Enhanced In Vitro Antiviral Activity of Hydroxychloroquine Ionic Liquids against SARS-CoV-2

The development of effective antiviral drugs against SARS-CoV-2 is urgently needed and a global health priority. In light of the initial data regarding the repurposing of hydroxychloroquine (HCQ) to tackle this coronavirus, herein we present a quantitative synthesis and spectroscopic and thermal characterization of seven HCQ room temperature ionic liquids (HCQ-ILs) obtained by direct protonation of the base with two equivalents of organic sulfonic, sulfuric and carboxylic acids of different polarities. Two non-toxic and hydrophilic HCQ-ILs, in particular, [HCQH2][C1SO3]2 and [HCQH2][GlcCOO]2, decreased the virus-induced cytopathic effect by two-fold in comparison with the original drug, [HCQH2][SO4]. Despite there being no significant differences in viral RNA production between the three compounds, progeny virus production was significantly affected (p < 0.05) by [HCQH2][GlcCOO]2. Overall, the data suggest that the in vitro antiviral activities of the HCQ-ILs are most likely the result of specific intra- and intermolecular interactions and not so much related with their hydrophilic or lipophilic character. This work paves the way for the development of future novel ionic formulations of hydroxychloroquine with enhanced physicochemical properties.


Introduction
The ongoing coronavirus disease 2019 (COVID-19) pandemic caused by the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has severely affected people's lives and the well-being of all societies around the world, posing unprecedented challenges to global public health. By mid-April 2022, there were almost 500 million people infected worldwide, of whom more than 6.1 million have died [1]. In addition, the pandemic has triggered a series of economic and social disruptions that have led towards an increase in extreme poverty and inequalities at a global scale, seriously jeopardizing people's livelihoods for years to come [2].
Despite the approval of limited drug treatments and the emergency use authorization of other drug molecules and monoclonal antibody preparations, antiviral therapy has had still little impact on COVID-19 clinical outcomes for most patients globally [3]. Remdesivir (Veklury ® ), the only antiviral drug approved by both the United States Food and Drug Administration (FDA) and the European Medicines Agency (EMA) for COVID-19 treatment [4,5], has been shown to shorten recovery times in hospitalized patients [6] but to have no effect on either the requirement for ventilation or patient survival [7]. Its intravenous  13 13   (0.894 mmol) of the corresponding organic acids (commercially available or prepared by reaction of the corresponding sodium salts with 1.3 mL of Amberlite 15 H + ) dissolved in distilled water or methanol were added dropwise under stirring at room temperature. After reacting for one hour, the solvent was removed in a rotary evaporator and the pure product was isolated in quantitative yield as a pale-yellow gel after drying under high vacuum for 24 h.  13 13 13 13 13

Water Solubility Studies
The

Octanol-Water Partition Coefficient Studies
The HCQ-ILs' octanol-water partition coefficients (Kow) corresponded to the distributions of the synthesized compounds between an aqueous phase and an n-octanol organic phase.
A small amount of each HCQ-IL was dissolved in previously prepared octanol-saturated water so as to produce a concentration of 1 mg/mL, and a sample was collected from this solution. To the remaining volume, equal parts of water-saturated octanol were added, followed by 2 h of vigorous stirring. After centrifugation at 5000 rpm for 10 min, a second sample was obtained from the aqueous phase. The two samples were analyzed via UV-Vis spectrophotometry to obtain the initial absorbance (Ai) and the final absorbance (Af), respectively, in the formula below [67]. Each experiment was performed in triplicate.
where Kow = octanol/water partition coefficient, Ai = initial absorbance, Af = final absorbance, df = dilution factor, Vwater = volume of water, Voctanol = volume of octanol. For lipophilic HCQ-ILs that failed to dissolve in octanol-saturated water, an analogous procedure was employed using water-saturated octanol to dissolve the compounds. In this context, the Kow formula was inverted.  13

Water Solubility Studies
The solubility of the hydrophilic HCQ-ILs (with the anions [C 1 SO 3

], [C 6 SO 3 ], [CampSO 3 ], [p-TolSO 3 ] and [GlcCOO]
) in water was determined by consecutively adding 5 to 10 µL of freshly double-distilled water to an Eppendorf tube containing ca. 30 mg of sample, precisely weighed, until a homogeneous solution was visually observed upon vortex mixing. For the lipophilic HCQ-ILs (with [DocSO 3 ] and [C 12 SO 4 ] anions), 1 mL volumes of water were added to a falcon tube containing ca. 1 mg of sample, until a maximum of 10 mL was reached.

Octanol-Water Partition Coefficient Studies
The HCQ-ILs' octanol-water partition coefficients (K ow ) corresponded to the distributions of the synthesized compounds between an aqueous phase and an n-octanol organic phase.
A small amount of each HCQ-IL was dissolved in previously prepared octanolsaturated water so as to produce a concentration of 1 mg/mL, and a sample was collected from this solution. To the remaining volume, equal parts of water-saturated octanol were added, followed by 2 h of vigorous stirring. After centrifugation at 5000 rpm for 10 min, a second sample was obtained from the aqueous phase. The two samples were analyzed via UV-Vis spectrophotometry to obtain the initial absorbance (A i ) and the final absorbance (A f ), respectively, in the formula below [67]. Each experiment was performed in triplicate.
where K ow = octanol/water partition coefficient, A i = initial absorbance, A f = final absorbance, df = dilution factor, V water = volume of water, V octanol = volume of octanol. For lipophilic HCQ-ILs that failed to dissolve in octanol-saturated water, an analogous procedure was employed using water-saturated octanol to dissolve the compounds. In this context, the K ow formula was inverted.

Critical Micelle Concentration
The critical micelle concentrations of the HCQ-ILs with surfactant-like anions were calculated using ionic conductivity measurements. A Crison Basic 30+ Radiometer Analytical conductivity meter was used to measure the ionic conductivities (µS/cm) of surfactant solutions in a glass cell at 20 • C containing a magnetic stirrer. For this method, a known amount of Milli-Q water was placed in a thermostated glass cell and the surfactant solution of known concentration was progressively added using a micropipette under constant stirring. Each conductivity value was measured at least three times.

Biological Studies
A preliminary assessment of antiviral activity, based on the capacity to inhibit the virus-induced cytopathic effect (CPE) on treated cells, was initially carried out for the HCQ-ILs, [HCQH 2 ][SO 4 ] and the corresponding anions as sodium or potassium salts. More comprehensive antiviral activity assays, intended for a direct measurement of SARS-CoV-2 replication (viral RNA transcription and production of infectious progeny viruses), were subsequently performed for the most promising HCQ-ILs. Three independent experiments with triplicate measurements were performed for all assays unless otherwise stated.
All work involving virus propagation and handling of viral cultures was performed in a biosafety level 3 (BLS-3) laboratory, following the WHO recommendations for laboratory biosafety guidance related to the SARS-CoV-2 virus [68].

Virus Stock
SARS-CoV-2 reference strain USA-WA1/2020 (catalog no. NR-52281) was obtained from the Centers for Disease Control and Prevention through the Biodefense and Emerging Infections Research Resources (BEI Resources), the National Institute of Allergy and Infectious Diseases (NIAID) and the National Institutes of Health (NIH) (Manassas, VA, USA) (www.beiresources.org, last accessed on 13 April 2022).
The strain was propagated in culture flasks of confluent Vero E6 cells (4 × 10 4 cells/cm 2 , cultured overnight) growing in DMEM supplemented with 2 mM L-glutamine, 1× NEAA, 24 mM HEPES, 50 µg/mL gentamycin, 2.5 µg/mL fungizone and 2% (v/v) FBS (herein designated as DMEM maintenance medium) at 37 • C in a 5% CO 2 atmosphere. Cultures were observed daily, and the virus was harvested when 80-90% of the cells manifested CPE. After the second passage, cell supernatants were collected and centrifuged at 4 • C 3000 rpm for 10 min to remove cell debris and stored at −80 • C in small aliquots as a working stock.
Stock viral titer was then determined by 50% Tissue Culture Infectious Dose (TCID 50 ) assay, using 10 replicates for each serial 10-fold dilution (from 10 −1 to 10 −8 ). Briefly, confluent Vero E6 monolayers (1.25 × 10 4 cells/well) cultured overnight in clear flat-bottom 96-well plates were infected with serial 10-fold dilutions of virus stock formerly prepared in DMEM supplemented with 2 mM L-glutamine, 1× NEAA and 24 mM HEPES (herein designated as DMEM base medium). After 1h of incubation at 37 • C (5% CO 2 ), the cells were washed with Dulbecco's phosphate-buffered saline (DPBS) (Gibco, Life Technologies Limited, Paisley, UK) and cultured with fresh DMEM maintenance medium for 72 h at previous incubation conditions. Virus-induced CPE was recorded under an inverted optical microscope, and viral titer, expressed as TCID 50 /mL, was calculated following the method of Reed and Müench [69].

Cell Viability Assay
The CellTiter-Glo ® Luminescent Cell Viability Assay (Promega, Madison, WI, USA) was used to measure the cytotoxic effects of the different compounds (HCQ-ILs and parental drug) and anions (as sodium or potassium salts) on Vero E6 cells. Briefly, confluent Vero E6 monolayers (1.25 × 10 4 cells/well) cultured overnight in white-wall clear-bottom 96-well plates were incubated with serial 2-and 10-fold dilutions (0.5 µM to 400 µM) of the compounds in DMEM maintenance medium at 37 • C and 5% CO 2 . Non-toxic and vehicle assay control wells were instead treated with equal volumes of, respectively, oseltamivir carboxylate (OSC; F. Hoffmann-La Roche Ltd., Basel, Switzerland; 2-fold serial dilutions-0.16 µM to 20 µM) and DMEM maintenance medium alone or with DMSO (0.8%). Wells with only vehicle medium (no cells) were used as background control. After 72 h, CellTiter-Glo ® Reagent was added to plates according to manufacturer's instructions, and luminescence was recorded using a FLUOstar OPTIMA plate reader (BMG Labtech, Madrid, Spain). A 10× lysis solution (Promega) was added to half of the vehicle-treated cells (2 times triplicate) 30 min before adding the assay reagent to include a positive toxic control in the assay.
Raw luminescence values, expressed in relative light units (RLUs), were analysed in a spreadsheet to calculate the percentage of cell viability relative to vehicle control and determine the 50% cytotoxic concentration (CC 50 ) (i.e., the concentration that reduces cell viability by 50%) of each compound and counterion by point-to-point curve fitting. After 1 h incubation at 37 • C (5% CO 2 ), cells were challenged with SARS-CoV-2 virus at 100 TCID 50 and returned to the incubator for another 1 h at the same conditions. Fresh DMEM maintenance medium containing the indicated concentrations of compounds/anions was added after washing cell monolayers twice with DPBS, and plates were incubated for 72 h at standard conditions (37 • C, 5% CO 2 ). DMEM maintenance medium alone or with DMSO (0.05%) was instead added to virus, vehicle and background (no cells) assay control wells. Virus-induced CPE was measured by CellTiter-Glo ® Luminescent Cell Viability Assay, following the manufacturer's instructions and using a Tecan Infinite M200 plate reader (TECAN, Männedorf, Switzerland) for recording luminescence.
Raw luminescence values were analysed in a spreadsheet to calculate the percentage of CPE inhibition relative to virus control (% = 100 × ((RLU infected treated cells − mean RLU virus control)/(mean RLU vehicle control − mean RLU virus control)); and determine the 50% (EC 50 ) and 90% (EC 90 ) effective concentration (i.e., the concentration that inhibits virus-induced CPE by 50% and 90%, respectively) of each compound by point-to-point curve fitting. Selectivity indexes (SI) were calculated from the relationship between CC 50 and EC 50 . Only two independent experiments were performed for the anions.

Inhibition of Viral RNA Transcription and Infectious Progeny Production
The pre-treatment, infection and treatment of cell monolayers were performed as described above for CPE inhibition, with only minor differences that included: the 96-well plates used (clear flat-bottom); the concentration range tested (1 µM to 50 µM); and the extent of final incubation (48 h). After 48 h post-infection (hpi), the cell supernatants of the three replicates under identical conditions were collected into the same tube and centrifuged at 3000 rpm for 10 min to remove the cell debris before being stored at −80 • C in small aliquots. Viral RNA (vRNA) transcription was assessed by quantification of virus yield in cell supernatants using a quantitative real-time RT-PCR (qRT-PCR) assay, while a TCID 50 assay was used to evaluate the production of infectious progeny viruses by determination of the infectious virus titer in the equivalent supernatants. Only qRT-PCR is detailed below, since the TCID 50 assay was performed as described above for virus stock titration. Data from both assays were analysed in a spreadsheet to calculate the percentage of inhibition relative to virus control and to determine the EC 50 , EC 90 and SI relative to each parameter as described for CPE inhibition. Four independent experiments were performed, as an additional assay was needed to clarify qRT-PCR results.

Statistical Analysis
Graphical representations and statistical analyses were performed using GraphPad Prism software version 9.2.0 for Mac (GraphPad Software, San Diego, CA, USA). Significant differences in CC 50 , EC 50 and EC 90 values were evaluated using a one-way analysis of variance (ANOVA) followed by Tukey's multiple comparison test. A p-value < 0.05 was considered significant.

Results and Discussion
We describe the synthesis of seven new ionic liquids containing hydroxychloroquine (HCQ-ILs) as dications by direct protonation with two equivalents of biocompatible anions. All prepared HCQ-ILs were characterized by spectroscopic techniques and their solubility in water and octanol-water partition coefficients were determined. The critical micelle concentrations of the most lipophilic HCQ-ILs were also determined. Lastly, evaluation of cytotoxicity and anti-SARS-CoV-2 activity in Vero E6 cells was performed.

Synthesis and Characterization
The synthesis of the HCQ-ILs was performed by the dropwise addition of the selected organic acids to a hydroxychloroquine base (HCQ), in an analogous fashion to previous works with other drugs (e.g., fluoroquinolones [51]). HCQ was previously prepared by passing hydroxychloroquine sulfate ([HCQH 2 ][SO 4 ], 1), the commercial form of HCQ, through the hydroxide exchange resin Amberlyst A-26 (OH). The obtained HCQ free base (2) was subsequently protonated at the alkyl secondary and tertiary amines in aqueous media with two equivalents of five sulfonic acids, one sulfuric acid and one carboxylic acid based on the following anions: methanesulfonate [C 1 SO 3 (9). Some of the organic acids were prepared from the corresponding sodium salts by previously undergoing ion exchange via the cationic resin Amberlite 15(H + ). Scheme 1 shows the employed synthetic methodology.
by passing hydroxychloroquine sulfate ([HCQH2][SO4], 1), the commercial form of HCQ, through the hydroxide exchange resin Amberlyst A-26 (OH). The obtained HCQ free base (2) was subsequently protonated at the alkyl secondary and tertiary amines in aqueous media with two equivalents of five sulfonic acids, one sulfuric acid and one carboxylic acid based on the following anions:  (9). Some of the organic acids were prepared from the corresponding sodium salts by previously undergoing ion exchange via the cationic resin Amberlite 15(H + ). Scheme 1 shows the employed synthetic methodology.

Scheme 1. Methodology for the synthesis of the HCQ-ILs.
All HCQ-ILs were isolated as gels and hence considered as room temperature ionic liquids (RTILs). They were characterized by FTIR and NMR ( 1 H, 13 C) spectroscopies, highresolution mass spectrometry (ESI-TOF) as well as elemental analysis. The thermal properties were studied by differential scanning calorimetry, and the critical micelle concentration of the most lipophilic compounds (5 and 8) were studied by electric conductivity measurements.
The FTIR spectra of the prepared HCQ-ILs (Figures S1-S7) supported the full ionization of the organic acids. In the case of compound 9, the presence of a carboxylate stretching band at 1578 cm −1 and the absence of one at ca. 1730 cm −1 , corresponding to the stretching of the carboxylic acid group, confirms the complete ionization of this HCQ-IL. In the case of the sulfonate-containing ILs 3-7, similar vibrational profiles within the 1350-850 cm −1 fingerprint zone were observed between the starting anion salts and corresponding products, which were strikingly different for HCQ and corresponding acids. Hence, this corroborated the presence of the anions and cations as ionized species. Figure 1 depicts a comparison of the FTIR spectra of [HCQH2][C1SO3]2 (3), hydroxychloroquine (2), potassium methanesulfonate (C1SO3K) and also methanesulfonic acid (C1SO3H). In more detail, the bands at 1333, 980 and 883 cm −1 are characteristic of non-ionized methanesulfonic acid, as they are absent from the spectra of the remaining compounds. In addition to the absence of these bands, the FTIR spectra of product 3 contains a band that appears at 1187 cm −1 in the spectrum of potassium methanesulfonate, ascribable to the stretching of the Scheme 1. Methodology for the synthesis of the HCQ-ILs.
All HCQ-ILs were isolated as gels and hence considered as room temperature ionic liquids (RTILs). They were characterized by FTIR and NMR ( 1 H, 13 C) spectroscopies, high-resolution mass spectrometry (ESI-TOF) as well as elemental analysis. The thermal properties were studied by differential scanning calorimetry, and the critical micelle concentration of the most lipophilic compounds (5 and 8) were studied by electric conductivity measurements.
The FTIR spectra of the prepared HCQ-ILs (Figures S1-S7) supported the full ionization of the organic acids. In the case of compound 9, the presence of a carboxylate stretching band at 1578 cm −1 and the absence of one at ca. 1730 cm −1 , corresponding to the stretching of the carboxylic acid group, confirms the complete ionization of this HCQ-IL. In the case of the sulfonate-containing ILs 3-7, similar vibrational profiles within the 1350-850 cm −1 fingerprint zone were observed between the starting anion salts and corresponding products, which were strikingly different for HCQ and corresponding acids. Hence, this corroborated the presence of the anions and cations as ionized species. Figure 1 depicts a comparison of the FTIR spectra of [HCQH 2 ][C 1 SO 3 ] 2 (3), hydroxychloroquine (2), potassium methanesulfonate (C 1 SO 3 K) and also methanesulfonic acid (C 1 SO 3 H). In more detail, the bands at 1333, 980 and 883 cm −1 are characteristic of non-ionized methanesulfonic acid, as they are absent from the spectra of the remaining compounds. In addition to the absence of these bands, the FTIR spectra of product 3 contains a band that appears at 1187 cm −1 in the spectrum of potassium methanesulfonate, ascribable to the stretching of the sulfonate group. Hence, these data indicate the ionization of methanesulfonic acid upon combination with hydroxychloroquine.   (Table S1). On the other hand, the negative mode spectra displayed the peak corresponding to [M−1] -at the expected m/z values, with an absolute error comprehended between −2.61 and −16.96 ppm (Table S2). The observed isotopic distribution was as expected for each case.

Thermal Properties
As previously mentioned, all HCQ-ILs were obtained as room temperature ionic liquids, as opposed to the starting [HCQH2][SO4], which is a solid that melts at 245 °C ( Figure  S36). Each sample was studied by differential scanning calorimetry ( Figures S37-S43) by sequentially heating and cooling the sample between −90 and 100 °C at 10 °C/min (two cycles) and 20 °C/min (one final cycle). One preliminary isotherm at 100 °C for 10 min was performed to remove residual water. In agreement with the amorphous state of the compounds, all thermograms displayed one glass transition temperature (Tg), which was calculated in the last heating cycle (Table 1).  (Table S1). On the other hand, the negative mode spectra displayed the peak corresponding to [M−1]at the expected m/z values, with an absolute error comprehended between −2.61 and −16.96 ppm (Table S2). The observed isotopic distribution was as expected for each case.

Thermal Properties
As previously mentioned, all HCQ-ILs were obtained as room temperature ionic liquids, as opposed to the starting [HCQH 2 ][SO 4 ], which is a solid that melts at 245 • C ( Figure S36). Each sample was studied by differential scanning calorimetry ( Figures S37-S43) by sequentially heating and cooling the sample between −90 and 100 • C at 10 • C/min (two cycles) and 20 • C/min (one final cycle). One preliminary isotherm at 100 • C for 10 min was performed to remove residual water. In agreement with the amorphous state of the compounds, all thermograms displayed one glass transition temperature (T g ), which was calculated in the last heating cycle (Table 1).  (9) 0.4 In addition, none of the compounds presented a tendency for crystallization in the several cooling cycles performed.
These data suggest that T g has a decreasing trend with increasing length of the sulfonate/sulfate alkyl chain. More specifically, the HCQ-IL with the smallest sulfonate anion, [C 1 SO 3 ] (3), presented a T g of 29.9 • C, which decreased to 9.9 for [C 6 SO 3 ] (4) and subsequently to −7.9 and −11.4 in the combinations with [DocSO 3 ] (5) and [C 12 SO 4 ] (8), respectively. This may be attributable to a more disorganized arrangement of the vitreous state in the HCQ-ILs bearing long alkyl chain anions.

Water Solubility and Octanol-Water Partition Coefficient Studies
On the one hand, the solubility of the prepared HCQ-ILs was determined by adding known volumes of freshly double-distilled water to fixed amounts of compounds at 37 • C until complete dissolution was observed by visual inspection. On the other hand, the octanol-water partition coefficients were measured by preparing ca. 1 mg/mL solutions of each HCQ-IL in octanol-saturated water, which were then thoroughly stirred with an equal volume of water-saturated octanol. UV-Visible absorption spectra were recorded for the initial and final aqueous or organic solutions, depending on the solubility of each HCQ-IL in the solvents. The results are presented as the logarithm of K ow , Log P. Table 2 compiles the data obtained from both experiments. Table 2. Water solubility (in mg/mL) and Log P values of the prepared HCQ-ILs.

HCQ-ILs
Water Solubility (mg/mL) Log P As expected, most HCQ-ILs are more soluble in water than the original drug (84 mg/mL), with the exception of the ones containing the lipophilic anions [DocSO 3 ] (5) and [C 12 SO 4 ] (8). While the latter are insoluble in water (lower than 0.5 mg/mL), the former displayed very high solubility comprehended between 910 ([C 6 SO 3 ], 4) and 2020 mg/mL ([CampSO 3 ], 6). Accordingly, the water-soluble compounds display lower Log P values than the water-insoluble ones. However, the correlation is not linear. In the particular case of [HCQH 2 ][C 1 SO 3 ] 2 (3), it showed the second highest solubility and also a Log P value near zero, consistent with only a slight preference for aqueous media in the presence of an apolar phase. The remaining salts displayed similar Log P values, independently of their water solubility profiles. While the latter followed the trend 6 > 7 > 9 > 4 with very different recorded values, their Log P followed a different trend of 4 > 7 > 6 > 9, nonetheless with very similar values.
Hence, the most soluble salts would rapidly and extensively dissolve after oral administration, which could mean that they have a higher oral bioavailability than the least soluble ones. In addition, positive or near positive Log P values may also lead to a higher degree of interaction with apolar structures, such as cellular membranes, which could also account for a potentially high drug uptake. An optimal balance between hydrophilic and lipophilic properties is required in order to enhance hydroxychloroquine oral bioavailability and consequently its therapeutic activity, while modulating its excretion and distribution throughout many different tissues and organs in order to reduce systemic toxicity.

Critical Micelle Concentration
The ability to form micelles of the lipophilic HCQ-ILs [DocSO 3 ] (5) and [C 12 SO 4 ] (8) was studied by ionic conductivity measurements. The inherent ability of these HCQ-ILs to self-aggregate into micelles can render adequate drug delivery properties. By measuring the conductivity of solutions with increasing concentrations of the compounds, two distinct slopes were observed for each case (see Figures S44 and S45 in ESI), consistent with the formation of micelles. The critical micelle concentration (cmc) values, calculated by resolving both equations, are given in Table 3, alongside the cmc values for the starting halide salts (Na[DocSO 3 ] and [C 12 SO 4 ]).  (8) 4.83 × 10 −6 -Na[C 12 SO 4 ] (8a) 9.60 × 10 −3 8.0 × 10 −3 [72] As expected, the cmc values of the starting halide salts are much higher (ca. three orders of magnitude) than those for the HCQ-ILs due to the higher degree of hydration of the former, which hinders the process of micelle formation, and hence higher concentrations of compounds are required. On the other hand, by adsorbing into the micellar surface in an easier fashion, the less hydrated HCQ-ILs decrease the charge repulsion between the polar heads and micelles are formed at lower concentrations.

Cytotoxicity in Vero E6 Cells
In order to assess the biocompatibility of the compounds under study, cytotoxicity assays were performed on Vero E6 cells, which are the gold standard for SARS-CoV-2 propagation studies. As shown in Figure 2A (Table 4). The only exceptions to this biocompatible profile were the lipophilic HCQ-ILs 5 and 8, which exhibited CC50 values of 69.8 and 73.2 µ M, respectively. This behaviour was supported by the high toxicity exhibited by the corresponding lipophilic anions 5a and 8a, the only ones found to be toxic to these cells, as evidenced in Figure 2B, with CC50 values of 142.2 µ M and 148.9 µM, respectively (Table 4). Table 4. Half-maximal cytotoxic activity (CC50) of [HCQH2][SO4] (1), HCQ-ILs and corresponding anions on Vero E6 cells. Results are presented as the mean  standard deviation of three independent experiments with triplicate measurements. Significant (SIG) differences were evaluated using a one-way analysis of variance (ANOVA) followed by Tukey's multiple comparison test. A p-value < 0.05 was considered significant, with *** p < 0.001 and **** p < 0.0001.

Compounds
CC50 (µM) 95% CI SIG Differences  The only exceptions to this biocompatible profile were the lipophilic HCQ-ILs 5 and 8, which exhibited CC 50 values of 69.8 and 73.2 µM, respectively. This behaviour was supported by the high toxicity exhibited by the corresponding lipophilic anions 5a and 8a, the only ones found to be toxic to these cells, as evidenced in Figure 2B, with CC 50 values of 142.2 µM and 148.9 µM, respectively (Table 4).
In other words, all HCQ-ILs were considered biocompatible with the exception of the latter two, which were found to be ca. three times more toxic than the original molecule in vitro. The enhanced lipophilic properties of these HCQ-ILs could promote disruption of the cell membrane and possibly lead to drug accumulation inside the cell beyond safety levels.  (9) resulted in more than 60% inhibition of CPE at 10 µM (70.5%, 60.8% and 62.5%, respectively), which contrasted with the remaining compounds, including 1 (lower than 20%; see Figure 3A).

In Vitro
The EC 50 values of these three novel formulations (8.1, 8.9 and 8.5 µM, respectively) were significantly lower than those of all other HCQ-ILs and differed by ca. two-fold from the EC 50 of 1 (16.5 µM) ( Table 5). No significant differences in EC 90 were observed between the seven novel HCQ-ILs and 1, with all values being registered beyond 20 µM.
The enhanced antiviral activity of 3 and 9 in this initial screening doubled their SI ratio (26.9 and 23.1, respectively) in comparison with 1 (13.0) ( Table 5), leading them to be identified as the most promising HCQ-ILs. Hence, these were selected for further and more comprehensive evaluation of their anti-SARS-CoV-2 activity by direct measurement of inhibition of virus infection (see Section 3.6.2). Despite its high activity, 8 (Ec 50 8.9 µM) presented one of the lowest SI ratios (7.8) given its high cytotoxicity towards Vero E6 cells (CC 50 69.8 µM) and was thus not selected for the subsequent inhibition studies. Also of note, none of the anions showed antiviral activity against SARS-CoV-2 ( Figure 4B).  (9) resulted in more than 60% inhibition of CPE at 10 µ M (70.5%, 60.8% and 62.5%, respectively), which contrasted with the remaining compounds, including 1 (lower than 20%; see Figure 3A). The EC50 values of these three novel formulations (8.1, 8.9 and 8.5 µ M, respectively) were significantly lower than those of all other HCQ-ILs and differed by ca. two-fold from the EC50 of 1 (16.5 µ M) ( Table 5). No significant differences in EC90 were observed between the seven novel HCQ-ILs and 1, with all values being registered beyond 20 µM.  (1) and HCQ-ILs 3-9 for the inhibition of the virus-induced cytopathic effect (CPE) on Vero E6 cells infected with SARS-CoV-2. The results are presented as the mean of three independent experiments with triplicate measurements. The 95% confidence interval (CI) is indicated in a separate column. SI represents the CC50/EC50 selectivity ratio. No inhibitory effects were found for the anions. Significant differences (SIG diff.) were evaluated using a one-way analysis of variance (ANOVA) followed by Tukey's multiple comparison test. A p-value < 0.05 was considered significant, with * p < 0.05, ** p < 0.01, *** p < 0.001. The enhanced antiviral activity of 3 and 9 in this initial screening doubled their SI ratio (26.9 and 23.1, respectively) in comparison with 1 (13.0) ( Table 5), leading them to be identified as the most promising HCQ-ILs. Hence, these were selected for further and more comprehensive evaluation of their anti-SARS-CoV-2 activity by direct measurement of inhibition of virus infection (see Section 3.6.2). Despite its high activity, 8 (Ec50 8.9 µ M)   (1) and HCQ-ILs 3-9 for the inhibition of the virus-induced cytopathic effect (CPE) on Vero E6 cells infected with SARS-CoV-2. The results are presented as the mean of three independent experiments with triplicate measurements. The 95% confidence interval (CI) is indicated in a separate column. SI represents the CC 50 /EC 50 selectivity ratio. No inhibitory effects were found for the anions. Significant differences (SIG diff.) were evaluated using a one-way analysis of variance (ANOVA) followed by Tukey's multiple comparison test. A p-value < 0.05 was considered significant, with * p < 0.05, ** p < 0.01, *** p < 0.001. presented one of the lowest SI ratios (7.8) given its high cytotoxicity towards Vero E6 cells (CC50 69.8 µ M) and was thus not selected for the subsequent inhibition studies. Also of note, none of the anions showed antiviral activity against SARS-CoV-2 ( Figure 4B).

Inhibitory Effects on vRNA Transcription and Progeny Production
The two most promising formulations, 3 and 9, were evaluated, together with the parental drug, for their capacity to inhibit vRNA transcription and the production of infectious progeny. In the case of 9, over 50% inhibition in both experiments was observed at 16.7 µ M (55.8% and 56.5%, respectively; Figure 4A,B). . Dose-response curves for the two HCQ-OSILs identified as promising and for the parental drug (1) based on viral RNA (vRNA) transcription (A) and progeny production (B). RNA transcription was assessed by quantification of virus yield using qRT-PCR assay (RdRp gene) and progeny production was determined by calculation of the infectious virus titer by TCID50 assay (TCID50/mL), both at 48 hpi in cell supernatants. Results are expressed in relative values compared to virus control. Data were plotted and generated as indicated in Figure 2. Four independent experiments were carried out on vRNA transcription assays (A).
On the other hand, 3 exhibited inhibitory profiles closer to the original molecule, causing a slightly lower inhibition (48.5%; 1: 34.0%) of vRNA transcription ( Figure 4A) and no further inhibition of progeny production at the same concentration ( Figure 4B). In fact, 9 presented very similar EC50 values for both experiments (16.3 and 16.9 µ M, respectively), and these were found to be the lowest for the three compounds. From both of these values, only the one regarding progeny production differed significantly from the EC50 values of 3 (21.8 µ M) and the parental drug 1 (21.7 µ M). No significant differences in EC90 . Dose-response curves for the two HCQ-OSILs identified as promising and for the parental drug (1) based on viral RNA (vRNA) transcription (A) and progeny production (B). RNA transcription was assessed by quantification of virus yield using qRT-PCR assay (RdRp gene) and progeny production was determined by calculation of the infectious virus titer by TCID 50 assay (TCID 50 /mL), both at 48 hpi in cell supernatants. Results are expressed in relative values compared to virus control. Data were plotted and generated as indicated in Figure 2. Four independent experiments were carried out on vRNA transcription assays (A).

Inhibitory Effects on vRNA Transcription and Progeny Production
The two most promising formulations, 3 and 9, were evaluated, together with the parental drug, for their capacity to inhibit vRNA transcription and the production of infectious progeny. In the case of 9, over 50% inhibition in both experiments was observed at 16.7 µM (55.8% and 56.5%, respectively; Figure 4A,B).
On the other hand, 3 exhibited inhibitory profiles closer to the original molecule, causing a slightly lower inhibition (48.5%; 1: 34.0%) of vRNA transcription ( Figure 4A) and no further inhibition of progeny production at the same concentration ( Figure 4B). In fact, 9 presented very similar EC 50 values for both experiments (16.3 and 16.9 µM, respectively), and these were found to be the lowest for the three compounds. From both of these values, only the one regarding progeny production differed significantly from the EC 50 values of 3 (21.8 µM) and the parental drug 1 (21.7 µM). No significant differences in EC 90 values were observed among compounds for either of the read-outs, with values varying between 22 µM and 25 µM ( Table 6). The SI ratios of both HCQ-ILs for the two read-outs were found to be slightly higher than that of the original molecule, as evidenced in Table 6.  4 ] and HCQ-ILs for the inhibition of viral RNA (vRNA) transcription and production of progeny infectious viruses on Vero E6 cells infected with SARS-CoV-2. The results are presented as the mean of four (vRNA) and three (progeny) independent experiments with triplicate measurements. The 95% confidence interval (CI) is indicated in a separate column. SI represents the CC 50 /EC 50 selectivity ratio. Significant differences (SIG diff.) were evaluated using a one-way analysis of variance (ANOVA) followed by Tukey s multiple comparison test. A * p-value < 0.05 was considered significant. These data indicate that there was a slightly significant (p < 0.05) enhanced inhibition of the production of virus particles capable of causing infection in the presence of 9 that was not observed for 3 and the parental drug 1. However, there was no significant difference in the production of viral RNA (from both infectious and non-infectious viruses) among the three compounds. These parameters yield a more robust estimation of the compounds' impact on virus infection than the one based on the inhibition of the virus-induced CPE, as in the latter case there was only an estimation based on the differentiation between viable and non-viable (dead) cells in the presence of SARS-CoV-2.

Conclusions
Since none of the anions (as sodium or potassium salts) possessed antiviral activity against the studied strain of SARS-CoV-2, the observed enhanced activities of the two most promising HCQ-ILs, [HCQH 2 ][C 1 SO 3 ] 2 (3) and [HCQH 2 ][GlcCOO] 2 (9), are suggestive of specific intramolecular (between cation and anions) and intermolecular (between the HCQ-ILs, cell organelles and/or viral structural components) interactions. Moreover, these activities seem not to correlate with the lipophilic or hydrophilic properties of the compounds. On one hand, both 3 and 9 possess water-solubility and Log P values similar to other tested HCQ-ILs that did not show improved antiviral activity. On the other hand, the highly lipophilic [HCQH 2 ][DocSO 3 ] (5) showed CPE inhibition similar to the two promising hydrophilic HCQ-ILs, while the analogously apolar [HCQH 2 ][C 12 SO 4 ] 2 (8) was even less effective in inhibiting the virus-induced CPE than the original drug. This work paves the way for the development of ionic formulations of hydroxychloroquine with enhanced physicochemical properties.