Lead optimization and biological evaluation of fragment-based cN-II inhibitors

Article abstract of DOI:10.1016/j.ejmech.2019.02.040

The development of cytosolic 5′-nucleotidase II (cN-II) inhibitors is essential to validate cN-II as a potential target for the reversion of resistance to cytotoxic nucleoside analogues. We previously reported a fragment-based approach combined with molecular modelling, herein, the selected hit-fragments were used again in another computational approach based on the Ilib-diverse (a software enabling to build virtual molecule libraries through fragment based de novo design) program to generate a focused library of potential inhibitors. A molecular scaffold related to a previously identified compound was selected and led to a novel series of compounds. Ten out of nineteen derivatives showed 50–75% inhibition on the purified recombinant protein at 200 μM and among them three derivatives (12, 13 and 18) exhibited K_i in the sub-millimolar range (0.84, 2.4 and 0.58 mM, respectively). Despite their only modest potency, the cN-II inhibitors showed synergistic effects when used in combination with cytotoxic purine nucleoside analogues on cancer cells. Therefore, these derivatives represent a family of non-nucleos(t)idic cN-II inhibitors with potential usefulness to overcome cancer drug resistance especially in hematological malignancies in which cN-II activity has been described as an important parameter.

Full text of DOI:10.1016/j.ejmech.2019.02.040

Accepted Manuscript
Lead optimization and biological evaluation of fragment-based cN-II inhibitors
Rémi Guillon, Rahila Rahimova, Preeti, David Egron, Sonia Rouanet, Charles
Dumontet, Nushin Aghajari, Lars Petter Jordheim, Laurent Chaloin, Suzanne
Peyrottes
PII:
S0223-5234(19)30154-0
DOI:
https://doi.org/10.1016/j.ejmech.2019.02.040
EJMECH 11132
Reference:
To appear in: European Journal of Medicinal Chemistry
Received Date: 14 November 2018
Revised Date: 21 January 2019
Accepted Date: 10 February 2019
Please cite this article as: Ré. Guillon, R. Rahimova, Preeti, D. Egron, S. Rouanet, C. Dumontet,
N. Aghajari, L.P. Jordheim, L. Chaloin, S. Peyrottes, Lead optimization and biological evaluation
of fragment-based cN-II inhibitors, European Journal of Medicinal Chemistry (2019), doi: https://
doi.org/10.1016/j.ejmech.2019.02.040.
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to
our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please
note that during the production process errors may be discovered which could affect the content, and all
legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT
Graphical Abstract
1

ACCEPTED MANUSCRIPT
Lead optimization and biological evaluation of fragment-based cN-II inhibitors
Rémi Guillon1, Rahila Rahimova2, Preeti3, David Egron1, Sonia Rouanet1, Charles Dumontet4,
Nushin Aghajari3, Lars Petter Jordheim4, Laurent Chaloin*,2 and Suzanne Peyrottes*,1
1 Institut des Biomolécules Max Mousseron (IBMM), UMR 5247 CNRS, Université de Montpellier,
ENSCM, Campus Triolet, cc1705, Place Eugène Bataillon, 34095 Montpellier, France
2 Institut de Recherche en Infectiologie de Montpellier (IRIM), CNRS, Université de Montpellier,
34293 Montpellier, France
3 Molecular Microbiology and Structural Biochemistry, UMR5086 CNRS – Université Lyon 1, 7
Passage du Vercors, 69367 Lyon, France
4 Université de Lyon, Université Claude Bernard Lyon 1, Centre de Recherche en Cancérologie de
Lyon, INSERM U1052, CNRS UMR 5286, Centre Léon Bérard, 69008 Lyon, France
* shared corresponding authorship
Keywords: 5’-nucleotidase, enzyme inhibitor, cancer
ABSTRACT: The development of cytosolic 5’-nucleotidase II (cN-II) inhibitors is essential to
validate cN-II as a potential target for the reversion of resistance to cytotoxic nucleoside analogues.
We previously reported a fragment-based approach combined with molecular modelling, herein, the
selected hit-fragments were used again in another computational approach based on the Ilib-diverse
(a software enabling to build virtual molecule libraries through fragment based de novo design)
program to generate a focused library of potential inhibitors. A molecular scaffold related to a
previously identified compound was selected and led to a novel series of compounds. Ten out of
nineteen derivatives showed 50 to 75% inhibition on the purified recombinant protein at 200 µM and
among them three derivatives (12, 13 and 18) exhibited Ki in the sub-millimolar range (0.84, 2.4 and
0.58 mM, respectively). Despite their only modest potency, the cN-II inhibitors showed synergistic
effects when used in combination with cytotoxic purine nucleoside analogues on cancer cells.
Therefore, these derivatives represent a family of non-nucleos(t)idic cN-II inhibitors with potential
usefulness to overcome cancer drug resistance especially in hematological malignancies in which
cN-II activity has been described.
1

ACCEPTED MANUSCRIPT
INTRODUCTION
5’-nucleotidases (EC 3.1.3.5) are a family of enzymes that catalyze the hydrolysis of 5’-nucleoside
monophosphates (5’-nucleotides) into their corresponding parent nucleosides and inorganic
phosphate, and thereby play a key role in the metabolism of nucleotides. The membrane-bound ecto-
5’-nucleotidase, CD73, is mainly responsible for the conversion of extracellular adenosine 5’-
monophopshate (AMP) into adenosine, whereas intracellular 5’-nucleotidases (cN-I, cN-II, cN-III,
cdN, mdN) regulate the intracellular pools of both purine- and pyrimidine-based 5’-nucleotides and
nucleosides, depending on their substrate specificity and intracellular compartments [1] [2]. cN-IA
uses both AMP and pyrimidines and is expressed in muscles, whereas its homologue cN-IB is AMP
specific and ubiquitously expressed. The cytosolic cN-II and cN-IIIA are GMP/IMP and pyrimidine
preferring, respectively. cN-II is ubiquitously expressed and at a higher level in highly proliferating
cells, whereas cN-IIIB is important in erythrocytes but has also been found in many other tissues.
More recently, a cN-IIIB was described and found to be specific to 7-methyl-GMP [3]. Finally, the
cytosolic cdN and the mitochondrial mdN are active on deoxyribonucleosides phosphorylated on 3’
or 5’, with a preference towards pyrimidine derivatives.
Several studies have shown that deficiencies of these proteins are associated to severe diseases. For
example, deficiency in cN-III due to genetic mutations leads to hemolytic anemia [4], CD73
deficiency is associated to arterial calcification [5], and the NT5C2 (coding for cN-II) containing
locus has been associated with autosomal recessive spastic paraplegia and psychiatric disorders [6-
8]. In addition, high expression level or activity of cN-II in tumor cells is associated with resistance
to cytotoxic nucleoside analogues and nucleobases [9-11]. Its direct role in this resistance was shown
by the sensitization of human cancer cells to purine nucleoside analogues after inhibition of its
expression by shRNA [12]. Based on these observations, cN-II represents an attractive target to
revert cancer resistance, and the use of cN-II inhibitors in combination with chemotherapy involving
cytotoxic nucleosides may be of great interest.
Purine ribonucleoside analogues modified at the 5’-position were the first inhibitors of cN-II
reported in the literature (Fig. 1) [13]. 5’-Deoxy-5'-isobutylthio derivatives of adenosine and inosine
inhibited 50% of the nucleotidase activity (extracted from rat liver) at 2-10 mM. Recently, it was
shown that fludarabine was able to interfere with human cN-II as a mixed-inhibitor (Ki 0.5 mM and
Ki' 9 mM) [14]. Over the last decade, we have also identified, developed and studied cN-II inhibitors
by using virtual screening [15], a substrate analogue-based approach [16-18], and finally a fragment-
based drug discovery (FBDD) strategy [19]. All these derivatives behaved as weak to modest cN-II
inhibitors with Ki in the millimolar range (Fig. 1). Since these compounds act as competitive
2

ACCEPTED MANUSCRIPT
inhibitors and taking into account that cN-II is also named high-KM 5’-nucleotidase, it is not
surprising to observe such high Ki values (KM for inosine 5’-monophosphate (IMP) is ~ 0.5 mM).
Despite this, co-exposure of cancer cells to nucleoside analogues and some of these cN-II inhibitors
potentiated the effect of cytotoxic nucleoside analogues [14, 15].
Figure 1. Chemical structures of the cN-II inhibitors reported in the literature. IBTA: 5’-Isobutyl
thioadenosine; IBTI: 5’-Isobutylthioinosine; AdiS: Anthraquinone-2,6-disulfonic acid, disodium salt;
C: cytosine; Y=H and X=OH, or Y=OH and X=H.
Herein, we pursue the development of cN-II inhibitors using a small set of fragment hits, which
have been previously selected by a FBDD approach [19], to build up a virtual library. The latter was
generated using the Ilib-diverse software (http://www.inteligand.com), and involved the random
association of various fragments and chemical groups in order to enhance binding affinity and/or
retain drug-like properties such as water solubility. These in silico approaches have been successfully
applied to the design of potential lead compounds such as creatine kinase [20] and phospholipase
[21] inhibitors. In silico virtual screening was performed on the basis of experimental cN-II crystal
structures [22] and compounds with the most favorable docking scores were carefully analyzed.
Thus, one derivative was hand-picked to guide structural optimization, on the basis of its
resemblance with a previously identified inhibitor.
3

ACCEPTED MANUSCRIPT
RESULTS AND DISCUSSION
In silico drug design, library generation and selection process. We combined in silico drug
design approaches including virtual library generation followed by in silico screening by molecular
docking to identify potential inhibitors [23, 24]. Briefly, virtual flasks containing the most active
fragments previously identified by nuclear magnetic resonance (NMR) screening and measurement
of the inhibition of the enzyme activity in vitro [19], were used for the combinatorial generation of
novel chemical scaffolds. An overview of the process is presented in Figure 2. Flask 1 contained 21
fragments, whereas flasks 2 and 3 contained 15 fragments each (structures of the components are
listed in table S1 and mainly include aromatic and bicyclic derivatives, nitrogen-containing
heterocycles such as benzimidazoles, purines…). Finally, flask 4 contained 43 polar or ionisable
building blocks (alcohols, carboxylic acids, phosphoesters…, see Table S1) available from the Ilib-
diverse software.
Figure 2. Schematic view of the focused library generation. Virtual flasks 1, 2 and 3 contain hit
fragments previously identified as cN-II binders by NMR. Combination of one fragment from each
flask with an additional functional polar group (flask 4) is achieved in silico using the Ilib-diverse
software and leading to a virtual library of 3,000 compounds. All obtained compounds were
evaluated for their theoretical binding affinity by docking on cN-II crystal structure.
The Ilib-diverse tool generated a focused virtual library by using a Monte-Carlo-based random
fragment connection including general rules of chemical synthesis feasibility. During this process,
the software is randomly selecting fragments from each flask and generates molecules with a broad
molecular diversity. In our setup, we defined a higher importance for experimentally validated hit
fragments (cN-II binders identified by NMR) meaning that the frequency of insertion of these
4

ACCEPTED MANUSCRIPT
fragments in the final library will be higher compared to the low-active fragments (underlined
fragments in Table S1). To further increase the diversity and also to improve the aqueous solubility
of compounds, combination of one fragment taken from each of the three flasks was followed by the
addition of a fourth ionisable chemical group (building blocks already included in the program that
can be also considered as a fourth flask). Thus, a total of 3,000 compounds was generated (Ilib-
diverse virtual library), all meeting the criteria of orally-delivered drug like compounds according to
Lipinski’s rule of five (one exception was the molecular weight with an upper limit at 600 Da instead
of 500 Da). Then, all compounds were evaluated by molecular docking on cN-II. The ones showing
a lower score than that of our previously identified lead compound (Fig. 1, UA2132) and hitherto
most potent cN-II inhibitor reported were discarded, leaving a total of 884 compounds for further
consideration. To narrow the number of molecules to be further examined for synthetic accessibility,
we focused on compounds with highest docking scores (using a cut-off value of 125, which is
approximately 20 points above the docking score of UA2132) resulting in a selection of 25 chemical
structures (Table S2).
The analysis of structure similarity and fragments composition of these 25 selected hits indicates
that some skeletons were privileged during the library generation and filtering process. Indeed,
numerous assemblies consisting of polycyclic aromatic structures linked to a phenyl or biphenyl
fragment were found to promote a favorable binding according to the docking results. Among these,
one derivative particularly attracted our attention (Fig. 3B, Ilib-72) due to its structural similarity
with the previously identified cN-II inhibitor UA2132 (Fig. 3A). In addition, several hit compounds
with docking scores being slightly lower than our cut-off value (125) were identified as structural
analogues of UA2132 or at least holding an adenine moiety linked to an aromatic cycle through an
amide link (see Table S2, hits highlighted in bold characters). Therefore, instead of performing the
chemical synthesis of all 25 best-ranked hit compounds, we prioritized the structure optimization of
our lead compound considering the structural variations observed in Ilib-72 as it was the best-ranked
analogue of UA2132.
In this respect, molecular docking studies were performed to predict and compare the potential
binding modes of Ilib-72 with UA2132.
5

ACCEPTED MANUSCRIPT
Figure 3. Chemical structures of the potential cN-II inhibitors. (A) Previously identified derivative
UA2132 with inhibition mode and constant; (B) Hand-picked derivative (Ilib-72) from the focused
library; (C) Generic structure proposed for structural optimization.
Since UA2132 is a competitive inhibitor, the docking of hit compounds was achieved in the
substrate binding site (Fig. 4). One should note that the crystal structure used for modelling (PDB
2XCW) was an inactive mutant in which one of the catalytic residues D52 was mutated to N.
Compared to other crystal structures, this mutation does not change the coordination of Mg2+ ion
(hexa-coordinated with D/N52, D54, D351, 2 water molecules and one phosphate oxygen from
IMP). This mutation was used by Wallden and colleagues in order to prevent substrate hydrolysis
during crystallization, thereby giving an image of the position of the substrate [25]. Similar docking
poses have been obtained using the wild-type enzyme (2JC9). For UA2132, the prediction suggests
only one binding mode with the adenine moiety near the catalytic residues which interact with the
magnesium ion (the two first via coordination bonds and the latter via a water molecule). In contrast
to UA2132, two binding orientations were predicted for Ilib-72 (Fig. 4A & 4B) with comparable
docking scores (126.9 and 126.1). The most favorable binding mode is shown in Figure 4A with the
purine and the benzyloxymethyl group near the magnesium ion. However, in the second binding
mode the α-hydroxyacetic acid chain is chelating the magnesium ion within a more compressed
structure (Fig. 4B). This latter may reflect the real binding orientation of this compound since the
6

ACCEPTED MANUSCRIPT
natural substrate (IMP) also carries a negative charge on the phosphate oxygen. Interestingly, Ilib-72
occupies a large space which extends beyond the substrate binding cavity (Fig. 4C).
Based on this docking information, we decided to keep the N-6 biaryl substituted adenine skeleton
(present in UA2132) and to add various groups either on the N-7 or N-9 position of the purine, and/or
a five-membered nitrogen containing heterocycle on the biaryl moiety (Fig. 3C and 5). Both
substituents were chosen for their ability to either increase affinity and/or improve solubility of the
final derivatives. Indeed, under physiological conditions tetrazole and phosphonate groups would be
partially or completely protonated (Fig. 5).
A
B
N52
D351
D356
C
Figure 4. A) Preferential binding mode predicted by molecular docking for UA2132 (orange sticks)
and Ilib-72 (cyan sticks). B) Second binding mode identified for Ilib-72 showing an inversed
orientation allowing the ionic interaction between carboxylic group and magnesium ion (depicted as
a green sphere). C) Same docking as in panel (B) with cN-II in surface representation to appreciate
the substrate binding site. Catalytic residues are shown in grey thin sticks.
7

ACCEPTED MANUSCRIPT
Figure 5. Structures of the studied compounds 6-13, 18-20, 24a-e and 25.
Synthesis. The targeted compounds, presented in Figure 5, may be divided into three groups
corresponding to: (i) N-7 or N-9 substituted derivatives of UA2132 (6-13), (ii) Ilib-72 analogues
incorporating a polar substituent on the imidazole ring (18-20) and (iii) analogues incorporating a
tetrazole ring as ionisable five-membered ring (24-25).
First, different substitutions on positions N-7 or N-9 of the purine heterocycle (Fig. 5) were
envisaged: (i) a hydroxyethyl chain (8 and 11), a methyl phosphonate (10 and 13) or a
methylcarboxylate (9 and 12) in order to improve the aqueous solubility of UA2132 as this was
identified as a limit for its in vivo evaluation (unpublished data); (ii) a benzyl (6a-b) or a
benzyloxymethyl (7a-b) as observed within the structure of Ilib-72 for better affinity to the target
protein. Then, due to synthetic issues we envisaged, as the closest analogues of Ilib-72, the
replacement of the α-hydroxyacetic acid chain by non-classical bio-isosteres such as a methyl
8

ACCEPTED MANUSCRIPT
phosphonate group (compounds 18, 19 and 20) mimicking the acetate anion and being partially
ionized at physiological pH. Another modification consisted in the replacement of both the imidazole
ring and the α-hydroxyacetic acid chain by a tetrazole ring (compounds 24a-e). Indeed, tetrazole is
known to present similar physicochemical properties as a carboxyl group, and it is increasingly used
in drug development. Furthermore, tetrazole can exist in both neutral and anionic states. The latter
would allow the formation of salts that may be valuable for water solubility.
The general synthetic approach used to prepare new bi-aryl carboxamide derivatives of adenine
incorporating various N-7/N-9-substituants (6-13), N-methylphosphonate and imidazole moieties
(18-20) or C-branched tetrazole (24a-e and 25) consists of coupling adenine or N-7/N-9-substituted
adenine (1-5) with commercially available biphenyl carbonyl chlorides (Scheme 1) or substituted
biphenyl carboxylic acids (17, 23a-e) (Schemes 2 and 3). When using carboxylic acids, an excess of
the activation reagent, N,N’-carbonyldiimidazole (CDI), was required as well as a catalytic amount
of 4-dimethylaminopyridine (DMAP), and the coupling reaction was performed in anhydrous
dimethylformamide (DMF) at 100 °C until starting material was consumed (HPLC monitoring). In
addition, due to the limited thermal stability of DMF at elevated temperature (leading to side
reactions such as dimethyl amination and/or formylation), DMF was replaced by N-
methylpyrrolidinone (NMP) for prolonged reaction time at temperatures over 100 °C. In a few cases,
low yields were observed for this coupling step and were associated to the fact that chromatographic
separation of the starting material and the desired compound was difficult. In addition, chemical
instability was observed for compound 7a, which was only obtained in 81.8% purity.
Alkylation reactions (Scheme 1) were carried out by reacting adenine with the appropriate
commercially available alkyl halide in DMF, using potassium carbonate as base. When using benzyl
bromide or benzylchloromethyl bromide, formation of both N-7/N-9-regioisomers (compounds 1a-b
and 2a-b) was observed and they were separated by silica gel chromatography. The N-7-derivatives
1a and 2a were obtained in 20 and 43% yield, respectively. The regioselectivity of alkylation was
unequivocally assigned by 1HNMR and UV spectra. In DMSO-d6, pronounced downfield shifts were
observed for the signals corresponding to the H-8 and NH2 of N-7 when compared with those of N-9
regioisomers. UV absorption spectra of N-7 isomers showed slight bathochromic shifts (2-3 nm)
compared to N-9 regioisomers. All these data are in agreement with data from the literature and
corresponding to similar N-7 and N-9 purine isomers [26, 27].
The synthesis of N-(phosphonomethyl) derivatives 5 (scheme 1 for N-9 adenine substitution) or 15
(scheme
2
for N-1-imidazole substitution) was performed employing diethyl p-
toluenesulfonyloxymethylphosphonate as required alkylating reagent [28].
9

ACCEPTED MANUSCRIPT
Scheme 1. Synthesis of N-7 or N-9 substituted derivatives of UA2132.
Reagents and conditions: (i) K2CO3, benzyl bromide (for 1a-b) or benzylchloromethyl ether (for
2a-b) or 2-bromoethylacetate (for 3) or ethyl-2-bromoacetate (for 4), DMF, rt, 1-48 h or NaH,
diethyl p-toluenesulfonyloxymethylphosphonate (for 5), DMF, 50 °C, 17 h; (ii) biphenyl-3-carbonyl
chloride, pyridine, rt or 100 °C, 1-4 h. (iii) NaOH 2M, MeOH, 0 °C, 30 min (for 11 and 12) or
trimethylsilyl bromide (TMSBr), DMF, rt, 48 h (for 13).
Cleavage of the phosphonate diethyl ether groups in the final step of synthesis of compounds 13
(scheme 1) and 20 (scheme 2) was carried out using trimethylsilyl bromine (TMSBr) and the desired
compounds were obtained as their corresponding phosphonic acids following purification on C18
reverse phase and eventually cation-exchange chromatography to isolate them as sodium salts. One
should notice that during the saponification of compound 16 (Scheme 2), concomitant loss of one
ethyl group occurred and led to the phosphonate monoethyl derivative 17. Moreover, attempts to
obtain N-imidazolylmethylphosphonate derivative 19 incorporating a benzyloxymethyl group
(BOM) on the N-7 position of adenine was tedious (yield 6%) due to a relative instability of the
BOM group.
For the series of compounds containing a tetrazole ring (compounds 24a-e and 25, scheme 3),
intermediates 22a-e have been prepared by 1,3-dipolar cycloaddition of nitrile precursors (21a-e)
with sodium azide and trimethylamine hydrochloride in NMP. The cyano-substituted biphenyl esters
21a-e were obtained beforehand according to the previously described Suzuki cross-coupling
protocol [19] using commercially available starting materials such as 2-, 3- or 4-bromobenzonitrile
10

ACCEPTED MANUSCRIPT
and 2-, 3- or 4-ethoxycarbonylbenzeneboronic acids. For compounds 24a, 24d, and 25, the coupling
reaction with the adenine moiety did not go to completion and low yields may be attributed to the
steric hindrance associated to the ortho orientation of the substituents.
Scheme 2. Synthesis of Ilib-72 analogues incorporating a polar substituent on the imidazole ring.
Reagents and conditions: (i) NaH, p-toluenesulfonyloxymethylphosphonate, DMF, 80 °C, 18 h; (ii)
3-ethoxycarbonylbenzeneboronic acid, Pd(PPh3)4, K2CO3, DMF, 100 °C; (iii) NaOH 2 M, EtOH,
1,4-dioxane, 80 °C. (iv) CDI, DMAP, adenine (for 18) or 2a (for 19), DMF, 100 °C. (v) TMSBr,
DMF, rt.
Scheme 3. Synthesis of derivatives incorporating a tetrazole ring.
Reagents and conditions: (i) 2-, 3- or 4-ethoxycarbonylbenzeneboronic acid, Pd(PPh3)4, K2CO3,
DMF, 100 °C; (ii) NaN3, NEt3.HCl, NMP, 150 °C (iii) NaOH 2 M, EtOH, 1,4-dioxane, 50 °C; (iv)
CDI, DMAP, adenine or 4 or 5, DMF or NMP, 80-150 °C.
Enzymatic studies. All synthesized compounds were evaluated for their inhibitory effect against
recombinant cN-II using the malachite green-based assay with quantification of inorganic phosphate
11

ACCEPTED MANUSCRIPT
produced from IMP (Tables 1-3) and the cLogP were calculated in order to estimate the impact of
the various modifications on aqueous solubility, in comparison to UA2132 used as a reference.
When introducing modifications in the N-7 position (Table 1), no real improvement of the
inhibitory effect could be observed (1a, 2a, 6a, 6b, and 7a) as compared to UA2132. An equivalent
inhibition was observed at 200 µM for derivatives 6a and 7a which were substituted at the N-7
position either by a benzyl or benzyloxymethyl group. This may be explained by a different
orientation adopted by these derivatives in the binding site when compared to the compounds with
N-9 modifications. Substitutions on the N-9 position of the purine heterocycle by aromatic
substituents was found to reduce the activity of several compounds (6b and 7b), whereas small and
linear substituents (ethyl chain for 8 and 11 or methylcarboxylate 9 and 12, or methylphosphonate 10
and 13) increased the inhibition of the enzyme. Indeed, for the best inhibitors in this series (>60% at
200 µM) substitutions were including a carboxylic (12) or a phosphonate (13) group, especially
when the negative charge was unmasked. Indeed, the presence of a carboxylate or a phosphonate
group is expected to increase the overall aqueous solubility as anticipated (cf. cLogP). The presence
of a biphenyl moiety was mandatory for the inhibition since smaller compounds lacking this group
were almost- or fully inactive (1b, 2b, and 3).
In addition, substitution of the biphenyl moiety by a five-membered ring seems to play a major
role as shown by the derivative 18 incorporating an imidazole and a (monoethyl)-methylphosphonate
group (without any modification in N-7 or N-9, derivative 19 was less efficient than 18).
Surprisingly, the fully unprotected compound 20 (with the closest similarity with Ilib-72) did not
inhibit cN-II activity suggesting that the negative charges affect the binding mode of this compound
(Table 2). According to the enzymatic inhibition, compound 18 appeared more advantageous (75%
at 200 µM) and this may be explained by the partial negative charge compared to a fully charged
compound 20 (harboring a double negative charge).
Table 1. Data for N-7 and N-9-substituted analogues of UA2132.
Inhibition (%) at 200 µM (or
N-7 / N-9
1 mM when no inhibition at 200
µM)
Compound
Structure
cLogP*
3.4
substituent
Ki (mM)
68 ± 5%
2132**
none
Ki = 0.89 ± 0.23
(Competitive)
12

ACCEPTED MANUSCRIPT
26 ± 10%
(1 mM)
1a
1b
2a
PhCH2-(N7)
PhCH2-(N9)
1.1
1.9
1.1
23 ± 4%
(1 mM)
BnOCH2-
(N7)
No inhibition
(1 mM)
BnOCH2-
(N9)
13 ± 3%
(1 mM)
2b
1.9
6a
6b
PhCH2-(N7)
PhCH2-(N9)
4.3
52 ± 2%
57 ± 9%
50 ± 1%
5.1
4.2
BnOCH2-
(N7)
7a***
BnOCH2-
(N9)
7b
5.0
34 ± 4%
13

ACCEPTED MANUSCRIPT
No inhibition
(1 mM)
3
(N9)
0.4
3 ± 3%
4
5
(N9)
(N9)
0.5
0.3
No inhibition
(1 mM)
8
(N9)
3.6
62 ± 4%
11
(N9)
2.8
46 ± 7%
60 ± 1%
9
(N9)
3.7
62 ± 4%
12
(N9)
2.7
Ki = 0.84 ± 0.16
(competitive)
47 ± 5%
10
(N9)
3.5
14

ACCEPTED MANUSCRIPT
n.a.
2.2
13
(N9)
Ki = 2.4 ± 0.41
(competitive)
*clogP values were calculated with MolInspiration (freely available online software:
http://www.molinspiration.com/cgi-bin/properties)
**Data for UA2132 are from [19]. n.a. when precipitation occurs upon addition of the green malachite
reagent. Compounds promoting significant enzyme inhibition are indicated in bold.
*** Compound 7a was 81.8% pure.
Table 2. Data for analogues incorporating a polar substituent on the imidazole ring
N-7 / N-9
Inhibition (%) at 200 µM
cLogP*
Compound
Structure
Ki (mM)
substituent
75 ± 5%
Ki= 0.58 ± 0.08
18
None
2.5
(competitive)
Single negative charge as
Ilib-72
20
None
1.8
49 ± 4%
BnOCH2-
(N7)
3.33
19
60 ± 5%
Compounds promoting significant enzyme inhibition are indicated in bold
*clogP values were calculated with MolInspiration (freely available online software:
http://www.molinspiration.com/cgi-bin/properties)
15

ACCEPTED MANUSCRIPT
Finally, tetrazolo derivative 24c (and to a lower extend 24b) was a good mimicry of carboxylate
group, when introduced in the meta or para orientation, confirming the improvement observed with a
carboxylate group which promotes an increased solubility of these compounds associated with an
enhanced activity (Table 3). According to the position of the tetrazolo group (ortho, meta or para
regarding to the biaryl group), there was a large difference in terms of enzymatic inhibition. The
most favorable substitution was in the para position (24c) followed by meta with less efficiency
(24b).
Table 3. Data for analogues incorporating a tetrazole as five-membered ring
N-7 / N-9
Inhibition (%) at 200 µM
cLogP*
Compound
Structure
substituent
Ki (mM)
24a
None
2.8
30 ± 4%
24b
24c
None
None
2.8
2.8
50 ± 2%
60 ± 3%
N
HN
N
N
O
N
HN
N
N
N
H
24d
24e
None
None
2.8
2.8
10 ± 2%
27 ± 5%
16

ACCEPTED MANUSCRIPT
25
(N9)
3.1
27 ± 6%
Compounds promoting significant enzyme inhibition are indicated in bold
*clogP values were calculated with MolInspiration (freely available online software:
http://www.molinspiration.com/cgi-bin/properties)
Thus, inhibition of the 5’-nucleotidase activity was observed for ten derivatives, showing 50 to 75%
inhibition of the purified recombinant protein at 200 µM. Among these, two derivatives (12 and 18)
were further studied to determine inhibition constants and mode of inhibition. They both exhibited Ki
in the sub-millimolar range, respectively 0.84 and 0.58 mM (Fig. 6) and were found as competitive
inhibitors (Figure S1). In addition, they appeared as potent as our reference compound UA2132 (Ki
=0.89 mM) with a slight advantage for 18 which is very close to the Ilib-72 chemical structure.
40
a
b
50
40
30
20
10
0
30
20
10
0
[Compound #12] = 0 mM
[Compound #12] = 1 mM
[UA2132] = 0 mM
[UA2132] = 1mM
0
1
2
3
4
0
1
2
3
4
[IMP] (mM)
[IMP] (mM)
0.12
0.1
0.2
0.15
0.1
0.08
0.06
0.04
0.02
0
0.05
0
-2
0
2
4
6
-2
0
2
4
6
1 / [IMP] (mM-1)
1 / [IMP] (mM-1)
17

ACCEPTED MANUSCRIPT
c
d
40
40
30
20
10
0
30
20
10
[Compound #18] = 0 mM
[Compound #18] = 1 mM
[Compound #13] = 0 mM
[Compound #13] = 1 mM
0
0
1
2
3
4
0
1
2
3
4
[IMP] (mM)
[IMP] (mM)
0.2
0.2
0.15
0.1
0.15
0.1
0.05
0
0.05
0
-2
0
2
4
6
-2
0
2
4
6
1 / [IMP] (mM-1)
1 / [IMP] (mM-1)
Figure 6. Michaelis-Menten kinetic profiles (with the corresponding Lineweaver-Burk plots)
obtained by using a competitive inhibition mode for compounds UA2132 (a), 12 (b), 13 (c) and 18
(d). Analysis of the kinetic inhibition curves (inhibition modes and equations used for fitting) are
detailed in supporting information.
Crystallography. A selection of the potential inhibitors (depending on their in vitro activity,
availability and solubility) was used for co-crystallization and soaking experiments at final ligand
concentrations of 5 and 10 mM. Diffraction data were collected on crystals soaked in solutions
containing UA2132, 2a, 7a, 12 and 20, respectively. All crystal structures solved within this study,
display the same overall fold as those previously reported with an α/β core-domain and a cap-domain
(Figure 7). Of the various studied crystal structures (Table S3) only those determined from crystals
soaked in a solution containing 10 mM of 2a or UA2132 displayed electron densities which partly
could be interpreted as being the two respective ligand (Fig. S2), at least in regions being visible
within the electron density map. Indeed, the region 401-416 is always reported as being highly
18

ACCEPTED MANUSCRIPT
disordered in form I crystals [15, 22, 25] whereas it was visible in the electron density of form II
crystals [19].
When taking the data from the previous enzyme kinetic studies into account, one cannot exclude
binding of the other flexible parts of the enzyme for the other molecules for which electron density is
lacking, at least for 2a, 13 and 18 for which no crystals could be obtained by co-crystallization in
either forms.
Figure 7. Overall three-dimensional structure of cN-II with the core-domain colored in red and the
cap-domain colored in teal. Binding sites observed so far within various crystal structures are in
purple the active site [22], in orange the “AdiS” binding site,[15] in dark blue effector site 1 and in
yellow effector site 2 [25], in wheat the “2c” site and in pink spanning over a longer range the “3c”
binding site [19].
Binding mode prediction to understand the inhibition mechanism. As no crystal structure with
a ligand bound in an unequivocal manner could be obtained, molecular docking was used to
understand the mechanism by which the most efficient compounds inhibit the enzyme. Compounds
12 and 13 exhibit very similar binding modes as predicted by docking (Fig. 8A). Both compounds
were found superimposed in the active site with the carboxylic or phosphoric acid chain pointing
towards the magnesium ion. This prediction cannot explain the difference in terms of enzymatic
inhibition (Ki of 0.84 and 2.4 mM, respectively). However, the docking score was a good indication
of the binding strength with 112.6 versus 145.9 for the carboxylic acid group and the phosphoric acid
chain, respectively. This stronger theoretical binding affinity for derivative 13 can be easily
explained by the overall negative charge of the compound (-2) and it should have been predicted as a
19

ACCEPTED MANUSCRIPT
better inhibitor. However, and according to the inhibition values, the inhibition efficiency seems to
be governed by a more complicated mechanism. One can imagine that the higher water solubility of
compound 13 may prevent or limit the access to the target binding site as the surrounding residues
are aromatic (F157, H209 and Y210) and also because best inhibitors were highly hydrophobic.
When comparing their binding mode with Ilib72 (Fig. 8B), no obvious similarity was observed with
a large shift of the biaryl moiety likely due to the differential positioning of the negative charge on
these compounds.
In contrast to compounds 12 and 13, the overlay between Ilib-72 and 18 or 20 (Fig. 8C) showed
interesting resemblances with nearby positioning of their adenine and biaryl groups. Intriguingly,
derivative 18 was determined more active in vitro than 20. This raises the question about the overall
negative charge of the compound and may indicate that a single negative charge is more suitable for
enzymatic inhibition (identically to Ilib-72). Interestingly, the coordination of the magnesium ion
changed depending on the compound (18 or 20), probably due to the different electron distribution
around the phosphate-bound oxygen atoms. Several water molecules were displaced and involved in
this coordination. Moreover, an important hydrogen bond network formed between the carbonyl
oxygen and water molecule is present for compound 18 while these interactions are absent for
compound 20 (Fig. 8D).
B
A
N52
D356
D351
20

ACCEPTED MANUSCRIPT
H209
C
D
Y210
N158
N52
T155
D351
D356
H352
Figure 8. Prediction of the binding modes of selected compounds by docking on 2XCW structure.
(A) Superimposition of UA2132 (orange), 12 (pink) and 13 (yellow) with negative charge near Mg2+
ion. (B) Same as in A with Ilib-72 (cyan sticks). (C) Binding modes of compounds 18 (blue), 20
(green) and Ilib-72 (cyan sticks). (D) Hydrogen bonding network with water molecules connecting
compound 18 (blue stick) to the magnesium ion (depicted as a purple sphere) whereas compound 20
is making only hydrogen bonds with N158 and T155.
Concerning derivatives 24a-e incorporating a tetrazole group, their binding orientation seems to be
governed by this group. Indeed, its interaction with the magnesium ion is thought to be strong and
the orientation of the rest of the molecule will depend on the ortho, meta or para position of this
substituent on the biaryl skeleton (Fig. 9A). Indeed, the five derivatives were predicted with the
tetrazolo group near the metallic ion with various orientation of the adenine moiety depending on
biaryl substitution. According to the docking predictions, compounds 24a, 24d and 24e bind to the
enzyme in a complete different manner, which may indicate a large variation in their binding
affinity. Attempting to explain the higher efficacy of compound 24c compared to 24b (Fig. 9B), a
similar hypothesis as previously expressed for compounds 18 and 20 can be postulated with the
implication of water molecules reinforcing the interactions between protein amino acids and the
inhibitor (Fig. 9B).
21

ACCEPTED MANUSCRIPT
A
Y21
N5
D35
N158
R20
D35
H35
Y15
B
Y21
N5
D35
N158
D35
R20
H35
Y15
Figure 9. (A) Effect of the tetrazole group (highlighted by a red circle) on the final orientation of
derivatives 24a (green), 24b (blue), 24c (yellow) and 24d (cyan) and 24e (white). (B) Close-up view
of the binding mode prediction of the two most active compounds from the tetrazolo series, 24b
(blue) and 24c (yellow). The magnesium ion is depicted as a green sphere and surrounding protein
amino acids are in thin sticks. Distances are color-coded for each compound.
In vitro biological evaluation. The effect of the newly synthesized compounds was evaluated on
the survival of different human cancer cell lines using a MTT (3-(4,5-dimethylthiazol-2-yl)-2,5-
diphenyltetrazolium bromide) based colorimetric assay (Table S4). IC50 values ranged from 36 µM
to >300 µM, similarly as for compounds previously reported [19]. Four N7- or N9-substituted
22

ACCEPTED MANUSCRIPT
compounds (6a, 6b, 7a and 10) and one tetrazolo-compound (24c) had IC50 below 100 µM on RL
cells. These derivatives showed cN-II inhibition of approximately 50% at 200 µM and cLogP values
allowing diffusion through cell membrane (>2.8), but there was no clear structural explanation of this
antiproliferative activity in regard to cN-II inhibition.
Intrinsic antiproliferative activity allowed us to perform synergy experiments with selected
compounds and purine nucleoside analogues as our hypothesis states that cN-II inhibitors would
enhance the effect of cytotoxic nucleoside analogues in co-exposure conditions (Fig. 10). Synergy
(defined as CI95<0.9) was observed with cladribine for 6a (RL and HL-60), 6b (CCRF-CEM and
HL-60), 7a (RL) and 10 (RL), with clofarabine for 6a (RL), 6b (HL-60) and 7a (RL) and with
fludarabine for 6a (RL and CCRF-CEM) and 6b (RL, CCRF-CEM and HL-60). The differences
between the cell lines is consistent with earlier observations showing that shRNA-mediated cN-II
inhibition is associated with a higher level of sensitization to purine nucleoside analogues in RL and
HL-60 cells as compared to CCRF-CEM cells [10]. Thus, based on these results, 6a could be
considered as a good candidate and clearly better than our previous inhibitor (UA2132). It is
however noteworthy that no co-incubation experiments with non-toxic compounds were performed
and could have allowed the identification of other active associations.
23

ACCEPTED MANUSCRIPT
5
4
3
2
1
0
A
B
C
Cladribine Clofarabine Fludarabine
2132
6a
6b
7a
10
Cladribine Clofarabine Fludarabine
2
1
0
2132
6a
6b
3
Cladribine Clofarabine Fludarabine
2
1
0
2132
6a
6b
Figure 10. Combination index 95 for associations of purine nucleoside analogues cladribine (black
bars), clofarabine (dark grey bars) or fludarabine (light grey bars) and cN-II inhibitors on RL (A),
CCRF-CEM (B) and HL-60 (C) cells. CI95 values (a parameter for the quantification of the
combinatory effect of the two compounds to achieve 95% inhibition of cell survival) were calculated
with the CompuSyn software based on the Chou and Talalay method [29] as indicated in Material
and Methods, and expressed as 10^(mean(logCI95)) from at least three independent experiments.
24

ACCEPTED MANUSCRIPT
Error bars indicate CI95-SEM and CI95+SEM. Drug combinations with CI95 < 0.9 are considered
synergistic, whereas they are additive for 0.9 < CI95 < 1.1 and antagonistic for CI95 > 1.1.
CONCLUSION
By using a combination of various in silico approaches, we designed a large chemical library of
potential cN-II inhibitors with numerous different scaffolds of which some leading to inaccessible
chemical synthesis. Consequently, we focused on the structural optimization of a previously
characterized inhibitor that was also part of the Ilib-generated compound (Ilib-72). This allowed us
to explore the chemical diversity of this lead compound in order to improve its affinity for the
targeted protein and/or physico-chemical properties such as solubility. Twenty-six derivatives were
synthesized, evaluated toward the purified recombinant enzyme. About ten compounds could inhibit
more than 50% of the enzymatic activity at a concentration of 200 µM. In comparison with our
reference compound (UA2132, a previously identified competitive inhibitor), three derivatives (12,
13 and 18) displayed similar potency, with Ki in the micromolar range and a competitive mechanism
of action. All three compounds have a polar and anionic substituent (either a carboxylate or a
phosphonate) and derivative 18 presented the closest structural similarity with Ilib-72, the derivative
selected from the library generation and in silico screening process. Molecular modelling and X-ray
crystallography studies were carried out to predict and assess the binding modes of these novel
derivatives. When tested in cell culture experiments, synergistic effects were observed with all
cytotoxic nucleoside analogues tested and compounds 6a-b. Some compounds had minor
cytotoxicity, and their complete toxicity profile on non-cancerous cells will be needed for further
development. These results are therefore encouraging to investigate further the potential of these
compounds as to our knowledge they appeared as the most potent cN-II inhibitors reported to date.
EXPERIMENTAL SECTION
Library generation and molecular docking studies. Using hit fragments previously identified by
NMR screening [19], a focused library of final compounds targeting cN-II was generated using Ilib-
diverse program (www.inteligand.com) by randomly assembling three different hit fragments. A
weighting percentage has been attributed for the fragments that were previously identified as cN-II
inhibitors (Table S1 in SI). In addition, several functional or polar groups were included to increase
the solubility of the final compounds (hydroxyl, sulfonate, phosphate and phosphonate groups). This
procedure allowed us to generate 3,000 compounds supposed to be chemically synthesizable (the
program includes general rules of chemical synthesis during the linkage process). All derivatives
25

ACCEPTED MANUSCRIPT
were analyzed and their coordination and geometry corrected by using the VegaZZ molecular
modelling program [30]. Atomic charges were assigned using the Gasteiger-Marsili empirical atomic
partial charges [31] and the potential energy of all compounds was minimized using 500 steps of
steepest descent followed by 5,000 steps of conjugate gradient (tolerance of 0.01 kcal/mol. Å). For
compounds comprising a tetrazole group, only one tautomer has been evaluated by docking.
Molecular docking was carried out with the GOLD v5.2 program (Genetic Optimization for Ligand
Docking) from Cambridge Crystallographic Data Centre (CCDC) Software Limited [32, 33] with
partial flexibility of protein amino acid sidechains and full flexibility of Ilib-diverse generated
compounds. Docking was performed on the cN-II crystal structure in which the nucleotidase was
complexed with IMP (PDB 2XCW) by applying 50 genetic algorithm steps for conformational poses
searching with a radius of 12 Å around the target atom, Mg2+ ion (located in the substrate binding
site). Although the 2XCW crystal structure harbors the D52N mutation, it was chosen among the
numerous available structures because of the accurate characterization of the substrate binding mode.
Our earlier docking studies performed using the 2JC9 structure without substrate bound in the active
site and compared to those performed using 2XCW in this study showed highly similar results, thus
giving a proof-of-concept for using this structure of an inactive mutant. Structural water molecules
present in the crystal structure near the IMP binding site were retained and allowed to contribute with
a 2 Å cut-off of translational and rotational freedom. Docking poses were analyzed by the clustering
method (complete linkage) from the rmsd matrix of ranking solutions. Solutions were classified
according to their respective scores calculated by the Goldscore scoring function. Structural analysis
and visualization of docking poses were achieved with the PyMOL Molecular Graphics System
(version 1.3, Schrödinger, LLC).
Enzyme inhibition and steady state kinetics assays. The nucleotidase activity was measured
using the Green Malachite Phosphate Assay kit (Gentaur) as previously described [16, 19]. Briefly,
in 96-wells plate arranged on ice, recombinant purified cN-II was added to a final concentration of
0.1 µM in 80 µL buffer containing 50 mM imidazole pH 6.5, 500 mM NaCl, 10 mM MgCl2 and then
incubated for 10 min with the different compounds to be evaluated as potential cN-II inhibitors. The
reaction was started by addition of the substrate (50 µM of IMP with an equimolar amount of Mg)
and incubated at 37° C for 5 min. The reaction was stopped by addition of 20 µL of Green malachite
reagent (strong acid) and free phosphate was quantified by reading the absorbance at 630 nm on a
plate reader (Tecan Sunrise). The percentage of enzymatic inhibition has been calculated by using
the following formula: % inhibition = 1 - [(x – Min) / (Max - Min)]. Min and Max refer to the
absorbance of full-inhibition and Max the signal in absence of inhibitor, respectively. For the most
26

ACCEPTED MANUSCRIPT
active compounds, Michaëlis-Menten kinetics were performed to determine the inhibition mode and
inhibition constants, as previously described [19]. Briefly, enzyme (alone or in presence of
compounds) and substrate (IMP at different concentrations varying from 0.2 to 3 mM) were mixed in
a thermostatically controlled beaker under magnetic stirring at 37 °C (same buffer as for the previous
assay). Every 7 s, samples were taken from the mixture and reaction was stopped by adding 10% of
perchloric acid. Quantification of IMP and inosine was performed by HPLC (Waters Alliance) using
a Partisphere 5-SAX column (AIT France) and 10 mM ammonium phosphate buffer pH 5.5 as
mobile phase. The raw data (IMP and inosine peaks area integrated with Empower software, Waters)
were analyzed using Grafit 7 (Erithacus Software) and fitted with equations describing either a
competitive, a non-competitive, an uncompetitive or a mixed inhibition mechanism. The equation
that reproduced best the experimental data was selected for the determination of the inhibition mode
by also taking into account the chi square value (confidence computed from the fitting giving the
independence statistical test between experimental and theoretical values).
Cell-based experiments. Cell survival in presence of cN-II inhibitors and purine nucleoside
analogues alone (intrinsic cytotoxicity) or in combination (synergy experiments) was determined
with human follicular lymphoma cells (RL), human acute lymphoblastic leukemia cells (CCRF-
CEM) and human acute promyelocytic leukemia cells (HL-60). as described earlier [15, 19]. Briefly,
ten thousand cells were seeded per well in 96-well plates and incubated with different concentrations
of drugs and inhibitors alone or in fixed ratio combinations for 72 hours, and cell survival was
determined with the MTT assay. Inhibitory concentrations 50 (IC50, concentrations associated with
50% cell survival as compared to unexposed cells) and combination index 95 (CI95, a parameter for
the quantification of the combinatory effect of the two compounds to achieve 95% inhibition of cell
survival) were calculated with the CompuSyn software based on the Chou and Talalay method. [29]
Drug combinations with CI95 < 0.9 are considered synergistic, whereas they are additive for 0.9 <
CI95 < 1.1 and antagonistic for CI95 > 1.1.
Crystallization, data reduction, structure determination and refinement. Protein used for X-
ray crystallography studies was prepared as described earlier [34] and concentrated to 8 mg/ml for
crystallization. All compounds used for crystallization were dissolved in 1% DMSO to a
concentration of 50 mM. Both nanoliter- (1:1 ratio of 200 nL protein: reservoir solution employing a
Mosquito™ robot (TTP Labtech) with 96 well plates and microliter crystallization experiments with
24 well plates were performed using the sitting drop vapor diffusion method, and “form I” [22]) and
“form II” [19] experimental conditions. Freshly purified C-terminal truncated cN-II (see ref. [15] 8
27

ACCEPTED MANUSCRIPT
mg/mL) was mixed with compounds UA2132, 2a, 7a, 12, 13, 18 and 20 to final concentrations of 5
and 10 mM of these latter and incubated at 4 °C for 20 minutes prior to co-crystallization setup.
Crystals grown in the presence of compounds (co-crystallization) appeared upon 1-2 weeks. Only
compounds 7a and 20 could be co-crystallized, and in forms II and I, respectively. For soaking
experiments (conducted in parallel), native crystals (both forms) appeared after approximately one
week at 4 °C. Only form I crystals were further studies due to an extreme fragility of form II crystals
which fell apart upon soaking. For cryo-protection, crystals were transferred to a drop containing the
crystallization reservoir solution (at the above mentioned conditions) supplemented with 10% and
15% glycerol for “form I” and “form II” crystals, respectively, and flash-frozen in liquid nitrogen.
Hereafter, X-ray diffraction data were collected at the European Synchrotron radiation facility
(ESRF Grenoble France), and integrated and scaled using programs from the XDS package [35]. All
crystal structures were determined using the molecular replacement method and the program
“Phaser” [36] as implemented in the “Phenix” software suite [37] and the previously reported cN-II
structures (2J2C) [22] and (5CQZ) [19] as search models for forms I and II, respectively. Refinement
was performed using the programs “Refmac5” [38] and “Phenix” [37], altering with iterative model
re-building and analysis with “COOT” [39]. Ligands and water molecules displayed 2Fo-Fc and Fo-
Fc electron densities contoured at the 1σ and 3σ level, respectively, prior to refinement. The pdb
coordinates and topology of the compounds tested for binding during refinement (though lacking in
final three-dimensional structures) were obtained using the PRODRG2 server [40] and the program
“eLBOW” as implemented in “Phenix”. Libraries for refinement of these molecules were generated
using the program LIBCHECK as implemented in the CCP4 program package [41]. Composite omit
maps were calculated (“Phenix”) in order to reduce model bias, and rmsd’s were calculated with the
program TM-score [42]. Data collection and refinement statistics are given in Table S3.
Coordinates and structure factors of the cN-II structures have been deposited at the RCSB
(Research Collaboratory for Structural Bioinformatics) Protein Data Bank (http://www.rcsb.org)
under the entry codes 6FIR (cN-II-2132, 5 mM), 6FXH (cN-II-2132, 10 mM), 6FIS (cN-II-2a),
6FIW (cN-II-20) and 6FIU (cN-II-12), respectively.
Description of general methods for chemical synthesis is already published [19]
Diethyl p-toluenesulfonyloxymethylphosphonate [28], compounds 1a-b, [43], 2a-b [44], 3 [45]
and 4 [46] were prepared according to previously published procedures.
28