Yuichiro
Akao
a,
Stacie
Canan‡
b,
Yafeng
Cao
c,
Kevin
Condroski§
b,
Ola
Engkvist
d,
Sachiko
Itono¶
a,
Rina
Kaki
e,
Chiaki
Kimura
e,
Thierry
Kogej
d,
Kazuya
Nagaoka
f,
Akira
Naito
e,
Hiromi
Nakai
e,
Garry
Pairaudeau
g,
Constantin
Radu||
h,
Ieuan
Roberts
g,
Mitsuyuki
Shimada**
a,
David
Shum
h,
Nao-aki
Watanabe
f,
Huanxu
Xie
c,
Shuji
Yonezawa
e,
Osamu
Yoshida
e,
Ryu
Yoshida
e,
Charles
Mowbray
i and
Benjamin
Perry
*i
aTakeda Pharmaceutical Company Limited, 26-1 Muraoka-Higashi 2-chrome, Fujisawa, Kanagawa 251-8555, Japan
bCelgene Corporation, Celgene Global Health, 10300 Campus Point Drive, San Diego, California 92121, USA
cWuXi AppTec Company Ltd., 666 Gaoxin Road, East Lake High-Tech Development Zone, Wuhan 430075, People's Republic of China
dAstraZeneca Discovery Sciences, R&D, AstraZeneca, Gothenburg, Sweden
eShionogi & Co., Ltd, 3-1-1, Futaba-cho, Toyonaka-shi, Osaka, Japan
fEisai Co., Ltd, 1-3,Tokodai 5-chome, Tsukuba, Ibaraki 300-2635, Japan
gAstraZeneca, Discovery Sciences, R&D, AstraZeneca, Cambridge, UK
hInstitut Pasteur Korea, 16, Daewangpangyo-ro 712 beon-gil, Bundang-gu, Seongnam-si, Gyeonggi-do, 13488 Republic of Korea
iDrugs for Neglected Diseases initiative, 15 Chemin Louis Dunant, Geneva 1202, Switzerland. E-mail: bperry@dndi.org
First published on 21st January 2021
An innovative pre-competitive virtual screening collaboration was engaged to validate and subsequently explore an imidazo[1,2-a]pyridine screening hit for visceral leishmaniasis. In silico probing of five proprietary pharmaceutical company libraries enabled rapid expansion of the hit chemotype, alleviating initial concerns about the core chemical structure while simultaneously improving antiparasitic activity and selectivity index relative to the background cell line. Subsequent hit optimization informed by the structure–activity relationship enabled by this virtual screening allowed thorough investigation of the pharmacophore, opening avenues for further improvement and optimization of the chemical series.
Although treatments for VL exist, these medications are rarely fit for purpose, with major issues linked to resistance, geographic variability in efficacy, toxicity, availability and affordability, and often long and painful treatment periods requiring many weeks of hospitalization.9–13 As a result, the past decade has seen increased efforts to develop new treatments for Leishmania related syndromes such as VL, with a primary focus on delivering pre-clinical candidates with improved profiles (broader geographic utility, reduced or no toxicity, oral dosing, up to 10 day treatment period).5 Target product profiles for these next generation VL treatments have been published and discussed widely.14,15
A combination of target-based and phenotypic-based drug discovery approaches have been applied extensively to identify new chemical matter for anti-leishmanial discovery projects, as well as the kinetoplastid field in general.9,16–18 Despite some excellent progress in target-based approaches, these often suffer from poor translation into efficacy in in vitro infection models,19 demonstrating the importance not only of a relevant mechanism of action but also the correct physicochemical properties to access the parasites once intracellular infection has taken hold. The phenotypic approach bypasses this issue by engaging cell-infection assays as the primary screen, however, these approaches often suffer from a combination of a low hit rate and identification of a high percentage of unsuitable chemical matter, with many hit lists comprising compounds rich in PAINS,20 toxicophores and intractable chemistries. Nevertheless, these approaches have yielded an interesting pipeline of novel potential treatments for VL, some of which have progressed to the early stages of clinical or preclinical evaluation.21 These include CRK12 inhibitor GSK3186899,22 the Leishmania proteasome inhibitors GSK3494245 and LXE-408,23,24 as well as DNDI-6174, DNDI-6148 and DNDI-0690, compounds identified via phenotypic drug discovery efforts.25
Despite this promising situation, the reality of clinical-level attrition in drug discovery R&D dictates that if the global health community are to ensure delivery of new, improved treatments to VL patients, further new chemical entities (NCEs) capable of tackling Leishmania infections still need to be identified and developed.26,27 Alongside the traditional target and phenotypic approaches mentioned above, recent reports have highlighted a number of efforts to identify novel anti-leishmanial NCEs via repurposing of existing drugs and clinical stage compounds,28 combined scaffold and parasite hopping,29 the screening of natural products,30 as well as investigation into host-mediated treatments for Leishmania infections.31
We report herein on early discovery efforts around an anti-leishmanial chemotype derived from the output of a kinetoplastid high throughput screening (HTS) campaign. Since 2012 we (DNDi) have tested over 2.2 million compounds in high throughput high content screening against kinetoplastid parasites including Leishmania donovani and Trypanosoma cruzi, the causative agent of Chagas disease,32 searching for novel and attractive small molecules as starting points for drug discovery campaigns.†† Seeking to improve the efficiency and throughput of the triaging of results from these phenotypic screening efforts led to the creation of the NTD drug discovery booster, a pre-competitive virtual screening model comprising key players from pharmaceutical R&D.33
The use of both orthogonal in silico techniques as well as orthogonally differentiated libraries, previously demonstrated to have minimal overlap,35 maximizes both the richness of the SAR explored around the initial seed phenotype, and increases the chances of identifying an interesting change of chemotype with similar or improved anti-parasitic activity, i.e. a scaffold hop. Furthermore, these newly identified scaffold hops can be subsequently mined using the same Booster approach, thus maximizing the value of the hits coming from HTS.
Combining the above findings with the nature of the assay used to identify the anti-leishmanial effect of the compound (high content imaging with parasite count) and the fact that the compound demonstrated anti-parasitic activity in more than one assay and against multiple related kinetoplastid parasites, we decided to cautiously investigate further. We were aware from the literature of the potential for a metal-chelation-driven pharmacological effect due to the bidentate heteroaryl-heteroaryl system in the core, indeed it could be hypothesized that this is the source of PAINS-like activity in other assays.39 Therefore, a key step for further investigation of 1 was to gain confidence via SAR and orthogonal assays that the observed anti-parasitic activity was unlikely to be driven by metal chelation, and that the PAINS flag could be legitimately discounted. The two key read-outs we targeted to achieve this were a) the ability of the chemotype to achieve a broader selectivity index for Leishmania donovani relative to the background host cell, and b) clear evidence that altering the core away from the PAINS-like chemotype maintained anti-parasitic activity, as suggested by the earlier paper from Marhadour et al.36
The partners performed two successive iterations of ligand-based virtual screening against their proprietary compound collections: cycle A and cycle B (see Fig. 1). Each distinct library was probed using a partner-specific in silico similarity search approach as described in Table 1 (see ESI† for detailed description). The compounds identified via these virtual screening efforts were investigated in a high content cell-based L. donovani infection assay using THP1 host cells. Assay read-out from this high content approach demonstrated the compound's efficacy at total parasite clearance as well as host cell cytotoxicity, yielding an anti-parasitic IC50, a host cell CC50 and a selectivity index (SI) between the two read-outs (SI = host cell IC50/parasite IC50; see ESI† for more details).
Company and computational approach | Number of compounds (actives)a | |
---|---|---|
On scaffoldb | Scaffold hopc | |
a Actives defined as compounds with SI > 5 against one or both parasites. b Compounds with a 2-substituted imidazo[1,2-a]pyridine core. c Compounds without a 2-substituted imidazo[1,2-a]pyridine core. | ||
A- ECFP4 (ref. 40) similarity (initial Tanimoto cut-off 0.7 descending incrementally until sufficient compounds had been identified) | 45 (12) | 15 (3) |
B- Series of substructure-based queries prioritized with Tanimoto similarity caluculated by use of Morgan fingerprint approach. Final selection cherry picked by eye | 69 (10) | 121 (18) |
C- ECFP4 (ref. 40) similarity (Tanimoto cut-off 0.6) and in-house fingerprint search (Tanimoto cut-off 0.7). Subsequent in-house “quality”, commercial availability, and IP filters | 85 (11) | 16 (0) |
D- Daylight and ChemAxon fingerprint similarity ranking,41,42 followed by Openeye pharmacophore alignment43,44 | 58 (12) | 56 (4) |
E- The top scoring 150 compounds were selected by Tanimoto similarity calculation using the FCFP4 fingerprint, followed by refinement to 96 compounds based on maximized diversity | 106 (11) | 86 (9) |
All companies | 363 (56) | 294 (34) |
Compounds with a SI > 5 were considered as demonstrating legitimate anti-parasitic activity. Results of this process are shown in Fig. 1. The first sequence of virtual screening (Fig. 1A) identified a total of 367 compounds from 4 partners, of which 25 demonstrated at least some degree of potency against L. donovani and shared the core imidazo[1,2-a]pyridine motif. Interestingly a new batch of compound 1, which displayed similar potency – cytotoxicity profile to the batch from the original HTS, appeared within this set (Fig. 2A). Compound 2 also appeared within this set, which demonstrated the potential of a new vector for exploration at the 8-position of the imidazo[1,2-a]pyridine core, as well as demonstrating efficacy with no background cell cytotoxicity. Parasitology data on each company's selection of compounds was fed back to the respective partner organizations. Parasitology data and the structure of compound 2 was shared with all partners and a second round of virtual screening was then completed using this elaborated parasitology data (Fig. 1B). The second round of in silico screening yielded a further 290 compounds from 5 participating organizations, of which 16 contained the imidazo[1,2-a]pyridine core and demonstrated some level of efficacy against L. donovani. Retrospective analysis of these two rounds of in silico screening demonstrated the power and versatility of this approach to hit to lead elaboration. Of the 657 compounds identified via in silico screening, almost 45% of the compounds represented a scaffold hop, defined in this instance as a compound not containing a 2-substituted imidazo[1,2-a]pyridine as the core. The hit rate of actives within this set of scaffold hops was only marginally lower than for those “on-scaffold” compounds which contained the original chemotype (11.5% compared to 15.4%) – yet these more structurally ambitious scaffold hop compounds would arguably not be included in a traditional hit to lead campaign due to the inherent increase in synthetic complexity of scoping multiple chemotypes. Compound 3 is a key example of such a scaffold hop originating from this approach, demonstrating significant potencies against both T. cruzi and L. donovani whilst representing a significant deviation away from the imidazo[1,2-a]pyridine core.§§
Combining the parasitology data for the key active compounds containing the original imidazo[1,2-a]pyridine core structure of 1, we collated data to form a hit series and move forward into detailed investigation and targeted analoguing as detailed below. An initial survey of the data generated by these rounds of virtual screening resulted in a number of initial findings: firstly, there appeared to be at least 5 vectors around the core which tolerated changes whilst retaining anti-leishmanial activity (positions 2, 3, 6, 7 and 8). Secondly, although certain structural changes in the chemotype induced significant cytotoxicity against the host THP1 cell line, we identified key examples which demonstrate acceptable differentiation between anti-parasitic activity and cell based cytotoxicity, with the most selective compounds achieving a selectivity index (SI) of >20 (4, 5, 6, 7). Analysis of the anti-parasitic activity and cell line cytotoxicity curves and high content images clearly demonstrated concentration points for most compounds where a full antiparasitic effect was observed with no cytotoxic observation on the background cell line (see Fig. 2B and C). Thirdly, we were able to identify compounds which achieved an antiparasitic effect with no background cell cytotoxicity whilst removing the 2-pyridyl functionality at the 2 position of the core (8), removing some suspicions about a metal-chelation driven SAR. These findings convinced us that the cytotoxicity observed for the series of compounds was not uniquely responsible for the observed anti-parasitic effect, and that the series merited further investigation. Finally, we also observed that the anti-leishmanial SAR around the scaffold was potentially non-additive, with the impact of certain functionality changes being influenced by substitution patterns elsewhere in the structure (Fig. 2A). This appeared particularly true for substitutions of the 6- and 7-position of the core imidazo[1,2-a]pyridine, as demonstrated by compounds 4–6, where the impact of a 6-methyl group improves potency when the 3-position is aryl substituted, but negatively impacts potency when the 3-position is alkylated. We therefore decided that rather than follow a rigid systematic investigation of each area of the molecule as an SAR follow-up strategy, we would use an empirical approach combining both single- and multi-dimensional SAR exploration. We made subtle variations around the 6- and 7-position substitution at the same time as exploring the 2- and 3-position. This was intended to give as broad a coverage of the chemotype as possible in a smaller number of compounds, and to avoid missing key improvements masked by the hypothesized non-additive SAR.
Compound | R3 | R6 | R7 | IC50a μM | CC50b μM | Clintc |
---|---|---|---|---|---|---|
a L. donovani parasite clearance in infected THP1 cells. b THP1 cytotoxicity. c Mouse microsomes, μL min−1 (mg.pr)−1. | ||||||
4 | N-Cyclopentylamino- | H | H | 1.80 | 42.5 | |
5 | N-(2,3-Dihydro-1,4-benzodioxin-6-yl)-amino | H | H | 1.16 | 25.6 | |
9 | N-2,2,2-Trifluoroethylamino- | H | H | 10.2 | >50 | |
10 | N-2-(N′,N'Dimethyl)-ethyl-1,2 diamino- | H | H | >50 | >50 | |
11 | N-2-Methyl-2-hydroxy-propylamino- | H | H | 30.4 | >50 | 171 |
12 | N-Cyclopropylmethylamino- | H | H | 3.57 | 19.4 | |
13 | N-Oxetan-2-ylmethylamino- | H | H | >50 | >50 | |
14 | N-Tetrahydrofuran-3-ylamino- | H | H | 29.4 | >50 | |
15 | N-Pyrolidin-3-ylamino- | H | H | >50 | >50 | |
16 | N-Morpholino- | H | H | >50 | >50 | |
17 | N-Benzylamino- | H | H | 9.34 | >50 | |
18 | N-Phenylamino- | H | H | 1.11 | 6.69 | |
19 | N-4-Fluorophenylamino- | Me | H | 2.71 | 9.61 | |
20 | N-(N-Methyl)-4-fluorphenylamino- | H | H | 3.9 | >50 | >1000 |
21 | N-3-Fluorophenylamino- | H | H | 3.70 | >50 | 242 |
22 | N-2-Fluorophenylamino- | H | OMe | 1.73 | 5.75 | 133 |
23 | N-3,5-Difluorophenylamino- | H | OMe | 1.91 | 4.3 | 62 |
24 | N-2,5-Dichlorophenylamino- | H | OMe | 3.91 | 8.93 | |
25 | N-5-Fluoropyridin-2-ylamino- | H | OMe | 1.71 | 16.0 | 81 |
26 | N-Pyridin-2-ylaminoamino- | H | Me | 4.53 | >50 | 130 |
27 | 4-Trifluoromethoxyphenylamino- | H | H | 3.49 | >50 | 223 |
28 | N-4-Cyananophenylamino- | H | H | 7.76 | >50 | |
29 | N-4-Methoxyphenylamino- | Me | H | 2.31 | 10.7 | |
30 | N-3-Methoxyphenylamino- | Me | H | 2.53 | 13.2 | |
31 | N-2-Methoxyphenylamino- | H | H | 2.61 | >50 | 326 |
Compound | R3 | R2 | R6 | R7 | IC50a μM | CC50b μM | Clintc |
---|---|---|---|---|---|---|---|
a L. donovani parasite clearance in infected THP1 cells. b THP1 cytotoxicity. c Mouse microsomes, μL min−1 (mg.pr)−1. d Value for library batch (value for resynthesized batch). | |||||||
4 | Cyclopentyl | Pyrid-2-yl | H | H | 1.80 | 42.5 | |
32 | 5-Methoxypyrid-2-yl | H | H | 5.26 | 19.3 | ||
33 | 4-Methoxypyrid-2-yl | H | H | 1.76 | 27.1 | 932 | |
34 | Pyrid-3-yl | H | H | >50 | >50 | ||
35 | 2-Methylpyrid-4-yl | H | H | 42.2 | >50 | ||
36 | 3,4-Dimethoxyphenyl | H | H | 16.2 | >50 | ||
37 | 4-(Dimethylamino)phenyl | H | Me | 7.37 (3.2)d | >50 (29.3)d | >1000 | |
38 | H | H | 7.9 | >50 | |||
39 | 4-Methoxyphenyl | H | Me | 10.2 | >50 | ||
40 | 2-Fluorophenyl | H | H | 25.9 | >50 | ||
41 | 2-Fluoro-4-methoxyphenyl | H | H | 22.0 | >50 | ||
42 | Thiazol-2-yl | H | H | >50 | >50 | ||
43 | N-Methylimidazo2-yl | H | H | >50 | >50 | ||
8 | Cyclohexyl | 4-N,N-Dimethylaminophenyl | Me | H | 5.93 | >50 | |
44 | 4-Methoxyphenyl | H | Me | 6.51 | >50 | ||
45 | 4-Ethoxyphenyl | H | H | 9.4 | >50 | ||
46 | 4-Tertbutylphenyl | H | H | 13.2 | >50 | ||
47 | 4-Fluorophenyl | 5-Methoxypyrid-2-yl | H | H | 2.5 | 15.4 | 126 |
48 | 4-Methoxypyrid-2-yl | H | H | 2.85 | 4.22 | >1000 | |
49 | 5-Chloropyrid-2-yl | H | H | 11.5 | >50 | ||
50 | N-Methylimidazo-2-yl | H | H | >50 | >50 | ||
51 | 3-Fluorophenyl | N-Methylimidazo-4-yl | H | OMe | 4.46 | 40 | <12 |
52 | Thiazol-4-yl | H | OMe | 15.7 | 40.6 | ||
53 | N-Methylpyrazol-3-yl | H | OMe | 36.2 | >50 |
Compound | R3 | R6 | R7 | IC50a μM | CC50b μM | Clintc |
---|---|---|---|---|---|---|
a L. donovani parasite clearance in infected THP1 cells. b THP1 cytotoxicity. c Mouse microsomes, μL min−1 (mg.pr)−1. d Library batch (resynthesized batch). e Library batch (initial HTS batch). | ||||||
19 | N-4-Fluorophenylamino- | Me | H | 2.71 | 9.61 | |
54 | CN | H | >50 | >50 | ||
55 | CF3 | H | 35.4 | >50 | ||
56 | MeO | H | 2.38 | 4.26 | 923 | |
57 | Cl | H | >50 | >50 | ||
58 | H | CN | >50 | >50 | 146 | |
59 | H | CF3 | 11.0 | 23.7 | ||
60 | H | MeO | 1.77 | 4.56 | 117 | |
61 | H | Me | 0.71 (2.45)d | 5.9 (11.8)d | 355 | |
21 | N-3-Fluorophenylamino | H | H | 3.7 | >50 | |
67 | H | Me | 2.47 | 14.24 | ||
62 | CH2OMe | H | 7.19 | 16.0 | ||
63 | CH2NMe2 | H | 5.12 | >50 | ||
64 | H | CH2OMe | 2.74 | 7.37 | 92 | |
65 | H | CH2NMe2 | 4.54 | 33.3 | ||
66 | H | OH | 16.87 | 43.3 | ||
1 | N-Cyclopentylamino- | Me | H | 4.67 (2.9)e | 28.0 (24.0)e | |
4 | H | H | 1.80 | 42.5 | ||
68 | H | Me | 1.88 | >50 | >1000 | |
5 | N-(2,3-Dihydro-1,4-benzodioxin-6-yl)-amino- | H | H | 1.16 | 25.6 | |
6 | Me | H | 0.62 (1.98)d | 16.0 (28.3)d | 411 | |
7 | Cl | H | 1.91 (6.33)d | >50 (17.4)d | >1000 | |
17 | N-Benzylamino- | H | H | 9.34 | >50 | |
69 | Me | H | 3.37 | >50 | >1000 | |
18 | N-Phenylamino- | H | H | 1.11 | 6.69 | |
70 | H | Me | 0.82 | 9.59 | ||
10 | N-2-(N′,N'Dimethyl)-ethyl1,2 diamino- | H | H | >50 | >50 | |
16 | N-Morpholino | H | H | >50 | >50 | |
71 | H | OMe | 17.9 | >50 | 137 |
Investigation of the 2-position of the core (Table 3) revealed that small substitutions such as chloro and methoxy on the 4- and 5-position 2-pyridyl were tolerated (32, 33, 42–47) and could be exploited to tune microsomal stability. Moving the position of the pyridyl nitrogen to the 3 or 4 position removed anti-leishmanial activity (34, 35), however replacement of the 2-pyridyl with 5-membered nitrogen-containing heteroaromatic rings was tolerated in certain instances (51), and this change also had a major impact on the metabolic stability of the compound. Interestingly, replacement of the pyridyl with electron enriched phenyls such as 4-anilino (37, 38) and 4-alkoxy substituted (44, 45) retained moderate activity and generally removed any evidence of cytotoxicity, suggesting that the anti-parasitic efficacy of the chemotype is likely not driven by a bidentate metal chelation effect. This observation of multiple non-2-pyridyl analogues further supported our earlier hypothesis regarding the redundancy of the PAINS flag for the initial hit 1.
Broader substitution of the 6- and 7-positions of the core imidazo[1,2-a]pyridine revealed a clear negative effect of withdrawing electron density from the core heterocycle, with cyano (54, 58) and trifluoromethyl (55, 59) substituents negatively impacting the antiparasitic effect (Table 4). Attempts to probe further away from the core negatively impacted potency and/or cytotoxicity (62–65), however incorporation of heteroatoms in this position could be used to improve microsomal stability (64). Scanning the impact of 6- and 7-substitution relative to the functionality at position 3 reconfirmed the previous observation that SAR appears non-additive, with changes having negative or positive impact on potency and cytotoxicity depending on the nature of the 3-position substitution.
Compound | Clint μL min−1 (mg.pr)−1 | Kinetic solubility μg mL−1 | ||
---|---|---|---|---|
Mouse | Human | pH 7.4 | pH 2.0 | |
6 | 411 | 95 | 8 | 72 |
7 | >1000 | >100 | <0.5 | >76 |
25 | 81 | 35 | >76 | >76 |
37 | >1000 | >100 | <0.5 | 67 |
61 | 354 | >100 | <0.5 | 73 |
Footnotes |
† Electronic supplementary information (ESI) available: Synthetic chemistry, in silico screening, and assay experimental details, including analytical data for all compounds. See DOI: 10.1039/d0md00353k |
‡ Current address: Elgia Therapeutics, Inc., San Diego, CA, USA. |
§ Current address: Loxo Oncology at Lilly, Boulder CO, USA. |
¶ Current address: Axcelead Drug Discovery Partners, Inc., Kanagawa, Japan. |
|| Current address: Molecular Devices, NYC, NY, USA. |
** Current address: The University of Tokyo, Tokyo, Japan. |
†† Compound libraries for HTS efforts at DNDi were sourced from a combination of commercial and proprietary collections. |
‡‡ Assignment of blinded names A–E is random. Specific contributions of each partner were, and remain, anonymous within the collaboration. |
§§ A full data set containing the compounds synthesized during this effort can be found in the supporting information and has been made fully available online: https://dndi.org/research-development/portfolio/drug-discovery-booster/ |
¶¶ Full synthetic details for all compounds in ESI.† |
|||| 3 was followed up further by subsequent rounds of in silico screening via this consortium approach, results to be communicated in future correspondence. |
This journal is © The Royal Society of Chemistry 2021 |