Anti-Neuraminidase Bioactives from Manggis Hutan (Garcinia celebica L.) Leaves: Partial Purification and Molecular Characterization

The neuraminidase enzyme (NA) from the influenza virus is responsible for the proliferation and infections of the virus progeny, prompting several efforts to discover and optimize effective neuraminidase inhibitors. The main aim of this study is to discover a new potential neuraminidase inhibitor that comes from Garcinia celebica leaves (GCL). The bioassay-guided isolation method was performed to obtain lead compounds. The binding interaction of the isolated compounds was predicted by using molecular docking studies. Friedeline (GC1, logP > 5.0), two lanastone derivatives (methyl-3α,23-dihydroxy-17,14-friedolanstan-8,14,24-trien-26-oat (GC2) and 24E-3a,9,23-trihydroxy-17,14-friedolanostan-14,24-dien-26-oate (GC3) with LogP > 5.0) and catechin (GC4, LogP = 1.4) were identified. The inhibitory potency of these four compounds on NA from C. perfringens and H1N1 was found to be as follows: GC4 > GC2 > GC3 > GC1. All compounds exhibited higher inhibitory activity towards C. perfringens NA compared to H1N1 NA. From the molecular docking results, GC4 favorably docked and interacted with Arg118, Arg371, Arg292, Glu276 and Trp178 residues, whilst GC2 interacted with Arg118, Arg371, Arg292, Ile222, Arg224 and Ser246. GC3 interacted with Tyr406 only. GC4 had potent NA inhibition with free energy of binding of −12 kcal/mol. In the enzyme inhibition study, GC4 showed the highest activity with an IC50 of 60.3 µM and 91.0 µM for C. perfringens NA and H1N1 NA—respectively.


Introduction
Neuraminidase (NA) is an enzyme that plays an essential role in the cleavage of sialic acid from the terminal receptors of cells, which will subsequently release new viruses from infected cells. NA can be found in many families of viruses, bacteria, protozoa, some invertebrates and some mammalian cells [1,2]. NA from different organisms has a different binding affinity or substrate preference, but they have some structural similarities, with several conserved domains and amino acid residues at the binding site [3]. Generally, the NA of all organisms will cleave the ketosidic bonds between the oligosaccharides of glycoproteins or glycolipids and the non-reducing end of sialic acid. [4]. The NA of the influenza virus specifically hydrolyses α-2,3-sialic acid from a galactose moiety at the site active and less efficiently the α-2,6-sialic acid-galactocyl moiety [5]. The extracts of n-hexane and EtOAc were further fractionated to enable the identification of specific active compounds. Fractionation of the n-hexane extract gave five fractions (F1-F5). F3, F4 and F5 showed the ability to inhibit both C. perfringens NA and H1N1-NA, as shown in Figure 2. F3, F4 and F5 were further fractionated to obtain pure compounds. Friedeline (labelled as GC1) was obtained from F3 (101.5 mg) and the identification of this compound was in concordance with several other previously reported studies (see Supplementary Materials) [19][20][21]. Then, (24E)-3a, 9, 23-trihydroxy-17,14-friedolanostan-14,24-dien-26-oate labelled as GC2 (25.4 mg) were isolated from F4 and the structural profile is similar, as reported by Rukachaisirikul et al. (2000). Methyl-3α,23dihydroxy-17,14-friedolanstan-8,14,24-trien-26-oat (labelled as GC3) was isolated from F4 (11.1 mg) and F5 (32.1 mg). The spectroscopy analysis of the third compound (methyl-3α,23-dihydroxy-17,14friedolanstan-8,14,24-trien-26-oat) confirmed the identity of the compound, based on data from the previous study (see Supplementary Materials) [22].
As illustrated in Figure 3, GC1 was not active against C. perfringens-NA, whilst GC2 and GC3 [23] showed inhibition against C. perfringens-NA with a maximum inhibition of 79% (IC 50 = 81.78 µg/mL) and 62% (IC 50 = 142.90 µg/mL), respectively ( Figure 3). With regards to the activity on H 1 N 1 -NA, both GC2 and GC3 did not show any significant activity. The EtOAc extract was found to be more active against NA than the n-hexane extract. Fractionation of this extract (19.69 g) gave four fractions (SF1: 0.9 g, SF2: 5.2 g, SF3: 2.8 g, SF4: 6.1 g), and it was found that SF4 showed good IC50 values against C. perfringens-NA and H1N1-NA, as shown in Figure 4. Subsequently, GC4 was isolated from this fraction (12.8 mg). A spectroscopy analysis of GC4 confirmed that GC4 is a compound called catechin. This compound showed good NA inhibition ability with the IC50 of 60.29 µM for C. perfringens-NA and 90.59 µM for H1N1-NA respectively, as shown in Figure 5.  The EtOAc extract was found to be more active against NA than the n-hexane extract. Fractionation of this extract (19.69 g) gave four fractions (SF1: 0.9 g, SF2: 5.2 g, SF3: 2.8 g, SF4: 6.1 g), and it was found that SF4 showed good IC 50 values against C. perfringens-NA and H 1 N 1 -NA, as shown in Figure 4. Subsequently, GC4 was isolated from this fraction (12.8 mg). A spectroscopy analysis of GC4 confirmed that GC4 is a compound called catechin. This compound showed good NA inhibition ability with the IC 50 of 60.29 µM for C. perfringens-NA and 90.59 µM for H 1 N 1 -NA respectively, as shown in Figure 5.  The EtOAc extract was found to be more active against NA than the n-hexane extract. Fractionation of this extract (19.69 g) gave four fractions (SF1: 0.9 g, SF2: 5.2 g, SF3: 2.8 g, SF4: 6.1 g), and it was found that SF4 showed good IC50 values against C. perfringens-NA and H1N1-NA, as shown in Figure 4. Subsequently, GC4 was isolated from this fraction (12.8 mg). A spectroscopy analysis of GC4 confirmed that GC4 is a compound called catechin. This compound showed good NA inhibition ability with the IC50 of 60.29 µM for C. perfringens-NA and 90.59 µM for H1N1-NA respectively, as shown in Figure 5.

Binding Interaction of the Isolated Compound from Garcinia celebica
As shown in Table 1, GC1 was found to be inactive against NA and this may be attributed to the absence of a hydrogen bond donor, such as hydroxyl moieties, in the molecular structure. This finding was parallel to the result of the molecular docking study, in which GC1 was found to have low docking favorability in NA. This will be further explained in the following subsection.

Binding Interaction of the Isolated Compound from Garcinia celebica
As shown in Table 1, GC1 was found to be inactive against NA and this may be attributed to the absence of a hydrogen bond donor, such as hydroxyl moieties, in the molecular structure. This finding was parallel to the result of the molecular docking study, in which GC1 was found to have low docking favorability in NA. This will be further explained in the following subsection.
The compounds GC2 and GC3 are friedolanostane derivatives. The presence of these compounds in Garcinia sp. has been reported previously in the literature [24,25]. Viera et al. (2004) reported on the isolation of 11 friedolanostane-related compounds from Garcinia speciosa leaves [26]. Five other friedolanostanes were isolated by Rukachaisirikul et al. (2005) from Garcinia hombroniana leaves [22], whilst two friedolanostane compounds were reported by Klaiklay et al. (2013) from the twigs of Garcinia hombroniana [24]. Nguyen et al. (2011), on the other hand, reported on the isolation of eight friedolanostane compounds from Garcinia benthami bark and leaves [27].
The ester functional group present in GC2 and GC3 might play an important role in increasing the activity of the molecules on NA. Experimentally, GC3 showed an IC 50 of more than 100 µg/mL. In the molecular docking study, the skeleton of GC3 (ring A, B, and C) was found to be positioned close to the hydrophobic pocket, as shown in Figure 6b. GC3 formed a hydrogen bond between 23-OH (from GC3) and Tyr406 from the enzyme, and did not form any interaction with the arginine triad. Thus, this compound is expected to be less active than the isolated flavonoid. GC2 showed a better interaction with NA as compared to GC3. It docked well with a free energy of binding, FEB of −10 kcal/mol. The presence of hydroxyl group at C-9 made the skeleton of triterpene more flexible. The ester group of GC2 interacted well with the arginine triad, as shown in Figure 6a. Two oxygens from the ester group accepted protons from Arg118, Arg371, and Arg292, while the 23-OH moiety donated a proton to Asp151 from Loop150. In addition, ring B and C of GC2 were positioned close to the hydrophobic pocket (Ile222, Arg224, and Ser246) and this is the reason why GC2 has the lowest FEB and high fit value to map with T2S202 model. Unfortunately, the activity of GC2 on C. perfringens-NA was classified as less active (IC 50 81.72 µg/mL or 123.26 µM). Similarly, as implied from the experimental IC 50 , the GC2 activity against H 1 N 1 -NA was not as good as the predicted activity based on the free energy of binding. One possible reason for this is the low solubility of this compound, which might have inversely affected the bioassay result. In drug discovery and development, the solubility of active compounds has a big influence on the administration, distribution, metabolism and excretion (ADME) characteristics of a particular compound [28,29]. Based on Table 2, LogP of GC2 and GC3 were 5.18 and 6.14 respectively, as predicted by the software (DS 2.5), and thus, they were categorized as having "poor" drug-like properties. Lipiski et al. (2012) [30] predicted that poor absorption or permeation is more likely once logP is greater than five [31,32]. In this study, GC2 and GC3 were dissolved in a slightly higher concentration of DMSO (2.5%), due to low solubility and precipitation that might have occurred when the MES buffer was added. Poor solubility may also cause other problems, including poor bioavailability in oral administration, difficulty in formulation, lack of efficacy, high toxicity, expensive and prolonged development, and the need for multiple daily doses [28,29].
The ester functional group present in GC2 and GC3 might play an important role in increasing the activity of the molecules on NA. Experimentally, GC3 showed an IC50 of more than 100 µg/mL. In the molecular docking study, the skeleton of GC3 (ring A, B, and C) was found to be positioned close to the hydrophobic pocket, as shown in Figure 6b. GC3 formed a hydrogen bond between 23-OH (from GC3) and Tyr406 from the enzyme, and did not form any interaction with the arginine triad. Thus, this compound is expected to be less active than the isolated flavonoid. GC2 showed a better interaction with NA as compared to GC3. It docked well with a free energy of binding, FEB of −10 kcal/mol. The presence of hydroxyl group at C-9 made the skeleton of triterpene more flexible. The ester group of GC2 interacted well with the arginine triad, as shown in Figure 6a. Two oxygens from the ester group accepted protons from Arg118, Arg371, and Arg292, while the 23-OH moiety donated a proton to Asp151 from Loop150. In addition, ring B and C of GC2 Catechin or GC4 was found to be the most active as a NA inhibitor, compared to the other three isolated compounds and DANA (2,3-didehydro-2-deoxy-N-acetylneuraminic acid), as a commercial inhibitor. The IC 50 of GC4 against C. perfringens-NA and H 1 N 1 -NA were 17.48 µg/mL (60.27 µM) and 26.29 µg/mL (90.95 µM) respectively and, thus, this compound can be classified as moderately active.
In this study, the molecular interaction of catechin and H 1 N 1 -NA (PDB code: 3B7E [33]) was investigated. Catechin favorably docked onto NA at the 2-catechol moiety (ring C), and it interacted well with the arginine triad through hydrogen bond and pi-cation interactions with binding energy −12 kcal/mol. As shown in Figure 7, it appeared that the compound did not form any interaction with the hydrophobic pocket (Ileu222, Arg224, and Ser246), but interacted through hydrogen bonding with Glu276. The 3-OH moiety of catechin formed a hydrogen bond with Trp178 (2.3 Å) and it is linked to 3-gallocyl to form epicatechin gallate (ECG).  [30] predicted that poor absorption or permeation is more likely once logP is greater than five [31,32]. In this study, GC2 and GC3 were dissolved in a slightly higher concentration of DMSO (2.5%), due to low solubility and precipitation that might have occurred when the MES buffer was added. Poor solubility may also cause other problems, including poor bioavailability in oral administration, difficulty in formulation, lack of efficacy, high toxicity, expensive and prolonged development, and the need for multiple daily doses [28,29]. GC2 compound [28,29]. Based on Table 2, LogP of GC2 and GC3 were 5.18 and 6.14 respectively, as predicted by the software (DS 2.5), and thus, they were categorized as having "poor" drug-like properties. Lipiski et al. (2012) [30] predicted that poor absorption or permeation is more likely once logP is greater than five [31,32]. In this study, GC2 and GC3 were dissolved in a slightly higher concentration of DMSO (2.5%), due to low solubility and precipitation that might have occurred when the MES buffer was added. Poor solubility may also cause other problems, including poor bioavailability in oral administration, difficulty in formulation, lack of efficacy, high toxicity, expensive and prolonged development, and the need for multiple daily doses [28,29]. logP is greater than five [31,32]. In this study, GC2 and GC3 were dissolved in a slightly higher concentration of DMSO (2.5%), due to low solubility and precipitation that might have occurred when the MES buffer was added. Poor solubility may also cause other problems, including poor bioavailability in oral administration, difficulty in formulation, lack of efficacy, high toxicity, expensive and prolonged development, and the need for multiple daily doses [28,29]. logP is greater than five [31,32]. In this study, GC2 and GC3 were dissolved in a slightly higher concentration of DMSO (2.5%), due to low solubility and precipitation that might have occurred when the MES buffer was added. Poor solubility may also cause other problems, including poor bioavailability in oral administration, difficulty in formulation, lack of efficacy, high toxicity, expensive and prolonged development, and the need for multiple daily doses [28,29]. Catechin or GC4 was found to be the most active as a NA inhibitor, compared to the other three isolated compounds and DANA (2,3-didehydro-2-deoxy-N-acetylneuraminic acid), as a commercial inhibitor. The IC50 of GC4 against C. perfringens-NA and H1N1-NA were 17.48 µg/mL (60.27 µM) and 26.29 µg/mL (90.95 µM) respectively and, thus, this compound can be classified as moderately active. In this study, the molecular interaction of catechin and H1N1-NA (PDB code: 3B7E [33]) was investigated. Catechin favorably docked onto NA at the 2-catechol moiety (ring C), and it interacted well with the arginine triad through hydrogen bond and pi-cation interactions with binding energy -12 kcal/mol. As shown in Figure 7, it appeared that the compound did not form any interaction with the hydrophobic pocket (Ileu222, Arg224, and Ser246), but interacted through hydrogen bonding with Glu276. The 3-OH moiety of catechin formed a hydrogen bond with Trp178 (2.3 Å) and it is linked to 3-gallocyl to form epicatechin gallate (ECG).

Discussion
Four compounds were isolated from GCL as listed in Table 1. They were obtained based on the results of NA inhibition by fractions recuperated from GCL extracts. Two of the three triterpenoids showed satisfactory inhibition against C. perfringens-NA, but they were less active against H1N1-NA (GC2 and GC3). This may be attributed to the nature of n-hexane extracts (hydrophobic molecules), which usually showed low activity against NA because of their low water solubility characteristics.

Discussion
Four compounds were isolated from GCL as listed in Table 1. They were obtained based on the results of NA inhibition by fractions recuperated from GCL extracts. Two of the three triterpenoids showed satisfactory inhibition against C. perfringens-NA, but they were less active against H 1 N 1 -NA (GC2 and GC3). This may be attributed to the nature of n-hexane extracts (hydrophobic molecules), which usually showed low activity against NA because of their low water solubility characteristics. Solubility is a very important factor that influences the inhibitory activity of compounds against NA [34]. Although this factor was recognized as a limitation in this current study, the discovery of active triterpenoids from GCL has never been reported before and this is an important finding that needs to be recorded and reported. The activity of these triterpenoids against NA is an important finding that could lead to the development of new actives through further simulations and synthetic chemistry.
GC4 showed moderate activity against both C. perfringens-NA and H 1 N 1 -NA. This flavonoid was confirmed as catechin and was obtained from EtOAc extracts. This compound was obtained from the fractionation results, and found that F4, which has inhibitory activity against NA, was best compared to other fractions. Catechin was already being reported as present in some Garcinia sp. such as G. kola [35] and G. penangiana [36]. However, to the best of the authors' knowledge, there are no reports precising the presence of this compound in G. celebica. Catechin was previously evaluated in vitro for its anti-influenza properties, and it showed good inhibition of influenza virus replication [37,38]. Another group of researchers investigated the ability of catechin-containing herbal tea to halt influenza virus infection in residents of a nursing home for the elderly, and they reported positive results [39]. Kuzuhara et al. (2009) in their publication explained that catechin inhibited the endonuclease activity of RNA polymerase in influenza A virus, thus this compound has a big potential to be further developed as an anti-influenza A drug [40]. Its action against the influenza virus could also be attributed to its antioxidant property. Liu et al. (2008) have discussed the anti-influenza activity of catechin, but the mechanism of action of this molecule at molecular level was not investigated [41]. Shan et al. (2012) proposed that the 4-chromanone moiety in catechin is responsible for its NA inhibition activity [42]. Uchide and Toyoda (2011) noted that the activity of ECG as an influenza virus inhibitor is contributed to mainly by the 3-gallocyl moiety of this compound, whereas the 5'-OH at the trihydroxybenzyl moiety at the 2-position plays a minor role. The presence of the hydroxyl group on C-5 played a critical role in the inhibition of NA [43].
The antioxidant property of catechin means that this molecule could scavenge the superoxide anion and hydroxyl radicals [44]. The orientation of the 4-chrommanone ring allowed catechol moiety to rotate, thus it could interact with the triad arginine residues (Asp151, Arg 292 and Arg 371). Three arginine residues (Arg 118, Arg 292, and Arg 371) and a glutamate residue (Glu 276) have an important role in the binding of sialic acid in the active site of NA [45]. These results are in line with a study conducted by Muller et al. [46], in which the phenyl ring of 4-chromanone moiety was favored by the Ile427 and Lys432 residues that formed the hydrophobic pocket of NA. However, this was not seen in the molecular docking conducted in this study, and instead it appeared to interact with Ileu222, Arg224, and Ser246 as the hydrophobic pocket.
The solvent was evaporated under reduced pressure to yield a concentrated methanol extract (179.8 g). The mixture of MeOH-water was filtered and further partitioned with n-hexane and EtOAc successively to give n-hexane and EtOAc fractions (26.9 g and 46.5 g, respectively).

Isolation of Compounds from EtOAc Extract of GCL
The crude extract from EtOAc (19.7 g) was subjected to gravity column chromatography (5 × 30 cm) with CHCl 3 /MeOH in a stepwise manner at 10%, producing four fractions (SF1, SF2, SF3, and SF4). The fractions were assayed for their NA inhibition, and F3 and F4 were selected because of their good activity (more than 50% inhibition) against C. perfringens NA. Further assay works with H 1 N 1 NA were conducted, in which SF4 showed up to 90% inhibition. SF4 (2.6 g) was subsequently subjected to small column chromatography (1 × 20 cm) with a CHCl 3 /MeOH (88:12) solvent system, to produce 76 fractions. Fraction 39-42 (831.1 mg) was further purified through semi-preparative liquid chromatography (three times) to obtain GC4 (12.8 mg).

General Experiments and Spectroscopy Methods
1 H NMR and 13 C NMR spectra were both recorded with a BRUKER AVANCE III 500 MHz spectrometer. Mass spectra were measured on an Agilent 1100 Series LC-MSD-Trap-VL spectrometer by using electrospray ionisation as the type of ion source. FTIR spectra were recorded using an IR-Prestige-21 (Shimadzu) spectrometer. Melting points were obtained by using an electrothermal melting point apparatus (STUART-SMP10). UV spectra were determined on an UV-Vis spectrophotometer (Analytical Jena, specord-200). Rotation index was determined using ADP 120 Bellingham (Stanley Ltd., Tokyo, Japan) The complete spectral data is provided in Supplementary Materials.

Neuraminidase (NA) Activity
NA was prepared in 2-(N-morpholino) ethanesulfonic acid (MES) buffer (Sigma ® ) to get a concentration of 0.3 µ/mL. The substrate MUNANA was prepared in the same buffer to get a concentration of 100 µM. The G. celebica leaves' extracts, fractions, and isolated compounds were prepared in 2.5% DMSO (Merck ® ), due to the solubility problem in concentrations between 7.8125 to 125 µg/mL. The times of incubation (agitated at 200 rpm, 37 • C) for the mixture of NA-coffee and NA-coffee-MUNANA were 30 min and 60 min, respectively, and the reaction was stopped by using glycine before reading. NA activity towards inhibitors was measured via a fluorogenic substrate, MUNANA, excited at 365 nm, with fluorescence emission at 450 nm, by using an ELISA microplate reader (Tecan-i-control infinite 200Pro) [47]. The data results were analyzed by GraphPad Prism 5.0.

Molecular Docking Simulation
Molecular docking methods were adopted from the previous study [48]. The NA protein of subtype N1 in complex with zanamivir (PDB code: 3B7E [33]) was used as the target. Molecular docking simulations were performed with AutoDock 4.2 [49].

Conclusions
Friedeline, catechin and two lanastone derivatives (methyl-3α, 23-dihydroxy-17,14-friedolanstan-8,14,24-trien-26-oat and 24E-3a,9,23-trihydroxy-17,14-friedolanostan-14,24-dien-26-oate) were obtained from G. celebica leaves by using bioassay-guided isolation. Based on the enzyme inhibition study, the two lanastone derivatives showed low activity on NA while friedeline was inactive. Catechin, on the other hand, showed the highest activity as a NA inhibitor compared to the other three compounds. On the contrary, a molecular docking study showed that the two lanastone derivatives have a good docking profile on the binding site of NA. This may be due to the solubility problem as discussed earlier, which may have inversely affected the assay performance of the compounds. Another possible reason may be the fact that although the compounds docked well, they may not have had sufficient time to exert their inhibitory effect on the enzyme, hence the IC 50 values of these compounds were high. From this study, it is suggested that the development of catechin as an anti-influenza agent would be valuable, but further structure modification may be needed to improve its inhibition activity.