Introduction

Deaths from tuberculosis (TB) caused mainly by Mycobacterium tuberculosis (Mtb) now exceed those of any other infectious disease: in 2016, there were 10.4 million cases and 1.3 million deaths1. Inadequate compliance with the current treatment regimens has led to the emergence of resistant Mtb strains ranging from multidrug2,3 to totally drug-resistant TB4,5. This aspect coupled with toxicities and protracted treatment regiments associated with the use of current TB drugs6,7 indicate new drugs are urgently needed8,9. For malaria, caused mainly by the apicomplexan parasite Plasmodium falciparum (Pf), there were 216 million cases and 445,000 deaths in 2016, of which 90% of cases and 91% of the deaths occurred in sub-Saharan Africa10. Although mortality has decreased in recent years10, the spread of parasites resistant to current drugs including artemisinins is especially problematic in South-East Asia10,11,12. Thus, emphasis is placed on new drug combinations that in addition to acting on asexual blood stage parasites must act on sexually differentiated gametocytes responsible for transmission. One of the most widespread of apicomplexan parasites is Toxoplasma gondii (Tg) that infects both humans and animal species including livestock13. Tg has a major economic impact, and toxoplasmic encephalitis, especially in immunocompromised patients, and ocular and pulmonary diseases are the most serious clinical manifestations. Current treatments based on combination therapies comprising sulfonamides and pyrimethamine or other antimicrobials are often associated with adverse side effects14. Thus, better drugs are required13. Neospora caninum (Nc) is closely related to Tg, and is a major cause of abortion and stillbirth in cattle. It has a significant veterinary and economic impact, but infection in humans has not been reported to date. No drugs or vaccines are on the market15.

Under a programme involving these diseases16, we are developing combinations of new artemisinins17 with redox or ‘pro-oxidant’ drugs18 coupled with a third drug. For the last, we focus on quinolones (Fig. 1, structure 1) that are active against all pathogens, but possess distinct mechanisms of action. Quinolones are in clinical use against TB and other infectious diseases19, the most prominent of which are fluoroquinolones, e.g. ciprofloxacin (Fig. 1)20. The antimalarial activity of quinolones is well established21. Newer compounds have enhanced efficacies22,23,24 including ELQ-300 (Fig. 1)25 that are also active against transmissible malaria parasites. Quinolones also display activities against Tg26, and here too newer compounds, e.g. ELQ-271 (Fig. 1)27 are notably potent.

Fig. 1
figure 1

Biologically active quinolones. The 4(1H)-quinolone scaffold 1, the fluoroquinolone ciprofloxacin active against TB (ref. 20), the endochin analogue EQ-300 and the quinoline O-carbonate prodrug ELQ-337 active against malaria (refs. 23,24), and ELQ-271 active against toxoplasmosis (ref. 27). Activity of decoquinate 2 against Mtb is unrecorded, but is active against malaria (refs. 31,32) and toxoplasmosis (ref. 33). Activities of the quinolones in vitro against Mtb, Pf and Tg are given. Pf 3D7 and D6 are chloroquine (CQ)-sensitive, W2 and Dd2 are CQ-resistant strains

Full size image

Thus, the quinolone decoquinate (DQ, 2) (Fig. 1) is chosen as the incipient template for this work. DQ is used in veterinary practice for treatment of coccidiosis28, has a good safety profile29 (see also https://www.drugs.com/pro/deccox.html), and is active against both asexual30 and sexually differentiated blood stages31 of Pf, and against liver stages of P. yoelii30 and P. berghei (Pb)31. DQ is active against Tg32 where like other quinolones25 it exerts activity on the cytochrome bc complex. DQ is economically available, but has the disadvantage of high lipophilicity coupled with exceedingly poor aqueous solubility (0.06 µg/mL). Overall, drug-likeness of DQ should be improved by attaching groups that improve physicochemical properties. However, we are also mindful that for TB, the DQ template is attractive, as the lipophilic decyl side chain should enhance passive diffusion through the mycolic acid cell wall of Mtb, and of considerable importance for shortening treatment regimens33,34, may assist penetration of infected macrophages and granulomatous lesions encapsulating Mtb in various metabolic states, for which indeed, lipophilic molecules appear to be preferred35. As far as we are aware, the activity of DQ itself against Mtb has not been recorded.

Here we describe the conversion of the ethyl ester unit in DQ into more polar alkoxy esters and less readily metabolized amides16,36,37,38, and attach groups to the quinolone nucleus at N-1 or O-4. In the last case, we focus here on carbamate derivatives: the role of carbamates in enhancing bioavailability of drugs is established39,40. Finally, in order to assess the effect of the decyl group on efficacy, we also prepare analogues that lack this lipophilic side chain. The compounds are evaluated in vitro against each of Mtb, and the apicomplexan parasites Pf, Pb, Tg and Nc, and cytotoxicities are assessed against a human fetal lung fibroblast cell line.

Results

Chemical derivatization of decoquinate

The methods used to prepare the DQ derivatives 3–27 are outlined in Fig. 2. Although DQ is relatively acidic (pKa 9.81), aminolysis with various basic alkyl amines proceeded without incident to provide the amides 3–10 (route i in Fig. 2), as reported previously38. For preparation of the N-alkyl DQ amides, alkylation of DQ at N-1 is first required. Non activated alkyl halides react via an SN2 process with the ambident nucleophile derived from the 4-pyridone system of DQ (Fig. 3), and for such electrophiles, N-alkylation predominates41,42. In the event DQ was treated with the 1° alkyl bromide and K2CO3 in DMF to give the N-alkyl intermediates which upon aminolysis provided the N-alkyl DQ amides 11–20 (Fig. 2).

Fig. 2
figure 2

Derivatives obtained from decoquinate 2. i. 1° or 2° amine, 1,8-diazabicyclo[5.4.0]undec-7-ene (DBU), CHCl3, reflux, 24–72 h (ref. 38); ii. 1° Alkyl halide, K2CO3, DMF, 10 h; iii. Carbamoyl chloride, DBU, CHCl3, reflux, 15 h; iv. 2-Methoxyethanol, conc. H2SO4, CHCl3, reflux, 72 h; R = amine component of 2° or 3° amide; R′ = alkyl group; R″ = carbamoyl

Full size image
Fig. 3
figure 3

Regiochemistry of reactions of the pyridone anion. Deprotonation of the 4-pyridone system in DQ 2 generates the pyridone anion that as an ambident nucleophile may react through N-1 or O-4. For SN2-reactive electrophiles 1° alkyl halides R′X, lower activation free energies are associated with reaction through N-1 leading to the N-1-alkyl 4-pyridone (ref. 41). However, O-alkylation is preferred for SN1-reactive electrophiles, and carbonyl electrophiles, e.g. acyl halides, that react via addition-elimination (R″X = activated alkyl or acyl halide)

Full size image

For conversion of DQ into derivatives bearing polar groups, reactions with carbamoyl halides were used. For SN1-reactive alkyl halides41,42 and acyl halides that react via addition-elimination43, the 4-pyridone anion (Fig. 3) predominantly reacts through O-4. Treatment of DQ with the carbamoyl chloride and DBU (route iii in Fig. 2) generated the quinoline ethyl ester carbamates 21 and 22. Transesterification of DQ with methoxyethanol (route iv in Fig. 2) gave the intermediate methoxyethyl ester that on treatment with carbamoyl chloride (route iii in Fig. 2) gave the quinoline carbamates 23 and 24. Finally, DQ was submitted to aminolysis and the amides were treated with the carbamoyl chlorides and DBU to generate the quinoline carbamates 25–27. For the analogues lacking the decyl side chain, each of aniline, 4-chloroaniline and 3,4-dimethoxyaniline was condensed with diethyl ethoxymethylenemalonate, and the intermediate acrylamide cyclized to the 2-carbethoxyquinolone44,45,46. Aminolysis then gave the products 28–33 (Fig. 4).

Fig. 4
figure 4

DQ chain-shortened amide analogues: v. Diethyl ethoxymethylenemalonate, acetonitrile, reflux, 10 h; vi. Eaton’s reagent, N2, 70 °C, 10 h; i. as route 1 in Fig. 2. 1° amine, DBU, CHCl3, reflux, 24–72 h

Full size image

NMR spectroscopy

Because of the regiochemical issue associated with the reaction of the 4-pyridone anion in DQ with electrophiles (Fig. 3), unambiguous identification of the products is required. Although 2D NMR spectroscopy has been used in related systems for this purpose47,48, structural assignments to the DQ derivatives are markedly complicated by the saturated decyloxy side chain. Thus, the N-ethyl DQ amide 16 (Fig. 4b) and the quinoline ester carbamate 22 (Fig. 2) were examined by 2D NMR spectroscopy using correlation (COSY), total correlation (TOCSY), nuclear Overhauser effect (NOESY), heteronuclear multiple-bond correlation (HMBC) and heteronuclear single-quantum correlation spectroscopy (HSQC, gHSQCAD) edited to provide CH2 signals in opposite phase to CH3 and CH signals. For 22, signals were also assigned by comparing with data in the Spectral Database for Organic Compounds49. The methods together are discussed in detail in the Supplementary Methods and 1H and 13C chemical shifts and assignments to compounds 16 and 22 are given in Supplementary Tables 1 and 2. Signals in the 1H NMR spectra of the other derivatives could now be assigned as set out in the Supplementary Methods; these allow for clear differentiation between the structural types of Fig. 2. The constitution of all compounds was also confirmed by high-resolution mass spectrometry. All compounds had greater than 95% purity as assessed by analytical HPLC. These were more soluble in common organic solvents such as ethanol, acetone or dimethyl sulfoxide and less lipophilic (cLogP < 8) than DQ. Full details of spectroscopic data including high-resolution mass spectrometric data are given in the Supplementary Methods.

X-ray crystallography

For structural proof and to obtain information required for eventual molecular modelling associated with putative biological targets, crystalline samples of the DQ amide 16 and the quinoline carbamate 22 were sought for X-ray structure determination. Crystal growth conditions for amide 16 were eventually established by solvent layer diffusion of hexane into a saturated acetone solution at ambient temperature wherein fine colourless needle-like crystals were obtained. Although carbamate 22 recrystallized easily from protic solvents such as ethanol, it was not possible to obtain sufficiently large crystal specimens by layering methods or by controlled evaporation via solvent vapour diffusion, although a microcrystalline sample was prepared from ethyl acetate in the latter manner. A single needle specimen of 16 with dimensions 0.35 × 0.06 × 0.02 mm was protected by immersion in Paratone and mounted in a Cryoloop on a Rigaku-Oxford Diffraction Supernova diffractometer equipped with copper micro-source and Atlas detector. X-ray diffraction intensity data were collected at −100 °C to a 2-theta maximum of 135° (Cu-Kα radiation). The structure of 16 (Fig. 5) was solved by the charge-flipping algorithm of Olex-solve, refined using full matrix least squares with anisotropic nonhydrogen atoms and in the Olex2 X-ray structure program suite50,51. The quinolone unit has all four substituents disposed in the molecular plane. The amide substituent at C(3) is coplanar with the ring and oriented so that the C(31) = O(32) is in an anti-orientation with respect to the quinolone keto group O(4). This places the amide N(33)H in an orientation to make a weak incipient intramolecular hydrogen bond to O(4). The pendant alcohol O(39)-H also forms an intermolecular hydrogen bond to O(4) of a neighbouring molecule displaced along the a-axis. O(39)---O(4) = 2.743(2)Å. The decyl group has a regular all-trans-torsional arrangement and this aligns in the molecular packing with symmetry inverted neighbours to give efficient packing of the distinctly antiparallel decyl chains (Fig. 5). Another key packing feature is that considerable pair-wise π−π interaction occurs between the quinolone ring and its a-displaced antiparallel neighbour. Details of the X-ray methods and structural summary are given in the Supplementary Methods and Supplementary Table 3. The CIF files appear in Supplementary Data 1.

Fig. 5
figure 5

X-ray crystallographic determination of structure of compound 16. Crystallographic numbering for 16, ORTEP plot with 50% probability displacement ellipsoids, and packing diagram viewed along the a-axis showing antiparallel arrangement of the n-decyl chains

Full size image

Activities against tuberculosis and preliminary mechanism of action

Compounds were screened against Mycobacterium tuberculosis (Mtb) H37Rv::gfp, an engineered reporter strain that constitutively expresses green fluorescent protein (GFP), in a broth microdilution assay with rifampicin as a reference. As detailed previously52, assays were performed in the glycerol-based GAST-Fe medium. Activities are reported as minimum inhibitory concentration (MIC) values; these are determined by quantitative fluorescence and presented as the compound concentrations which inhibit the fluorescence signal by 90% (MIC90) or 99% (MIC99) relative to the untreated bacillary control at day 14 postinoculation. Data are in Table 1. DQ was inactive against Mtb. In contrast, 21 of the derivatives exhibited antimycobacterial activity, the most potent being the N-ethyl DQ amides 15–17 (Fig. 2); MIC99 values ranged from 1.25 µM for 16 to 3.64 µM for 17. The N-propyl counterpart 11 of compound 16 and the other N-propyl quinolones were less active, and the quinoline carbamates even less so, with the quinoline ethyl ester carbamate 22 displaying an MIC99 of >125 μM. Replacement of the ethyl ester with an amide in the latter compounds restored relatively good anti-Mtb activity: the quinoline amide carbamate 25 possessed an MIC99 of 4.89 μM. In notable contrast, the analogues 30–35, wherein the ethoxy and decyloxy chains of DQ were replaced with less bulky units (Fig. 4) were inactive against Mtb H37Rv (e.g. 3 MIC90 17.5 µM, cf. 30, MIC90 >125 µM). Most of the decoquinate derivatives exhibited low to moderate cytotoxicities against the normal human fetal lung fibroblast WI-38 cell line with IC50 values greater than 30 µM (Table 1).

Table 1 In vitro antimycobacterial activities and cytotoxicities of DQ derivatives
Full size table

Compounds 15–17 were tested in assays designed to provide an early indication of the potential target. The first utilizes an M. tuberculosis cytochrome bd oxidase deletion mutant cydKO, which is hypersusceptible to compounds such as imidazopyridines that target the QcrB subunit of the qcrCAB-encoded cytochrome bc1 complex53,54,55. No hypersensitivity was observed when the cydKO mutant was tested against 15–17. Similarly, there was no MIC shift observed when the compounds were tested against a derivative of the cydKO mutant carrying an additional Ala317Thr point mutation in qcrB53, supporting the notion that QcrB does not constitute the primary target. These results were confirmed for compound 16 using the XF96 Extracellular Flux Analyser56, in which the change in oxygen consumption rate (OCR) following the addition of a compound to bacterial cultures is determined at three different concentrations. Notably, no significant change in the OCR was observed following treatment with 16, also reinforcing the notion that QcrB is not the primary target (Supplementary Methods and Supplementary Figure 1). Bioluminescence reporter assays57 were used to establish if the compounds 15–17 disrupt cell wall homoeostasis or are genotoxic in Mtb. All three compounds elicited a strong signal using the PiniB-LUX reporter (Supplementary Figure 2), triggering an immediate luminescence signal on day 1 which was sustained throughout the 10-day assay, as was observed for the isoniazid control. This indicates that an aspect of cell wall metabolism constituted the primary target. When tested against the DNA damage reporter PrecA-LUX, the compounds did not elicit a signal above background for the first 4 days of the assay (Supplementary Figure 2). Interestingly, however, at days 7 and 10, a definite luminescence signal was detected, suggesting that DNA metabolism might constitute a secondary or downstream target. A similar phenotype has been observed with 5-fluorouracil58.

Activities against apicomplexan parasites

For the malaria parasite Plasmodium falciparum (Pf), the DQ derivatives with artemether, artesunate, and chloroquine as reference drugs were assayed against drug-sensitive Pf NF54 and drug-resistant K1 and W2 intraerythrocytic asexual parasites using the SYBR Green I fluorescence proliferative readout59. The DQ amides38 and the N-alkyl DQ amides 11–18 active against Mtb (Table 1) show weak or no significant activity against Pf in comparison with DQ, the reference drugs and the quinoline ester derivatives (Table 2). Likewise, the chain shortened amide analogues 28–33 (Fig. 4) when applied at 1 μM to CQ-sensitive Pf NF54 cultures showed poor activities in vitro with <50% growth inhibition. Although DQ displays good activities, the quinoline ester carbamates 21 and 22 (Fig. 2) are appreciably more reactive, and are approximately equipotent with artemether and artesunate (Table 2), widely regarded as among the most efficacious of antimalarial drugs. Compounds 22, 23, and 27 all have resistance indices <1, indicating that they display higher activities against multidrug-resistant than CQ-sensitive Pf strains.

Table 2 In vitro antimalarial activities and cytotoxicities of DQ derivatives (determined by SYBR Green I fluorescence viability readout, data from three independent biological replicates, n  =  3, each performed in technical triplicates, ±SE)
Full size table

Activity of the best compound 22 with chloroquine, artesunate, and atovaquone as comparator drugs was assessed against asexual blood stages of Pf NF54 and K1 (see above) and TM90-C2B resistant to atovaquone using an assay that measures parasite replication in erythrocytes by incorporation of [3H]-hypoxanthine into parasite DNA60; this assay thus uses a different readout to the above method (Table 3 and Supplementary Methods). Both assay methods indicate that compound 22 displays no cross-resistance with CQ-resistant strains. As noted above, quinolones act against the Pf cytochrome bc1 complex and display cross-resistance with atovaquone-resistant Pf25,61; for DQ, this is attributed to partial binding within the ubiquinol-binding site carrying one or more of the resistance mutations30. Thus, according to Table 3, cross-resistance is also observed for the DQ derivative 22. It is nevertheless anticipated that such an issue will not be of consequence once combinations of one or more of these or newer derivatives with artemisinins that do not display cross-resistance with atovaquone-resistant parasites (cf. Table 3) are developed.

Table 3 In vitro antimalarial activities of 22 (determined by [3H]-hypoxanthine incorporation; data from 2–3 independent experiments)
Full size table

In relation to potential for blocking transmission of parasites from human to mosquito, whereas DQ 2 showed no inhibition (cf. ref. 31), compounds 21, 22, and 24 showed 20–40% inhibition of gametocyte viability when treated at 1 µM against early (I−III) and late-stage (IV−V) Pf NF54 gametocytes, activities similar to those of atovaquone62 (Supplementary Methods and Supplementary Figure 3). Thus, these compounds are less active as transmission blocking agents than endochin-like quinolones such as ELQ-300 (Fig. 1) that arrest development of late-stage gametocytes at 1 µM25. Nevertheless, we describe elsewhere that newer artemisinins and redox active drugs to be used in combination with the quinolones are potently active in this regard17.

Compounds were tested in vitro for causal prophylactic effect against P. berghei sporozoites expressing luciferase in a liver stage malaria model63. Dose response analyses of compound 22 in 12-point titrations starting from 50, 10, and 2.5 µM showed that 22 maintained 100% inhibition of liver stage parasites down to 0.01 µM (Supplementary Figure 4) and was 154-fold and ~2-fold more potent respectively than puromycin and atovaquone. To establish the parasite selectivity of this compound in the liver stage of infection, cytotoxicity was determined by measuring bioluminescence from the reaction between free ATP of compound-treated, uninfected hepatocytes and a luciferase reporter system. The average maximum hepatocyte inhibition of compound 22 was 28.7 ± 11.2%, with no measurable IC50 up to 50 µM (Table 4). Thus, as in the above cases, compound 22 is highly selective for the malaria parasite. In comparison, in a related screening protocol using GFP-expressing liver stage sporozoites infecting the human hepatocellular carcinoma Huh7 cell line, DQ 2 itself expresses an activity of 2.6 nM31.

Table 4 In vitro activities of 22 against liver stage P. berghei and cytotoxicities (luciferase-expressing P. berghei ANKA GFP-Luc-SMcon sporozoites; data ± SD from biological duplicate (n = 2) and technical quadruplicate measurements)
Full size table

For the other apicomplexan parasites Toxoplasma gondii (Tg) and Neospora caninum (Nc), DQ 2, the N-alkyl DQ amides 12 and 19, the quinoline ester carbamates 21–24, the quinoline amide carbamates 26 and 27 and others were tested in vitro against Tg and Nc beta-galactosidase reporter strains grown in human foreskin fibroblasts (HFF)64,65. Prescreening at concentrations of 1000, 100, and 10 nM enabled preselection of compounds 22–24. Next, proliferation of Tg and Nc was tested in the presence of a concentration series of these compounds ranging from 0 to 20 nM. Toxicity against HFF host cells was determined using the Alamar Blue vitality test with a concentration series ranging from 0 to 10 µM. IC50 values and error margins were determined after logit-log-transformation of the respective values corrected for blanks. The data are summarized in Table 5. The compounds are appreciably more reactive than the bumped kinase inhibitor BP-1294, a pyrazolopyrimidine that inhibits calcium-dependent protein kinases of these parasites65. With respect to cytotoxicities against host HFF cells, compounds 22–24 were highly selective for the parasites. According to Table 2, the compounds displayed negligible cytotoxicities towards the human fetal lung fibroblasts WI-38 cell line, with IC50 values >100 µM.

Table 5 In vitro activities against T. gondii and N. caninum and cytotoxicities (determined against T. gondii and N. caninum tachyzoites in infected human foreskin fibroblasts (HFF), and cytotoxicities against host HFF)
Full size table

Assessment of cytotoxicities

The sulforhodamine B (SRB) assay was used to assess cytotoxicities against the normal human fetal lung fibroblast WI-38 cell line (HFLF WI-38) with emetine as standard. Activities are reported as IC50 values—the concentration required to reduce the viability of the cell line by 50%. The selectivity index (SI) is obtained by dividing the IC50 value by the MIC90 value for the mycobacterial assays and by the IC50 value for antimalarial activities. The biological activities of the compounds are in Tables 1, 2. The most active antimycobacterial compounds 15–17 had selectivity indices 11–69 (Table 1) representing an acceptable margin for carrying forward the compounds for development as TB drugs. For the apicomplexan parasites, DQ and the active compounds showed no appreciable toxicity to mammalian WI-38 cells.

Discussion

It is notable that the quinolone amide analogues 28–33 (Fig. 4) lacking the lipophilic decyloxy side chain of the DQ amides (Fig. 2) were inactive against Mtb. This vindicates the premise that one or more lipophilic side chains corresponding at least to the decyloxy side chain of DQ is a prerequisite for the development of new quinolones with antimycobacterial activities. It is further noted that the DQ amides 3, 4, 6, and 9 containing ethanolamino- or ethylenediamino-linked amide side chains elicit activities superior to the other DQ amides. In contrast, all such compounds are essentially inactive against malaria. The most active DQ amides 15–17 additionally contain an alkyl group attached to N-1. For mechanism of action of these compounds against Mtb, the results obtained from triage assays53,54, the lack of hypersensitivity observed in a cydKO mutant, and measurement of OCR eliminate mycobacterial respiration as target. In contrast, the immediate and sustained luminescence signal in the PiniB-LUX reporter assay strongly implicates cell wall homoeostasis as a likely primary target57. Moreover, the detection of late-stage (after day 4) triggering of recA promoter activity suggests some effect on DNA metabolism as well. In some respects, this is similar to the polypharmacology profile of 5-fluorouracil57. Further work is required to determine the precise mechanism, and will focus initially on DprE166 and MmpL367 given the frequency with which these have been identified in other in vitro screens for potential antimycobacterial compounds68. To summarize, the N-alkyl DQ amide derivatives 15–17 are easily accessible, with 16 being approximately equipotent in vitro with the fluoroquinolones ciprofloxacin (MIC90 1.5–12 μM, Fig. 1), gatifloxacin (0.66–1.3 μM) and moxifloxacin (0.62–1.3 μM) used clinically against TB20. Overall, the selectivities and novel mode of action with respect to disruption of cell wall homoeostasis indicate that these accessible quinolones have the potential to supplement other drugs for treatment of TB.

For malaria, even though in contrast to DQ, the N-alkyl DQ amide derivatives 11, 12, 17, and 18 (Fig. 2) are soluble in polar solvents, these are less active in vitro than DQ (Table 2). In notable contrast, the quinoline amide carbamates 26 and 27 are approximately equipotent with, and the quinoline ester carbamates 21, 22, and 24 are substantially more active than, DQ. The compounds may be crystallized from ethanol, are stable as pure substances and are soluble in common organic solvents. It is of relevance to note that the quinoline carbonate ELQ-337 (Fig. 1) serves as a prodrug for ELQ-300 from which it is chemically derived23. However, carbamates are expected on electronic grounds to be more stable to hydrolysis under neutral conditions than carbonates; clearly the question as to whether these compounds act as prodrugs for DQ must be resolved by assessment of pharmacokinetic parameters. In any event, the differing activities in vitro of each of 21 bearing a morpholino carbamoyl group and 22 bearing the more polar methanesulfonyl-imidazolidinone carbamoyl group must reflect enhanced uptake for the latter compound. Overall, the compounds 21, 22, and 24 that are economically obtained in one chemical step from DQ display in vitro efficacies markedly superior to DQ itself. For the other apicomplexan parasites (Table 5), it is noteworthy that the quinoline ester carbamates 22–24 are the most active compounds. Overall, the best compounds 21–24, in particular compound 22, have efficacies and selectivity indices with respect to normal cell lines that indicate considerable potential for use as therapies against apicomplexan parasites.

It is of no small consequence that the quinolone decoquinate, itself long used for treatment of parasites that cause coccidiosis in livestock28,29, can be economically converted into derivatives that are more potent than DQ itself, or display activities against a pathogen, namely Mtb, against which DQ appears to be inactive and wherein a principal mode of action differing to that of the antitubercular quinolones is operating. Moreover, the IC50 values for all compounds against normal human fetal lung fibroblast cell line WI-38 ranged from 40 to >100 μM, indicating good selectivity for the bacterial, and in particular, for the parasite, pathogens. The overall conversion of decoquinate into two of the optimum compounds 16 and 22 and their biological activities are summarized in Supplementary Figure 5. The next phases of the programme involving target profile evaluation and assessment of pharmacokinetic properties coupled with the combination studies as foreshadowed in the Introduction are underway, and will be reported upon in due course.

Methods

Preparation of decoquinate derivatives and analogues

The amines, the carbamoyl chlorides, the alkyl halides, and 1,8-diazabicyclo[5.4.0]undec-7-ene (DBU) from Sigma Aldrich or AKSci were used as supplied. 1H and 13C NMR spectra (NWU) of the products were obtained on a 600 MHz Bruker Avance™ III spectrometers as solutions in CDCl3, or DMSO-d6. Chemical shifts (δ) are reported in parts per million (ppm) values. 1H chemical shifts are reported downfield of tetramethylsilane, and were referenced to solvent signals in CDCl3 (7.26 ppm) or DMSO-d6 (2.50 ppm). 13C chemical shifts were referenced to CDCl3 (δ 77.16 ppm) or DMSO-d6 (δ 39.51 ppm). Peak multiplicities are abbreviated as follows: s (singlet), d (doublet), dd (doublet of doublet), t (triplet), q (quartet), and m (multiplet). Coupling constants J are in Hz. NMR data were analysed using MestReNova Software, version 5.3.2-4936. The NMR methods (HKUST) for compounds 16 and 22 employed a Varian VNMRS-500 spectrometer, with summary of methods given in the Supplementary Methods.

Exact mass measurements were recorded on Bruker MicroTOF Q II mass spectrometer that has an APCI/ESI source set at 300 °C. A full scan, ranging between m/z 50 and 1500, was generated at a capillary voltage of 4500 V, end plate offset voltage of −500 V, and a collision cell RF voltage of 100 V. Melting points were determined with a Büchi melting point B-545 instrument and were uncorrected. Thin layer chromatography (TLC) was performed using Merck silica gel plates (60F254), and components were visualized by ultraviolet light (254 nm). Silica gel 230–400 (particle size 40–63 μm) mesh was used for column chromatography. For HPLC analyses, an Agilent 1100 instrument equipped with a gradient pump, autosampler, diode array UV detector and OpenLab CDS Chemstation Rev.C.01.07 SR3 data acquisition software was used (Agilent Technologies, Palo Alto, CA, USA). The column was a Venusil XBP C18(2) -column, 150 × 4.6 mm, 5 µM spherical particles, 100 Å pore size (Agela Technologies, Newark, DE, USA). The mobile phase was acetonitrile and 0.1% H3PO4 acid in water with a linear gradient from 30% acetonitrile to 85% after 5 min and holding until 15 min before equilibrating with 30% acetonitrile to 20 min. The flow rate was set at 1 mL/minute and the injection volume was 10 µL. The UV signal was monitored at 210 nm.

DQ amides 3–10

A mixture comprising the 1° or 2° amine (5 eq.), DQ (1 g, 2.34 mmol) and DBU (0.33 g, 320 µL, 2.1 mmol, 0.9 eq.) in chloroform (25 mL) was stirred under reflux for 24–72 h. Progress of the reaction was monitored by TLC. Upon completion, the reaction mixture was concentrated by evaporation under reduced pressure. The residue was dissolved in chloroform (20 mL) and washed with distilled water (4 × 20 mL). The organic layer was dried (Na2SO4), filtered and the filtrate was evaporated to dryness to leave the crude product. This was submitted to chromatography. Elution with dichloromethane-methanol (10:1 v/v) provided the fraction containing the product that was isolated by evaporation of solvent. The product was recrystallized from ethyl acetate. Data are given in Supplementary Methods.

N-alkyl DQ amides 11–20

A mixture of DQ (0.5 g, 1.19 mmol, 1 eq.), K2CO3 (0.69 g, 5.0 mmol,) and the alkyl halide (5 eq.) in DMF (10 mL) was heated under gentle reflux for 10 h, after which the mixture was evaporated to dryness under reduced pressure. The residue was treated with chloroform, and filtered. The filtrate was washed with distilled water (4 × 20 mL). The organic phase was evaporated to dryness to obtain the crude N-alkyl product which was used in the ensuing step without purification.

A mixture of the crude N-alkyl product (0.5 g, 1 eq.), DBU (0.33 g, 320 µL, 2.1 mmol), the amine (5 eq.), and chloroform (15 mL) was stirred under reflux for 24–72 h. Progress of the reaction was monitored by TLC. Upon completion, the reaction mixture was treated with additional chloroform (20 mL) and washed with distilled water (4 × 20 mL). The organic layer was dried (Na2SO4), filtered and the filtrate was evaporated to dryness to leave the crude product. This was submitted to chromatography. Elution with dichloromethane-methanol (10:1 v/v) provided the fraction containing the product that was isolated by evaporation of solvent. The product was recrystallized from ethyl acetate. Data are given in Supplementary Methods.

Quinoline ester carbamates 21–24

To mixture of DQ (1 g, 2.34 mmol, 1 eq.) and 2-methoxyethanol (1 mL, 5 eq.) in chloroform (20 mL) was added conc. H2SO4 (0.33 g, 320 µL, 2.1 mmol, 0.9 eq.). The mixture was stirred under reflux for 72 h. The reaction mixture was washed with aqueous saturated NaHCO3 solution (3 × 20 mL). The organic layer was dried (Na2SO4), filtered and the filtrate was evaporated to dryness to leave the crude product. This was submitted to chromatography on silica gel. Elution with dichloromethane-methanol (20:1 v/v) provided the fraction containing the intermediate that was isolated by evaporation of solvent. This intermediate was used in the next step without further purification.

A mixture comprising DQ (0.5 g for preparation of 21, 22), or the above crude product (0.5 g for 23, 24), chloroform (15 mL), DBU (100 µL, 0.7 mmol), and the carbamoyl chloride (5 eq.) was stirred under reflux for 15 h, and then washed with saturated aqueous NH4Cl (3 × 20 mL). The organic layer was dried (Na2SO4), filtered and the filtrate was evaporated to dryness to obtain the crude product which was purified by chromatography. Elution with dichloromethane-methanol (20:1 v/v) provided the fraction containing the product that was isolated by evaporation of solvent. Derivatives 21–24 were thereby obtained in 50–80% yields. Data are given in Supplementary Methods.

Quinoline amide carbamates 25–27

The amine (5 eq.) was added to a mixture of DQ (0.5 g, 1.19 mmol) and DBU (0.33 g, 320 µL, 2.1 mmol, 0.9 eq.) in chloroform (25 mL) and the mixture was stirred under reflux for 72 h. Progress of the reaction was monitored by TLC. Upon completion, the reaction mixture was treated with additional chloroform (20 mL) and washed with distilled water (4 × 20 mL). The organic layer was dried (Na2SO4), filtered and the filtrate was evaporated to dryness to leave the crude product, which was used in the next step without further purification.

A mixture of the above crude product (0.5 g), chloroform (15 mL), DBU (0.7 mmol, 100 µL), and the carbamoyl chloride (5 eq.) was stirred under reflux for 15 h. It was then cooled and washed with saturated aqueous NH4Cl solution (3 × 20 mL). The organic layer was dried (Na2SO4), filtered and the filtrate was evaporated to dryness to obtain the crude product which was purified by chromatography. Elution with dichloromethane –methanol (20:1 v/v) provided the fraction containing the product that was isolated by evaporation of solvent. Derivatives 26–29 were thereby obtained in 50–80% yields. Data are given in Supplementary Methods.

Chain-shortened DQ analogues 30–35

A solution of the 3- or-4 substituted aniline and diethyl ethoxymethylenemalonate (1 eq.) in acetonitrile (10 mL) was stirred under reflux for 5–10 h. Progress of the reaction was monitored by TLC. Upon completion, the mixture was evaporated to dryness to leave a crude product, which was dissolved in Eaton’s reagent (10 mL) and the resultant mixture heated at 50–70 °C under nitrogen for 10–20 h. Thereupon the mixture was cooled to room temperature and then added portionwise to saturated aqueous NaHCO3. The resulting precipitate was collected by filtration, dried overnight and subjected to aminolysis as below.

The foregoing quinoline (0.5 g) and DBU (0.33 g, 320 µL, 2.1 mmol) in chloroform (25 mL) was treated with the amine (5 eq.), and the resulting mixture was stirred under reflux for 24 h. Progress of the reactions was monitored by TLC. Upon completion, the reaction mixture was treated with additional chloroform (20 mL) and washed with distilled water (4 × 20 mL). The organic layer was dried (Na2SO4), filtered and the filtrate was evaporated to dryness to leave the crude product. This was submitted to chromatography. Elution with dichloromethane-methanol (10:1 v/v) provided the fraction containing the product that was isolated by evaporation of solvent. The product was recrystallized from ethyl acetate and air dried. Data are given in Supplementary Methods.

In vitro antimycobacterial assays

The broth microdilution method was used to assess the MIC90. The procedure involves using a range of concentrations of the test compound in one 96-well microtitre plate. A culture of mutant Mtb (H37RvMA::gfp) strain expressing recombinant GFP off a plasmid integrated at the attB locus was grown to OD600 between 0.6 and 0.7. This culture was diluted 1:100 in GAST/Fe medium. A total volume of 50 µL of the GAST/Fe medium was added to rows 2–12 of the plate. To row 1, 640 µM of the compound was added in duplicate. In addition to the test compound in row 1, other wells contained GAST/Fe medium, 5% DMSO, rifampicin and kanamycin as controls. From there two-fold serial dilution were performed by transferring 50 µL solution from each preceding row to the next ensuring that all the wells contained 50 µL (the dilution was exclusively for compound containing wells from row 1, and not for controls). The plate was incubated at 37 °C (humid conditions) for 14 days. MIC90 values were scored by quantitative fluorescence using a Fluostar Optima microplate reader (BMG Labtech) at 7 and 14 days postinoculation. Digital images were captured and stored.

In vitro blood stage asexual P. falciparum parasite assays

Asexual P. falciparum parasites were maintained in human O+ erythrocytes at 37 °C with 90% N2, 5% CO2 and 5% O2 in RPMI 1640 medium supplemented with Albumax 1. Parasite growth was determined microscopically using Giemsa-stained smears. Synchronized (>95%) ring stage parasites were obtained using 5% d-sorbitol (Sigma-Aldrich). Drug solutions were prepared from a 10 mM stock solution 100% DMSO (Sigma-Aldrich) in supplemented RPMI 1640 media (final DMSO concentration <0.1% v/v). Drug-sensitive NF54 and drug-resistant K1 and W2 (chloroquine, quinine, pyrimethamine and cycloguanil) Pf parasites were treated using a twofold drug serial dilution for 96 h. The dose−response (IC50 values) of the DQ derivatives and reference drugs (Table 3) were determined using the SYBR Green I based readout for parasite proliferation59. Data analysis was performed using GraphPad Prism 6. Data are from three independent biological replicates, performed in technical triplicates.

In vitro liver stage P. berghei parasite assays

A. stephensi mosquitoes infected with P. berghei-ANKA GFP-Luc-SMcon (reference clone 15cy1) sporozoites were purchased from the Insectary Core at New York University. Salivary glands were dissected and suspended in cold DMEM (Invitrogen, Carlsbad, USA) before homogenization with a glass Dounce tissue grinder (Wheaton Industries). Liberated sporozoites were passed through 20 µM (Millipore SCNY00020) then 11 µM (Millipore NY1104700) nylon net filters, and counted using a Neubauer hemocytometer (C-Chip, InCyto, Korea). The hepatoma cell line HepG2-A16-CD81EGFP was used for liver stage cytotoxicity and infection assays. Hepatocyte cultures were maintained using DMEM containing phenol red (Invitrogen cat# 11965–092), 10% FBS (Corning cat# 35-011-CV), and 1× Pen Strep Glutamine (100 Units/mL penicillin, 100 µg/mL streptomycin, and 0.292 mg/mL l-glutamine) (Invitrogen cat# 10378-016). Screening assays were performed using phenol red free DMEM (Invitrogen cat# 31053-028), 5% FBS (Corning), and 5× Pen Strep Glutamine (Invitrogen). Both maintenance and screening were at 37 °C and 5% CO2.

Less than 24 h prior to the addition of hepatocytes, 50 nL of compounds solubilized in DMSO (0.5% final DMSO concentration) were stamped into empty 1536-well plates (GNF ref#789173-F, Greiner Bio-One) using the Gen 4 Plus Acoustic Transfer System (Biosero, San Diego, USA). Dose−response titrations were made in 12-points at a 1:3 dilution ratio, with atovaquone and puromycin used as positive controls for the infected and cytotoxicity assays, respectively. DMSO (0.5%) was used as a negative control in both assays. To each well, 3×103 of the HepG2-A16-CD81EGFP cells were dispensed in 5 µL and incubated for 24 h at 37 °C and 5% CO2. Following the dissection, purification and counting of sporozoites (spz), a solution containing 200 spz/µL was made in screening assay media (DMEM (Invitrogen cat# 31053-028)), 5% FBS (Corning), and 5× Pen Strep Glutamine (Invitrogen). Each well received 5 µL of the spz solution (1×103 spz/well). All parasite and cell solution dispensing was performed using the GNF Systems Washer/Dispenser II (GNF). Infected plates were centrifuged at 330 × g for 3 min immediately after the addition of sporozoites to aid with hepatocyte invasion. In the cytotoxicity studies, 5 µL of screening assay media was used only to match the volume of infected plates (10 µL). All plates were then incubated for 48 h at 37 °C and 5% CO2 (high humidity).

Following the 48 h incubation, plates were inverted and centrifuged at 150 × g for 30 s to remove well media. To infected plates, 2 µL of BrightGlo (Promega) was dispensed into each well using the MicroFlo (BioTek) liquid handler and read immediately using the EnVision Microplate Reader (PerkinElmer). CellTiterGlo (Promega) was reconstituted and then diluted 1:2 with deionized water prior to adding 2 µL/well for the cytotoxicity assay. Plates with CellTiterGlo were left out for 2 min before reading with the EnVision to allow for a plateau of the bioluminescent signal. Data analysis was performed using GraphPad Prism 7. Data points were normalized to positive and negative controls and a nonlinear regression curve was fit to determine IC50 values.

In vitro assays of T. gondii and N. caninum tachyzoites

Compounds were dissolved in DMSO to yield 1 mM stock solutions. Tachyzoites of each of T. gondii and N. caninum NC-1 expressing E. coli beta-galactosidase were used65. Confluent HFF previously seeded into 96-well-plates were treated with the compounds or with DMSO as a control, and were infected with 103 tachyzoites per well and incubated at 37 °C, 5% CO2. Beta-galactosidase assays were performed when the control wells were thoroughly infected, namely, after 2–3 d for Tg, and after 3–4 d for Nc. At these times, the medium was removed, specimens were washed with PBS, and beta-galactosidase was assayed and IC50 values were calculated after logit-log-transformation of the original data. Toxicity against HFF host cells was determined using the Alamar Blue vitality test with a concentration series ranging from 0 to 10 µM.

In vitro cytotoxicity assays

The WI-38 cell line (normal human fetal lung fibroblast) from ECACC was routinely maintained as a monolayer cell culture at 37 °C, 5% CO2, 95% air and 100% relative humidity in EMEM containing 10% fetal bovine serum, 2 mM l-glutamine and 50 µg/mL gentamicin. For the screening experiments, the cells (21–50 passages) were inoculated in 96-well microtiter plates at plating densities of 10,000 cells/well and incubated for 24 h. After 24 h, the cells were treated with the compounds dissolved in DMSO and diluted in medium to produce five concentrations. Neat cells served as the control. The blank contained complete medium without cells. Parthenolide was used as a standard. The plates were incubated for 48 h after addition of the compounds. Viable cells were fixed to the bottom of each well with cold 50% trichloroacetic acid, washed, dried and dyed by SRB. Unbound dye was removed and protein bound dye was extracted with 10 mM Tris base for optical density determination at a wavelength of 540 nm using a multiwell spectrophotometer. Data analysis was performed using GraphPad Prism software. Fifty per cent of cell growth inhibition (IC50) was determined by nonlinear regression.

Preparation and characterization of decoquinate derivatives and analogues

See Supplementary Methods.