Exploring Organosilane Amines as Potent Inhibitors and Structural Probes of Influenza A Virus M2 Proton Channel

Abstract

We describe the use of organosilanes as inhibitors and structural probes of a membrane protein, the M2 proton channel from influenza A virus. Organosilane amine inhibitors were found to be generally as potent as their carbon analogs in targeting WT A/M2, and more potent against the drug resistant A/M2-V27A mutant. In addition, intermolecular NOESY spectra with dimethyl substituted organosilane amine inhibitors clearly located the drug binding site at the N-terminal lumen of the A/M2 channel close to V27.

Influenza A virus is a serious human health threat that urgently requires development of small molecule antiviral agents.¹ Drug resistance is the predominant issue in influenza pharmaceutical research, due to the rapid mutational rate and high tendencies to reassortment.² Currently, there are four small molecule drugs used for the prevention and treatment of influenza A virus infections in the U.S. Oseltamivir and zanamivir are neuraminidase inhibitors that block the release of progeny viruses from the host cells; amantadine and rimantadine are M2 channel blockers that inhibit the viruses' uncoating process by preventing the acidification of endosomally entrapped viruses. However, most currently circulating viruses are resistant to amantadine and rimantadine, and the number of viruses resistant to oseltamivir and zanamivir is on the rise.³ Thus there is clearly a need to develop novel antivirals that are able to combat drug resistant viruses.

A/M2 forms a homotetrameric proton selective channel in viral membranes and plays an essential role in mediating viral uncoating⁴ and budding.⁵ Additionally, it equilibrates the pH across the Golgi apparatus to prevent the premature conformational change of hemagglutinin.^6-8 A/M2 is more conserved than other drug targets of influenza A virus with only three predominant drug resistant mutations S31N, V27A and L26F observed in widely circulating viruses,^{9, 10} all of which are located in the transmembrane domain drug binding site.

A carbon to silicon switch is a widely explored strategy in developing and marketing organosilane pesticides.^{11, 12} There is also a continuing interest in the pharmaceutical industry to fine tune the pharmacological or pharmacokinetic properties of marketed drugs using the same strategy.^13-16 Silicon-containing compounds generally have no heavy metal associated toxicities and have similar metabolic profiles as their carbon analogs.^{13, 14} Apart from the increased size and hydrophobicity of silicon compared to the corresponding carbon counterpart, organosilanes can also be designed to mimic high-energy tetrahedral intermediates or novel scaffolds that are not accessible to carbon analogs.^{17, 18} The most common carbon to silicon switch strategies fall into one of the two classes (Scheme 1). In the first class, a quaternary carbon is replaced with a silicon to increase hydrophobicity¹⁹ (Scheme 1 a). In the second class, a carbonyl is replaced with a sterically hindered silanediol to mimic the high-energy intermediate of an amide bound hydrolysis, provides opportunities to inhibit proteins such as proteases²⁰ (Scheme 1 b).

Scheme 1. Common strategies of carbon to silicon switch in drug design.

(a) ^at-Butyl is substituted by trimethylsilane in p38 MAP kinase inhibitor; (b) silanedioles were designed to mimic the tetrahedral hydrolysis intermediate in angiotensin-converting enzyme (ACE) inhibitor.

Hydrophobicity plays a critical role in improving antiviral potency in designing A/M2 inhibitors as anti-influenza drugs.^21-23 The pore facing residues of A/M2 channel (V27, A30, S31 and G34)^{24, 25} form a hydrophobic binding pocket that favors binding of hydrophobic molecule such as adamantane or spirane amines. Our previous structure-activity relationship (SAR) studies of spirane amine compounds showed a positive correlation between hydrophobicity and antiviral potency.^{22, 23} Considering the increased size and hydrophobicity of silicon compared with its carbon analog, we therefore rationalized that replacement of the quaternary carbon in spirane amine inhibitors with silicon would increase their potency.

Another interesting aspect of organosilane compounds is that CSi bond is polarized towards carbon due to the higher electronegativity of carbon (2.50 for carbon and 1.74 for silicon). This results in an upfield chemical shift of silicon α-protons to ∼ 0 ppm, which is generally well separated from the protein background signals. In light of this unique property, we envisioned that organosilane compounds can be applied to map the A/M2 channel drug binding site using simple nuclear overhauser enhancement spectroscopy (NOESY), thereby obviating the need for more technically demanding and less sensitive half-filtered experiments.²⁶

The minimal requirements for potent inhibition of A/M2 to consist of a basic group attached to an alkyl moiety of at least 8 carbon atoms with good steric fit to the A/M2 binding cavity.²¹ To test the effectiveness of a carbon to silicon switch in A/M2 inhibitor design, we first examined a commercially available organosilane amine, amine-3-aminopropyl trimethylsilane hydrochloride 2 (Figure 1), which only contains six carbons and one silicon. Encouragingly, it showed a 75.5 ± 0.5% inhibition against the wild type (WT) A/M2 at 100 μM concentration in two electrode voltage clamp (TEVC) assay, which is comparable with the eight carbon analog 1 (86.6 ± 1.1%).²¹ This is by far the smallest molecule (less than 8 carbons) that is active against A/M2.

Figure 1.

The smallest known organic M2 inhibitor.

In this study, two silaspirane amines 5 and 6 were designed and synthesized based on the previously identified spirane amine scaffold (Scheme 2). The synthesis started from dichlorosilane, double substitution of the chlorides with vinyl magnesium bromide yielded the divinyl cyclosilane.^{27, 28} Without purification, it underwent an hydroboration with 9-borabicyclo[3.3.1]nonane (9-BBN), followed by an exchange reaction with borane-methyl sulfide complex (BMS) and subsequent methanolysis gave the silaborinane intermediate.²⁹ Next, following Brown's dichloromethyl methyl ether (DCME) process and oxidation with 30% H₂O₂ furnished the spirosilicon ketone intermediate, 3 and 4 in a one pot process with ∼ 40-50% overall yield.^{27, 28} Finally, the ketone was converted to amine 5 and 6 using the two step hydroxylamine condensation/reduction with ∼ 75% yield.²³

Scheme 2. Synthesis of silaspiran amines.

The synthesized inhibitors were tested in a two-electrode voltage clamp assay using Xenopus laevis frog oocytes microinjectected with RNA expressing either the WT A/M2 or A/M2-V27A mutant protein.³⁰ The potency of the inhibitors was expressed as the percentage inhibition of A/M2 current observed after 2 min of incubation with 100 μM compounds, and IC₅₀ values were determined for selected potent compounds. As discussed previously, the potency in this assay reflects primarily the kinetics of binding rather than true equilibrium due to the difficulty of maintaining the oocytes for extended periods at low pH.²¹ Thus, the IC₅₀ values reflect upper limits of the true dissociation constant.

As expected, both silaspirane amines 6 and 5 showed similar potencies as their carbon analogs 7 and 8 in inhibiting WT A/M2 channel activity. All were more active than amantadine. Noteworthy was an increase in antiviral potency of silaspirane amine inhibitors against A/M2-V27A compared to their carbon analogs. The IC₅₀ of 6 against A/M2-V27A was 31.1 μM, which is more than 2.7 fold more potent than the previously identified weak A/M2-V27A inhibitor 7. Similarly, a 3.3 fold potency increase against the V27A mutant was seen when the quaternary carbon in 8 was switched to silicon to give 5. The dramatic antiviral potency increase against V27A by switching to silicon might be due to the larger size and higher lipophilicity of silicon compared with carbon, thus providing better hydrophobic contact between the drug and the channel.^{31, 32}

Membrane proteins are characterized by high content of aliphatic residues (Ala, Val, Leu, Ile),³³ and this results in crowded signal overlap at 0.5-1 ppm in the proton dimension of their NMR spectra, also their large size and rapid relaxation render traditional half-filtered experiments difficult. To map the drug binding sites in membrane proteins, it is desired to have a small molecule inhibitor which shows characteristic signals beyond the normal range of protein signals. To achieve this goal, two 4,4-disubstituted silacyclohexane amine derivatives 10 and 14 and one 4,4-dimethyl-1,4-azasilepane 13 were designed and synthesized (Table 2 and Supporting Scheme S1).

Table 2. Antiviral activities of organosilane structural probes against the WT A/M2 and A/M2-V27A mutant.

% WT A/M2 inhibition at 100 μM inhibitor conc	95.3 ± 1.1%	95.4 ± 1.4%	95.4 ± 0.4%	75.4 ± 1.8%	78.5 ± 1.2%
IC50 (μM)	2.4 ± 0.2	2.6 ± 0.2	5.7 ± 0.3	26.1 ± 2.2	19.4 ± 1.6
% A/M2-V27A inhibition at 100 μM inhibitor conc	0	7.9 ± 1.1%	19.1 ± 1.0%	2.5 ± 0.4%	0
IC50 (μM)	N.Aa	N.A	N.A	N.A	N.A

^aN.A = not available.

All compounds were also tested against S31N mutant, and found to have less than 20% inhibition at 100 μM.

4,4-Dimethyl silacyclohexane amines 10 and 14 showed a high potency against the WT A/M2 and only minimal inhibition against A/M2-V27A mutant (Table 2), which was expected, as A/M2-V27A prefers binding molecules with extended conformations.^{23, 32} Compound 13 was found to be less active than 10 and 14, but slightly more active than its carbon analog 16. The chemical shifts of the methyl protons in all three organosilane amine compounds were close to 0 ppm, which is distinguished from the protein signals and is ideal for serving as structural probes for intermolecular NOESY experiments.

With the structural probes (10, 14 and 13) in hand, we next pursued ¹³C-edited NOESY spectra with WT M2TM (22-46) peptide reconstituted in DPC micelles. To facilitate assignment, residues at positions V27, A30 and G34 were selectively uniformly ¹⁵N and ¹³C labeled. When two equivalents of compound 10 (1 mM) was added to M2TM tetramer (0.5 mM), two strong NOE cross-peaks were observed between one of the two methyl groups from 10 and both of the γCH₃ of V27 (Figure 2b); weaker NOEs were observed for the other methyl group of 10 and both of the γCH₃ groups of V27. Additionally, this methyl group also shows an NOE to the βCH₃ of Ala30 (Figure 2b). The nonequivalent crosspeaks between the two methyl groups of 10 reflects the nonsymmetrical binding geometry of the drug inside the channel. The substituent with stronger NOE signals to V27 is assigned to the axial from 10, which in models is closer to V27 than A30;^{34, 35} the methyl group that showed crosspeaks to Ala30 was assigned to the equatorial methyl from 10. No NOEs were found between the methyl groups of 10 and Gly34, which lies lower in the structure of the channel as viewed in Fig. 2d. This binding model is consistent with our earlier studies showing drug binds inside the channel with amine pointing down towards H37 (Figure 2d). ³⁶ The same NOE cross peak patterns were observed when the experiment was repeated with only one equivalent of 10 (0.5 mM) (Supporting Figure S1). In comparison with 10, the two methyl groups in 13 are equivalent, showing one single peak at 0.07 ppm (indicative of a lower barrier for interconversion of the 7-membered ring). As a weaker inhibitor against WT A/M2 compared with 10, when two equivalents (1 mM) was added to M2TM tetramer (0.5 mM), two strong NOE cross-peaks were observed and assigned to the γCH₃ of V27 and the methyl protons of 13. In addition, an unambiguous NOE was assigned to the βCH₃ of A30 and methyl protons of 13 (Figure 2f).

Figure 2.

¹³C-edited ¹H NOESY spectrums of M2TM(22-46)-V₂₇A₃₀G₃₄ in DPC micelles in the presence of either 10 (spectra a, b and c) or 13(spectra e, f and g). 10 and 13 are bound in the pore, as assessed from intermolecular NOEs circled in red lines. (a and e) Upfield region of 1D ¹H NMR. (b and f) Upfield region of 2D ¹³C-edited ¹H NOESY spectra with 100 ms mixing time. (c and g) Upfield region of 2D ¹³C-HSQC, complete assignment was reported earlier.³⁶ (d and h) Docking models of compound 10 and 13 in the A/M2 channel. The poses of the drugs reflect the relative NOEs signal intensities. Sample conditions: 2 mM A/M2(22-46) with residues V₂₇A₃₀G₃₄ double ¹⁵N, ¹³C labeled, 100 mM DPC, 1mM 10 or 13, and 50 mM pH 7.5 phosphate buffer in 10% D₂O, 90% H₂O. Spectra were recorded at 313K on a Varian Inova 600 MHz (for 13) and a Bruker Avance 500 MHz spectrometer (for 10).

Both 16 and its silicon analog 13 were crystallized by solvent evaporation in CH₂Cl₂/CH₃OH (3:1 v/v) and their structures determined by X-ray crystallography. The Si-C bond length (1.87 Å) was found on average 22% longer than the corresponding C-C bond length (1.54 Å), which results in nearly 50% volume increase at the quaternary center (Supporting Table S1). The size expansion effect of carbon to silicon switch might contribute to the antiviral potency increase of organosilane against A/M2-V27A, as the expanded organosilane amines is able to fill in the extra space created by the bulky valine 27 to smaller alanine mutation (Table 1). Moreover, silanes are more lipophilic than the corresponding alkanes; the ClogPs for 16 and 13 are 2.99 and 4.15 respectively. Thus, the improved antiviral potency of organosilanes compared with their carbon analogs might be due to the synergetic effects of size expansion and increased lipophilicity.

Table 1. Antiviral activities of silaspirane amines against WT A/M2 and A/M2-V27A mutant.

% WT A/M2 inhibition at 100 μM inhibitor conc	90.8 ± 2.5%	89.0 ± 1.5%	94.5 ± 0.6%	95.9 ± 0.9%	93.9 ± 1.8%
IC50 (μM)	16 ± 1.2	12.6 ± 1.1	13.7 ± 1.7	3.3 ± 0.2	7.8 ± 0.6
% A/M2-V27A inhibition at 100 μM inhibitor conc	0	53.2 ± 2.3%	67.4 ± 1.1%	25.2 ± 0.9%	47.8 ± 0.5%
IC50 (μM)	N.Aa	84.9 ± 13.6	31.3 ± 2.3	318.6 ± 57.3	96.3 ± 13.4

^aN.A = not available.

All compounds were also tested against S31N mutant, and found to have less than 20% inhibition at 100 μM.

Biologically active organosilanes, discovered through either rational design or high-throughput screening, are attractive analogs of their carbon counterparts due to their unique properties. Carbon to silicon switch retains the overall 3D conformation of the inhibitor, thus providing a minimal perturbation of the binding between the ligand and the receptor. Additionally, the larger covalent radius of silicon (1.17Å) compared to that of carbon (0.77Å) (Supporting Table S1), and the higher hydrophobicity renders the organosilane based inhibitors with improved properties. Organosilanes are also often easier to synthesize than their carbon analogs and in certain cases, allow access to novel scaffolds not accessible by standard carbon chemistry.¹⁷ In this report, we explored a carbon to silicon switch in A/M2 inhibitors design. The silaspirane amines were as potent as their carbon counterparts against WT, and were more potent in targeting drug resistant V27A, which highlights their promise for further optimization. Moreover, this replacement shows promise for NMR spectroscopy; three organosilane structural probes were designed to map the A/M2 drug binding site. Previously, a debate concerning the location of the pharmacologically relevant A/M2 drug binding site(s) was settled in favor of the pore-binding model using the intermolecular deuterium-¹³C dipolar dephasing experiments in solid state NMR³⁷ as well as the solution NOESY difference experiments.³⁶ The organosilane structural probes designed in this study provide a direct measurement of NOESY cross peaks between the organosilane drug and the A/M2 protein. The ¹³C-edited ¹H NOESY clearly showed the drug binds inside the A/M2 channel close to the N-terminal lumen with its positively charged ammonium pointing down towards His37. In conclusion, our studies not only provide a strategy for improving the potency of A/M2 inhibitors, but also demonstrate the utility of organosilanes as structural probes for drug binding site mapping, which can be similarly applied to other proteins. This work shows the potential of using organosilane analogous for direct detection of drug-protein interactions, using the highly sensitive NOESY experiment, thereby alleviating frequently encountered problems associated with half-filtered experiments.