Hydroxy-xanthones as promising antiviral agents: Synthesis and biological evaluation against human coronavirus OC43

Graphical abstract

Although the first coronavirus was discovered in the 1930s, they have played a major role in contributing to severe infections, causing outbreaks of Severe Acute Respiratory Syndrome Coronavirus (SARS-CoV) in 2003 and Middle East Respiratory Syndrome Coronavirus (MERS-CoV) in 2012. 1 In December 2019, Wuhan, Hubei Province, China reported the first cases of unexplained respiratory viral illness and, by March 11th 2020, coronavirus disease 2019 (COVID-19) caused by SARS-CoV-2 was declared a global pandemic by the World Health Organisation (WHO). COVID-19 has resulted in over 600 million confirmed infections and more than 6 million deaths worldwide to date, becoming the biggest global health crisis since the influenza pandemic of 1918. [2][3] In the interim few years, the emergence of new and vaccine-escaping SARS-CoV-2 variants, and other potentially pandemic-causing coronaviruses, threaten the efficacy of COVID-19 vaccines and current antivirals, highlighting the demand for new, effective antiviral medication and advancement for existing therapies. [4][5] It is estimated that over 85% of all biologically-active compounds contain a heterocyclic motif, providing a useful handle for lead optimisation. [6][7] Of particular note are the xanthone class, naturally occurring compounds that can be found as secondary metabolites in diverse terrestrial and marine plants, fungi, and lichen. 8 Chemically, xanthones (9H-xanthen-9-ones) encompass a family of compounds with an oxygencontaining dibenzo-γ-pyrone heterocyclic scaffold, [9][10] and provide a wide range of different substitutions modulating several biological responses, thus being considered a promising and interesting structure for drug development. 11-12 . Within the subset of naturally occurring xanthones, the hydroxy-xanthones, prevalent in Swertia, have been considered promising antiviral small molecules, potentially inhibiting multiple coronavirus target proteins; [13][14] and a number of these xanthones have been isolated, Fig  1. 15 Furthermore, mangiferin (Fig. 2), a hydroxy-xanthone glucoside, and the main active component in Mangifera indica, has previously shown to have broad-spectrum antiviral activity against a range of viruses including human immunodeficiency virus 1 (HIV-1) and herpes simplex virus type 1 (HSV-1). 16,17 More recently, mangiferin was identified as a potential inhibitor of the SARS-CoV-2 main protease (Mpro) in in silico studies, 18 and possess good pharmacokinetic properties, [19][20] suggesting inhibition of a conserved viral protein, valuable in the production of a broad-spectrum anti-coronavirus treatment. Further studies exhibited that mangiferin has significant binding affinity towards the spike glycoprotein of SARS-CoV-2 and ACE2 receptor and that it may be useful as a therapeutic and/or prophylactic agent for restricting viral attachment to the host cells. 21 Given the need for the development of antiviral treatments against possible emerging coronavirus variants in the future, coupled with the preliminary evidence that hydroxy-xanthones appear to possess antiviral activity, we wished to ascertain if synthesised derivatives would also show potential against human coronavirus (HCoV) OC43. HCoV-OC43 is a seasonal coronavirus that causes common colds, belonging to the same betacoronavirus genus as SARS-CoV-2, and has been previously used as a model virus in drug screening studies. 15,[22][23] Using HCoV-OC43 allowed us to rapidly screen the antiviral activity of our synthesised xanthones, and identify hit compounds for optimisation into a series of leads as potential broad-spectrum pan-coronavirus antivirals.
To this end, Scheme 1 delineates the route used for the preparation of a series of nine hydroxy-xanthones (4) through the union between salicylic acid derivatives (1) and a functionalised phenol (2), promoted by phosphorus(V) oxychloride and zinc(II) chloride at 70 • C. 24 A limitation to this route is its benzophenone intermediate (3), which, for the reaction to proceed directly to the desired xanthone, necessitates an additional hydroxyl moiety ortho to the carbonyl group. If this limitation is not fulfilled, then an additional cyclization of the benzophenone intermediate is needed to obtain the xanthone. 25 In general, the reactions proceeded smoothly within two hours with modest to high conversions, but isolation of these known compounds, [26][27][28][29][30] in the high yields often reported, proved difficult; however, up to 30% isolated yields were achieved, in-line with the experiences of others when making these, and similar, compounds. 24 Nevertheless, the reactions could be carried out on a gram scale providing enough material for characterisation and biological evaluation. Fig. 2 shows the structures of the nine xanthones prepared using this method. Interestingly, the synthesis of derivatives using dihydroxylated salicylic acids resulting in hydroxyl groups on the 'left-hand' B-ring (as drawn above), except in position-8, could not be achieved using this method. However, analogues (10)(11)(12)(13)(14), containing hydroxyl group isosteres (-CH 3 , -OCH 3 , -F), 31 could be made in similar yields, although the electronics of the functional group made an impact on isolated yields here. For example, 6-OCH 3 derivative (12) was only isolated in 6% yield, whereas the 7-OCH 3 derivative (13) was isolated in a more modest 28% yield; presumably, the electron-donating capability of the 6-OCH 3 group directly onto the carbonyl of the acid reduced it electrophilicity in the Friedel-Crafts acylation step, thereby reducing reactivity of this isomer. Contrastingly, the 7-F derivative (14) could only be isolated in low yield (5%), whereas the 6-F derivative could not be isolated at all, despite the fact that the 6-F derivative could, by the same argument above, increase reactivity in the Friedel-Crafts step, suggesting that factors other than electronics alone are at play.
Furthermore, analogues 7, 8 and 9 from pyrogallol and resorcinol, respectively, initially halted at the benzophenone stage (3 in Scheme 1) under the same reaction conditions, a phenomenon with some precedent. 32 Nevertheless, the reaction did provide related, more flexible, benzophenone analogues (15 and 16 (R = H)) for biological evaluation. Furthermore, benzophenones 15 and 16 (R = OH) could be cyclised to the analogous xanthones (7 and 8) by heating the benzophenone in methanol/water at reflux, although, 3-hydroxyxanthone (9) could not be cyclised from benzophenone (16 (R = H)). A possible explanation for this difference in reactivity could be that in the cases of pyrogallol and resorcinol derivatives, intramolecular H-bonding to the newly installed ketone (e.g. 17 in Fig. 3) can take place, and the conditions of the reaction are not conducive to cyclisation from this conformation and so halt at the benzophenone stage. Alternatively, with phloroglucinol (e.g. 18 in Fig. 3), whilst similar H-bonding can take place, there always exists a proximal hydroxyl group for cyclisation under the conditions of the reaction.
Having access to a number of xanthones (and benzophenones) now in hand, we set about evaluating their antiviral activity against HCoV- OC43 virus. All of the compounds prepared were fully characterised and were purified to over 90% purity (determined by HPLC) prior to biological evaluation.
The compounds were evaluated for antiviral activity using Baby Hamster Kidney fibroblast (BHK-21) cell monolayers infected with HCoV-OC43 virus, by determining viral infectivity after a three-day treatment.
In order to ensure that any observed antiviral activity was due to interaction of the compound with the virus, and not through cytotoxic behaviour, all compounds screened were initially assessed for their cytotoxicity (Fig. 4). None of the xanthones tested showed significant changes in cell viability, and were deemed to be non-toxic (≥85% cell viability) at the concentration tested.
An inhibition of viral infectivity was observed in cells treated with the xanthone derivatives compared to the negative (DMSO) control, with a log 10 reduction ranging from 1.95 to 2.79 (Table 1). Given the structural modifications in the panel of xanthones, the differences in activity between the compounds were limited, compared to previously observed data in anticancer assays., [33][34][35] For example, dihydroxyxanthones showed different levels of growth inhibitory activities (2 to 10-fold differences) in human cancer cell lines compared to dimethoxy derivatives at the same position. 34 On the contrary, the structural changes made here have not contributed to drastic changes in the viral infectivity. Nevertheless, efforts have been made below to decipher trends within the data.
Xanthone itself showed a log 10 reduction in infection of 2.04 (Table 1). This was a larger reduction than expected considering the lack of functional groups on the periphery of the compound, although xanthone has been shown to be bioactive as an ovicide and larvicide of the codling moth, 36 and as a suppressor of allergic contact dermatitis. 37 Interestingly, 1,3-dihydroxyxanthone (5, log 10 reduction 2.50) showed a significant (p ≤ 0.05) log 10 reduction compared to xanthone, suggesting that the addition of two hydroxyl groups is able to potentially improve the binding capability of the compound, and is mirrored with other 1,3dihydroxyxanthones 6, 10-14. Comparable to this, 3,8-dihydroxyxanthone (8) also showed a significant (p ≤ 0.001) decrease in infectivity (log 10 reduction 2.20). However, 1,3,8-trihydroxyxanthone (6) showed similar efficacy as xanthone (log 10 reduction 2.00) despite the additional hydroxyl groups, suggesting that a fine balance of the number of hydroxyl groups and their position is important for antiviral activity, but more work needs to be done to understand this effect through structureactivity relationship investigations.
The methyl substituted xanthones were expected to demonstrate greater biological activities due to their increased lipophilicity, which can improve receptor binding. 39 1,3-Dihydroxy-7-methylxanthone (11) demonstrated the highest antiviral activity within the panel of xanthones with a significant (p ≤ 0.05) log 10 reduction of 2.79 compared to the negative control, and the methoxyxanthones (12 and 13) showed a similar log 10 reduction of 2.67. The lack of any difference in the activity of 1,3-dihydroxy-6-methoxyxanthone (12) and 1,3-dihydroxy-7methoxyxanthone (13) again suggests that the position of the functional groups has little effect on their antiviral efficacies.

Table 1
Inhibition of viral infectivity. BHK-21 cells were infected with HCoV-OC43 (MOI = 0.1), and treated with compounds (10 μM) for 3 days. The table shows log 10 reduction in TCID 50 /ml compared to the DMSO (negative) control (n = 3; mean ± standard error). Significance of differences was determined by ANOVA with Tukey's multiple comparisons; ns, not significant; *, p ≤ 0.05; **, p ≤ 0.001. dihydroxy-6-fluoroxanthone, the effect of changing the position of the fluorine group could not be further evaluated. We were also interested to investigate the antiviral activity of benzophenones (15 and 16), due to the increased flexibility of benzophenones compared to xanthones. However, the similar log 10 reduction of benzophenone (15) and 3,4-dihydroxyxanthone (7) at 1.95, and the similar viral reduction exhibited by benzophenone (16 (R = H), log 10 reduction 2.45) and 1,3-dihydroxyxanthone (5, log 10 reduction 2.50) indicates that this increased flexibility of benzophenones does not improve biological activity. Although no change in biological activity was observed, the solubility of the benzophenones was greater than that of the xanthone derivatives, which suggests that the benzophenones may be easier to formulate into pharmaceuticals. 39,38 In our assays, mangiferin showed a higher reduction in HCoV-OC43 infectivity than the synthesised xanthone derivatives, showing a significant (p ≤ 0.001) log 10 reduction of 3.00 compared to the negative control. Mangiferin includes two hydroxyl groups at position-1 and position-3, identical to 1,3-dihydroxyxanthone (5), with additional hydroxyl groups at position-6 and position-7 and the presence of glucose at position-2 (Fig. 2). As a lower range of activity was observed with the synthesised xanthones, such as 5 (log 10 reduction 2.50), 6 (log 10 reduction 2.00), 7 (log 10 reduction 1.95) and 8 (log 10 reduction 2.20), the initial results of mangiferin (log 10 reduction 3.00) may indicate that the sugar is important for biological activity against HCoV-OC43. Further research to determine the 50% inhibitory concentrations (IC 50 ) of mangiferin and the hydroxylated xanthones will help delineate the contribution of the sugar moiety.
To further explore the translational potential of these synthetic xanthones from the cellular experiments, simple property predictions were performed using Molinspiration Cheminformatics. 40 Unfavourable physicochemical properties linked to poor pharmacokinetic profiles are responsible for the failure of many drug candidates, 41 hence, evaluation earlier in the drug design process is advisable, with computational predictions providing a good starting point. Table 2 shows the results of property predictions for the compounds synthesised.
The predicted values should be considered approximate, with some property predictions considerably dependent on conformation, nevertheless, they provide useful information to be considered in future ligand optimisations. The first test of drug-likeness we looked at was Lipinski's Rule of Five, 42 with compounds having fewer violations of these rules expected to have better oral bioavailability. None of our synthesised compounds demonstrated any violations of these rules. The Rule of Three provides an alternative test of drug-likeness, 43 from which hit compounds should be biased toward lower molecular weight and lipophilicity, making it easier to optimise drug development candidates that are also drug-like. All compounds have at least one violation of the rule of three, typically due to the number of hydrogen-bond acceptors, which should be ≤3, 44 but given that these xanthones are beyond the definition of 'fragments', and that mangiferin itself is the most potent xanthone, yet has the most violations across the board, this is less of a concern. Additionally, Veber et al. suggested that a PSA of <140 Å 2 is desirable for oral bioavailability, 44 and all synthesised compounds adhere to this rule.
In conclusion, nine hydroxy-xanthones and two benzophenone intermediates have been prepared and biologically evaluated against human coronavirus OC43. Overall, the results of the initial in vitro screening of the test compounds show promising biological activity against HCoV-OC43, as well as possessing favourable property predictions making them interesting lead compounds for further development. Generally, the addition of functional groups increases the biological activity of the xanthone core, as demonstrated by compounds 5, 7, 8, 11, 12, 13, 15, 16 (R = H), showing a statistically significant reduction in infectivity of virus (p ≤ 0.05). Nevertheless, further research is required to determine the optimal position and number of functional groups, as well as to understand their mechanism of action, particularly compared to that of mangiferin, previously shown to have broad-spectrum antiviral activity.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Data availability
No data was used for the research described in the article.