family consists of three genera—the genus Flavivirus (including human pathogens such as dengue virus, West Nile virus, and yellow fever virus), the genus Hepacivirus (hepatitis C virus (HCV)), and the genus Pestivirus (including veterinary pathogens such as the bovine viral diarrhea virus (BVDV), and the classical swine fever virus (CSFV)). Pestivirus infections of domesticated livestock (e.g., cattle, pigs, and sheep) cause significant economic losses worldwide, [1
BVDV is ubiquitous and causes a range of clinical manifestations (including abortion, respiratory problems, chronic wasting disease, immune system dysfunction, and predisposition to secondary viral and bacterial infections). BVDV-1 and -2 strains can cause acute fatal disease with mortality rates of 17–32% [4
]. BVDV is also able to establish a persistent infection in fetuses [7
]. When born, these persistently infected animals remain viremic throughout their lifespan and serve as continuous virus reservoirs. Persistently infected animals might also succumb to fatal mucosal disease if they are superinfected with a closely related BVDV strain. Vaccines are used in some countries in an attempt to control the pestivirus disease, with varying degrees of success [3
]. Other containment strategies comprise quarantine and persistent-infected animal culling [3
]. Currently, there are no approved antiviral drugs to control pestivirus infections. Such drugs might be an important tool to control BVDV on infected farms.
Classical swine fever is a highly infectious viral disease that affects domestic and wild pigs. CSFV is included in the list of diseases notifiable to the OIE (www.oie.int
). CSFV is considered to cause one of the most devastating diseases for the pig industry, throughout the world, both from an economical and sanitary point of view [8
]. The disease is endemic in Asia and is prevalent in many countries of Central and South America. In contrast to North America, where CSFV was eradicated several decades ago, the European Union (EU) still has an ongoing progressive eradication program that started in the early 1990s [8
]. The most efficient vaccines currently available against CSFV are live attenuated vaccines [8
]. However, many efforts have recently been put into the development of new and safer marker vaccines against CSFV, along with improved diagnostic tools [9
]. Broad-spectrum pestivirus inhibitors might also be considered to control outbreaks with CSFV, in otherwise disease-free areas. Other possible uses of anti-pestivirus drugs might be (i) to treat valuable animals infected with pestiviruses in zoologic collections, (ii) to treat expensive animals in breeding programs and in vitro
embryo production [11
], to (iii) cure established cell lines from contaminating pestiviruses [12
Several classes of pestivirus inhibitors [14
] have been reported. They either target a cellular protein/enzyme, i.e., α-glycosidase (which is involved in the maturation of virions [15
]), as well as viral encoded enzymes such as the NS3 protease and helicase/NTPase [16
], or the NS5B RNA-dependent RNA polymerase (RdRp). Polymerase inhibitors include nucleoside [14
] and non-nucleoside inhibitors, such as N
]indol-3-ylthio)ethyl]-1-propanamine (VP32947) [17
], a thiazole urea derivative [18
], a cyclic urea derivative [19
], imidazo-pyridines (BPIP) [20
], ethyl 2-methylimidazo[1,2-a
]pyridin-8-carboxylate (AG110) [21
], pyrazolotriazolopyrimidinamine (LZ37) [22
], 2-(2-benzimidazolyl)-5-[4-(2-imidazolino)phenyl]furan (DB772) [23
], 5,6-dimethoxy-1-indanone [24
], 2-phenylbenzimidazole [26
], substituted 2,6-bis(benzimidazol-2-yl)pyridines [27
], benzimidazole derivative [28
], and arylazoenamine derivatives [29
BVDV strains that are resistant to the majority of these non-nucleosidic RdRp inhibitors all carry mutations in the fingertip domain of the viral RdRp. However, most of these inhibitors do not inhibit the in vitro
activity of the recombinant viral polymerase but are able to inhibit the activity of the BVDV replicase complex (RC), in a dose dependent manner [20
]. The fingertip domain of the polymerase is crucial for the function of the polymerase and the viral RC. This domain is, thus, apparently a “hot spot” binding site for selective inhibitors of pestivirus replication [20
Here, we report on the antiviral characteristics and mode of action of a series of quinolinecarboxamide analogues as a new class of chemicals that inhibit the replication of pestiviruses.
3.1. In Vitro Antiviral Activity of Quinolinecarboxamides
Previously, we performed a large-scale, CPE-based antiviral screen [21
], during which we identified two quinolinecarboxamides [TO502-2403 (CSFCII)/ TO505-6180 (CSFCI), Figure 1
] that resulted in selective in vitro
inhibition of BVDV-1 replication. Both analogues inhibited the replication of the BVDV-1 (strain NADL) in a MDBK cell, in a dose-dependent manner. The EC50
for TO502-2403 (CSFCII) was 0.2 µM (SD 0.06 µM) (Figure 2
A) and 0.07 µM (SD 0.02 µM) for TO505-6180 (CSFCI, Figure 2
A). Both compounds were devoid of cytotoxicity (CC50
> 100 µM). The anti-BVDV activity of both molecules was further corroborated by measuring the effect of the compounds on infectious virus yield and on viral RNA synthesis. The EC50
for inhibition of infectious virus yield in the culture supernatant was 0.2 µM (SD 0.09 µM) for TO502-2403 (CSFCII) (Figure 2
B) and 0.6 µM (SD 0.1 µM) for TO505-6180 (CSFCI, Figure 2
B). The EC50
for inhibition of viral RNA formation in cells and in the culture supernatant for TO502-2403 (CSFCII) was 0.06 µM (SD 0.01 µM) and 0.05 µM (SD 0.02 µM), respectively, showing a very similar pattern of inhibition (Figure 2
C,D). The EC50
for the inhibition of intra- and extracellular viral RNA formation for TO505-6180 (CSFCI, Figure 2
C,D) was 0.03 µM (SD 0.01 µM) and 0.06 µM (SD 0.005 µM), respectively.
Both quinolinecarboxamide analogues modestly inhibited the in vitro replication of the classical swine fever virus (CSFV) and proved inactive in vitro, against the hepatitis C virus that belongs to BVDV, and CSFV that belongs to the family of Flaviviridae. Furthermore, quinolinecarboxamide analogues proved inactive against a panel of unrelated RNA and DNA viruses (data not shown).
3.2. 3D-QSAR Analysis of All Tested Analogues of CSFC Lead Molecules
To better understand the relationships between the structural properties of the CSFC analogues tested and their biological activity, a 3D Quantitative structure-activity relationships (QSAR) model was build, following previously described methodology. From the 145 tested compounds, 23 were removed from the modeling process due to undetermined EC50
values. In addition, 17 more compounds were removed to eliminate the activity cliffs generators from the dataset. The final model contained four principal components and showed a good statistical performance with R2Train
= 0.82, q2LOO
= 0.66, and q2Ext
= 0.73. The detailed dataset composition and partitioning, including the compounds removed due to any of the reasons listed above, as well as model predictions are provided as the Supplementary Table S2
According to the 3D-QSAR model, the steric effect contributed to 72% of bioactivity, while the contribution of the electrostatic effect was 28%. This model is summarized in Figure 3
for the CSFCII and CSFCI derivatives. In both cases, bulky substituents at the R1
-position, positively influenced the bioactivity, and their presence could be considered critical for the anti-BVDV bioactivity. This effect could be seen from the comparison of compounds 43
, which lacked a substituent at R1
, with compounds 85
. As a result of the surrounding electrostatic field of position R1
, negative charged moieties in the para position of aromatic rings used in R1
could increase activity.
The positioning of an electronegative molecular interaction field (MIF) close to R1, suggests that non-phenyl aromatic substituents at R1, such as furan or pyridine with electronic clouds displaced toward the N or O atoms could increase bioactivity (see compounds 3, 36, 110 and 132).
Substituents at R3 could have a negative impact on the anti-BVDV bioactivity, due to the presence of a steric unfavorable region close to this position. Small substituents are well tolerated, however, larger atoms such as I (compound 22 with a larger halogen compared to other analogues that carry a halogen substituent at this position) decreased activity. A bulky substituent at position R3 had a major negative impact on bioactivity (compound 98). Additionally, a steric favorable region was observed around R4. Small substituents at this position improved the bioactivity of the compounds (compound 20, 24 and 25).
For the CSFCII analogues, bulky substituent at position R2 also had a positive effect on the bioactivity (critical for bioactivity). For example, the lack of such a moiety at R2 in analogues 44 and 47, compared to compounds 1–9 abolished the bioactivity of these compounds. However, too large substituents like those present in compounds 57 (6-CF3) and 59 (6-CH3O2S) had a steric unfavorable effect, as suggested by the existence of a delimited steric favorable region near position R2. A para-negative substituent in R2 could have a negative effect on bioactivity, given that it occupied a region with positive electrostatic MIF.
Several positive and negative electrostatic fields perpendicular to each other were observed around the R1 and R2 position. These fields were orientated as such that they favored stacking of the pi-electrons from the aromatic rings between them. For this reason, substituents with aromatic properties could be preferable in position R1 and R2.
In addition to the previously discussed influence of substitutions at R1, R3, and R4, for the CSFCI derivatives, small non-bulky substitutions at any position—R21, R22, R23 and R24—contributed to improved bioactivity. In contrast, bulky substitutions at the later positions highly decreased bioactivity. This effect could be observed from the comparison of compounds 51, 57, 59 and 112. According to the steric MIF surrounding positions R21, R22, R23, and R24, the simultaneous substitution of these positions in the same compound could increase bioactivity. Additionally, electronegative substitutions were favored at position R24 and disfavored at R23. Finally, the presence of the pyrazol moiety in the CSFCII derivatives and of the benzothiazole group in the CSFCI analogues attached to the quinoline-4-carboxamide scaffold were essential for bioactivity.
3.3. Isolation and Characterization of Drug-Resistant Viruses
To decipher the mechanism through which the CSFC analogues inhibited viral replication, escape mutants to the drugs were selected by propagating BVDV (strain NADL) for 25 passages, in increasing concentrations of the drug (from 1.8 to 30 µM) and were characterized. The TO502-2403/CSFCII drug-resistant virus variants (TO502-2403res
) proved more than 500-fold less susceptible to the inhibitory effect of TO502-2403/CSFCII, than the parent wild-type strain and about 1580-fold less susceptible to BPIP, 11-fold less susceptible to LZ37, >63-fold less susceptible to AG110, 200-fold less susceptible to BBP, and >1429-fold less susceptible to TO505-6180/CSFCI (Table 1
). The TO505-6180/CSFCI drug resistant variants (TO505-6180 res
) were >1429-fold less susceptible to the inhibitory effect to TO505-6180/CSFCI, compared to the parent wild-type strain and about 900-fold less susceptible to BPIP, 11-fold less susceptible to LZ37, 16-fold less susceptible to AG110, >333-fold less susceptible to BBP, and 160-fold less susceptible to TO502-2403/CSFCII (Table 1
). Previously selected variants that were resistant to different inhibitors showed cross-resistance to TO505-6180/CSFCI and to TO502-2403/CSFCII, ranging from 19-fold (i.e., TO502-2403/CSFCII against AG110res
) to >1429-fold (Table 1
). To identify the molecular changes that are responsible for the drug-resistant phenotype, we compared the TO502-2403res
genome sequences to the parental wild-type strain (GenBank accession no. AJ781045). For TO502-2403res
, we identified a transition of T to the C mutation, at position 10,862, which resulted in an amino acid change of phenylalanine (F) to proline (P) at amino acid residue 224 in the NS5B gene.
, two mutations in the NS5B polymerase gene were identified, one at the position 10,862 [T to A substitution (F224Y)] and the other at position 10,982 [A to G substitution (N264D)]. Interestingly, the mutation at position F224 was also identified in the genome of the BPIPres
] and the LZ37res
3.4. Effect of the CSFC Analogues on the BVDV RdRp and Replication Complexes
As the mutations identified in the drug-resistant virus variants were all located in the NS5B gene, that encoded the viral RdRp, we next studied the inhibitory effect of both CSFC analogues on the in vitro
polymerase activity of the enzyme. TO502-2403/CSFCII, TO505-6180/CSFCI, and the nucleotide analogue 3′-dGTP (which was included as a positive control) were tested for potential inhibitory activity against the highly purified BVDV RdRp, by using poly(C) as a template. The 50% inhibitory concentrations for the BVDV polymerase activity were < 1 µM for 3′-dGTP. TO502-2403/CSFCII and TO505-6180/CSFCI had no effect on the activity of the viral polymerase (Figure 4
Since TO502-2403/CSFCII and TO505-6180/CSFCI did not inhibit the activity of the purified BVDV RdRp, we tested the effect of the compound on viral RCs isolated from MDBK cells that had been infected with the wild-type virus or with the selected TO502-2403res
virus BVDV strain. TO502-2403/CSFCII inhibited the activity of the BVDV WT replication complexes in a dose-dependent manner, with a maximum inhibition ~90% at 10 µM (Figure 5
A,B). In contrast, the activity of RCs isolated from MDBK cells that had been infected with the TO502-2403res
strain were not susceptible to the inhibitory effect of TO502-2403/CSFCII (Figure 5
A,B). TO505-6180 inhibited the activity of the BVDV wt RCs in a dose-dependent manner (Figure 5
C,D), whereas RCs isolated from MDBK cells that had been infected with the TO505-6180res
virus was 3 to 4-fold less susceptible to the inhibitory effect of TO505-6180, as compared to BVDV wt RCs (Figure 5
3.5. Computational Docking of CSFC Analogues in the BVDV RdRp Crystal Structure
Based on the crystal structure of the BVDV RdRp (PDB 1S48) [24
], the amino acid position F/P/Y224 was located in a small cavity near the fingertip domain of the BVDV polymerase. The F224 position was already shown to be implicated in resistance of BVDV to VP32947, BPIP, and LZ37 [17
]. Position N264 was located in the conserved motif I of the finger domain of the BVDV polymerase.
Docking of TO502-2403/CSFCII in this cavity revealed the following possible interactions between the polymerase and the compound: (i)
a hydrogen bond between the N in the middle of the scaffold of TO502-2403/CSFCII and the residue A392 via the main chain O and (ii)
a stacking interaction between the aromatic ring of the F224 and the aromatic ring of the quinoline-4-carboxylic acid scaffold (Figure 6
). When the phenylalanine mutated into a proline the stacking interaction would be lost, thereby, also destabilizing the hydrogen bond with A392. Hence, when the F224P substitution occurred, the interaction between the TO502-2403/CSFCII inhibitor and the RdRp would no longer be possible
Docking studies of TO505-6180/CSFCI in the same cavity of the BVDV polymerase, as described above, resulted in a model that revealed the following possible interactions between the polymerase and the inhibitor. Stacking interactions with F224 were observed, together with a hydrogen bond from the O in the furan ring of the inhibitor and the amide N of Asn264. In such cases where one or both of these interactions were lost due to the mutation N264D or F224Y, binding of this inhibitor was no longer favourable. Here, we speculated that the presence of the OH group in the side chain of the mutated Y224 might interact with the surrounding protein matrix, for instance forming an H-bond with the A392 backbone CO group and forcing the Y224 into another rotameric state, where the stacking interaction with the inhibitor is not possible.
During the course of a large screening effort dedicated to identifying pestivirus inhibitors, two quinolinecarboxamide analogues (i.e., TO502-2403/CSFCII and TO505-6180/CSFCI) were identified as selective in vitro
inhibitors of the replication of pestiviruses. Both independently identified hit compounds inhibited the in vitro
BVDV-1 and CSFV replication but proved inactive against related viruses (the HCV) and a panel of unrelated RNA and DNA viruses. Based on the hit molecule TO502-2403/CSFCII, a series of 104 new analogues were synthesized in an attempt to optimize this compound class for inhibitory properties and the selectivity on the replication of BVDV. A 3D-QSAR model, Figure 3
, was derived from the antiviral activity of these analogues evaluated against BVDV. From this model, we could extrapolate that the bio-activity of the lead molecule could be improved by introducing bulky aromatic groups with negatively charged substituents in para in R1
, and by adding bulky aromatic groups that lacked a negatively charged, small substituent in para at R2
of the CSFCII analogues. Position R3
could only tolerate small substituents, while small substituents in position R4
favored bioactivity. In the case of the CSFCI derivatives, only small non-bulky substitutions were allowed at the R21
, and R24
Parallel drug-resistant BVDV variants for both hit molecules were selected and a geno- and phenotypic analysis was performed. In vitro
carried an amino acid transition from phenylalanine (F) in the WT to proline (P), at position 224, the TO505-6180res
virus carried a phenylalanine (F) to tyrosine (Y) mutation at position 224 (F224Y), and an asparagine (N) to aspartic acid (D) mutation at position 264 (N264D) of the viral RNA-dependent RNA polymerase. Interestingly, amino acid position F224 was previously shown to be involved in the antiviral resistance of BVDV to VP32947 [17
], BPIP [20
], and LZ37 [22
], whereas mutation N264D was reported to be key in the phenotypic resistance against DB772 [23
], 5,6-dimethoxy-1-indanone [26
], and 2-phenylbenzimidazole [28
]. Here, for the first time we reported the selection of a drug-resistant virus that combines both the F224P/Y and the N264D mutation. Both TO502-2403res
were resistant to inhibition by other compounds (i.e., LZ37, BPIP, AG110, and BBP) that target the same domain within the RdRp. However, TO505-6180res
proved relatively more susceptible to inhibition by AG110 than TO502-2403res
(16-fold vs. 63-fold). Likewise, BVDV variants resistant to various inhibitors targeting the same hotspot in the RdRp were to a much lesser extent inhibited by both CSFC analogues tested, as compared to WT BVDV. Remarkably, AG110res
were relatively more susceptible to inhibition by TO502-2403. An explanation for the peculiar (cross)-resistance pattern observed (Table 1
), regardless of the chemical structure of these inhibitors, could be that three common interaction sites constituted the pocket that binds all inhibitors described above. The location of the first interaction site of the pocket is based on the AA position F224, which is located in the fingertip region of the RdRp. The fingertip region interacts with the thumb domain, with which it encloses the active site of the RdRp, forming an entrance to the template-binding channel. The fingertip region is presumptive to be engaged in the flexibility of the finger domain responsible for template/product translocation [42
], protein–protein interactions, or dimerization of the RdRp in the replication complex [40
], enabling the assembly of an active replication complex. Mutation (F224P/Y) present in the resistance virus variants selected under antiviral pressure with CSFC-analogues, thus, might cause conformational changes in the fingertip region of RdRp, disrupting one of the above-mentioned functions.
Furthermore, the F224 position in the RdRp was located at a distance of only 14 Ǻ from E291 (which is mutated in strains resistant to AG110) which was located in the conserved motif II of the RdRp. A second site of interaction, based on the location of mutation I261M [27
], P262A [45
], and N264D [45
], were located in motif I of the RdRp, close to the NTP i
+ 1 binding site, indicating a tentative role in binding the incoming NTP [40
]. The third interaction site was based on the E291G mutation reported for AG110res
], which was part of the RdRp motif II that, through contacts with the phosphate backbone and bases, might have been involved in template binding [40
]. Thus, the fact that the E291G mutation (identified in AG110res
viruses) was located in motif II of the RdRp and given the importance of the interaction of AG110 with residue(s) from motif II, might explain the relative increased susceptibility of TO505-6180res
to AG110 and that of AG110res
to TO502-2403. Especially since AG110res
combines a mutation in motif II with a different mutation on the fingertip of the polymerase and the BBPres
has no mutations in the fingertip region of RdRp, but only in motif I. Given the flexibility of the finger domain of the RdRp, it might as well be that mutations in one binding site of the pocket influences the conformation of the entire pocket, possibly mediated through additional interactions with the RNA template.
For TO502-2403res, a single mutation in this region of the RdRp proved sufficient to confer complete resistance.
Even if this mutation provided strong evidence that the viral polymerase was the target of interaction with the class of quinolinecarboxamides, similar to the reference compound BPIP [20
], AG110 [21
], and LZ37 [22
], none of these compounds had any inhibitory effect on the highly purified BVDV RdRp.
Since, the quinolinecarboxamides did not inhibit the activity of the purified BVDV RdRp in in-vitro enzymatic assays, we tested the effect of the compound on viral RCs. Indeed, within the intact cell, the RdRp functions in the context of membrane-bound RCs, which consist of several virus proteins, host proteins, and various forms of viral RNA. Both CSFC analogues inhibited the function of the wild-type RC but not that of the mutant strain. The observation that quinolinecarboxamides inhibited the function of RC but not that of the purified RdRp, might be explained by the fact that, following binding to the NS5B, the compound might disrupt the protein–protein interactions and the stability between the several virus-encoded proteins, and might interfere with the function and the proper formation of a functional replication complex. Another possible explanation might be that the binding of quinolinecarboxamides to the polymerase, results in reduced finger flexibility or impairment of the ability of the polymerase to translocate its template/product during polymerization.
In conclusion, a class of quinolinecarboxamides that are selective inhibitors of the replication of pestiviruses was identified. Quinolinecarboxamides were cross-resistant with a number of earlier reported pestivirus replication inhibitors, such as BPIP, AG110, BBP, and LZ37. All these inhibitors were selected for mutations within a region of the finger domain of the RdRp, which was only 7 Å across. Quinolinecarboxamides can interact with the fingertip of the BVDV-RdRp region, confirming that this cavity is a “hot spot” for inhibition of pestivirus replication.