The bacterial species used in this study were Mycobacterium aurum (ATCC 23366) and M. bovis BCG Pasteur (ATCC 35734). The mammalian cell lines used for cytotoxicity assays were the murine macrophage cell line RAW 264.7 (ATCC-71) and human leukemia monocytic cell line THP-1 (ATCC TIB-202). Mycobacteria were cultured in Middlebrook 7H9 broth (BD Difco™) or Middlebrook 7H10 agar media (BD Difco™) supplemented with albumin/dextrose/catalase (ADC, Remel™) or oleic acid/albumin/dextrose/catalase (OADC, BD Difco™) enrichments, respectively. Mammalian cell lines were cultured in Roswell Park Memorial Institute (RPMI) 1640 media with L-glutamine (Gibco^®) supplemented with fetal bovine serum (FBS, Sigma Aldrich, Non-US origin, sterile filtered, heat-inactivated).

The procedures employed for chemical synthesis are reported in the Supplementary Information together with the relevant characterization data.

All compounds were dissolved in dimethyl sulfoxide (DMSO) to a concentration of 50 mg/ml. 2 µl aliquots from a 2-fold serial dilution of this stock were transferred into a 96-well microtiter plate. Isoniazid was included as a positive control and DMSO as a negative control. Middlebrook 7H10 agar medium containing 0.5% glycerol and 10% (v/v) OADC (200 µl) was dispensed into each well and allowed to solidify. A mid-logarithmic phase culture of M. aurum or M. bovis BCG was diluted in Middlebrook 7H9 broth to give a cell density of 1 x 10⁶ cells/ml (Gupta and Bhakta, 2012) and then dispensed (2 µl, ~ 2000 cells) into each well aseptically. The plates were then incubated at 35 °C (for M. aurum) and 37 °C (for M. bovis BCG). The results were observed after 5 days for M. aurum or 14 days for M. bovis BCG. MICs were visually determined by observing the lowest concentration of each compound at which no bacterial growth could be observed. These assays were performed in biological triplicate.

All compounds were dissolved in DMSO to a concentration of 50 mg/ml. Aliquots (2 µl) from a 2-fold serial dilution of this stock were transferred into a 96-well microtiter plate, with a DMSO-only well included as a negative control. Cultures of RAW 264.7 or THP-1 cells (5 ml), were diluted to a cell density of 5 x 10⁵ cells/ml in RPMI and aliquots (100 µl) were then transferred into each well of the 96-well plate. One row of the 96-well plate was left without cells, as a sterility control. The cells were then incubated without shaking (37 °C, 5% CO₂) for 48 hours. The spent media was removed by pipetting and the media was replaced with complete RPMI (170 µl) and resazurin solution (0.01%, 30 µl). For non-adherent THP-1 cells, the cells were pelleted by centrifugation of the 96-well plates (270 x g, 2 min) before aspiration of the media. The plates for both cell lines were then incubated for a further 24 h (37 °C, 5% CO₂). Growth inhibitory concentrations (GICs) were visually determined by observing the lowest concentration of each compound at which the resazurin dye was not reduced from blue to pink. Visual determinations were confirmed by fluorescence readings (Excitation wavelength: 560 nm, Emission wavelength: 590 nm) for each plate using a BioTek^® Synergy 2™ Multi-Detection Plate Reader. These assays were performed in biological triplicate.

The assay was modified from a previously published protocol (Rodrigues et al., 2008). Early log phase cells of M. aurum (OD₆₀₀ ∼ 0.8) were harvested by centrifugation and resuspended in 1 x PBS to an OD₆₀₀ of 0.4. The test samples contained 4−6 × 10⁷ CFU/ml in PBS, glucose (0.4%), ethidium bromide (0.5 mg/l), and the compounds of interest at ¼ × MIC. Blank samples contained all of the components mentioned above, except the bacterial suspension, which was replaced with 1 × PBS. Verapamil, a known efflux pump inhibitor, was used as the positive control at a concentration of 125 µg/ml. The experiment was performed in a fluorimeter (FLUOstar OPTIMA, BMG Labtech) programmed with the following parameters: wavelengths of 544 nm for excitation and 590 nm for detection of fluorescence, gain 2200, a temperature of 37°C, and a cycle of measurement every minute for a total period of 60 min. The accumulation or efflux of ethidium bromide was monitored on a real-time basis. The ability of compounds to inhibit ethidium bromide efflux was quantified by determining the relative final fluorescence (RFF) of each compound. This was calculated according to Formula 1:

RF_treated is the relative fluorescence (RF) at the last time point of the EtBr accumulation assay for the experimental sample, while RF_control is the RF of the untreated control sample at the last time point of the assay (Machado et al., 2011; Coelho et al., 2015).

The synergistic effect of THIQs were examined in combination with isoniazid, ethambutol and rifampicin against M. aurum. The assay was conducted in a 96-well microtiter plate using a SPOTi checkerboard distribution as reported previously (Guzman et al., 2013). The compounds were serially diluted in DMSO across a range of concentrations spanning both above and below the MIC of each compound/drug. Each row of the 96-well plate contained different concentrations of the THIQ derivatives, and each column contained different concentrations of the first-line anti-TB drug. The checkerboard was constructed by adding 1 µl of each of the stock concentrations to the corresponding well and then dispensing 200 µl of warm Middlebrook 7H10 agar medium. The plates were then spotted with M. aurum. The plates were incubated in sealed bags at 35° C for 5 days and Fractional Inhibitory Concentration Indices A FICI value were calculated using Formula 2:

An FICI value ≤ 0.5 indicates synergism; a value between 0.5 and 4.0 indicates no interaction; and a value higher than 4.0 indicates antagonism (Odds, 2003).

An initial group of C-1 substituted THIQs (Group I) was prepared via a phosphate-mediated, biomimetic Pictet-Spengler reaction between dopamine hydrochloride and a range of aldehydes (Table 1) (Pesnot et al., 2011). Both the major (6,7-, 1-10a) and minor (7,8-, 1-10b) regioisomers were isolated by preparative HPLC (Guzman et al., 2015). The C-1 alkyl chain length was systematically explored, as previous work had highlighted a longer alkyl chain at this position as having promising activities. C-1 cyclopentyl, 2-methyl benzyl and 3,4-methylenedioxybenzyl moieties were also explored, the latter of which had been present in several active analogues identified previously (Guzman et al., 2015). Moreover, a hydroxyl group at C-8 had also been associated with higher activities in some cases, so both 6,7- and 7,8-regioisomers were generated (Guzman et al., 2015). All compounds were prepared using commercially available aldehydes with the exception of 9a and 9b, for which (m-tolyl)acetaldehyde was prepared from the commercially available alcohol by a Parikh-Doering oxidation (Parikh and Doering, 1967), and 10a and 10b, for which 3,4-(methylenedioxy)phenylacetaldehyde was prepared as previously described (Guzman et al., 2015). Regioisomers 11a and 11b were prepared from a Pictet-Spengler reaction between m-tyramine and hexanal. THIQs 4a.HCl and 11a.HCl were prepared as the hydrochloride salts of their respective free-bases. The enantio-enriched THIQs (1S)-4a.HCl and (1S)-11a.HCl were synthesized using norcoclaurine synthase (TfNCSΔ33) in a biocatalytic Pictet-Spengler reaction between hexanal and dopamine or m-tyramine, respectively (Table 1) and were included in Group I to explore C-1 stereochemistry (Pesnot et al., 2012; Ruff et al., 2012; Lichman et al., 2015).

Group II was designed primarily to further the exploration of A-ring substituents, including hydroxy substitutions unexplored in Group I; C-5 bromo substitution and A-ring methoxy substituents. A C-1 pentyl chain was retained for all compounds as the most potent and selective compounds in Group I featured this substituent (see below). The initial phenethylamines required for the synthesis of this series were either acquired commercially or prepared by reduction of the appropriate phenylacetonitrile (see supplementary information). Brominated phenethylamines were prepared by bromination and subsequent demethylation of the appropriate methoxy phenethylamines (Scheme S1). THIQs 12-15 were then prepared via a phosphate-mediated Pictet-Spengler reaction (Table 2) (Pesnot et al., 2011). THIQ 16 with a C-7 hydroxyl group was also prepared to compare to 11a which features a C-6 hydroxyl group. The synthesis of 16 (Scheme 1A) began with an initial amide coupling reaction between 4-methoxyphenyethylamine and hexanoic acid to give 17. This was followed by Bischler-Napieralski cyclisation (Heravi et al., 2014) and subsequent imine reduction to afford 18 and 19 as a mixture of regioisomers. These compounds were then demethylated and separated by preparative HPLC to isolate 16 (and 11a; Scheme 1A). Methoxy-substituted THIQs at C-6; C-6 and C-7; or C-6 and C-8 (18, 20 and 21) were synthesized in a similar manner (Scheme 1B) to compare to their hydroxy-substituted counterparts 4a, 11a and 12. First, amides 22-24 were prepared by a Sheppard amidation (Sabatini et al., 2017), before Bischler-Napieralski cyclisation to afford dihydroisoquinolines 25-27, which then underwent imine reduction to give the desired THIQs 18, 20 and 21.

To explore the influence of B-ring aromaticity, Group III was prepared. This group contained fully aromatic isoquinolines rather than THIQs. The A-ring-unsubstituted isoquinoline 28 was prepared by first synthesizing β-hydroxy amide 29 via a Sheppard amidation (Sabatini et al., 2017) between 2-amino-1-phenylethanol and hexanoic acid (Scheme 2A). This was followed by a Pictet-Gams cyclisation (Wang, 2010) to directly afford the fully aromatic isoquinoline. A-ring substituted isoquinolines were prepared by oxidation of previously prepared dihydroisoquinolines 25-27 with Pd/C under air to afford the methoxy-substituted compounds 30-32 (Scheme 2B) (Awuah and Capretta, 2010). These were then demethylated to afford their hydroxy-substituted counterparts 33-35. Compound 36 (Scheme 2C) was synthesized as previously reported, for use in synergism studies (Guzman et al., 2015).

Generally, for the 6,7-dihydroxy-substituted THIQs, 1a-7a, potency against M. aurum was found to increase with increasing chain length (Table 3). Toxicity to RAW 264.7 cells was high for THIQs 1a, 2a, 4a, 6a and 7a. However, 3a and 5a displayed lower cytotoxicity, with GIC values of ≥ 62.5 µg/ml, indicating a non-linear relationship between carbon chain length and cytotoxicity. The same trends were not observed for the 7,8-dihydroxy-substituted THIQs, 1b-7b, all of which had MIC values of ≥ 125 µg/ml, with no clear relationship between chain length and activity (Table 3). All of these compounds had high cytotoxicity against the RAW 264.7 cell line with GIC values of 15.6 µg/ml, suggesting no relationship between chain length and cytotoxicity. Of the alkyl chain-bearing THIQs, the C-1 pentyl THIQs, 4a and 4b, appeared to confer the best balance of antimycobacterial potency and mammalian cell cytotoxicity. For the THIQs with cyclic C-1 substituents, it was observed that the 7,8-hydroxy-substituted examples had greater potency against M. aurum than their 6,7-substituted counterparts (Table 3). However, these compounds were also more toxic to RAW 264.7 cells and did not show improved selectivity. Only the 3,4-methylenedioxybenzyl moiety (10a & 10b) was found to confer comparable potency and selectivity to the C-1 pentyl group (4a & 4b). The 6- and 8-mono-hydroxy-substituted THIQs (11a & 11b) were observed to have similar activity to 4a and 4b (Table 3). Both 4a.HCl and 11a.HCl were found to have comparable activity to their respective free-bases, 4a and 11a, however the hydrochloride salts were found to exhibit reduced cytotoxicity against the RAW 264.7 cell line. Many THIQs and isoquinolines in Group II and all of Group III were therefore formulated as their HCl salts to reduce cytotoxicity and to improve the solubility of the more lipophilic compounds. Interestingly, (1S)-4a.HCI and (1S)-11a.HCI, were found to have low potency and cytotoxicity. When compared to the equivalent racemic THIQs, these results suggest that the (1R)-THIQs may be more potent and cytotoxic than the corresponding (1S)-THIQs. All compounds in this series displayed selectivity for toxicity to the mammalian RAW 264.7 cells over M. aurum; an issue which would require optimization to overcome.

In THIQ Group II (Table 2 and Scheme 1), it was observed that the transition away from catechol-bearing THIQs generally resulted in increased potency and selectivity (Table 4). Dimethoxy-substituted THIQ 21 was the most promising compound in this series, with the highest potency observed against both M. aurum and M. bovis BCG. The high cytotoxicity of this compound against RAW 264.7 resulted in SI values of 1 or 2 (M. aurum or M. bovis BCG, respectively). These SI values are in line with those observed for the most selective THIQs reported in our previous work (Guzman et al., 2015). The two dimethoxy-substituted THIQs, 20 and 21, were also among the most selective compounds in the series, suggesting that A-ring methoxy-substitution may improve selectivity.

The isoquinolines in Group III (Schemes 2A, B) were found to have similar activity to their THIQ counterparts (Table 5), suggesting that B-ring aromaticity had a limited influence on the activity of these compounds. A notable exception to this trend is 33, which was observed to be significantly more antimycobacterial and selective than the corresponding THIQ, 11a. This compound is also notable for having an SI of 2 for both M. aurum and M. bovis BCG over RAW 264.7 cells and is the only compound in this study with any selectivity for toxicity to M. aurum over the RAW 264.7 cell line. Compound 32 was the most potent compound in this series against both mycobacterial species. It was not found to be selective for toxicity to M. aurum but had an SI of 2 for M. bovis BCG over RAW 264.7.

The compounds in Group III were also screened for cytotoxicity against the human THP-1 cell line, alongside selected compounds from Groups I and II (Table S1). In general, these compounds showed similar cytotoxicity against this cell line, however, some displayed lower cytotoxicity, resulting in increased SI values. Notably, Compound 11a had an SI value of 2 for activity against M. bovis BCG over THP-1 cells, while 21 and 30 had SI values of 4.

The 6,7-dihydroxy-THIQs (2a, 5a & 6a) displayed little to no inhibition of ethidium bromide efflux (Table 6). However, all 7,8-dihydroxy-THIQs (2b-5b), with the exception of 6b, displayed efflux inhibition. 5b resulted in the greatest accumulation of ethidium bromide of the THIQs tested, with an RFF of 1.42, demonstrating some potency as an efflux inhibitor, but still significantly less than the RFF of 2.09 observed for the positive control Verapamil. The activity of the 7,8-dihydroxy substituted THIQs alongside the inactivity of the 6,7-dihydroxy-subsituted analogues, suggested that either a C-8 hydroxy substituent was a requirement for efflux pump inhibition, or that a C-6 hydroxy substituent abolishes this activity. The activity of the 5-bromo-substituted THIQ 13 suggested that a bromide atom at this position may confer efflux pump inhibition, while the low activity of 36 demonstrated that the C-1 substituent is also an important factor in efflux pump inhibition. Interestingly, no correlation was observed between efflux pump activity and compound MIC.

The most notable trend from the synergism studies (Table 7) is that the 5-bromo-substituted THIQs, 15 and 36, were both able to affect a 2-fold increase in the potency of rifampicin at a concentration of 0.1 µg/ml, with FICIs of 0.5, indicating synergism. Compound 7b was also found to have the same activity alongside rifampicin. None of the compounds assayed were found to have synergistic activity with isoniazid, and only compound 5b was found to have any synergistic activity with ethambutol, causing a 2-fold increase in potency, at a concentration of 0.1 µg/ml.