A combined in silico/in vitro approach unveils common molecular requirements for efficient BVDV RdRp binding of linear aromatic N-polycyclic systems

Article abstract of DOI:10.1016/j.ejmech.2016.03.080

In this work, we present and discuss a comprehensive set of both newly and previously synthesized compounds belonging to 5 distinct molecular classes of linear aromatic N-polycyclic systems that efficiently inhibits bovine viral diarrhea virus (BVDV) infe

Full text of DOI:10.1016/j.ejmech.2016.03.080

Accepted Manuscript
A combined in silico/in vitro approach unveils common molecular requirements for
efficient BVDV RdRp binding of linear aromatic N-polycyclic systems
A. Carta, I. Briguglio, S. Piras, P. Corona, R. Ibba, E. Laurini, M. Fermeglia, S. Pricl,
N. Desideri, E.M. Atzori, P. La Colla, G. Collu, I. Delogu, R. Loddo
PII:
S0223-5234(16)30255-0
10.1016/j.ejmech.2016.03.080
EJMECH 8505
DOI:
Reference:
To appear in: European Journal of Medicinal Chemistry
Received Date: 20 January 2016
Revised Date: 16 March 2016
Accepted Date: 25 March 2016
Please cite this article as: A. Carta, I. Briguglio, S. Piras, P. Corona, R. Ibba, E. Laurini, M. Fermeglia,
S. Pricl, N. Desideri, E.M. Atzori, P. La Colla, G. Collu, I. Delogu, R. Loddo, A combined in silico/in
vitro approach unveils common molecular requirements for efficient BVDV RdRp binding of linear
aromatic N-polycyclic systems, European Journal of Medicinal Chemistry (2016), doi: 10.1016/
j.ejmech.2016.03.080.
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

ACCEPTED MANUSCRIPT
A combined in silico/in vitro approach unveils
common molecular requirements for efficient
BVDV RdRp binding of linear aromatic N-
polycyclic systems.
A. Carta†,■,*, I. Briguglio†, S. Piras†, P. Corona†, R. Ibba†, E. Laurini§,■,#, M. Fermeglia§,#, S.
Pricl§,#,*, N. Desideri♦, E. M. Atzori♦, P. La Colla○, G. Collu○, I. Delogu○, R. Loddo○,■
† Dipartimento di Chimica e Farmacia, Università degli Studi di Sassari, via Muroni 23A, 07100
Sassari (SS)
§
Molecular Simulations Engineering (MOSE) Laboratory, DEA, University of Trieste, Piazzale
Europa 1, 34127 Trieste, Italy
#
National Interuniversity Consortium for Material Science and Technology (INSTM), Research
Unit MOSE-DEA, University of Trieste, Piazzale Europa 1, 32127 Trieste, Italy
♦ Dipartimento di Chimica e Tecnologie del Farmaco, Sapienza, Università di Roma, P.le Aldo
Moro, 5, 00185 Roma, Italy
○
Dipartimento di Scienze Biomediche, Sezione di Microbiologia e Virologia, Università
Cagliari, Cittadella Universitaria, 09042 Monserrato, Italy
1

ACCEPTED MANUSCRIPT
KEYWORDS: Bovine viral diarrhea virus (BVDV); RNA-dependent RNA polymerase (RdRp)
inhibitors; Imidazo[4,5-g]quinoline, Pyrido[2,3-g]quinoxaline, Aromatic N-polycyclic systems
ABSTRACT: In this work, we present and discuss a comprehensive set of both newly and
previously synthesized compounds belonging to 5 distinct molecular classes of linear aromatic
N-polycyclic systems that efficiently inhibits bovine viral diarrhea virus (BVDV) infection. A
coupled in silico/in vitro investigation was employed to formulate a molecular rationale
explaining the notable affinity of all molecules to BVDV RNA dependent RNA polymerase
(RdRp) NS5B. We initially developed a three-dimensional common-feature pharmacophore
model according to which two hydrogen bond acceptors and one hydrophobic aromatic feature
are shared by all molecular series in binding the viral polymerase. The pharmacophoric
information was used to retrieve a putative binding site on the surface of the BVDV RdRp and to
guide compound docking within the protein binding site. The affinity of all compounds towards
the enzyme was scored via molecular dynamics-based simulations, showing high correlation
with in vitro EC50 data. The determination of the interaction spectra of the protein residues
involved in inhibitor binding highlighted amino acids R295 and Y674 as the two fundamental H-
bond donors, while two hydrophobic cavities HC1 (residues A221, I261, I287, and Y289) and
HC2 (residues V216, Y303, V306, K307, P408, and A412) fulfill the third pharmacophoric
requirement. Three RdRp (K263, R295 and Y674) residues critical for drug binding were
selected and mutagenized, both in silico and in vitro, into alanine, and the affinity of a set of
selected compounds towards the mutant RdRp isoforms was determined accordingly. The
agreement between predicted and experimental data confirmed the proposed common molecular
rationale shared by molecules characterized by different chemical scaffolds in binding to the
2

ACCEPTED MANUSCRIPT
BVDV RdRp, ultimately yielding compound 6b (EC50 = 0.3 µM; IC50= 0.48 µM) as a new,
potent inhibitor of this Pestivirus.
Introduction
DNA and RNA viruses represent a major cause of disease in humans and animals. Some of
them, notably the Ebola and, previously, the H1N1 viruses expressed their ability to create a
pandemic scenario across globe, highlighting our unpreparedness to face infections caused by
neglected pathogens that can easily spread worldwide. In this panorama, our multiannual
program of antiviral research has focused on the design and synthesis of a wide series of small
molecules endowed with antiviral activity. Various lead compounds emerged during the years,
mainly active against representative viruses of Flaviviridae family.
The Flaviviridae is a large family of related positive-strand RNA viruses (ssRNA+) that consists
of three genera: Flavivirus, Pestivirus and Hepacivirus. Several globally important human and
animal pathogens[1] are comprised within this group. The family members are genetically
diverse but share similarities in genome organization, mechanisms of gene expression, and
replication strategies.[2]
The Flavivirus genus currently contains 73 viruses classified as 53 species [3] divided into four
groups: the tick-borne Flaviviruses, the mosquito-borne Flaviviruses, the not-known-vector
Flaviviruses, and the non-classified Flaviviruses. A number of the arthropod-borne Flaviviruses
cause disease in humans. Particularly, Dengue Fever (DFV), Yellow Fever (YFV), West Nile
(WN), and Japanese Encephalitis (JE) viruses cause acute febrile illness, encephalitis, and
hemorrhagic fevers, against which effective vaccines or established drug treatments for related
3

ACCEPTED MANUSCRIPT
disease prevention and cure are still lacking. Attenuated YFV-based vaccines represent a
noteworthy exception in this setting, although, despite their effectiveness, they sometimes shows
toxic side effects that can be fatal,[4] and their utilization is incomplete in many areas.[5]
The Pestivirus genus includes animal pathogens of major economic impact for the livestock
industry, such as Swine Fever virus (SFV), Border Disease virus (BDV), and Bovine Viral
Diarrhea virus (BVDV).[6]. Particularly, BVDV establish a persistent infection in cattle that
remain viremic throughout life and often succumb to fatal mucosal disease.[7] Furthermore,
BVDV shows the ability to cross the placenta of susceptible animals causing a variety of fetal
infections.[8] In the United States alone, a recent study revealed that the cost of exposing calves
to BVDV in the feed-yard was $67.49/head. The vast majority of this amount, $58.83, was due
to the loss in performance, primarily decreased efficiency. The remainder ($8.66) reflected an
increase in mortality. The bottom line of this report is that, while few in number, persistently
infected calves can have an economically devastating effect in feedlot cattle.[9] BVDV
represents the prototypical Pestivirus virus and therefore is the best-characterized member of this
group. Because of the genome organization, translation, replication strategy, and protein
functions of Pestiviruses, BVDV is more closely related to Human Hepatitis C (HCV) virus than
any of the classical Flaviviruses. As such, it has been widely adopted as a surrogate model for
anti-HCV drug design and development.[10, 11] Lastly, BVDV is easily cultured in vitro, where
it undergoes a complete replication cycle, and molecular clones are available for genetic studies.
The Hepacivirus genus features only HCV, the major cause of human hepatitis, globally.[12]
The most recent World Health Organization (WHO) report estimates that 130–150 million
people globally have chronic hepatitis C infection.[13]
4

ACCEPTED MANUSCRIPT
At the moment, patients with severe cases of Flavivirus infections are treated only by
supportive care, which includes intravenous fluids, hospitalization, respiratory support, and
prevention of secondary infections.[14] In fact, despite worldwide extensive research and public
health concern associated with Flavivirus diseases, no specific/routine treatment is available
today to combat any Flavivirus infections, the sole exception being constituted by new drugs
(e.g., simeprevir and sofosbuvir) currently in clinical trials for hepatitis C.[15, 16] Consequently,
there is an urgent need for new lead compounds that efficiently target distinct stages of the viral
replication cycle in a virus-specific way.
In this context, and in the framework of a long-lasting antiviral research program, our group
designed and synthesized a series of new angular and linear N-polycyclic systems against
representative ssRNA+ viruses. Small molecules inhibitors of BVDV has been reported in the
literature with various scaffolds [17].
We recently reported a bicyclic system such as (E)-2-(1-(2-(2,4-dimethoxyphenyl)-1H-
benzo[d]imidazol-5-yl)ethylidene)hydrazine-1-carbothioamide (227G, Fig. 1) that inhibits, in
cell-based assays, both BVDV and HCV replication in the low micromolar range (EC50 values of
0.80 and 1.11 ꢀM, respectively) [18].
5

ACCEPTED MANUSCRIPT
Fig. 1. Chemical structure and biological activity of benzimidazole derivative 227G.
The molecular target was identified in the viral RNA dependent RNA polymerase (RdRp) [19]
and dose−response curves indicate that 227G potently inhibits (IC50 =0.0020 ± 0.0004 ꢀM) the
BVDV RdRp activity [20]. In an attempt to widen the knowledge on the anti BVDV activity of
nitrogen-containing heterocyclic compounds we complicate the bicyclic system and we
synthesized a series of [4,7]phenantroline derivatives [21] (Fig. 2). At first, we demonstrated that
some [4,7]phenantroline derivatives were fairly active against BVDV in vitro, compound A1
being particularly promising against both BVDV and HCV (Fig. 2). Next, in silico experiments
on the putative final target, the HCV RNA-dependent RNA-polymerase (RdRp) NS5B,
contributed to develop a molecular-based rationale for the structure-activity relationship (SAR)
for these compounds [21]. The target enzyme was identified in the light of some similarities
between a number of our active compounds and the non-nucleoside inhibitor of HCV NS5B
((2s)-2-[(2,4-dichloro-benzoyl)-(3-trifluoromethyl-benzyl)-amino]-3-phenyl-propionic
hereafter named Shire A [22].
acid),
6

ACCEPTED MANUSCRIPT
Fig. 2. Chemical structures of [4,7]phenantroline derivatives A.[21]
Unfortunately, the most potent angular derivatives turned out to be non-selective inhibitors of
BVDV. Compound A1, in fact, resulted active against non only BVDV but also against CVB-2
and Sb-1 (EC50 = 5 and 17 µM, respectively). With the aim of improving both their potency and
selectivity towards BVDV, we next conceived and synthesized three new classes of linear
aromatic N-tricyclic compounds derived from the expansion of the quinoline nucleus with 1,2,3-
triazole, imidazole or pyrazine (Fig. 3).[23] Thus, triazolo[4,5-g]quinoline (B), imidazo[4,5-
g]quinoline (C), and pyrido[2,3-g]quinoxaline derivatives (D) were tested in cell-based assays
for both cytotoxicity and antiviral activity against representative members of several virus
families, and found endowed with promising antiviral activity and selectivity index against
several pathogenic RNA viruses.[23]
7

ACCEPTED MANUSCRIPT
Fig. 3. Chemical structures of triazolo[4,5-g]quinoline (B), imidazo[4,5-g]quinoline (C), and
pyrido[2,3-g]quinoxaline derivatives (D).
Interestingly, while selective anti-BVDV activity was reported for all three classes of
compounds in Fig. 3, derivatives endowed with substantial anti-BVDV activity were more
numerous among imidazoquinolines and pyridoquinoxalines than among triazoloquinolines.
Nevertheless, each class contained at least one potent and selective molecule. On this basis,
selected leads from these series were then evaluated for anti-HCV activity in a replicon system
and tested in enzyme assays for activity against the RdRps of BVDV and HCV. 4-Chloro-2-(4-
nitrophenyl)-3H-imidazo[4,5-g]quinoline (C2, Fig. 4) thus emerged as the most potent and
selective compound against BVDV (EC50 = 1.2 µM, CC50> 100 µM). [32] It also proved to be
active, although only moderately selective, against HCV. These results suggested that, although
with different efficiency, compound (C2) (Fig. 4) targets both viral RdRps. Consequently, to
broad our knowledge concerning the SAR of imidazo[4,5-g]quinolines, the chemical structure of
C2 was simplified via the elimination of the central ring or the opening of the imidazole ring,
leading to various imidazopyridines (E) and N-benzylidenequinolinamines (F), respectively (Fig.
4). Remarkably, just a few imidazopyridine derivatives resulted active against BVDV, while all
N-benzylidenequinolinamines maintained their antiviral behavior. Particularly, compounds F1
8

ACCEPTED MANUSCRIPT
and F2 showed EC50 values of the same order of magnitude of the reference drug 2′-C-methyl-
guanosine (1.1 µM) [24] - as shown in Fig. 4.
Fig. 4. Chemical structure of lead compound C2 and the rationale leading to its subsequent
modifications.
The selective anti-BVDV activity of pyridoquinoxaline-based compounds (D)[23] was also
further investigated by synthesizing further derivatives, among which the thienyl derivative (D1)
9

ACCEPTED MANUSCRIPT
showed selectivity, good potency and an interesting balance between activity and cytotoxicity
(Fig. 5).
Fig. 5. Chemical structures of lead compounds D1 and D2.
Furthermore, the derivatives bearing different substituents on the N atom at position 1 or the O
atom at position 2 were tested in cell-based assays for cytotoxicity and antiviral activity. The 5-
chloro-2-methoxy-3-(thiophen-2-yl)pyrido[2,3-g]quinoxaline D2, totally devoid of cytotoxicity,
emerged for its selective and potent antiviral activity against BVDV (EC50= 1.3 ꢀM, Fig. 5).
Molecular modeling studies confirmed the BVDV RdRp as the target enzyme for these
compounds and allowed the identification of important molecular determinants for their NS5B
inhibition.[25]
To expand the SAR investigation on the above linear N-tricyclic systems, in the present paper we
initially designed and synthesized new imidazo[4,5-g]quinolines derived by further development
of the basic compound 2-phenyl-3H-imidazo[4,5-g]quinoline (C1) (EC50 = 4 µM, CC50 > 100
µM).[32] To this purpose, we modulated substituents at positions R1, R2, and/or R3, or replaced
10

ACCEPTED MANUSCRIPT
the phenyl ring with a heterocyclic moiety, as illustrated in Fig. 6. Specifically, substituents were
chosen with the aim of evaluating the influence of electron donating groups (EDGs) or electron
withdrawing groups (EWGs) on the phenyl group (R3 in Fig. 6), of the Cl atom at position R1,
and/or general steric hindrance of substituents in R2 on the antiviral activity of the resulting
molecules.
Fig. 6. Proposed rational modifications of lead compound C1 (R1 = Cl, R2 = H, R3 = H; EC50 =
4.0 ꢀM, CC50>100ꢀM).
Thus, we synthesized compounds bearing either chlorine or hydrogen at position R1 to
quantify the specific contribution of the halogen to the compound affinity for the binding site of
the BVDV NS5B. In addition, a large moiety such as methyl cyclohexyl was inserted in R2 to
test the tolerance of the viral polymerase binding pocket towards substituent bulkiness. In
combination with the two modifications mentioned above, EDGs or EWGs were inserted at the
4′ position or at the 2’ and 4’ positions of the phenyl ring (R3 in Fig. 6). Lastly, we synthesized
and assayed a furan(3)-yl derivative since no compound bearing a heterocyclic moiety at position
2 was previously reported. All modifications are indicated in details in Scheme 1.
11

ACCEPTED MANUSCRIPT
Scheme 1. Selected chemical modifications of lead compound C1 and synthetic pathways
leading to the compounds 2-5a, 6-11b, 12-13c, and 14b.
In the light of the encouraging results obtained, we next proceeded by performing a parallel in
silico/in vitro analysis of the best compounds of all previous (B-F)[23, 24] and present series and
determined the common molecular determinants leading to their successful BVDV RdRp
inhibition.
Chemistry
The synthetic route to obtain the designed 3H-imidazo[4,5-g]quinoline derivatives (2-5a, 6-11b,
12-13c, and 14b) is shown in Scheme 1. The diamines 1a-c react with appropriate bisulphite
compounds (15 and 17) in refluxed ethanol or in dimethylformammide (DMF) at 130 °C for 2-12
12

ACCEPTED MANUSCRIPT
h, to give the corresponding 3H-imidazo[4,5-g]quinolines (2-5a, 6-10b, 12-13c, and 14b) in
good yield (54-90%), with the exception of 13c (25%). Compound 4-(4-chloro-3H-imidazo[4,5-
g]quinolin-2-yl)benzenesulfonamide (11b) was in turn prepared by condensation of the
diaminoquinoline 1b with 4-sulfamoylbenzoic (16) and polyphosphoric acid for 4 h at 200 °C
(25% yield). 6,7-Diaminoquinolines 1a-c intermediates were prepared following the procedure
previously described.[23, 26] Bisulphite compounds (15 and 17), were obtained by reacting the
aldehydes with Na2S2O5 in hydroalcoholic solution.[23] Aldehydes and inorganic reagents were
commercially available.
Results and discussion
As discussed above, the original hypothesis motivating the current work proposed that specific
substituents at different positions of the scaffold of the lead compound C1 (Fig. 6) would be
selective and effective inhibitors of the BVDV RdRp. To investigate this proposition,
compounds 2-5a, 6-11b, 12-13c, and 14b were synthesized and evaluated for activity and
selectivity against BVDV and other ssRNA+, ssRNA-, dsRNA and DNA viruses.. Data from
these experiments are presented in Tables 1 and S1, respectively.
Table 1. Cytotoxicity and antiviral activity of 3H-imidazo[4,5-g]quinoline derivatives (2-5a,
6-11b, 12-13c, and 14b) against BVDV. Data represent mean values of three independent
determinations. Variation among duplicate samples was less than 15%.
MDBKa BVDVb
CC50 (µM) EC50(µM)
MDBKa
CC50 (µM)
BVDVb
EC50(µM)
Compound
S.I.
Compound
S.I.
>100
>100
>100
>100
4.0
7.0
3.4
6.0
>25.0
>14.2
>29.4
>16.7
>100
>100
20
6.1
>100
5.0
>16.4
-
2a
3a
4a
5a
10b
11b
12c
13c
4.0
33
5.0
>6.6
13

ACCEPTED MANUSCRIPT
>100
>100
>100
>100
>333.3
>50.0
>15.9
>17.2
>100
>10
5.0
2.0
1.1
1.3
>50.0
>9.1
3.8
6b
7b
8b
9b
0.3
2.0
6.3
5.8
14b
2’-MeGua
Ribavirine
a
Compound concentration (ꢀM) required to reduce the viability of mock-infected MDBK cells by 50%, as
determined by the MTT method.
b
Compound concentration (ꢀM) required to achieve 50% protection of MDBK cells from BVDV-induced
cytopathogenicity, as determined by the MTT method.
2’-MeGua = 2’-C-methyl-guanosine.
From the results shown in Table 1 it is apparent that all compounds, with the singular
exception of 11b, were significantly active, with EC50 ranging between 0.3 and 7 µM along with
low cytotoxicity (CC50 > 100 µM, except for compounds 12c and 13c, for which CC50 = 20 and
33 µM, respectively). The selectivity index (S.I.), which is a parameter of preferential antiviral
activity of a compound in relation to its cytotoxicity (CC50/EC50), of all new compounds is also
tabulated in Table 1. As can be seen, the SI index values of the title compounds were
substantially better than those of the positive controls, 2’-C-methyl-guanosine (S.I. = 9.1) and of
the gold-standard reference ribavirin, for which S.I. = 3.8.
Preliminary SARs shows that the presence of a chlorine substituent on the scaffold of
compound C1 (R1 in Fig. 6) is slightly more beneficial to the antiviral activity of the relevant
compounds (e.g. 6b, EC50 = 0.3 µM; 7b, EC50 = 2.0 µM; 8b, EC50 = 6.3 µM) with respect to
those featuring an hydrogen atom at the same position (2a, EC50 = 4.0 µM; 3a, EC50 = 7.0 µM;
4a, EC50 = 3.4 µM). Within the same molecular series, the presence of either EWGs [e.g., -CN
(2a and 6b, EC50 = 4.0 µM and 0.3 µM, respectively), –CF3 (3a and 7b, EC50 = 7.0 µM and 2.0
µM, respectively), –COCH3 (9b, EC50 = 5.8 µM), –F (5a, EC50 = 6.0 µM) and –NO2 (14c, EC50
= 5.0 µM)] or EDGs [e.g., 2,4-OCH3 (4a and 8b, EC50 = 3.4 µM and 6.3 µM, respectively), and -
14

ACCEPTED MANUSCRIPT
SCH3 (10b, EC50 = 6.1 µM)] retains the antiviral activity with respect to the lead compounds C1
(EC50 = 4 µM). Finally, although some new compounds showed a modest and scattered activity
against other than BVDV viruses (Table S1), we can reasonable identify these new quinolone
derivatives as selective and potent anti-BVDV agents.
The occurrence of a bulky substituent in R2 does not result in an appreciable effect on the
activity of the relevant compound 13c with respect to the analogous, unsubstituted molecule C2
(EC50 = 5.0 µ and 1.2 µM, respectively), implying that the target BVDV RdRp can accommodate
a large group at that position with minimum steric hindrance. At the same time, however, while
compound C2 was devoid of cytotoxicity (CC50 > 100 µM), the presence of the cycloaliphatic
moiety on 13c seems to endow this compound with some toxic effects (CC50 = 33 µM). An
analogous behavior can be observed in the case of compound 12c.
Finally, compound 14b, featuring the replacement of the phenyl ring on the scaffold of C1
with a heterocyclic moiety (i.e., furane), showed better antiviral activity (EC50 = 2.0 µM) than
the C1 derivative, preserving at the same time its non-toxic character (CC50> 100 µM).
Globally, the combination between an electron withdrawing group such as –CN on phenyl
position 4’ and a chlorine substituent on position 4 of the imidazo[4,5-g] quinoline scaffold led
to define compound 6b as new lead compound, bearing EC50 = 0.3 µM, no cytotoxicity and a S.I.
> 333.3.
Ordinarily, the development of a pharmacophore model aims at creating a powerful tool for the
design of new and possibly more potent molecules with respect to a specific target. In this work,
however, the common feature pharmacophore approach was exploited to identify the molecular
requirements shared by all series of compounds (i.e., previously synthesized B-F series [23, 24],
and newly synthesized compounds 2-5a, 6-11b, 12-13c, and 14b) in determining their inhibitory
15

ACCEPTED MANUSCRIPT
activity against the BVDV RdRp. Accordingly, a selection of 31 compounds among the entire
sets were selected, along with their experimentally BVDV EC50 values, as the training set for the
development of the corresponding three-dimensional (3D) pharmacophore model (Fig. S1).
The calculated best-fit values ranked Hypo1, which consists of two Hydrogen Bond Acceptor
features (HBA) and one Hydrophobic Aromatic feature (HY_AR) (Fig. 7A), as the best model.
Accordingly, the most active compounds of training set belonging to the 4 different scaffolds
considered in this paper were mapped onto Hypo1, as shown in Fig.7B-F.
Fig. 7. Geometrical relationships (A) among the features of the top-scoring 3D pharmacophore
Hypo1, and pharmacophore mapping of the training set compounds B1 (B), 6b (C), D2 (D), E2
(E), and F2 (F). The hypothesis features are portrayed as meshed spheres, color-coded as
follows: cyan, HY_AR; green, HBA.
In the successive step, taking advantage of (i) the ligand-binding pharmacophoric requirements
determined above, and (ii) the binding mode of compound D2 onto the BVDV RdRp described
in our previous work,[25] we docked and scored the free energy of binding ∆Gbind for compounds
16

ACCEPTED MANUSCRIPT
B1, 6b, E2, and F2 towards the viral target protein, using these 4 additional compounds (together
with D2) as proofs-of-principle.
Briefly, for compound D2 we already located a putative binding pocket located between the
BVDV RdRp fingers domains 1 and 2 (residues 139–313 and 351–410, respectively).[25] By
applying a consolidated molecular dynamic (MD) computational recipe,[27, 28] some specific
interactions between the pyridoquinoxaline derivative D2 and the critical RdRp binding site
residues were identified. In details, the aromatic portions of D2 are encased in two receptor
cavities comprising residues A221, I261, I287, Y289 and V216, Y303, V306, K307, P408 and
A412, respectively. Moreover, two stable hydrogen bonds (HBs) were detected between the N-
pyrido and the quinoxaline nitrogen atom of D2 and the donor atoms of the side chains of R295
and Y674 of the BVDV RdRp, respectively (Fig. 8). For the record, the thiophene ring of D2
was involved - through its sulfur atom - in a further HB with S411; however, since the strength
of this additional HB was weaker compared to the other two polar interactions,[25] it was
classified as a secondary requirement for binding optimization.
Therefore, a thorough search for a residue sequence satisfying the 3D pharmacophoric
requirements described above was performed and successfully retrieved in this specific protein
region. As clearly shown in Fig. 8, the three features identified by our 3D pharmacophore model
find the corresponding counterparts in the binding mode interactions collected by the MD
procedure.
17

ACCEPTED MANUSCRIPT
Fig. 8. Equilibrated MD snapshot of the RdRp of BVDV in complex with D2.
The image is a zoomed view of the polymerase binding site. The ligand is portrayed in atom-
colored sticks-and-balls (gray, C; blue, N; green, Cl; yellow, S), while the secondary structure of
the protein is shown as a gray transparent ribbon. The residues satisfying the 3D pharmacophoric
requirements in the interaction with D2 are evidenced as green (HBA) and cyan (HY_AR) sticks
and labeled. The cyan area represents the hydrophobic cavity in the protein binding site which
encases the heterocyclic moiety of the compound (see Fig. 6). Dark gray labeled residues in
sticks within the dark gray area represent the second protein hydrophobic cavity next to the
pyridoquinoxaline ring of D2. Hydrogen atoms, water molecules, ions and counterions are
omitted for clarity.
In the light of these encouraging results, following the same procedure, compounds B1, 6b,
E2, and F2 were also docked and scored for affinity towards the binding site of the BVDV
RdRp. Pleasingly, for each compound a solution was found in the corresponding set of docked
ligand conformations that best reproduced the key 3D pharmacophore requirements, as shown in
Fig. 9. Each resulting receptor/ligand complex was then relaxed by energy minimization,
followed by MD simulations. Finally, the relevant values of the free energy of binding ꢁGbind
18

ACCEPTED MANUSCRIPT
between all compounds and the polymerase were evaluated by applying the molecular
mechanics/Poisson−Boltzmann surface area (MM/PBSA) computational ansatz (see
Computational Details), as listed in Table 2.
Fig. 9. Equilibrated MD snapshot of the RdRp of BVDV in complex with B1, 6b, E2, and F2.
The image is a zoomed view of the polymerase binding site. Ligands are portrayed in atom-
colored sticks-and-balls (gray, C; blue, N; green, Cl; yellow, S), while the secondary structure of
the protein is shown as a gray transparent ribbon. The residues satisfying the 3D pharmacophoric
requirements in the interaction with D2 are evidenced as green (HBA) and cyan (HY_AR) sticks
and labeled. The cyan area represents the hydrophobic cavity in the protein binding site which
encases the phenyl/heterocyclic moiety of the compounds (see Fig. 6). Dark gray labeled
residues in sticks within the dark gray area represent the second protein hydrophobic cavity next
19

ACCEPTED MANUSCRIPT
to the pyridoquinoxalinic ring of the compounds. Hydrogen atoms, water molecules, ions and
counterions are omitted for clarity. The intensity of the surface color for the hydrophobic cavities
is proportional to the corresponding strength of ligand/protein interactions.
Table 2. Free energy components (ꢁHbind and –TꢁSbind) and total binding free energies (ꢁGbind
)
for compounds D2, B1, 6b, E2, and F2 in complex with the BVDV RdRp. The IC50(calcd) values
were obtained from the corresponding ∆Gbind values using the relationship: ∆Gbind = -
RTln(1/IC50).
-
∆Hbind
(kcal/mol)
Compound
∆Gbind(kcal/mol)
IC50(calcd)(µM)
T∆Sbind(kcal/mol)
-21.56(0.17)
-19.75(0.15)
-22.09(0.15)
-20.42(0.16)
-21.48(0.17)
13.37(0.25)
12.40(0.19)
13.68(0.26)
12.96(0.21)
13.39(0.19)
-8.19(0.30)
-7.32(0.24)
-8.41(0.26)
-7.46(0.25)
-8.09(0.25)
0.99
4.3
D2
B1
6b
E2
F2
0.69
3.5
1.2
Our qualitative and quantitative computational analysis confirms the experimental trend in
terms of correlation between binding capability and inhibitor activity against the viral
polymerase. Indeed, all tested compounds exhibit a good affinity against their target protein in
the sub- and low-micromolar range (Table 2). Moreover, these molecules share a common
binding mode, although slight differences are found for compounds B1 and E2, as expected. In
fact, from their mapping onto the 3D pharmacophore model it could be anticipated that, at
variance with the other ligands, both these compounds would not perfectly map all molecular
requirements, although for two opposite reasons. The triazoloquinoline derivative B1, lacking a
bulky aromatic substituent on the 5-membered ring, cannot fill the second hydrophobic cavity of
20

ACCEPTED MANUSCRIPT
the protein binding site; however, it can partially compensate this missing interaction by
establishing a further H-bond with the side chain of K263 (Fig. 9A). On the other hand, the
smaller dimensions of the heterocyclic scaffold of imidazopyridine E2 does not allow this
compound to assume an optimal conformation in the binding site; consequently E2 cannot
perform stable interactions within the first hydrophobic cavity (Fig. 9C). Conversely, the new
imidazoquinoline derivative 6b and the N-benzylidenequinolinamine F2, featuring a distribution
of functional groups similar to D2, fit very well all the pharmacophore requirements (Fig.s 9B
and 9D); as a result, their optimal binding conformations reflect into highly favorable ꢁGbind
values (Table 2).
MD results are further corroborated by the quantitative analysis of the H-bonds between the
compounds and the polymerase amino acids mainly involved in binding (Table S3) and by a per-
residue deconvolution of the enthalpic contribution to binding afforded by each single residue of
the BVDV binding site (Fig. 10). To facilitate reading, the values of ꢁHbind,res for the critical
RdRp residues are clustered according to a specific underlying interaction type as follows: A221,
I261, I287, and Y289 belong to the first hydrophobic cavity (HC1), and are considered for
hydrophobic and π-interactions. Analogously, amino acids V216, Y303, V306, K307, P408, and
A412 are clustered in the alternative hydrophobic cavity (HC2), and considered for analogous
interaction types.
21

ACCEPTED MANUSCRIPT
Fig. 10. Per-residue binding enthalpy decomposition ꢁHbind,res for compounds D2, B1, 6b, E2,
and F2 in complex with the BVDV RdRp.
The collected data support the indications derived from the overall MM/PBSA energetic
analysis: the best compounds D2, 6b, and F2 have indeed quite similar interaction spectra as the
main H-bonds involving residues R295 and Y674 persist during the entire simulation trajectory,
and their contribution to binding amount to the substantial value of -2 kcal/mole for each amino
acid (Fig. 10). Furthermore, the aromatic moiety of the molecules are harbored in both
hydrophobic cavities HC1 and HC2; accordingly, the overall corresponding interactions afford
an extensive favorable contribution to binding, equal to ꢁHbind,res= -5.95 kcal/mol, -6.00
kcal/mol, and -5.86 kcal/mol for D2, 6b, and F2, respectively (Fig. 10). Finally, even the small
stabilization afforded by the supplementary H-bond with S411 is confirmed by the energy
decomposition analysis.
22

ACCEPTED MANUSCRIPT
In this scenario, our computational approach allows to better understand the reasons of the low
yet significant inhibitory efficacy of the two remaining compounds B1 and E2. Although the
ꢁHbind,res value of the interaction of the triazoloquinoline derivative with the HC2 residues is
essentially negligible (-0.12 kcal/mol, Fig. 10), the presence of the permanent H-bond, with an
Average Dynamic Length (ADL) of 1.99 ± 0.05 Å, detected between the nitro group of B1 and
the side chain of K263 partially offsets this contribution; concomitantly, the interactions of B1
with the side chains of residues lining the HC1 are comparable to the ones characterizing the best
binder (Fig. 10). The binding mode of the imidazopyridine E2, instead, satisfies the same
requisites underlying the binding of best RdRp inhibitors; however, its smaller scaffold is not
able to preserve the same intensity in these interactions, particularly in the case of the
hydrophobic contribution from HC1 (ꢁHbind,res = -0.90 kcal/mol, Fig. 10).
The computational results discussed above yielded a precise scheme of the principal
intermolecular interactions sustaining the binding between the BVDV RdRp and the present,
specific classes of its inhibitors. To validate these indications we next performed a combined in
silico/in vitro mutagenesis study of those protein residues involved in the binding site, mainly via
highly stabilizing H-bonds with the ligands (Fig. 10). To this purpose, the following polymerase
residues: R195, Y674 (two key ligand/protein anchor points for all tested compounds) and K263
(exerting an exclusive role in the binding mode of the triazoloquinoline derivatives B1) were all
mutated into alanine. Then, the affinity of the mutated viral RdRp towards all compounds D2,
B1, 6b, E2, and F2 were concurrently evaluated in silico and in vitro. The aggregate results are
listed in Table 3, while all dose–response curves obtained from in vitro enzyme assays are
showed in Fig. S2.
23

ACCEPTED MANUSCRIPT
Table 3. In vitro/in silico alanine mutagenesis of the RdRp of BVDV residues K263, R295, and
Y674.According to its definition (∆∆Gbind=∆Gbind(WT)-∆Gbind(MUT)), positive values calculated for
∆∆Gbind indicate that the considered amino acid substitution at a given position of BVDV RdRp
is favorable in terms of interaction with the inhibitors, whereas negative values indicate that the
corresponding mutation is unfavorable to binding. n.e. = not evaluable (see Fig. S2).
IC50(calcd)
(µM)
IC50(exp)
(µM)
RdRp
residue
∆Gbind
(kcal/mol)
∆∆Gbind
(kcal/mol)
Compound
D2
WT
-8.19 ±0.30
-7.98 ± 0.25
-3.69 ± 0.27
-3.93 ± 0.23
-
0.99
1.4
2.0 × 103
1.3 × 103
0.64
0.96
n.e.
K263A
R295A
Y674A
-0.21
-4.50
-4.26
n.e.
WT
-7.32±0.25
-5.42 ± 0.24
-3.75 ± 0.27
-3.59 ± 0.26
-
4.3
9.0
140
n.e.
n.e.
K263A
R295A
Y674A
-1.90
-3.57
-3.73
107
B1
6b
E2
F2
1.8 × 103
2.3 × 103
WT
-8.41±0.26
-8.32 ± 0.24
-3.83 ± 0.23
-3.99 ± 0.25
-
0.69
0.80
1.6 × 103
1.2 × 103
0.48
0.75
n.e.
K263A
R295A
Y674A
-0.09
-4.58
-4.42
n.e.
WT
-7.46±0.25
-7.51 ± 0.24
-3.80 ± 0.26
-4.01 ± 0.23
-
3.4
11
15
K263A
R295A
Y674A
0.05
-3.66
-3.45
3.2
1.6 × 103
1.2 × 103
n.e.
n.e.
WT
-8.09 ±0.25
-7.91 ± 0.25
-3.87 ± 0.22
-4.11 ± 0.24
-
1.1
1.6
1.5 × 103
0.98 × 103
0.81
1.0
K263A
R295A
Y674A
-0.18
-4.22
-3.98
n.e.
n.e.
24

ACCEPTED MANUSCRIPT
Inspection of Table 3 immediately reveals the excellent agreement between computational
prediction and experimental determination of IC50 values for all compounds considered, this
evidence thereby constituting a solid validation of the in silico strategy presently adopted. Next,
our combined approach was able to dissect the effect of each mutation on the binding strength of
all tested compounds against the polymerase, ultimately supporting our description of their
binding mode onto the viral protein. In detail, the substitution of the polar residue R295 and
Y674 with alanine is predicted to dramatically decrease the BVDV RdRp affinity of all 5
molecules, as testified by the corresponding ∆∆Gbind values in Table 3: the loss of at least 3.5
kcal/mol indeed means that these specific residues are very important not only for their single
point interaction but also for the overall conformation and stability of the entire polymerase
binding site. Since the exerted effect of the two mutations is comparable for all 5 compounds, for
the sake of brevity we will discuss here in details the results achieved for the mutated proteins in
complex with the new imidazoquinoline derivatives 6b. The top panel of Fig. S3 nicely shows
how the binding site for this compound is deeply affected by both R295A and Y674A mutations:
the absence of the relevant drug/protein H-bonds forces the compound to adopt a different
conformation upon binding and this, in turn, reflects into a general weakening of the entire
network of stabilizing intermolecular interactions.
In keeping, the calculated ꢁGbind values plummet from -8.41 kcal/mol for the wild-type protein
to -3.83 kcal/mol and -3.99 kcal/mol for the R295A and Y674A mutant polymerases,
respectively (Table 3). Consistently, the relevant IC50(calcd) values increase more than 4 orders of
magnitude, reaching values as high as 1.6 × 103µM (R295A) and 1.2 × 103µM (Y674A). The
compound activity tested in vitro parallels this computational prediction: indeed, the BVDV
25

ACCEPTED MANUSCRIPT
RdRp inhibitory activity of 6b at all compound concentrations considered was so low that the
corresponding IC(50,exp) could not be determined (Table 4 and Fig. S2).
Lastly, the BVDV RdRp K263A mutant protein was studied with the purpose of and validating
the specific binding mode predicted for compound B1; as such, lysine residue 263 in fact does
not constitute a topic molecular determinant for the polymerase binding of any other compounds
considered. In fact, taking again derivative 6b as proof-of-concept, mutating K into A at that
position results in a negligible modification of the drug binding configuration and intermolecular
interactions (Fig. S3C) and, hence, in the corresponding ∆Gbind value (Table 3). On the contrary,
for compound B1 this anchor point instrumental to protein binding is lost in the presence of the
A623 mutation (Fig. S3D), and the calculated affinity against the BVDV RdRp plunges one
order of magnitude, i.e., from -7.32 kcal/mol for the WT to -5.42 kcal/mol for the K632A mutant
(Table 3). Accordingly, the predicted and experimental IC50s both increase from a low
micromolar value for the WT (IC(50,calcd) = 4.3 µM and IC(50,exp) = 9.0 µM) to > 100µM for the
K263A mutated protein(IC(50,calcd) = 107µM and IC(50,exp) = 140µM), as shown in Table 3 and
Fig. S2.
Conclusions
In our long-standing research activity in the design, synthesis, characterization, and testing of
antiviral compounds towards viruses of the Flavivirus genera, we produced a plethora of
molecular classes among which several compounds were found concomitantly endowed with
high potency towards BVDV and negligible toxicity.
Along our antiviral research activity, linear aromatic N-tricyclic systems emerged for their
interesting anti-Flaviviridae activity. From the combination of the quinoline nucleus with 1,2,3-
26

ACCEPTED MANUSCRIPT
triazole, imidazole or pyrazine, best results were obtained for imidazo[4,5-g]quinoline
derivatives as anti-BVDV agents. In this paper we expanded the SAR analysis through rational
modifications of lead compound 2-phenyl-3H-imidazo[4,5-g]quinoline C1 (EC50 = 4 ꢀM, CC50 >
100 ꢀM).
Diverse substituents at positions R1, R2, and/or R3 of proposed model in Fig. 6, and replacement
of phenyl ring with a furane moiety led to 12 new imidazo[4,5-g]quinoline derivatives. Influence
on the anti-BVDV activity of the Cl atom at position R1, of steric hindrance of substituents in R2,
of electron donating groups (EDGs) or electron withdrawing groups (EWGs) on the phenyl
group (R3 in Fig. 6) and of heterocyclic moiety on quinoline scaffold was evaluated.
The SAR analysis of this molecular series highlighted how the combined presence of a chlorine
substituent and an electron withdrawing group such as –CN on position 4′ of phenyl moiety
retains and enhances the antiviral activity of the corresponding derivative compared to the lead
compounds C1. From this series, compound 6b emerged as new lead compound, showing EC50
value of 0.3 ꢀM and S.I. > 333.3.
With the purpose of finding a sensible molecular rationale underlying the common affinity
towards the RdRp BDVD of these different molecular classes, we then performed a jointed in
silico/in vitro study. To the purpose, we started with the development of a common-feature three-
dimensional pharmacophore model, which allowed us to determine two H-bond acceptors and
one aromatic hydrophobic feature as the fundamental interactions leading to efficient
drug/protein binding. Exploiting this information, the binding site and the binding mode of all 5
classes of compounds onto the BVDV RdRp were retrieved; next, the activity of all compounds
was scored using the MM/PBSA approach, obtaining a good agreement between calculated free
27

ACCEPTED MANUSCRIPT
energies of binding ∆Gbind and the corresponding experimental EC50 values. The per residue
deconvolution of the enthalpic component of ∆Gbind identified two RdRp amino acids – namely
R295 and Y674, as the two fundamental H-bond donors, while two hydrophobic cavities HC1
(residues A221, I261, I287, and Y289) and HC2 (residues V216, Y303, V306, K307, P408, and
A412) were determined to fulfill the third pharmacophoric requirement. Bases on this
information, three RdRp residues critical for drug binding were selected and mutagenized, both
in silico and in vitro, into alanine, and the affinity of a set of selected compounds towards the
mutant RdRp isoforms was determined accordingly. The agreement between predicted and
experimental data confirmed the proposed common molecular rationale shared by molecules
characterized by different chemical scaffolds in binding to the BVDV RdRp.
Ultimately, the information yielded by our combined in silico/in vitro approach could provide an
interesting tool for the efficient development of new, powerful compounds against this important
Pestivirus.
Experimental Section
General directions. Melting points are uncorrected and were measured in open capillaries in a
1
Digital Electrothermal IA9100 melting point apparatus. H NMR spectra were recorded on a
Varian BrukerAvance II 600-MHz spectrometer, using TMS as internal standard. The chemical
shift values are reported in ppm (δ) and coupling constants (J) in Hertz (Hz). Signal multiplicities
are represented by: s (singlet), d (doublet), dd (double doublet), t (triplet), q (quadruplet) and m
(multiplet). Column chromatography was performed using 70-230 mesh (Merck silica gel 60).
Light petroleum refers to the fraction with b.p. 40-60°C. TLC using Merck F-254 commercial
28