Probing the B- & C-rings of the antimalarial tetrahydro-β-carboline MMV008138 for steric and conformational constraints

Abstract

The antimalarial candidate MMV008138 (1a) is of particular interest because its target enzyme (IspD) is absent in human. To achieve higher potency, and to probe for steric demand, a series of analogs of 1a were prepared that featured methyl-substitution of the B- and C-rings, as well as ring-chain transformations. X-ray crystallography, NMR spectroscopy and calculation were used to study the effects of these modifications on the conformation of the C-ring and orientation of the D-ring. Unfortunately, all the B- and C-ring analogs explored lost in vitro antimalarial activity. The possible role of steric effects and conformational changes on target engagement are discussed.

Graphical Abstract

Malaria was estimated to be responsible for 405,000 deaths worldwide in 2018.¹ Many prevention methods and drug treatment protocols are available, but emerging resistance to artemisinin and its partner drugs is of great concern. Thus there is a pressing need to develop antimalarials that possess new mechanisms of action.² Malaria is caused by Plasmodium parasites, of which P. falciparum is the most prevalent, accounting for more than 96% of the malaria cases worldwide.¹ Plasmodium sp. contain a relict organelle termed the apicoplast, which is responsible for the biosynthesis of the critical isoprenoid precursors isopentenyl diphosphate (IPP) and dimethylallyl diphosphate (DMAPP).³ Whereas Plasmodium sp. synthesize these compounds via the methylerythritol phosphate (MEP) pathway, the mevalonate pathway is used to synthesize them in humans.³ This biochemical divergence commends the MEP pathway as a target for antimalarial drug development,⁴ since inhibitors of MEP target enzymes would not adversely affect IPP biosynthesis in humans.

Our initial work in this area⁵ identified MMV008138 as a MEP pathway inhibitor, by performing a phenotypic screen of the 400-compound Malaria Box⁶ with the IPP rescue protocol.³ MMV008138 is a tetrahydro-β-carboline, and differentially-functionalized examples of this scaffold are found in a number of other antimalarials,⁷ and compounds directed towards other indications.⁸ Subsequent resistance selection studies by Wu et al. demonstrated that 2-C-methyl-D-erythritol-4-phosphate cytidylyltransferase (IspD, E.C.2.7.7.60), the third enzyme in the MEP pathway, is the target of MMV008138.⁹ Initially, the absolute configuration of MMV008138 was unknown, since it was not disclosed in the Malaria Box; subsequently three independent investigations^9–11 demonstrated that the active stereoisomer is (1R, 3S)-configured, as depicted in 1a in Figure 1. Kinetic studies established that 1a competes with cytidine triphosphate (CTP) in its IspD-catalyzed reaction with 2-C-methyl-D-erythritol-4-phosphate.¹¹

Figure 1.

Lead compound 1a and tight D-ring SAR. Note that the X= 2ʹ-Br, 4ʹ-Cl and X= 2ʹ-Cl, 4ʹ-F analogs are also potent in both assays.¹⁰

A collection of D-ring analogs of 1a was prepared by the Pictet-Spengler (PS) reaction of l-Trp-OMe·HCl with various benzaldehydes, separation of diastereomers, and hydrolysis.¹⁰ Examination of these analogs in both in vitro growth inhibition (SYBR Green) and P. falciparum IspD (PfIspD) inhibition assays show a very close correlation between growth inhibition (EC₅₀) and target engagement (IC₅₀).^10b These data also demonstrate a very tight SAR on the D-ring. At least one halogen is required on the 2ʹ- or 4ʹ- position to retain potency in both assays, as shown in Figure 1; substitution at other D-ring positions is not tolerated. It thus appears that the D-ring of 1a binds within a snug, well-defined pocket of PfIspD. In the absence of an X-ray structure for this species of IspD,¹² we have speculated^10b that halogen-bonding¹³ contributes to the affinity of 1a for its target.

Since the in vitro potency of 1a was not improved by modulation of the D-ring, we sought to probe the steric requirements for binding around the B- and C-rings, in the hope of identifying more potent analogs. Previously we disclosed analogs 6b and 6d (Scheme 1),^10b which feature methyl substitution at C1, but with non-optimal substitution of the D-ring (X = H (6b), X = 4ʹ-Cl (6d)). We attributed the low growth inhibition potency of these compounds to the absence of 2ʹ,4ʹ-dichloro substitution. Synthesis of such C1-Me analogs of 1a requires PS reaction of l-Trp-OMe·HCl 2 with acetophenones 3, which are significantly less electrophilic than benzaldehydes. Thus, the ester precursors to 6b and 6d were prepared by PS reaction with acetophenones 3b and 3d, according to Horiguchi’s protocol:¹⁴ ketimine formation in neat Ti(OⁱPr)₄, followed by treatment with TFA and TFAA, all at 70 °C. As we noted in our earlier publication,^10b application of the Horiguchi protocol to ortho-substituted acetophenones 3a and 3c did not give the expected products. However, we subsequently found that if ketimine formation was followed by treatment with TFA/TFAA at 0 °C to room temperature, the desired trans-esters 5a and 5c could be isolated in 20% and 17% yield respectively (isopropyl esters result from Ti(OⁱPr)₄-mediated transesterification). As detailed by Horiguchi,¹⁴ the trans-relative configuration of 5a and 5c was established by the absence of an NOE correlation between H3 and the C1-methyl; this correlation is visible in the corresponding cis-isomers 4a and 4c (Supplementary Material, Figures S2-S3). Hydrolysis of 5a and 5c afforded the desired amino acids 6a and 6c.

Scheme 1.

Synthesis of C1-Me analogs of 1a.

Unfortunately, the presence of a 2ʹ-Cl substituent in the D-ring of these C1-methyl analogs did not restore antimalarial activity (Table 1, entries 6, 8). Compared to our lead 1a (EC₅₀ = 250 ± 70 nM), C1-methyl analog 6a shows no growth inhibition at 10,000 nM. Similarly, the weakly potent 2ʹ-Cl substituted 1c (EC₅₀ = 3,280 ± 990 nM) loses all growth inhibition potency upon C1-methylation (6c, no growth inhibition at 10,000 nM). Since the mere addition of a methyl group at C1 should not drastically affect permeability or transport of these compounds, we conclude that the loss of growth inhibition potency is due to reduced affinity for PfIspD, the target of 1a (and its potent analogs, cf. Figure 1). In particular, it appears that either there is no room in the PfIspD binding pocket for a methyl group at C1, or that the C1-methyl in 6a induces a conformational change that disfavors binding. We were fortunate to obtain a crystal of 7a (Figure 2), the methyl amide derivative of 6a, and to compare it to 8a, the methyl amide analog of 1a, which we previously crystallized.^10a Since 8a is equipotent (Dd2 strain growth EC₅₀ = 190 ± 30 nM) with 1a, comparison of the conformations of 7a and 8a could be informative.

Table 1.

P. falciparum growth inhibition by 1a-d, f, and indicated B- and C-ring analogs.

Entry	Compound	Dd2 strain P. falciparum Growth EC50 (nM)
1	1a	250 ± 70a, c
2	1b	>10,000a
3	1c	3,280 ± 990a, d
4	1d	1,170 ± 60a, e
5	1f	700 ± 90a, c
6	6a	>10,000
7	6b	>10,000b
8	6c	>10,000
9	6d	>10,000b
10	8a	190 ± 30a
11	12f	>10,000
12	12i	>10,000
13	16a	65% inhibition at 10 μM
14	(±)-20a	>10,000
15	24a	~8,000
16	25a	>10,000
17	26a	>10,000

^aReported previously.^10a

^bReported previously.^10b

^c100% rescued by 200 μM IPP @ 10 μM.

^d60% rescued by 200 μM IPP @ 10 μM.

^e50% rescued by 200 μM IPP @ 10 μM.

^f100% rescued by 200 μM IPP @ 2.5 μM.

Figure 2.

A. Comparison of τ (R-C1-C1ʹ-C2ʹ) dihedral angles in 7a and 8a. B) PyMOL¹⁵ overlay of X-ray structures of 7a (cyan: carbon; blue: nitrogen; green: chlorine) and 8a (red). A thermal ellipsoid depiction of 7a is provided in the Supporting Information, Figure S1).

As can be seen in Figure 2, the tetrahydropyridine C-rings of 7a and 8a adopt very similar conformations, featuring a pseudoequatorial C(O)NHMe group and an apparent electrostatic interaction between the amide NH and the tetrahydropyridine nitrogen N2 (Figure 2). There are 4 molecules of 7a in the unit cell, and the average RMSD of the 6 C-ring atoms of them from 8a is 0.041 Å (individual values 0.030, 0.033, 0.046, 0.057 respectively). However, steric strain between the C1-Me and the C2ʹ-Cl in 7a causes the D-ring to adopt a different orientation. We define τ as the dihedral angle between the C1 substituent (CH₃ for 7a, H for 8a), C1, C1ʹ, and C2ʹ. For 7a (C1-Me), the average τ value is −63.6° (individual values −59.15°, −62.29°, −64.00°, and −68.91°, respectively), whereas the τ value in 8a (C1-H) is nearly 30° smaller, at −36.5°.

Given the extreme sensitivity of growth and PfIspD inhibition to substitution of the D-ring (see Table 1), it is possible that this dihedral angle change alone, apart from the added steric bulk at C1, could be deleterious to potency. To enforce a smaller τ dihedral angle, we thus proposed to connect C1 and C2ʹ with an ethylene bridge as shown in 12f and 12i (Scheme 2), imparting a cipargamin^7b-like spiro-fusion. Compound 12f would thus be a conformationally-constrained mimic of 2ʹ-Me,4ʹ-Cl-substituted 1f, which has significant potency (EC₅₀ = 700 ± 90 nM). The ketone PS reaction between l-Trp-OMe·HCl 2 and indanones 9f/9i was performed according to the original Horiguchi protocol;¹⁴ diastereomer separation gave esters 11f and 11i, which mimic the trans-orientation of 5a-d. The trans-configuration of 11f and cis-configuration of 10f was confirmed via 1D NOE experiments (Figure 3).

Scheme 2.

Synthesis of spiro-fused analogs 12f, 12i.

Figure 3.

1D NOE observed in 10f and 11f.

Irradiating C3-H revealed NOE to C7ʹ-H for the trans-isomer 11f, but not for the cis-isomer 10f. When one of the diastereotopic C2ʹ-H is irradiated in 11f, NOE to the N9-H is seen. In contrast, irradiation of one of the diastereotopic C2ʹ-H in the cis-isomer 10f transfers NOE to C3-H (Supplementary Material, Figures S4-S7). Interestingly, ¹H NMR analysis of 11f, 11i and their cis-diastereomers (10f, 10i) demonstrated one large and one small coupling constant of H3 to the H4 protons, indicating a near antiperiplanar relationship of H3 and H4β. Thus, regardless of cis- or trans-orientation, the 3-CO₂ⁱPr group is pseudoequatorial, and H3 is pseudoaxial. The trans-esters 11f and 11i were then hydrolyzed to give the desired spiro-fused analogs 12f and 12i. Unfortunately, neither compound was potent for growth inhibition (Table 1, entries 11–12). Compound 12f (EC₅₀ > 10,000 nM) differs from 1f (EC₅₀ = 700 ± 90 nM) by substitution of the C2ʹ-CH₃ group with an ethylene bridge to C1. Molecular mechanics-based conformational analysis^{16, 17} of the methyl amide of 11f (11f NHMe), demonstrates that the spiro-ring fusion generates two conformer ensembles defined by narrow ranges in the τ dihedral angle (here C2ʹ-C1-C7aʹ-C3aʹ, Figure 3). One ensemble (13 conformers) features τ = −19 ± 1°, more similar to that of 8a (τ = −36.5 °) than of 7a (τ = −63.8°). The other ensemble (16 conformers) features τ = +17 ± 2°. Since the lowest energy conformers in each ensemble are similar in energy (Supplementary Material, Table S1), both conformers are expected to be populated, and be available to bind PfIspD. Thus, unless the τ dihedral angle exhibited by 8a (and presumably 1a) precisely matches the steric requirement of PfIspD, it seems most likely that the low potency of spiro-fused analog 12f, is due to steric bulk at the C1 position, as we concluded for C1-Me analogs 6a and 6c.

To probe the effect of substitution at other positions, we prepared analogs featuring methylation at N2, C3, and N9 (Scheme 3). Since we have shown that antimalarial potency of 1a and analogs requires trans-(1R,3S)-configuration,¹⁰ we took special pains to ensure that we had isolated the correct stereoisomer in each case. The N2-Me analog 16a was prepared from PS reaction of N_α-methyl-l-tryptophan methyl ester 13 and 2,4-dichlorobenzaldehyde 3a. The PS adduct 14a was isolated as an inseparable mixture of diastereomeric methyl esters, hydrolyzed, and separated by reverse phase prep-HPLC into the cis-isomer 15a and the trans-isomer 16a. The relative configuration of these isomers was assigned on the basis of the H3-H4α and H3-H4β coupling constants (³J_34α and ³J_34β). As Van Linn et. al have shown for 1,2,3-trisubstituted PS adducts, 1,3-trans-configured compounds have ³J_34α ≈ ³J_34β = 4.4–4.9 Hz.¹⁸ In contrast, 1,3-cis-configured-1,2,3-trisubstituted PS adducts have differentiated values for these coupling constants (³J_34α =3.8–5.2 Hz and ³J_34β = 6.5–8.0 Hz).

Scheme 3.

Synthesis of 16a, (±)-20a, and 24a.

The racemic C3-Me analog (±)-20a was prepared from α-methyl-DL-tryptophan (±)-17. Esterification with SOCl₂/MeOH followed by PS reaction with 2,4-dichlorobenzaldehyde 3a gave cis-ester (±)-18a and trans-ester (±)-19a; the relative configuration of these compounds was established by 1D NOE experiments (Figure 4). Irradiating the C3-Me in (±)-18a shows NOE to H1, but no such NOE is seen on irradiation of the C3-Me in (±)-19a. In addition, irradiation of the C3-Me (±)-18a shows NOE signal to H4α, but not H4β. In contrast, irradiation of the C3- Me in (±)-19a shows NOE to both H4α and H4β (Supplementary Material, Figures S8-S9). Hydrolysis of the trans-ester (±)-19a was again accomplished with the catch and release protocol,¹⁹ but required elevated temperature, possibly due to steric hindrance caused by the C3-Me.

Figure 4.

Assignment of relative configuration in (±)-18a/(±)-19a and 22a/23a by 1D NOE.

The N9-Me analog 24a was prepared by esterification of 1-Me-l-tryptophan 28, PS reaction with 2,4-dichlorobenzaldehyde 3a, separation of diastereomers 22a and 23a, and hydrolysis of 23a. (Scheme 3). The relative configuration of the diastereomeric esters 22a and 23a was again assigned by 1D NOE spectroscopy. Irradiating H3 of the cis-isomer 22a shows an NOE signal to H1, but this correlation is not seen in trans-isomer 23a (Figure 4). In addition, irradiation of H3 in 23a shows NOE signals to both H4α and H4β, indicating the 3-CO₂Me group is pseudoaxial orientation. For 22a a strong correlation of H3 to H4α is observed, as shown in Figure 4 (Supplementary Material, Figures S11-S12).

As can be seen in Table 1, methyl substitution at N2, C3 and N9 (compounds 16a, (±)-20a and 24a, respectively) unfortunately abrogates P. falciparum growth inhibition (EC₅₀ ≥ 8,000 nM, entries 13–15). As we concluded for methylation at C1 (e.g. 6a, entry 6), we consider it unlikely that addition of a methyl group would significantly impact permeability or transport, and we conclude that these modifications reduce affinity for PfIspD. Again, steric hindrance to binding caused by methylation at N2, C3 and N9 may be responsible. However, since studies of the stereoisomers^9–10,11 and C3-variants¹⁰ of 1a indicate that interaction of the 3-CO₂⁻ group with PfIspD is important for affinity, we thought it is important to rule out less direct explanations for the low potency of these compounds. In particular, we sought to determine whether substitution at C1, N2, C3, and N9 could strongly bias a pseudoequatorial- (ψ_eq-) or pseudoaxial- (ψ_ax-) orientation of the 3-CO₂⁻ group in the tetrahydropyridine ring.

Thus, for 1a, 6a, 12f, 16a and 24a, we examined ¹H-¹H coupling constants between H3 and H4α, and between H3 and H4β (Table 2). Since (±)-20a lacks a proton at C3, we used NOE to deduce the preferred orientation of the 3-CO₂⁻ group, as was done for the methyl ester precursor (±)-19a (Figure 4). Due to reasons of solubility, these experiments were carried out in CD₃OD; we do recognize that the conformational thermodynamics could vary somewhat in water. However as described in the supporting information (Supplementary Material, Section F), in large part the conformer preferences seen comport with the principles of maximizing ψ_eq-substitution, and relief of torsional strain. These effects are summarized in Figure 5.

Table 2.

Predominant orientation of the 3-CO₂⁻ group in 1a and analogs, as judged by ¹H NMR spectroscopic coupling constants ³J₃₄ and NOE.

Entry	Cpd	R1	R2	R3	R9	3J34a (Hz)	3-CO2−
1	1a	H	H	H	H	8.5, 5.5	ψeq & ψax
2	6a	CH3	H	H	H	12.0, 5.1	ψeq
3	12f	na	na	na	na	11.7, 5.0	ψeq
4	24a	H	H	H	CH3	10.3, 4.9	ψeq
5	16a	H	CH3	H	H	5.0, 5.0	ψax
6	(±)-20a	H	H	CH3	H	NOEb	ψax

^a1H NMR spectra measured in CD₃OD.

^bPredominant conformation determined by ¹H-¹H NOE experiments, since R³ = CH₃, see text.

Figure 5.

Effects of C1-, N2, C3- and N9 substitution on the preferred conformation of the tetrahydropyridine C-ring of 1a.

Thus, as anticipated, substitution at C1, N2, C3, and N9 does affect the preferred conformations of these analogs of 1a. However, since neither enforced ψ_eq-CO₂⁻ -orientation (e.g. 6a, 12f, 24a) nor enforced ψ_ax-3-CO₂⁻-orientation (e.g. 16a, (±)-20a) conferred potency, it appears that the low potency of these compounds is indeed steric in origin.

Finally, in addition to installing substituents on the B- and C-rings, we also investigated the shifted C-ring analog 25a and the open C-ring analog 26a (Scheme 4). The reductive amination of l-Trp-OMe·HCl 2 with 3a and sodium cyanoborohydride gives intermediate 27a. Subsequent treatment of 27a with dimethoxymethane and TFA followed by hydrolysis affords the shifted C-ring analog 25a; hydrolysis of 27a gives open C-ring analog 26a. Unfortunately, neither 25a nor 26a was potent for growth inhibition of P. falciparum (Table 1).

Scheme 4.

Synthesis of shifted and open C-ring analogs of 1a.

To conclude, placement of a methyl group on the B-ring (N9, e.g. 24a), C-ring (C1: 6a; N2: 16a; C3: (±)-20a) and installation of a spiro-fusion between the C- and D-rings (e.g. 12f) all drastically reduced in vitro antimalarial potency. In addition to the obvious consequence of increased steric bulk at these positions, these modifications also affected the preferred ψ_eq- or ψ_ax- orientation of the C3-CO₂⁻ group in the tetrahydropyridine C-ring (Table 2), and the orientation of the D-ring (Figure 2 and text). Lastly, shifted and open C-ring analogs of 1a (25a and 26a, respectively) were found to lack in vitro activity. These structural modifications of 1a and 1f are not expected to significantly impact permeation or transport, since they comprise formal addition of a CH₂ unit (e.g. 6d, 12f, 12i, 16a, 24a), isomerization (25a), or addition of H₂ (26a). We conclude therefore that the loss of P. falciparum growth inhibition potency is due to lack of target engagement. In particular, we propose that the binding pocket of PfIspD features very close contact with the B- and C-rings of 1a, as we had proposed earlier for the D-ring.^10b Whether this close contact also extends to all positions of the A-ring of 1a, and whether A-ring modifications could improve potency, work is in progress, and will be reported in due course.

Abbreviations

PS

Pictet-Spengler

MEP

methylerythritol phosphate

IPP

isopentenyl diphosphate

DMAPP

dimethylallyl diphosphate

Pf

Plasmodium falciparum

RMSD

root-mean-square deviation