OH group is free to form a
hydrogen bond to water.
In contrast to 10, the 2-substituents in 9 (2-amino-
substituted, Table 2, Figure 5A) and 20 (2-hydrazinyl, Table
2) resulted in increased affinity. For 20, the binding mode
closely matches that of AOPCP and an additional hydrogen
bond is formed between the hydrazinyl substituent and the
side chain oxygen atom of N390 (Figure 6B), which probably
accounts for the improved Ki value (6-fold increase in potency
toward human CD73). The electron density in the C2-pocket
It also appears likely that 29 binds via the same flipped
binding mode of 21, as modeling indicates that the bulky
2950
https://dx.doi.org/10.1021/acs.jmedchem.9b01611
J. Med. Chem. 2020, 63, 2941−2957

Journal of Medicinal Chemistry
pubs.acs.org/jmc
Article
Figure 7. Metabolic stability of CD73 inhibitor 15 in comparison to AOPCP (A) in human blood plasma and B in rat liver microsomes. Results are
presented as percentage of remaining compound measured by LC-MS at different time intervals. Data points are from three separate experiments
performed in duplicates.
cyclohexylethylthio in the 2-position cannot be accommodated
by the C2-pocket due to its size (data not shown). The flipped
binding mode is an interesting starting point for further
inhibitor development.
of AOPCP derivatives. This has also been observed for 2-
substituted adenosine derivatives, e.g., 2-chloroadenosine,
which are the poor substrates of adenosine deaminase.42
Selectivity. Selected potent compounds with diverse
structural modifications were investigated in terms of their
selectivity against different ecto-nucleotidases. The lead
compound AOPCP (II) had been reported to block human
ecto-nucleotide pyrophosphatase/phosphodiesterase1 (NPP1)
with a Ki value of 16.5 μM.19 We therefore investigated the
AOPCP derivatives 8, 15, 16, and 29 with NPP1 and its
isoenyzmes NPP2 and NPP3, however, none of the
compounds showed a significant inhibition of these ecto-
nucleotidases at a high concentration of 10 μM (see
Supporting Information, Table S3). Because AOPCP is an
analogue of ADP, we also studied whether selected derivatives
(8, 16, 29) were able to activate the ADP-activated P2Y
receptor subtypes P2Y1 and P2Y12. In contrast to ADP and 2-
methylthio-ADP,19 AOPCP and its 2-substituted derivatives
did not significantly activate these receptors at a high
concentration of 10 μM (see Supporting Information, Table
S4). These results indicate that 2-substituted AOPCP
derivatives show a high degree of selectivity for CD73, similar
to what has previously been reforted for N6-substituted
AOPCP derivatives.19,20
Metabolic Stability. The presented AOPCP derivatives
display high aqueous solubility due to deprotonation under
physiological conditions and are therefore suitable for
application by injection or infusion. Peroral bioavailability
cannot be expected in their native form, but lipophilic ester
prodrugs may allow peroral application as well. Compound 15,
which is the most potent CD73 inhibitor of the present series,
was investigated for its metabolic stability in human blood
plasma and rat liver microsomes compared to AOPCP.19,20
Incubation was performed at 37 °C, after which the solutions
were analyzed by LC-MS. Incubation in human blood plasma
for 5 h showed that inhibitor 15 was more stable than AOPCP
(II), with less than 25% degradation under the applied
conditions, while AOPCP showed almost 50% degradation
under the same conditions (Figure 7A). Incubation with rat
liver microsomes showed that inhibitor 15 was quite stable and
much more so relative to AOPCP (Figure 7B). While AOPCP
was completely degraded within 30 min, inhibitor 15 was still
almost completely intact even after 2 h of incubation, and less
than 40% of the compound was metabolized under the same
conditions. This shows that a simple substituent in the 2-
position (2-iodo in 15) is sufficient to prevent fast metabolism
CONCLUSIONS
■
To probe the effects of diverse substituents in the 2-position of
AOPCP on CD73 inhibiton, and to study their interactions
with CD73 on a molecular level, we designed, synthesized, and
analyzed a series of 2-substituted adenosine-5′-O-
[(phosphonomethyl)phosphonic acid] derivatives. Substitu-
tion with small polar or lipophilic residues (e.g., −NH2, Cl, I)
already led to highly efficient CD73 inhibitors with low
nanomolar potency, resulting in high ligand efficiency.
Compared to the recently published clinical candidate I,18
which is structurally based on the N6-substituted AOPCP
derivatives previously described by our groups,19,20 the present
compounds are easier to synthesize requiring less reaction
steps, display significantly lower molecular weights, and are in a
low nanomolar affinity range, which is ideal for their
application as pharmacological tools or drugs. In contrast,
compound I is a superpotent inhibitor with a very slow onset
requiring careful kinetic studies.18
We obtained cocrystals structures for four different
inhibitors of the present series with human CD73 and
characterized the water structure of the C2-pocket at very
high resolution. Surprisingly, we observed diverse binding
modes for these competitive inhibitors, depending on the
nature of the 2-substituent.
Size and flexibility of single substituents at the C2-position
of AOPCP derivatives are of much greater importance for the
inhibitory potency of CD73 inhibitors, as compared to those at
the N6-position, where small as well as bulky hydrophobic
substituents are tolerated and enhance the inhibitory
potency.19 High-resolution structures together with SAR data
demonstrate the importance of the water structure of the C2-
pocket and of substitution effects on the π−π interactions with
the nucleobase clamp for understanding the affinity improve-
ments caused by the substituents at the C2-position. While the
relatively small single halogen substituents and single hydro-
philic substituents currently result in the best potency, we
assume that further structure-based exploration of the C2-
pocket by designing substituents forming favorable contacts to
the hydrophobic base might result in further affinity improve-
ments. This may lead to potent competitive CD73 inhibitors,
potentially without the charged bisphosphonate group. Overly
large and inflexible hydrophobic moieties such as the
piperazinyl substituent resulted in a flipped binding mode of
2951
https://dx.doi.org/10.1021/acs.jmedchem.9b01611
J. Med. Chem. 2020, 63, 2941−2957

Journal of Medicinal Chemistry
pubs.acs.org/jmc
Article
to confirm the structures of the synthesized new compounds and all
final products.
the AOPCP scaffold and loss in affinity. Surprisingly, this
binding mode only resulted in a 2-fold loss in potency against
human CD73 and it might therefore be a starting point for
further inhibitor development.
General Procedure for the Synthesis of Nucleotides. The
nucleotides (7−10, 15, 16, 20, 21, 28, and 29) were synthesized by a
previously reported method.19 An ice-cooled solution of
methylenebis(phosphonic dichloride) (3 mmol) in trimethyl
phosphate (5 mL) was added to a suspension of the appropriate
nucleoside (1 mmol) in trimethyl phosphate (2 mL) at 0 °C kept also
on ice. After stirring for 15−25 min, upon disappearance of the
nucleoside as indicated by TLC, 10 mL of cold 0.5 M aqueous TEAC
solution (pH 7.4−7.6) was added. The mixture was further stirred at
0 °C for 30 min, followed by stirring at rt for 2 h. Next, trimethyl
phosphate used for the synthesis was extracted using 2 × 100 mL of
tert-butylmethyl ether, and the aqueous layer was subsequently
collected and lyophilized for 3 days. The obtained mixture of the
desired nucleotide analogue and side products were separated by ion-
exchange chromatography on DEAE Sephadex (A-25, HCO−3 form),
using a linear gradient of 0−800 mM of TEAC in distilled water.
Fractions containing the product were pooled and evaporated to
dryness in vacuo, while adding ethanol repeatedly to remove the
TEAC buffer. Thus obtained liquid was lyophilized by adding 10 mL
of water for 3 days to give glassy solids. The compound was then
purified by RP-HPLC using a gradient of 50 mM ammonium
hydrogencarbonate/acetonitrile from 100:0 to 40:60, and suitable
fractions were pooled and lyophilized several times by adding 20 mL
of water each time to finally obtain white coarse powder of the final
products.
In addition, while previous studies (mainly for N6-
substituted AOPCP derivatives) reported only minor species
differences (typically >3-fold higher potency at human as
compared to rat CD73), we observed very large species
differences (of up 12.5-fold) between rat and human CD73-
inhibitory potency for many of the 2-substituted AOPCP
derivatives. This is likely related to the amino acid replace-
ments N499S and F500Y in rat CD73, which influence the
water structure of the C2-pocket, the π−π interactions with the
nucleobase, and also the conformational freedom of residue
500, as demonstrated for the CD73 × 21 complex. Compound
21 exhibited a very different binding mode compared to other
AOPCP derivatives due to its large, unflexible 2-substituent
and also displayed a particularly large species difference, being
12.5-fold more potent toward human CD73 compared to the
rat enzyme. The present study provides valuable information
for the further improvement of these inhibitors, which have
great potential as future drugs.
EXPERIMENTAL SECTION
■
Guanosine-5′-O-[(phosphonomethyl)phosphonic Acid] (7). 1H
NMR (600 MHz, D2O) δ 8.28 (s, 1H, C8-H), 5.95 (dd, J = 5.5 Hz,
1H, C1′-H), 4.57−4.51 (m, 1H, C2′-H), 4.40−4.37 (m, 1H, C3′-H),
4.37−4.33 (m, 1H, C4′-H), 4.25−4.08 (m, 2H, C5′-H2), 2.20 (t, J =
19.6 Hz, 2H, P-CH2-P). 13C NMR (151 MHz, D2O) δ 165.76 (1C,
C6), 161.43 (1C, C2), 156.96 (1C, C4), 140.49 (1C, C8), 120.15
(1C, C5), 90.01 (1C, C1′), 86.76 (1C, C4′), 76.54 (1C, C2′), 73.06
(1C, C3′), 66.39 (1C, C5′), 49.52 (1C, P-CH2-P). 31P NMR (243
MHz, D2O) δ 18.69 (d, J = 9.5 Hz, Pα), 15.46 (d, J = 9.5 Hz, Pβ). LC-
MS (m/z): negative mode 440 [M − H]−, positive mode 442 [M +
H]+. Purity by HPLC-UV (254 nm)-ESI-MS: 100%. HRMS (ESI):
m/z [M − H]− calcd for C11H17N5O10P2 440.0371, found 440.0381.
2-Amino-6-chloropurine Riboside-5′-O-[(phosphonomethyl)-
General. All reagents used for the synthesis and biological studies
were commercially obtained from various producers (Acros, Aldrich,
Fluka, Merck, and Sigma) and used without further purification and
drying. The purity of all starting materials used was greater than 95%.
Thin layer chromatography (TLC) using aluminum sheets with silica
gel 60 F254 (Merck) were used to monitor the reactions. Silica gel
0.060−0.200 mm, pore diameter ca. 6 nm, was employed for column
chromatography. LC/ESI-MS (HPLC analysis coupled to electro-
spray ionization mass spectrometry) was used to determine the final
compounds’ purity and was found to be greater than 95%. For LC/
ESI-MS measurement, the samples were prepared by dissolving 1 mg/
mL of compound in H2O/MeOH (1:1) containing 2 mM ammonium
acetate. Subsequently, a sample of 10 μL was injected into an HPLC
instrument and elution was performed with a gradient of water/
methanol (containing 2 mM ammonium acetate) from 90:10 to 0:100
for 15 min at a flow rate of 250 μL/min. The instrument was an API
2000 (Applied Biosystems, Darmstadt, Germany) mass spectrometer
(turbo ion spray ion source) coupled with a Waters HPLC system
1
phosphonic Acid] (8). H NMR (500 MHz, D2O) δ 8.28 (s, 1H,
C8-H), 6.11 (dd, J = 5.7 Hz, 1H, C1′-H), 5.24−5.20 (m, 1H, C2′-H),
4.67−4.57 (m, 1H, C3′-H), 4.30−4.28 (m, 1H, C4′-H), 4.27−4.13
(m, 2H, C5′-H2), 2.16 (t, J = 19.9 Hz, 2H, P-CH2-P). 13C NMR (126
MHz, D2O) δ 155.11 (1C, C2), 153.87 (1C, C6), 151.33 (1C, C4),
142.82 (1C, C8), 132.05 (1C, C5), 91.59 (1C, C1′), 86.57 (1C, C4′),
74.05 (1C, C2′), 72.59 (1C, C3′), 66.39 (1C, C5′), 49.51 (1C, P-
CH2-P). 31P NMR (202 MHz, D2O) δ 18.76 (d, J = 9.5 Hz, Pα),
15.37 (d, J = 9.5 Hz, Pβ). LC-MS (m/z): negative mode 458 [M −
H]−, positive mode 460 [M + H]+. Purity by HPLC-UV (254 nm)-
ESI-MS: 100%. HRMS (ESI): m/z [M − H]− calcd for
C11H16ClN5O9P2 458.0032, found 458.0038.
1
(Agilent 1100) using a Phenomenex Luna 3 μ C18 column. H, 31P,
and 13C NMR spectra were performed on a Bruker Avance 500 and
600 MHz spectrometer and were recorded at room temperature using
deuterated solvents (DMSO-d6, MeOD-d4, or D2O). High-resolution
mass spectroscopy (HRMS) analysis was performed using a Bruker
Daltonik micrOTOF-Q. For lyophilization and for drying of the
nucleosides before the final reactions, a freeze-dryer (Christ Alpha 1−
4 LSC) was used. For purification of the final nucleotide analogues, a
preparative HPLC purification system (Knauer Smartline 1050), and
for anion exchange chromatography, an FPLC instrument (Amersham
Biosciences) were used.
2,6-Diaminopurine Riboside-5′-O-[(phosphonomethyl)-
1
phosphonic Acid] (9). H NMR (500 MHz, D2O) δ 8.22 (s, 1H,
C8-H), 5.92−5.88 (m, 1H, C1′-H), 4.70−4.65 (m, 1H, C2′-H),
4.51−4.48 (m, 1H, C3′-H), 4.35−4.33 (m, 1H, C4′-H), 4.22−4.18
(m, 2H, C5′-H2), 2.21−2.15 (t, J = 19.8 Hz, 2H, P-CH2-P). 13C NMR
(126 MHz, D2O) δ 165.69 (1C, C6), 156.87 (1C, C2), 150.67 (1C,
C4), 142.30 (1C, C8), 120.33 (1C, C5), 90.19 (1C, C1′), 86.73 (1C,
C4′), 76.74 (1C, C2′), 72.94 (1C, C3′), 66.47 (1C, C5′), 46.67 (1C,
P-CH2-P). 31P NMR (202 MHz, D2O) δ 18.36 (d, J = 7.4 Hz, Pα),
16.90 (d, J = 9.8 Hz, Pβ). LC-MS (m/z): negative mode 439 [M −
H]−, positive mode 441 [M + H]+. Purity by HPLC-UV (254 nm)-
ESI-MS: 100%. HRMS (ESI): m/z [M − H]− calcd for
C11H18N6O9P2 439.0530, found 439.0539.
Isoguanosine-5′-O-[(phosphonomethyl)phosphonic Acid] (10).
1H NMR (500 MHz, D2O) δ 8.27 (s, 1H, C8-H), 5.88 (dd, J = 5.2
Hz, 1H, C1′-H), 4.70−4.68 (m, 1H, C2′-H), 4.51−4.47 (m, 1H, C3′-
H), 4.39−4.31 (m, 1H, C4′-H), 4.18−4.11 (m, 2H, C5′-H2), 2.24 (t,
J = 19.9 Hz, 2H, P-CH2-P). 13C NMR (151 MHz, D2O) δ 165.97
Triethylammoniumhydrogen carbonate buffer (TEAC, 1 M) was
prepared as previously described19 by adding dry ice slowly to 1 M
triethylamine solution in water with stirring for 2−3 h until a pH
value of approximately 7.4−7.6 was indicated (pH meter).
Altogether, 10 different 2-substituted AOPCP derivatives and
analogues were synthesized and investigated, eight of which are new
compounds not previously described in literature. Although the
synthesis of guanosine-5′-O-[(phosphonomethyl)phosphonic acid]
(7) and of 2-chloroadenosine-5′-O-[(phosphonomethyl)phosphonic
acid](16) had previously been described,43,44 their interaction with
CD73 is reported for the first time in this publication. Some of the
prepared intermediate nucleoside, which were not previously
1
described, were also fully analyzed in the present study. H NMR,
13C NMR, DEPT-135 31P NMR spectroscopy, and HRMS were used
2952
https://dx.doi.org/10.1021/acs.jmedchem.9b01611
J. Med. Chem. 2020, 63, 2941−2957

Journal of Medicinal Chemistry
pubs.acs.org/jmc
Article
(1C, C2), 154.91 (1C, C6), 152.66 (1C, C4), 141.36 (1C, C8),
120.33 (1C, C5), 90.86 (1C, C1′), 87.56 (1C, C4′), 77.10 (1C, C2′),
73.45 (1C, C3′), 66.48 (1C, C5′), 49.22 (1C, P-CH2-P). 31P NMR
(243 MHz, D2O) δ 18.22 (d, J = 9.4 Hz, Pα), 17.35 (d, J = 9.5 Hz,
Pα). LC-MS (m/z): negative mode 440 [M − H]−, positive mode 442
[M + H]+. Purity by HPLC-UV (254 nm)-ESI-MS: 100%. HRMS
(ESI): m/z [M − H]− calcd for C11H17N5O7P2 440.0371, found
440.0378.
167.39 (1C, C2), 158.06 (1C, C6), 152.94 (1C, C4), 142.12 (1C,
C8), 136.87 (1C, Allyl), 120.55 (1C, C5), 119.27 (1C, Allyl), 90.08
(1C, C1′), 86.38 (1C, C4′), 76.71 (1C, C2′), 73.02 (1C, C3′), 66.38
(1C, C5′), 49.50 (1C, P-CH2-P), 36.51 (1C, Allyl). 31P NMR (202
MHz, D2O) δ 19.33 (d, J = 8.5 Hz, Pα), 14.47 (d, J = 8.6 Hz, Pβ). LC-
MS (m/z): negative mode 496 [M − H]−, positive mode 498 [M +
H]+. Purity by HPLC-UV (254 nm) ESI-MS: 100%. HRMS (ESI):
m/z [M − H]− calcd for C14H21N5O9P2S 496.0455, found 496.0463.
2-Cyclohexylethylthioadenosine-5′-O-[(phosphonomethyl)-
phosphonic Acid] (29). A simplified purification procedure was
developed for this compound after synthesis of a second batch. The
obtained mixture of the crude nucleotide analogue was directly
purified by RP-HPLC using a gradient of water/acetonitrile
(containing 0.05% TFA) from 20:80 to 80:20, and suitable fractions
were pooled and lyophilized to obtain the desired product as a white
powder. Prior used ion-exchange chromatography is not required
using this optimized procedure. 1H NMR (500 MHz, D2O) δ 8.37 (s,
1H, C8-H), 6.11 (dd, J = 4.7 Hz, 1H, C1′-H), 4.55−4.48 (m, 1H,
C2′-H), 4.32−4.26 (m 1H, C3′-H), 4.28−4.21 (m, 1H, C4′-H),
4.19−4.08 (m, 2H, C5′-H2), 3.16−3.01 (m, 2H, −CH2-CH2-), 2.19
2-Iodoadenosine-5′-O-[(phosphonomethyl)phosphonic Acid]
1
(15). H NMR (600 MHz, D2O) δ 8.43 (s, 1H, C8-H), 6.05 (dd, J
= 5.4 Hz, 1H, C1′-H), 4.76−4.33 (m, 1H, C2′-H), 4.54−4.52 (m,
1H, C3′-H), 4.39−4. 36 (m, 1H, C4′), 4.22−4.09 (m, 2H, C5′-H2),
2.20 (t, J = 19.9 Hz, 2H, P-CH2-P). 13C NMR (151 MHz, D2O) δ
155.38 (1C, C6), 149.51 (1C, C4), 139.57 (1C, C8), 119.44 (1C,
C2), 118.50 (1C, C5), 87.05 (1C, C1′), 83.90 (1C, C4′), 74.29 (1C,
C2′), 70.13 (1C, C3′), 63.42 (1C, C5′), 46.62 (1C, P-CH2-P). 31P
NMR (243 MHz, D2O) δ 18.67 (d, J = 9.7 Hz, Pα), 15.38 (d, J = 9.8
Hz, Pβ). LC-MS (m/z): negative mode 550 [M − H]−, positive mode
552 [M + H]+. Purity by HPLC-UV (254 nm)-ESI-MS: 100%.
HRMS (ESI): m/z [M − H]− calcd for C11H16IN5O9P2 549.9388,
found 549.9394.
(t, J = 19.8 Hz, 2H, P-CH2-P), 1.77−1.43 (m, 9H, -CH2-CH2-
cyclohexane-H), 1.34−1.02 (m, 4H, −CH2-CH2-cyclohexane-H). 13
C
2-Chloroadenosine-5′-O-[(phosphonomethyl)phosphonic Acid]
1
(16). H NMR (500 MHz, D2O) δ 8.46 (s, 1H, C8-H), 6.01 (dd, J
NMR (126 MHz, D2O) δ 167.64 (1C, C2), 157.32 (1C, C6), 152.90
(1C, C4), 142.11 (1C, C8), 118.83 (1C, C5), 89.77 (1C, C1′), 86.48
(1C, C4′), 76.82 (1C, C2′), 73.05 (1C, C3′), 66.43 (1C, C5′), 39.46
(1C, P-CH2-P), 35.46 (1C, CH2-CH2−), 31.60 (1C, cyclohexane),
30.34 (1C, −CH2-CH2-cyclohexane), 29.06 (2C, cyclohexane), 28.71
(3C, cyclohexane). 31P NMR (202 MHz, D2O) δ 18.68 (d, J = 9.8 Hz,
Pα), 15.27 (d, J = 9.4 Hz, Pβ). LC-MS (m/z): negative mode 566 [M
− H]−, positive mode 568 [M + H]+. Purity by HPLC-UV (254 nm)-
ESI-MS: 100%. HRMS (ESI): m/z [M − H]− calcd for
C19H31N5O9P2S 566.1238, found 566.1248.
= 5.0 Hz, 1H, C1′-H), 4.71−4.69 (m, 1H, C2′-H), 4.53−4.50 (m,
1H, C3′-H), 4.31−4.26 (m, 1H, C4′-H), 4.16−4.12 (m, 2H, C5′-H2),
2.07 (t, J = 19.5 Hz, 2H, P-CH2-P). 13C NMR (126 MHz, D2O) δ
159.10 (1C, C6), 156.60 (1C, C2), 152.91 (1C, C4), 142.91 (1C,
C8), 120.50 (1C, C5), 90.04 (1C, C1′), 86.63 (1C, C4′), 77.09 (1C,
C2′), 72.84 (1C, C3′), 66.05 (1C, C5′), 49.48 (1C, P-CH2-P). 31P
NMR (202 MHz, D2O) δ 21.25 (d, J = 9.6 Hz, Pα), 12.84 (d, J = 9.7
Hz, Pβ). LC-MS (m/z): negative mode 458 [M − H]−, positive mode
460 [M + H]+. Purity by HPLC-UV (254 nm)-ESI-MS: 100%.
HRMS (ESI): m/z [M − H]− calcd for C11H16ClN5O9P2 458.0032,
found 458.0040.
CD73 Assay. Recombinant soluble rat CD73, recombinant soluble
human CD73, or membrane preparations of MDA-MB-231 cells
natively expressing human CD73 were prepared as previously
described.21,37 The enzyme inhibition assays were performed
essentially as described.21,36 Solutions of test compound of varying
concentrations were prepared in demineralized water, and an aliquot
(10 μL) was added to 70 μL of assay reaction buffer containing 25
mM Tris, 140 mM sodium chloride, and 25 mM sodium dihydrogen
phosphate, pH 7.4. Then, 10 μL of CD73-containing solution or
suspension (rat CD73, 1.63 ng; human CD73, 0.365 ng; membrane
preparation of MDA-MB-231 cells expressing CD73, 7.4 ng of protein
per vial) was added, followed by the addition of 10 μL of
[2,8-3H]AMP (specific activity: 20 mCi/ mmol, American Radio-
labeled Chemicals, MO, USA, obtained via Hartmann Analytic,
Braunschweig, Germany) for the initiation of the reaction (the final
substrate concentration was 5 μM). The mixture was incubated for 25
min at 37 °C in a shaking water bath, then 500 μL of cold
precipitation buffer (100 mM lanthanum chloride, 100 mM sodium
acetate, pH 4.0) were added to stop the reaction and to facilitate the
precipitation of unreacted substrate. The formed precipitate was
separated by filtration through GF/B glass fiber filters using a cell
harvester (M-48, Brandel, Gaithersburg, MD, USA) and washed three
times with 400 μL of cold (4 °C) demineralized water per vial. The
filtrate was collected, 5 mL of the scintillation cocktail (ULTIMA
Gold XR, PerkinElmer, MA, USA) was added to each vial, and
radioactivity was measured by scintillation counting (Tri-Carb 2900
TR, Packard/PerkinElmer; counting efficacy: 49−52%). The resulting
data were plotted, and concentration−inhibition curves were fitted
with GraphPad Prism 7 (GraphPad Software, La Jolla, USA). The
mean IC50 standard error of the mean (SEM) from duplicate or
triplicate determinations obtained from three independent experi-
ments was used to calculate the Ki value using the Cheng−Prusoff
equation.
2-Hydrazinyladenosine-5′-O-[(phosphonomethyl)phosphonic
1
Acid] (20). H NMR (600 MHz, D2O) δ 8.30 (s, 1H, C8-H), 5.98
(dd, J = 5.1 Hz, 1H, C1′-H), 5.04−4.94 (m, 1H, C2′-H), 4.62−4.48
(m, 1H, C3′-H), 4.38−4.30 (m, 1H, C4′-H), 4.27−4.14 (t, 2H, C5′-
H2), 2.28−2.13 (t, J = 19.5 Hz, 2H, P-CH2-P). 13C NMR (151 MHz,
D2O) δ 159.06 (1C, C6), 157.10 (1C, C2), 152.80 (1C, C4), 142.75
(1C, C8), 120.14 (1C, C5), 90.89 (1C, C1′), 86.69 (1C, C4′), 75.81
(1C, C2′), 73.11 (1C, C3′), 66.45 (1C, C5′), 49.51 (1C, P-CH2-P).
31P NMR (243 MHz, D2O) δ 18.49 (d, J = 9.5 Hz, Pα), 16.14 (d, J =
9.4 Hz, Pβ). LC-MS (m/z): negative mode 454 [M − H]−, positive
mode 456 [M + H]+. Purity by HPLC-UV (254 nm) ESI-MS: 100%.
HRMS (ESI): m/z [M − H]− calcd for C11H19N7O9P2 454.0639,
found 454.0651.
2-(1-Piperazinyladenosine-5′-O-[(phosphonomethyl)-
1
phosphonic Acid] (21). H NMR (600 MHz, D2O) δ 8.15 (s, 1H,
C8-H), 6.02 (dd, J = 5.3 Hz, 1H, C1′-H), 4.88−4.84 (m, C2′-H),
4.55−4.51 (m, 1H, C3′-H), 4.34−4.29 (m, 1H, C4′-H), 4.19−4.08
(m, 2H, C5′-H2), 3.83−3.76 (m, 4H, piperazinyl-H), 2.90−2.89 (m,
4H, piperazinyl-H), 2.13 (t, J = 19.3 Hz, 2H, P-CH2-P). 13C NMR
(151 MHz, D2O) δ 161.50 (1C, C2), 158.82 (1C, C6), 154.30 (1C,
C4), 141.32 (1C, C8), 120.14 (1C, C5), 90.00 (1C, C1′), 86.07 (1C,
C4′), 76.05 (1C, C2′), 73.06 (1C, C3′), 66.58 (1C, C5′), 47.53 (1C,
P-CH2-P), 45.89 (2C, piperazine), 44.46 (2C, piperazine). 31P NMR
(243 MHz, D2O) δ 19.73 (d, J = 8.9 Hz, Pα), 14.49 (d, J = 9.0 Hz,
Pβ). LC-MS (m/z): negative mode 508 [M − H]−, positive mode 510
[M + H]+. Purity by HPLC-UV (254 nm) ESI-MS: 100%. HRMS
(ESI): m/z [M − H]− calcd for C15H25N7O9P2 508.1109, found
508.1100.
2-Allylthioadenosine-5′-O-[(phosphonomethyl)phosphonic
1
Acid] (28). H NMR (500 MHz, D2O) δ 8.34 (s, 1H, C8-H), 6.09
(dd, J = 5.2, 1.8 Hz, 1H, C1′-H), 6.06−5.99 (m, 1H, S-CH2-CH
CH2), 5.38−5.33 (dq, J = 17.0, 1.5 Hz, 1H, S-CH2-CHCH2), 5.17
(dt, J = 10.2, 1.3 Hz, 1H, S-CH2-CHCH2), 4.84−4.79 (m, 1H, C2′-
H), 4.56−4.53 (m, C3′-H), 4.44−4.38 (m 1H, C4′-H), 4.22−4.01
(m, 2H, C5′-H2), 3.82 (dt, J = 6.8, 1.4 Hz, 2H, S-CH2-CHCH2),
2.15 (t, J = 19.8 Hz, 2H, P-CH2-P). 13C NMR (126 MHz, D2O) δ
Determination of Compound Stability in Human Blood
Plasma. Blood plasma from healthy donors was obtained from the
blood bank, University Clinic Bonn, and the study was performed as
previously described.19 Human plasma was added to an aqueous
solution of inhibitor to obtain a final concentration of 200 μM
2953
https://dx.doi.org/10.1021/acs.jmedchem.9b01611
J. Med. Chem. 2020, 63, 2941−2957

Journal of Medicinal Chemistry
pubs.acs.org/jmc
Article
(volume: 2 mL) and incubated at rt. Then, 200 μL of reaction mixture
were withdrawn at different time points (1, 5, 30, 60, 120, and 300
min intervals) and quenched with 400 μL of ice-cold acetonitrile to
stop the reaction, followed by centrifugation at 8000g for 10 min. The
supernatants were then transferred to clean LC-MS (liquid
chromatography−mass spectroscopy) vials and analyzed by HPLC.
Three separate experiments were performed in duplicate for each
compound.
Determination of Compound Metabolism in Rat Liver
Microsomes. Rat liver microsomes were prepared as previously
described.19 Compounds were incubated with rat liver microsomes (2
mg) at a concentration of 200 μM in a final volume of 2 mL with an
NADPH regenerative system. The NADPH regenerative system
consists of NADP (0.57 mM), NADH (0.57 mM), isocitrate (6.4
mM), isocitrate dehydrogenase (0.57 mM), and MgCl2 (23.4 mM)
(pH 7.2). The samples were incubated for different time intervals at
37 °C in a water bath. The reaction was stopped by adding ice-cold
acetonitrile, followed by centrifugation at 8000g. The supernatant was
analyzed by LC-MS.
Expression, Purification, and Crystallization of Refolded
Human CD73. For cocrystallization experiments with AOPCP
derivatives in the closed enzyme state, construct no. 3 of human
CD738 was recombinantly expressed in Escherichia coli and refolded
from inclusion bodies. This construct encodes for the mature human
CD73 protein sequence (residues 27−549) and contains the T376A
natural variation (database entry P21589/VAR_022091 of Uni-
ProtKB/Swiss-Prot) as well as four asparagine to aspartate mutations
at the natural glycosylation sites and a C-terminal His6 tag. For the
crystal structures of the open form of CD73 construct no. 2 of human
CD73 was used, which differs from construct no. 3 in the following
additional surface mutations: K145S/K147S/K478S. The general
procedure for the production of active human CD73 was performed
according to Knapp et al.45 Briefly, prepurified inclusion bodies were
renatured by a fast dilution protocol in an arginine−glycerol buffer at
13 °C. Renatured, active protein was separated from misfolded species
by an AMP-agarose affinity chromatography, followed by two steps of
size exclusion chromatography. The protein was stored at −80 °C in
aliquots until further use. For crystallization, vapor diffusion trials
were set up at 19 °C. Then 1 μL of the protein solution containing 1
mM of the respective inhibitor and 100 μM of ZnCl2 were mixed with
an equal amount of reservoir buffer solution to give concentrations
listed in Table 2. Crystals were stepwise transferred to a cryo buffer
composed as reported in Table 2 and finally flash frozen in liquid
nitrogen.
Crystallization, X-ray Data Collection and Structure Determi-
nation. X-ray data collection was carried out at 100 K on beamline
14.1 of the Berlin Synchrotron (BESSY, Berlin, Germany) equipped
with a PILATUS 6 M detector (Dectris, Baden, Switzerland). The
diffraction data were indexed, integrated, and scaled with XDS46 or
XDSAPP47 as well as aimless48 of the CCP4 suite (Collaborative
Computational Project, no. 4, 1994). Coordinates of human CD73 in
the closed conformation (PDB 4H2I)8 or open state (PDB 4H2G)8
were taken as starting models for rigid body refinement under
REFMAC5.49,50 The models were further improved by iterative cycles
of TLS refinement using BUSTER51 and manual rebuilding with
COOT.52 The two domains of residues 26−335 and 336−551 were
used as TLS groups. Structures were validated by MOLPROBITY.53
Coordinate files and restraint dictionaries for inhibitors were
generated using the GRADE web server (http://grade.
globalphasing.org/). Final structure building and interpretation were
aided by feature enhanced maps calculated with PHENIX.54,55
Relevant crystallographic parameters of the structures are listed in
Table 2. Figures were prepared using PYMOL (http://pymol.org).
Synthesis of 2-amino- and 2-oxo-substituted purineribo-
side-5′-O-[(phosphonomethyl)phosphonic acid] deriva-
tives; synthesis of 2-chloro-, 2-iodo-, 2-hydrazinyl-, and
2-piperazinyl-substituted adenosine-5′-O-
[(phosphonomethyl)phosphonic acid] derivatives; syn-
thesis of 2-allylthio- and 2-cyclohexylethylthio-substi-
tuted adenosine-5′-O-[(phosphonomethyl)phosphonic
acid] derivatives; yields and purities of nucleosides;
yields and purities of nucleotides; procedure for the
synthesis of 2′,3′,5′-tri-O-acetylguanosine (2); proce-
dure for the synthesis of 2-amino-6-chloro-(2′,3′,5′-tri-
O-acetyl-β-D-ribofuranosyl)-9H-purine (3); procedure
for the synthesis of 2-amino-6-chloropurine riboside
(4); procedure for the synthesis of 2,6-diaminopurine
riboside (5); procedure for the synthesis of isoguanosine
(6); procedure for the synthesis of 6-chloro-2-iodo-
2′,3′,5′-triacetyl-purine riboside (11); procedure for the
synthesis of 2-iodoadenosine (12); procedure for the
synthesis of 2,6-dichloro-2′,3′,5′-triacetyl-purine riboside
(13); procedure for the synthesis of 2-chloroadenosine
(14); procedure for the synthesis of 2-hydrazinyladeno-
sine (17); procedure for the synthesis of 2-Boc-(N-
piperazinyl)adenosine (18); procedure for the synthesis
of 2-(N-piperazinyl)adenosine (19); general procedure
for the synthesis of 2-thioadenosine derivatives (26, 27);
LC/ESI-MS spectra of the synthesized nucleotide
analogue 15; LC/ESI-MS spectra of the synthesized
nucleotide analogue 29; sections of the 31P NMR spectra
for the purity control of compounds 15 and 20; sections
of the 31P NMR spectra for the purity control of
compounds 21 and 29; activity of compounds at human
NPPs and ADP-activated P2 receptor (P2Y1 and P2Y12)
NPP1−3 assays; potency of selected derivatives at
human NPP1−3; calcium mobilization assay for P2Y1
receptor P2Y12 receptor β-arrestin assay; potency of
selected derivatives as agonists at ADP-activated P2Y
receptors; comparison of human CD73 and rat CD73;
water structure in the C2-pocket of unliganded CD73 in
the open conformation; changes in the water structure of
the C2-pocket upon binding of AOPCP and derivatives;
superposition of the binding modes of AOPCP (gray,
4H2I) and of 21 (cyan) to CD73 (PDF)
Molecular formula strings of final compounds (CSV)
Accession Codes
Final coordinates and structure factors of cocrystals with
inhibitors PSB-12676 (9), PSB-12690 (10), PSB-12646 (20),
and PSB-12604 (21) and the cocrystal structures in the open
form have been deposited in the Protein Data Bank (www.rcsb.
org) with the corresponding accession codes listed in Table 2.
Authors will release the atomic coordinates and experimental
data upon article publication.
AUTHOR INFORMATION
Corresponding Authors
■
̈
Norbert Strater − Institute of Bioanalytical Chemistry, Center
for Biotechnology and Biomedicine, Leipzig University, D-04103
Leipzig, Germany; orcid.org/0000-0002-2001-0500;
Phone: +49 341 9731311; Email: strater@bbz.uni-
leipzig.de; Fax: +49 341 9731319
ASSOCIATED CONTENT
■
sı
* Supporting Information
Christa E. Muller − PharmaCenter Bonn, Pharmaceutical
̈
The Supporting Information is available free of charge at
https://pubs.acs.org/doi/10.1021/acs.jmedchem.9b01611.
Institute, Department of Pharmaceutical & Medicinal
Chemistry, University of Bonn, D-53121 Bonn, Germany;
2954
https://dx.doi.org/10.1021/acs.jmedchem.9b01611
J. Med. Chem. 2020, 63, 2941−2957

Journal of Medicinal Chemistry
pubs.acs.org/jmc
Article
orcid.org/0000-0002-0013-6624; Phone: +49-228-73-
2301; Email: christa.mueller@uni-bonn.de.; Fax: +49-228-
73-2567
beam time and assistance during synchrotron data collection
and to the Helmholtz Zentrum Berlin for traveling support.
ABBREVIATIONS USED
■
Authors
AOPCP, α,β-methylene-ADP, [{5-(6-aminopurin-9-yl)-3,4-
dihydroxyoxolan-2-yl}methoxy-hydroxyphosphoryl]-
methylphosphonic acid; AP, alkaline phosphatase; Boc, 1-tert-
butoxycarbonyl; CD73, cluster of differentiation 73; COPCP,
α,β-methylene-CDP; DEAE, diethylaminoethyl; DMSO-d6,
deuterated dimethyl sulfoxide; D2O, deuterated water; FPLC,
fast protein liquid chromatography; LC-MS, liquid chromatog-
raphy coupled to mass spectrometry; GPI, glycosylphosphati-
dylinositol; MeOD-d4, deuterated methanol; eNPP, nucleotide
pyrophosphatase; eNTPDase, ecto-nucleoside 5′-triphosphate
diphosphohydrolase; PSB, Pharmaceutical Sciences Bonn; RP-
HPLC, reverse phase-high performance liquid chromatogra-
phy; SARs, structure−activity relationships; Sf9, Spodoptera
frugiperda 9; TEAC, triethylammonium hydrogencarbonate;
UOPCP, α,β-methylene-UDP
Sanjay Bhattarai − PharmaCenter Bonn, Pharmaceutical
Institute, Department of Pharmaceutical & Medicinal
Chemistry, University of Bonn, D-53121 Bonn, Germany
Jan Pippel − Institute of Bioanalytical Chemistry, Center for
Biotechnology and Biomedicine, Leipzig University, D-04103
Leipzig, Germany
Emma Scaletti − Institute of Bioanalytical Chemistry, Center for
Biotechnology and Biomedicine, Leipzig University, D-04103
Leipzig, Germany
Riham Idris − PharmaCenter Bonn, Pharmaceutical Institute,
Department of Pharmaceutical & Medicinal Chemistry,
University of Bonn, D-53121 Bonn, Germany
Marianne Freundlieb − PharmaCenter Bonn, Pharmaceutical
Institute, Department of Pharmaceutical & Medicinal
Chemistry, University of Bonn, D-53121 Bonn, Germany
Georg Rolshoven − PharmaCenter Bonn, Pharmaceutical
Institute, Department of Pharmaceutical & Medicinal
Chemistry, University of Bonn, D-53121 Bonn, Germany
Christian Renn − PharmaCenter Bonn, Pharmaceutical
Institute, Department of Pharmaceutical & Medicinal
Chemistry, University of Bonn, D-53121 Bonn, Germany
Sang-Yong Lee − PharmaCenter Bonn, Pharmaceutical
Institute, Department of Pharmaceutical & Medicinal
Chemistry, University of Bonn, D-53121 Bonn, Germany
Aliaa Abdelrahman − PharmaCenter Bonn, Pharmaceutical
Institute, Department of Pharmaceutical & Medicinal
Chemistry, University of Bonn, D-53121 Bonn, Germany
Herbert Zimmermann − Institute of Cell Biology and
Neuroscience, Goethe-University, D-60438 Frankfurt am Main,
Germany
REFERENCES
■
(1) Burnstock, G. Physiology and pathophysiology of purinergic
neurotransmission. Physiol. Rev. 2007, 87, 659−797.
(2) Muller, C. E.; Jacobson, K. A. Recent developments in adenosine
̈
receptor ligands and their potential as novel drugs. Biochim. Biophys.
Acta, Biomembr. 2011, 1808, 1290−1308.
(3) Burnstock, G.; Verkhratsky, A. Long-term (trophic) purinergic
signalling: purinoceptors control cell proliferation, differentiation and
death. Cell Death Dis. 2010, 1, No. e9.
(4) Zimmermann, H. 5′-Nucleotidase: molecular structure and
functional aspects. Biochem. J. 1992, 285, 345−365.
̈
(5) Strater, N. Ecto-5′-nucleotidase: structure-function relationships.
Purinergic Signalling 2006, 2, 343−350.
̈
(6) Zimmermann, H.; Zebisch, M.; Strater, N. Cellular function and
molecular structure of ecto-nucleotidases. Purinergic Signalling 2012,
8, 437−502.
Ali El-Tayeb − PharmaCenter Bonn, Pharmaceutical Institute,
Department of Pharmaceutical & Medicinal Chemistry,
University of Bonn, D-53121 Bonn, Germany
(7) Heuts, D.P.H.M.; Weissenborn, M. J.; Olkhov, R. V.; Shaw, A.
M.; Gummadova, J.; Levy, C.; Scrutton, N. S. Crystal structure of a
soluble form of human CD73 withecto-5′-nucleotidase activity.
ChemBioChem 2012, 13, 2384−2391.
Complete contact information is available at:
https://pubs.acs.org/10.1021/acs.jmedchem.9b01611
(8) Knapp, K.; Zebisch, M.; Pippel, J.; El-Tayeb, A.; Muller, C. E.;
̈
̈
Strater, N. Crystal structure of the human ecto-5′-nucleotidase
(CD73): Insights into the regulation of purinergic signaling. Structure
2012, 20, 2161−2173.
Author Contributions
(9) Thompson, L. F.; Eltzschig, H. K.; Ibla, J. C.; Van De Wiele, C.
J.; Resta, R.; Morote-Garcia, J. C.; Colgan, S. P. Crucial role for ecto-
5′-nucleotidase (CD73) in vascular leakage during hypoxia. J. Exp.
Med. 2004, 200, 1395−1405.
S.B., J.P., and E.S. contributed equally.
Notes
The authors declare the following competing financial
interest(s): N.S. and C.E.M. have given scientific advice to
Arcus Biosciences (Arcus Biosciences, Inc. is a publicly traded
biotechnology company).
́
(10) Antonioli, L.; Pacher, P.; Vizi, E. S.; Hasko, G. CD39 and
CD73 in immunity and inflammation. Trends Mol. Med. 2013, 19,
355−367.
(11) Bours, M. J. L.; Swennen, E. L. R.; Di Virgilio, F.; Cronstein, B.
N.; Dagnelie, P. C. Adenosine 5′-triphosphate and adenosine as
endogenous signaling molecules in immunity and inflammation.
Pharmacol. Ther. 2006, 112, 358−404.
ACKNOWLEDGMENTS
■
S.B. and C.E.M. were supported by the Ministry for
Innovation, Science, Research and Technology of the State
of North-Rhine-Westfalia (NRW International Graduate
Research School BIOTECH-PHARMA). C.E.M. was sup-
ported by the Deutsche Forschungsgemeinschaft (SFB1328).
We thank Marion Schneider for LCMS analyses and Sabine
Terhart-Krabbe and Annette Reiner for NMR spectra. Prof.
Dr. Henrik Ditzel is acknowledged for the gift of MDA-MB-
231 cells. N.S. thanks the SFB 1052 (Deutsche Forschungsge-
meinschaft) for support. J.P., E.S., and N.S. are grateful to the
Joint Berlin MX-Laboratory at BESSY II, Berlin, Germany, for
(12) Zhang, B. CD73: a novel target for cancer immunotherapy.
Cancer Res. 2010, 70, 6407−6411.
(13) Stagg, J.; Divisekera, U.; McLaughlin, N.; Sharkey, J.; Pommey,
S.; Denoyer, D.; Dwyer, K. M.; Smyth, M. J. Anti-CD73 antibody
therapy inhibits breast tumor growth and metastasis. Proc. Natl. Acad.
Sci. U. S. A. 2010, 107, 1547−1552.
́
(14) Antonioli, L.; Fornai, M.; Blandizzi, C.; Pacher, P; Hasko, G.
Adenosine signaling and the immune system: When a lot could be too
much. Immunol. Lett. 2019, 205, 9−15.
(15) Vijayan, D.; Young, A.; Teng, M. W. L.; Smyth, M. J. Targeting
immunosuppressive adenosine in cancer. Nat. Rev. Cancer 2017, 17,
709−724.
2955
https://dx.doi.org/10.1021/acs.jmedchem.9b01611
J. Med. Chem. 2020, 63, 2941−2957

Journal of Medicinal Chemistry
pubs.acs.org/jmc
Article
́
(16) Allard, B.; Longhi, M. S.; Robson, S. C.; Stagg, J. The
ectonucleotidases CD39 and CD73: Novel checkpoint inhibitor
targets. Immunol. Rev. 2017, 276, 121−144.
venient access to 6-halo and 2,6-dihalopurine nucleosides and 2-
deoxynucleosides with acyl or silyl halides. J. Org. Chem. 2003, 68,
666−669.
(17) Hay, C. M.; Sult, E.; Huang, Q.; Mulgrew, K.; Fuhrmann, S. R.;
McGlinchey, K. A.; Hammond, S. A.; Rothstein, R.; Rios-Doria, J.;
Poon, E.; Holoweckyj, N.; Durham, N. M.; Leow, C. C.; Diedrich, G.;
Damschroder, M.; Herbst, R.; Hollingsworth, R. E.; Sachsenmeier, K.
F. Targeting CD73 in the tumor microenvironment with MEDI9447.
OncoImmunology 2016, 5, No. e1208875.
(18) Bowman, C. E.; da Silva, R. G.; Pham, A.; Young, S. W. An
exceptionally potent inhibitor of human CD73. Biochemistry 2019, 58,
3331−3334.
(19) Bhattarai, S.; Freundlieb, M.; Pippel, J.; Meyer, A.;
Abdelrahman, A.; Fiene, A.; Lee, S-Y.; Zimmermann, H.; Yegutkin,
(33) El-Tayeb, A.; Gollos, S. Synthesis and structure-activity
relationships of 2-hydrazinyladenosine derivatives as A2A adenosine
receptor ligands. Bioorg. Med. Chem. 2013, 21, 436−447.
(34) Abiru, T.; Miyashita, T.; Watanabe, Y.; Yamaguchi, T.;
Machida, H.; Matsuda, A. Nucleosides and nucleotides. 107. 2-
(Cycloalkylalkynyl)adenosines: adenosine A2 receptor agonists with
potent antihypertensive effects. J. Med. Chem. 1992, 35, 2253−2260.
(35) El-Tayeb, A.; Iqbal, J.; Behrenswerth, A.; Romio, M.; Schneider,
̈
M.; Zimmermann, H.; Schrader, J.; Muller, C. E. Nucleoside-5′-
monophosphates as prodrugs of adenosine A2A receptor agonists
activated by ecto-5′-nucleotidase. J. Med. Chem. 2009, 52, 7669−
7677.
̈
̈
G. G.; Strater, N.; El-Tayeb, A.; Muller, C. E. α,β-Methylene-ADP
derivatives and analogs: development of potent and selective ecto-5′-
nucleotidase (CD73) inhibitors. J. Med. Chem. 2015, 58, 6248−6263.
(20) Bhattarai, S.; Pippel, J.; Meyer, A.; Freundlieb, M.; Schmies, C.;
Abdelrahman, A.; Fiene, A.; Lee, S-Y.; Zimmermann, H.; El-Tayeb,
(36) Freundlieb, M.; Zimmermann, H.; Muller, C. E. A new,
̈
sensitive ecto-5′-nucleotidase assay for compound screening. Anal.
Biochem. 2014, 446, 53−58.
̈
(37) Servos, J.; Reilander, H.; Zimmermann, H. Catalytically active
̈
̈
A.; Yegutkin, G. G.; Strater, N.; Muller, C. E. X-ray co-crystal
structure guides the way to subnanomolar competitive ecto-5′-
nucleotidase (CD73) inhibitors for cancer immunotherapy. Adv.
Therap. 2019, 2, 1900075.
soluble ecto-5′-nucleotidase purified after heterologous expression as
a tool for drug screening. Drug Dev. Res. 1998, 45, 269−276.
(38) Cheng, Y.-C.; Prusoff, W. H. Relationship between the
inhibition constant (Ki) and the concentration of inhibitor which
causes 50% inhibition (IC50) of an enzymatic reaction. Biochem.
Pharmacol. 1973, 22, 3099−3108.
(21) Junker, A.; Renn, C.; Dobelmann, C.; Namasivayam, V.; Jain,
S.; Losenkova, K.; Irjala, H.; Duca, S.; Balasubramanian, R.;
̈
Chakraborty, S.; Borgel, F.; Zimmermann, H.; Yegutkin, G. G.;
Muller, C. E.; Jacobson, K. A. Structure-activity relationship of purine
(39) Qiao, Z.; Li, X.; Kang, N.; Yang, Y.; Chen, C.; Wu, T.; Zhao,
M.; Liu, Y.; Ji, X. A novel specific anti-CD73 antibody inhibits triple-
negative breast cancer cell motility by regulating autophagy. Int. J.
Mol. Sci. 2019, 20, 1057.
̈
and pyrimidine nucleotides as ecto-5′-nucleotidase (CD73) inhib-
itors. J. Med. Chem. 2019, 62, 3677−3695.
(22) Dumontet, C.; Peyrottes, S.; Rabeson, C.; Cros-Perrial, E.;
(40) Schiebel, J.; Gaspari, R.; Wulsdorf, T.; Ngo, K.; Sohn, C.;
Schrader, T. E.; Cavalli, A.; Ostermann, A.; Heine, A.; Klebe, G.
Intriguing role of water in protein-ligand binding studied by neutron
crystallography on trypsin complexes. Nat. Commun. 2018, 9, 3559.
(41) Salonen, L. M.; Ellermann, M.; Diederich, F. Aromatic rings in
chemical and biological recognition: energetics and structures. Angew.
Chem., Int. Ed. 2011, 50, 4808−4842.
(42) Maguire, M. H.; Sim, M. K. Studies on adenosine deaminase 2.
Specificity and mechanism of action of bovine placental adenosine
deaminase. Eur. J. Biochem. 1971, 23, 22−29.
(43) Engelsma, S. B.; Meeuwenoord, N. J.; Overkleeft, H. S.; van der
Marel, G. A.; Filippov, D. V. Combined phosphoramidite-
phosphodiester reagents for the synthesis of methylene bisphospho-
nates. Angew. Chem., Int. Ed. 2017, 56, 2955−2959.
(44) Gough, G.; Maguire, M. H.; Penglis, F. Analogues of adenosine
5′-diphosphate-new platelet aggregators. Influence of purine ring and
and phosphate chain substitutions on the platelet-aggregating potency
of adenosine 5′-diphosphate. Mol. Pharmacol. 1972, 8, 170−177.
(45) Knapp, K. Structural characterization of the human ecto-5′-
nucleotidase: implications for purinergic signalling. Doctoral dis-
sertation. Leipzig University, 2011; http://research.uni-leipzig.de/
straeter/h5nt/
́
Geant, P. Y.; Chaloin, L.; Jordheim, L. P. CD73 inhibition by purine
cytotoxic nucleoside analogue-based diphosphonates. Eur. J. Med.
Chem. 2018, 157, 1051−1055.
(23) Baqi, Y.; Lee, S.-Y.; Iqbal, J.; Ripphausen, P.; Lehr, A.; Scheiff,
̈
A.; Zimmermann, H.; Bajorath, J.; Muller, C. E. Development of
potent and selective inhibitors of ecto-5′-nucleotidase based on an
anthraquinone scaffold. J. Med. Chem. 2010, 53, 2076−2086.
(24) Ripphausen, P.; Freundlieb, M.; Brunschweiger, A.;
Zimmermann, H.; Muller, C. E.; Bajorath, J. Virtual screening
̈
identifies novel sulfonamide inhibitors of ecto-5′-nucleotidase. J. Med.
Chem. 2012, 55, 6576−6581.
(25) Iqbal, J.; Saeed, A.; Raza, R.; Matin, A.; Hameed, A.; Furtmann,
́
N.; Lecka, J.; Sevigny, J.; Bajorath, J. Identification of sulfonic acids as
efficient ecto-5′-nucleotidase inhibitors. Eur. J. Med. Chem. 2013, 70,
685−691.
(26) Uchino, K.; Matsuo, T.; Iwamoto, M.; Tonosaki, Y.; Fukuchi,
A. New 5′-nucleotidase inhibitors, NPF-86IA, NPF-86IB, NPF-86IIA,
and NPF-86IIB from Areca catechu; Part I. Isolation and biological
properties. Planta Med. 1988, 54, 419−422.
(27) Al-Rashida, M.; Batool, G.; Sattar, A.; Ejaz, S. A.; Khan, S.;
́
Lecka, J.; Sevigny, J.; Hameed, A.; Iqbal, J. 2-Alkoxy-3-(sulfonylar-
ylaminomethylene)-chroman-4-ones as potent and selective inhibitors
(46) Kabsch, W. XDS. Acta Crystallogr., Sect. D: Biol. Crystallogr.
2010, D66, 125−132.
(47) Krug, M.; Weiss, M. S.; Heinemann, U.; Muller, U. XDSAPP. A
̈
of ectonucleotidases. Eur. J. Med. Chem. 2016, 115, 484−494.
́
(28) Corbelini, P. F.; Figueiro, F; das Neves, G. M.; Andrade, S.;
Kawano, D. F.; Oliveira Battastini, A. M.; Eifler-Lima, V. L. Insights
into ecto-5′-nucleotidase as a new target for cancer therapy: a
medicinal chemistry study. Curr. Med. Chem. 2015, 22, 1776−1792.
(29) Wan, Z.-K.; Binnun, E.; Wilson, D. P.; Lee, J. A highly facile
and efficient one-step synthesis of N6-adenosine and N6-2′-
deoxyadenosine derivatives. Org. Lett. 2005, 7, 5877−5880.
(30) Middleton, R. J.; Briddon, S. J.; Cordeaux, Y.; Yates, A. S.; Dale,
C. L.; George, M. W.; Baker, J. G.; Hill, S. J.; Kellam, B. New
fluorescent adenosine A1-receptor agonists that allow quantification of
ligand-receptor interactions in microdomains of single living cells. J.
Med. Chem. 2007, 50, 782−793.
graphical user interface for the convenient processing of diffraction
data using XDS. J. Appl. Crystallogr. 2012, 45, 568−572.
(48) Evans, P. R.; Murshudov, G. N. How good are my data and
what is the resolution? Acta Crystallogr., Sect. D: Biol. Crystallogr.
2013, 69, 1204−1214.
(49) Murshudov, G. N.; Vagin, A. A.; Dodson, E. J. Refinement of
macromolecular structures by the maximum-likelihood method. Acta
Crystallogr., Sect. D: Biol. Crystallogr. 1997, 53, 240−255.
́
(50) Murshudov, G. N.; Skubak, P.; Lebedev, A. A.; Pannu, N. S.;
Steiner, R. A.; Nicholls, R. A.; Winn, M. D.; Long, F.; Vagin, A. A.
REFMAC5 for the refinement of macromolecular crystal structures.
Acta Crystallogr., Sect. D: Biol. Crystallogr. 2011, 67, 355−367.
(51) Blanc, E.; Roversi, P.; Vonrhein, C.; Flensburg, C.; Lea, S. M.;
Bricogne, G. Refinement of severely incomplete structures with
(31) Nair, V.; Young, D. A. A new synthesis of isoguanosine. J. Org.
Chem. 1985, 50, 406−408.
(32) Francom, P.; Robins, M. J. Nucleic acid related compounds.
118. Nonaqueous diazotization of aminopurine derivatives. Con-
2956
https://dx.doi.org/10.1021/acs.jmedchem.9b01611
J. Med. Chem. 2020, 63, 2941−2957

Journal of Medicinal Chemistry
pubs.acs.org/jmc
Article
maximum likelihood in BUSTER-TNT. Acta Crystallogr., Sect. D: Biol.
Crystallogr. 2004, D60, 2210−21.
(52) Emsley, P.; Lohkamp, B.; Scott, W. G.; Cowtan, K. Features
and development of Coot. Acta Crystallogr., Sect. D: Biol. Crystallogr.
2010, 66, 486−501.
(53) Chen, V. B.; Arendall, W. B.; Headd, J. J.; Keedy, D. A.;
Immormino, R. M.; Kapral, G. J.; Murray, L. W.; Richardson, J. S.;
Richardson, D. C. MolProbity: all-atom structure validation for
macromolecular crystallography. Acta Crystallogr., Sect. D: Biol.
Crystallogr. 2010, 66, 12−21.
́
(54) Adams, P. D.; Afonine, P. V.; Bunkoczi, G.; Chen, V. B.; Davis,
I. W.; Echols, N.; Headd, J. J.; Hung, L.-W.; Kapral, G. J.; Grosse-
Kunstleve, R. W.; McCoy, A. J.; Moriarty, N. W.; Oeffner, R.; Read, R.
J.; Richardson, D. C.; Richardson, J. S.; Terwilliger, T. C.; Zwart, P.
H. PHENIX: a comprehensive Python-based system for macro-
molecular structure solution. Acta Crystallogr., Sect. D: Biol. Crystallogr.
2010, 66, 213−221.
(55) Afonine, P. V.; Moriarty, N. W.; Mustyakimov, M.; Sobolev, O.
V.; Terwil-Liger, T. C.; Turk, D.; Urzhumtsev, A.; Adams, P. D. FEM:
feature-enhanced map. Acta Crystallogr., Sect. D: Biol. Crystallogr.
2015, 71, 646−666.
2957
https://dx.doi.org/10.1021/acs.jmedchem.9b01611
J. Med. Chem. 2020, 63, 2941−2957