4-Aminoquinolines Modulate RNA Structure and Function: Pharmacophore implications of a conformationally restricted polyamine

Abstract

RNA structure plays an important role in regulating cellular function and there is a significant emerging interest in targeting RNA for drug discovery. Here we report the identification of 4-aminoquinolines as modulators of RNA structure and function. Aminoquinolines have a broad range of pharmacological activities, but their specific mechanism of action is often not fully understood. Using electrophoretic mobility shift assays and enzymatic probing we identified 4-aminoquinolines that bind the stem-loop II motif (s2m) of SARS-CoV-2 RNA site-specifically and induce dimerization. Using fluorescence-based RNA binding and T-box riboswitch functional assays we identified that hydroxychloroquine binds the T-box riboswitch antiterminator RNA element and inhibits riboswitch function. Based on its structure and riboswitch dose-response activity we identified that the antagonist activity of hydroxychloroquine is consistent with it being a conformationally restricted analog of the polyamine spermidine. Given the known role that polyamines play in RNA function, the identification of an RNA binding ligand with the pharmacophore of a conformationally restricted polyamine has significant implications for further elucidation of RNA structure-function relationships and RNA-targeted drug discovery.

1. Introduction

RNA structure plays an important role in regulating cellular function and there is a significant emerging interest in targeting RNA for drug discovery [1–4]. However, even with the increased number of studies on RNA-small molecule complexes, there are relatively few reports of extensive, comprehensive drug discovery efforts targeting noncoding RNA (compared to reports on protein targets) and of these, most have focused on targeting small loops and bulges in RNA [4]. As part of a comprehensive RNA-targeted drug discovery project, we have been focused on investigating ligand-RNA binding interactions involving larger, more dynamic RNA structural motifs.

One RNA structural motif that we have investigated is the bacterial T-box riboswitch [5–19]. The T-box riboswitch, located in the 5′-untranslated region of mRNA, controls expression of many essential genes in Gram-positive bacteria by structurally sensing tRNA aminoacylation charging ratios to turn on transcription (or translation) [20]. The highly conserved antiterminator element in the T-box riboswitch is an ideal target for antibacterial drug discovery [3,10,21]. We have identified small molecules that bind the antiterminator RNA in a structure-specific manner and modulate T box riboswitch activity [7,10,12,17–19,22].

Another RNA structural motif that we have investigated is the stem-loop II motif RNA [23]. The stem-loop II motif (s2m) is a highly conserved RNA structural element found in the 3′-untranslated region of coronaviruses such as SARS-CoV-1, SARS-CoV-2 and other related viruses [24,25]. While the functional role remains unknown, evolutionary conservation of the s2m structural element indicates that it is likely critical [25]. In addition, disruption of the structure by interfering ligands inhibits viral replication [26], further substantiating its important role for viruses that have it.

A classic strategy for rational drug design is to make analogs of an endogenous ligand that binds the biomolecular target. Some common analog design strategies include conformational restriction, functional group isosteric replacement and scaffold extension to make additional contacts with the target biomolecule. This strategy has been effectively applied to make analogs of cognate ligands that inhibit metabolite-sensing riboswitches [27]. One common class of endogenous RNA binding ligands that has not yet been investigated for RNA targeted drug discovery are the aliphatic polyamines. Polyamines are known to serve a functional role interacting with bulged regions of RNA [28]. In addition, we [17] and others [29] identified the polyamine spermidine as a positive modulator of T-box riboswitch function and we have shown that spermidine binds the T-box riboswitch antiterminator RNA element in a structure-specific manner [18].

Here we report on 4-aminoquinolines as RNA binding ligands that contain the pharmacophore of a conformationally-restricted spermidine. We identified structure specific interactions of these compounds with the SARS-CoV-2 s2m RNA and the T-box riboswitch antiterminator RNA, in addition to inhibition of T-box riboswitch function. Given the known role that polyamines play in RNA function, the identification of an RNA binding ligand with the pharmacophore of a conformationally restricted polyamine has significant implications for further elucidation of RNA structure-function relationships and RNA-targeted drug discovery.

1. Materials and Methods:

1.1. Reagents & Computational Resources

All reagents were molecular biology grade, RNase/DNase-free or highest purity grade possible. RNase T1 was from ThermoFisher Scientific. Ligands were from Sigma-Aldrich with purity >90%. Tilorone was purchased as the HCl salt. Custom synthesized HPLC-purified RNAs (unlabeled and fluorescently labeled) were obtained fully deprotected from Horizon Discovery. RNAs were dialyzed in 5 mM MOPS, pH 7.0, 0.01 mM EDTA and renatured prior to use. Computational work was done using the Owens Cluster at the Ohio Supercomputer Center [30] and a MacPro running the Schrödinger Suite of software.

1.2. Electrophoretic mobility shift assay

Electrophoretic mobility shift assays (EMSA) were conducted using unlabeled RNA with the sequences shown in Figures 2a and 3b. EMSA experiments were run on 10% 29:1 Acrylamide:Bisacrylamide gels in 1X Tris-HEPES (34 mM Tris, 66 mM HEPES), pH 7.5, 50 mM NaCl, 0.1 mM MgCl₂ at 0, 0.1 and 1.0 mM ligand with RNA (for AM, 6 μM; for s2m, 3 μM) at 4 °C. Gels were stained with SYBER Green II (ThermoFisher).

Figure 2.

Ligand effect on s2m model RNA. a) SARS-CoV-2 stem-loop II motif (s2m) model RNA (black) with palindromic sequence indicated in bold. Nucleotides shown in grey are as designed in previous studies [23]. The secondary structure indicated is based on NMR, probing and molecular dynamics studies [35,48]. b & c) Electrophoretic mobility shift assays (EMSA) in the presence of ligands (1.0 and 0.1 mM). d) Effect of aminoquinolines on RNase T1 probing of SARS-CoV-2 s2m.

Figure 3.

T-box riboswitch mechanism and ligand binding to antiterminator model RNA. a) T-box riboswitch mechanism for transcriptional / translational regulation of gene expression. b) T-box riboswitch antiterminator model RNA, AM and AM_Δ [37] c) Ligand specificity assay for binding model RNAs. d) Electrophoretic mobility shift assay of unlabeled AM in the presence of ligand (1.0 and 0.1 mM).

1.3. Enzymatic Probing

RNase T1 probing of SARS-CoV-2 s2m was conducted as previously described on a DY-547 fluorescently labeled RNA with the sequence shown in Figure 2 [23]. Enzymatic probing conditions were 10 mM Tris, pH=7, 100 mM KCl, 0.1 mM MgCl₂, 4.0 μM DY547-labeled s2m. 0.004 U/μl RNase T1, [ligand] = 0.1, 0.5 and 1.0 mM.

1.4. Computational ligand docking

Ligand docking studies were performed using the Glide [31] module in the Schrödinger Small Molecule Drug Discovery suite. Docking studies for s2m were as described previously [23]. Docking studies for the T-box riboswitch antiterminator model RNA were done using the previously described protocol [12].

1.5. T-box riboswitch antiterminator model RNA specificity and functional assays

T-box riboswitch antiterminator specificity assays were conducted as previously described [19]. Final reaction conditions were 55 mM MOPS, pH= 6.5, 50 mM NaCl, 0.01 mM EDTA, 100 nM 5′-TAMRA-model RNA, 0.01 mM ligand.

T-box riboswitch function (transcription readthrough) assays were conducted as previously described [17]. Reactions were run in duplicate with final reaction conditions of 20 mM Tris-HCl, pH 7.4, 40 mM KCl, 4 mM MgCl₂, 1 mM DTT, 5 mM spermidine, 0.1 μM RTprb (readthrough molecular beacon probe), 10 nM glyQS DNA template (or the transcription conrol DNA template that lacks the riboswitch), 0.23 U/μl RNase inhibitor, 0.05 U/μl E. coli RNA polymerase (Holoenzyme), 0.01 mM CTP, GTP, UTP & 0.4 mM ATP, with and without tRNA (30 nM). The sequence composition of the molecular beacon probe (RTprb), glyQS DNA template and the transcription control DNA template are as described previously [17]. Data were fit to a sigmoidal dose-response using Prism (Graphpad). The data point for 10 mM hydroxychloroquine was excluded from the line fit, but the unusual trend observed at this high ligand concentration is plotted and discussed in the results.

The molecular beacon probe control assay was conducted as described previously [17] by monitoring the fluorescence of the molecular beacon in the presence of the probe’s target RNA sequence and ligand. Reactions were run in duplicate with final reaction conditions of 20 mM Tris-HCl, pH 7.4, 40 mM KCl, 4 mM MgCl₂, RTprb (100 nM), target RNA sequence (40 nM) with the range of ligand concentrations indicated in Figure 4c. The sequence of the target RNA sequence is as described previously [17].

Figure 4.

Aminoquinoline effect on T-box riboswitch function. a&b) Ligand effect on glyQS T-box riboswitch transcription readthrough with tRNA (filled square) and without tRNA (open square) and on the transcription control lacking the riboswitch (filled triangle). c) Control assay for monitoring ligand effect on fluorescence of molecular beacon probe for tilorone (triangle) and hydroxychloroquine (diamond) d) Comparison of atomic scaffold of hydroxychloroquine vs. spermidine.

3. Results and Discussion:

3.1. Aminoquinolines alter SARS-CoV-2 s2m electrophoretic mobility and confer site-specific protection from chemoenzymatic probing

We previously used computational methods to screen a compound library of FDA approved drugs to identify specific binders of the s2m RNA element in SARS-CoV-2 and characterized the computationally predicted best binders (select aminoglycosides and polymixins) using chemo-enzymatic probing methods [23]. In that same docking study, we were intrigued to notice that a few ligands were predicted to bind s2m within 5% of the best docked poses from the entire library, but had no violations to the Lipinski Rule of 5 (hydroxychloroquine, chloroquine and tilorone). Compounds that satisfy the Lipinski Rule of 5 tend to be orally bioavailable with features characteristic of pharmaceutically useful drugs (molecular weight, hydrogen bond acceptors/donors and hydrophobicity) making them good starting scaffolds for further optimization [25]. These compounds also had the physicochemical features characteristic of RNA binding ligands (heterocycles, rod-like structure and protonatable amines) [32].

We initially investigated possible RNA-ligand interactions using electrophoretic mobility shift assays (EMSA) and enzymatic probing of SARS-CoV-2 s2m model RNA (Figure 2a). Tilorone resulted in significant band smearing at 1.0 mM (characteristic of ligand-induced RNA aggregation) and at 0.1 mM slight formation of one slower migrating band (Figure 2b). In contrast, incubation with chloroquine and hydroxychloroquine both resulted in the formation of two slower migrating bands (Figure 2c). The relative migration of these bands with respect to the parent band and to one another is similar to the EMSA of the kissing dimer and the extended duplex of s2m previously observed by others [33].

To identify possible sequence-specific contacts, we used enzymatic probing in the presence of chloroquine and hydroxychloroquine (Figure 2d). Ligand-induced changes in the enzymatic cleavage profile were observed primarily at G28 with 60% reduction in band intensity at 0.5 mM ligand, but only 10% or less change in the band intensity at G22 and G24. The G28 residue is directly 3′ to the palindromic 5′GUAC3′ sequence that forms the intermolecular base pairing of the kissing dimer of s2m [33]. At 1 mM hydroxychloroquine, overall protection from RNaseT1 cleavage was observed. Formation of an extended duplex (consistent with the EMSA results) could explain this observed protection from the single-stranded nuclease.

The EMSA and enzymatic probing results indicate likely specific binding interactions between s2m and the 4-aminoquinolines. This is consistent with the recent identification of an aminopyridine which is structurally similar to the 4-aminoquinolines and binds s2m with low micromolar affinity [34]. However, since the exact function of the s2m RNA element in viruses remains unknown [25] and the full dynamic structural details are still being elucidated [35], we decided instead to further investigate 4-aminoquinolines and tilorone as possible RNA-binding ligands using the T-box riboswitch antiterminator model RNA as the target RNA.

3.2. Aminoquinolines exhibit RNA structure-dependent effects with T-box Riboswitch Antiterminator Model RNAs

We focused our ligand binding studies on the T-box riboswitch antiterminator RNA element. In this riboswitch, the charging state of the tRNA (aminoacylated vs. non-aminoacylated) is sensed by the discriminator domain which consists of the antiterminator (or antisequestrator) element and the 5′adjacent Stem III (Figure 3a) [20]. The non-aminoacylated tRNA 5′NCCA3′ acceptor end nucleotides base pair with the first four nucleotides in the antiterminator bulge forming a helix that coaxially stacks on the antiterminator helix A1 with the resulting complex further stabilized by a pseudohelix and A-minor clamp formed with Stem III nucleotides [36].

We used a fluorescence-monitored 1° screening RNA specificity assay previously developed in our lab [19,37] and identified that of the structurally related aminoquinolines tested, hydroxychloroquine had the largest effects on the antiterminator model RNA (AM) compared to the control model RNA (AM_Δ) that lacks the 7 nucleotide bulge essential for antiterminator function [37] (Figure 3c). Tilorone, however, had a similar effect on both model RNAs. These results indicate that the aminoquinolines, especially hydroxychloroquine, likely bind AM in a structure-specific manner whereas tilorone is likely nonspecific.

We also investigated the effect of ligand-induced changes in the electrophoretic mobility of AM. A slower migrating band was observed at 1 mM for all the ligands, most prominently with tilorone (Figure 3d). For the aminoquinolines, the shifted band was most intense for amodiaquine followed by hydroxychloroquine and then chloroquine. Given the subtle structural differences between the aminoquinolines, these results are consistent with a structure-specific aminoquinoline-RNA interaction.

3.3. Hydroxychloroquine inhibits T-box riboswitch function

We next investigated hydroxychloroquine and tilorone using the fluorescently monitored T-box riboswitch transcription readthrough assay [17]. This molecular beacon fluorescence-based assay monitors production of transcription products of the T-box riboswitch glyQS template (in the presence or absence of tRNA) and of a transcription control template lacking the T-box riboswitch sequence. Tilorone had an apparent non-specific effect since it inhibited both the transcription control and the T-box riboswitch readthrough (Figure 4a). Tilorone also diminished the fluorescent signal in the probe-target RNA control assay [17] (Figure 4c), likely indicating a non-specific effect on RNA structure and/or the molecular beacon fluorophore. In contrast, hydroxychloroquine inhibited both the basal readthrough (IC₅₀ = 1 mM) and tRNA-induced readthrough (IC₅₀ = 3 mM) of the T-box riboswitch without inhibiting the transcription control (Figure 4b). In addition, hydroxychloroquine did not interfere with the fluorescent signal in the probe-target RNA control assay (Figure 4c). We were unable to assay amodiaquine in a similar manner due to it causing precipitation in the dose-response reactions at the concentrations tested.

The identification of hydroxychloroquine as an antagonist of T box riboswitch function is consistent with the antiterminator model RNA binding specificity results in that hydroxychloroquine has a specific effect on the T-box riboswitch, but not general RNA transcription. In addition, since hydroxychloroquine inhibited basal level transcription readthrough (i.e., without tRNA) its mechanism of inhibition may involve the antiterminator. In the absence of tRNA, the relative stability of the antiterminator versus the terminator [10] is the key conformational switch that determines readthrough versus termination.

Interestingly, the apparent effect of hydroxychloroquine on the riboswitch changed at higher concentrations of ligand (10 mM, Figure 4b). This change in reactivity trend at higher concentrations and the observed IC₅₀ value of 1 mM are analogous to what we previously reported for the effect of spermidine on T-box riboswitch function [17]. This similarity made us reexamine the structure of hydroxychloroquine and related aminoquinolines to recognize it as a potential, conformationally restricted analog of spermidine (Figure 4d).

3.4. Hydroxychloroquine Docks in A1 Helix of T-box Antiterminator

Since the T-box riboswitch functional assay results were consistent with hydroxychloroquine inhibition involving the antiterminator, we computationally docked hydroxychloroquine to the NMR-derived structure of AM using our previously validated docking protocols [12,18,19]. The lowest energy docked structure of hydroxychloroquine (E_model = −161 Kcal/mol, Figure 5) spans the A1 helix with good shape complementarity and extensive contacts (pi-pi stacking, cation-pi stacking, hydrogen bonding and salt bridges). This is the same region where the additional contacts between the T-box riboswitch RNA and tRNA form to stabilize the antiterminator-tRNA base pairing (i.e., the A-minor latch and pseudohelix) [36]. Consequently, hydroxychloroquine’s mechanism of inhibition may involve formation/stabilization of a non-functional conformation in the antiterminator RNA (and/or its complex with tRNA).

Figure 5.

Hydroxychloroquine docking with antiterminator model RNA. a) Lowest Glide E_model pose of hydroxychloroquine bound to T-box riboswitch antiterminator model RNA, AM, b) detail of hydroxychloroquine (thick tube) molecular interactions with AM nucleotides (wireframe).

4. Conclusion

While functionalized quinolines are known to have a wide range of therapeutic activities [38,39] there are relatively few reports of quinolines binding RNA [40]. This is in surprising contrast to studies of the oxidized form of the heterocycle, quinolones, which have frequently been reported as RNA binding ligands [2,41–43]. Consequently, the antagonist activity of hydroxychloroquine on the T-box riboswitch function has significant implications for potential effects of aminoquinolines on the function of other RNA targets. Consistent with this is the observed structure-specific interaction of 4-aminoquinolines with the SARS-CoV-2 s2m. As discussed above, the RNA structure-activity profile and chemical structural similarity led us to hypothesize that hydroxychloroquine may be acting as a conformationally restricted analog of spermidine.

For the T-box riboswitch antiterminator RNA, the pharmacophore model of a conformationally restricted polyamine is intriguing. Spermidine is known to play an important role in T-box riboswitch transcription readthrough [17,29], and we previously identified that spermidine binds T-box riboswitch antiterminator model RNA [18]. In addition, a polyamine is used in the crystallization method of the tRNA-discriminator region complex, but is not present as a co-ligand in the crystal structure [36]. One possible explanation is that the polyamine assists in the folding of the functional complex between tRNA and the T-box riboswitch, but is not required for final overall stabilization of the complex (i.e., functions as a molecular chaperone). A mechanistic step that could be affected by this is the formation of the A-minor latch and pseudohelix that occurs after formation of the tRNA-antiterminator base pairing and finally commits the complex to antitermination. This mechanistic step has been proposed previously as a good modulatory target based on sequence-function variation studies [36]. Conformationally restricted analogs of spermidine (e.g., hydroxychloroquine) might compete with this favorable chaperone effect. Turning an agonist into an antagonist through conformational restriction is a very effective drug design strategy [44]. Additional studies will help determine the specific mechanism by which hydroxychloroquine inhibits the T-box riboswitch.

The implications of 4-aminoquinolines as possible conformationally restricted analogs of spermidine on the SARS-CoV-2 s2m is more challenging to discern since the exact function of the RNA is still unknown [33]. Furthermore, while in vitro studies indicated antiviral activity of hydroxychloroquine against SARS-CoV 1 and 2, clinical studies resulted in cardiotoxicity [45]. Consequently, our interest in the interaction of hydroxychloroquine with s2m is focused on the insight it provides with respect to how 4-aminoquinolines affect structured RNAs. The observed ligand-induced EMSA bands are consistent with formation of the s2m kissing dimer and extended duplex. The enzymatic probing results indicate ligand binding adjacent to the palindromic kissing dimerization site. Others have previously determined that the viral N protein acts as a molecular chaperone to form the s2m kissing dimer and its conversion to the stable extended duplex [33]. Together the results indicate that the observed additional bands in the presence of a 4-aminoquinoline may be due to a similar molecular chaperone effect.

Quinolines are considered natural privileged scaffolds in drug discovery with a broad range of pharmacological activities [46]. Based on the observed activities with two different RNAs, 4-aminquinolines are likely privileged scaffolds for binding structured RNAs. Even with the modest IC₅₀ value observed in the T-box riboswitch function (1 mM), the results indicate that 4-aminoquinolines may be useful starting scaffolds for designing more potent RNA ligands or for targeting RNAs in areas where aminoquinolines are known to accumulate (e.g., endosomes [47]). The effect on RNA conformational transitions indicates that 4-aminoquinolines may act as an RNA molecular chaperone. This identification of an RNA structural modulator with the pharmacophore of a conformationally restricted polyamine has significant implications for understanding the functional role of polyamine-RNA interactions, for RNA-targeted drug discovery and for elucidating additional mechanisms of action for quinolines.

Figure 1.

Structures of 4-aminoquinolines and tilorone