Anti-SARS-CoV-2 Activity and Cytotoxicity of Amaryllidaceae Alkaloids from Hymenocallis littoralis

The Amaryllidaceae species are well-known as a rich source of bioactive compounds in nature. Although Hymenocallis littoralis has been studied for decades, its polar components were rarely explored. The current phytochemical investigation of Amaryllidaceae alkaloids from H. littoralis led to the identification of three previously undescribed compounds: O-demethyl-norlycoramine (1), (−)-2-epi-pseudolycorine (2) and (+)-2-epi-pseudolycorine (3), together with eight known compounds: 6α-hydroxyhippeastidine (4), 6β-hydroxyhippeastidine (5), lycorine (6), 2-epi-lycorine (7), zephyranthine (8), ungeremine (9), pancratistatin (10) and 9-O-demethyl-7-O-methyllycorenine (11). Among the eight previously reported compounds, five were isolated from H. littoralis for the first time (compounds 4, 5, 7, 8, and 9). Compounds 1, 4, 5, 7, 8, and 11 exhibited weak anti-SARS-CoV-2 activity (EC50 = 40–77 µM) at non-cytotoxic concentrations. Assessment of cytotoxicity on the Vero-E6 cell line revealed lycorine and pancratistatin as cytotoxic substances with CC50 values of 1.2 µM and 0.13 µM, respectively. The preliminary structure-activity relationship for the lycorine-type alkaloids in this study was further investigated, and as a result ring C appears to play a crucial role in their anti-SARS-CoV-2 activity.


Introduction
Amaryllidaceae alkaloids (AAs) belong to the large group of isoquinoline alkaloids, and they are almost exclusively found in the Amaryllidaceae family [1,2]. These alkaloids show a wide variety of pharmacological activities, such as anticancer, antiplasmodial, anti-inflammatory, antioxidant, etc. One of them, galanthamine, is a worldwide registered drug for the treatment of cognitive decline in mild to moderate Alzheimer's disease and various other memory impairments [3,4]. To date, over 600 Amaryllidaceae alkaloids have been reported up to the end of 2018, and the number is still increasing [5].
Hymenocallis littoralis (Jacq.) Salisb. or 'spider lily' (Amaryllidaceae) is widely distributed by the sea and in swamps in tropical, sub-tropical and temperate climates throughout the world [6]. Alongside its medicinal value, it is also known for its horticultural and ornamental appeal [7]. This plant species has been the subject of research for decades and up to now, to our knowledge, 37 alkaloids have been isolated from H. littoralis, including four lycorine-type alkaloids (lycorine, littoraline, diacetyllycorine, norpluviine), four lycorenine-type alkaloids (lycorenine, homolycorine, hippeastrine, 6-O-methyllycorenine), One of the common 1 H NMR spectroscopic patterns for Amaryllidaceae alkaloids is the presence of a methylenedioxy proton resonance around 5.95 ppm (-O-CH 2 -O-signal) [26,27]. In the absence of this characteristic signal, substituents such as hydroxy-or methoxy groups are prevalent instead of the methylenedioxy moiety. Using this NMR pattern as a starting point, five compounds (6-10) were found to possess the methylenedioxy moiety, and five compounds (2)(3)(4)(5)11) were found to bear other substituents ( Figure 1). The detailed structure elucidation procedure of the three unreported compounds (1-3) is described hereafter, and the NMR data of known compounds can be consulted in the Supporting Information.
Compound 1 was obtained as an off-white and amorphous solid. HRMS-based elemental analysis suggested a chemical formula of C 15 15 H 20 NO 3 ). Its 1 H NMR spectrum showed two ortho-oriented aromatic hydrogens (δ H 6.51 and δ H 6.55, J = 8.0 Hz), which are typical for the galanthamine-type skeleton [27]. Indeed, the 1 H and 13 C NMR spectra were similar to those reported for galanthamine, but two clear differences could be observed: (1)-Compound 1 does not have two olefinic protons around 6 ppm like galanthamine, implying the absence of the double bond in the structure of galanthamine, and (2)-Compound 1 does not show any signals typical for -OCH 3 or -NCH 3 groups, indicating the absence of the methoxy and the N-methyl groups in the structure of galanthamine. Therefore, compound 1 was identified as O-demethyl-norlycoramine (or 1,2-dihydro-O-demethyl-N-demethylgalanthamine). Correlations of the 2D NMR spectra were in agreement with the proposed structure ( Figure 2). Four methylene groups resonated in the upfield region (seven protons in the range of 1.4-2.0 ppm). In COSY, correlations between H-1 and H-2, H-2 and H-3, H-3 and H-4, and H-4 and H-4a confirmed that they are part of the same spin system (see Figure 2). Similarly, another spin system was observed for H-11 and H-12 on COSY. These COSY correlations were crucial to correctly assign the positions of the overlapping proton signals (see Table 1). Assignment of H-7 was confirmed by the NOE correlations between H-7 and H-6. Key HMBC correlations listed as follows aided in confirming the 2D structure ( Figure 2): C-10b with H-2, H-4a, H-12; C-1 with H-4a; C-12 with H-6; C-7 with H-6; C-9 with H-7; C-10a with H-7. Finally, the molecular formula C 15 H 19 NO 3 was confirmed by HRMS-based elemental analysis. Remarkably, a search based on the chemical structure of this compound gave an exact match on Scifinder (checked on 24 March 2023), but not a single reference reporting this compound could be found. Thus, to the best of our knowledge, compound 1 was reported for the first time as a natural product in this study. Biosynthetically, since the substituent of C-9 is a phenolic hydroxyl instead of a methoxy, compound 1 could originate directly from a para-ortho' oxidative phenol coupling of norbelladine [5,28]. Enzymes responsible for methylation might not be involved in its biosynthetic pathway ( Figure 3).   Owing to the stereocenter at C-10b, the 3D structure of compound 1 does not show a planar scaffold, from which two sides of a plane can easily be distinguished, like compounds 2 and 3 (see Figure 2). The NOESY spectrum of compound 1 showed strong cross-peaks between δ H -1.67, δ H -1.85, δ H -3.97 and δ H -3.19, indicating their spatial proximity. Next, the NOE effect observed between H-4a (δ H -4.31) and δ H -1.75 implied that they are spatially close to each other. Another important NOE effect was observed between δ H -1.67 and H-3 (δ H -4.09), indicating δ H -1.67, δ H -4.09, δ H -4.31 and δ H -1.75, and δ H -3.26 are co-facial (see Table 1 and Figure 2). Finally, the configuration of the 3-OH was assigned on the same side with δ H -1.67 due to a weak NOESY cross-peak between them. Combining all data together, the relative configuration of compound 1 was proposed as (3S,4aS,10bR)-Odemethyl-norlycoramine or its enantiomer.
As a follow-up, DP4+ probability was assessed to compare the closeness between the experimental and calculated chemical shifts of all possible diastereomers. The computed diastereomers were chosen as follows: For compound 1, the presence of three stereocenters infers eight possible configurations (four pairs of enantiomers). As the DP4+ method can only be used to assign a relative configuration, due to the incapacity of NMR to determine absolute configuration, the chemical shift values of only one member of every pair of enantiomers needed to be computed. As shown in Table 2, the highest probability was found for diastereomer (4aS,3S,10bR), which is in line with the configuration deduced from NOE correlations. Therefore, the relative configuration of compound 1 was confirmed, and the absolute configuration is either (4aS,3S,10bR) or (4aR,3R,10bS). It is important to state that the probability of 73% alone is not conclusive and validation by interatomic distances obtained through the NOESY experiment was required. This practice was recommended by the authors of a recent review on the topic of DP4+ probability [29]. As listed in Table 1, calculated chemical shifts of the diastereomer (4aS,3S,10bR) showed a strong resemblance with the experimental ones. In addition, this ascertained that the proposed planar structure was correct. Table 2. DP4+ probabilities of compounds 1-3 and computed J-couplings of compounds 2 and 3 estimated using CD 3 OD as solvation model. For compound 1, stereocenters' order is (4a,3,10b). For compounds 2 and 3, stereocenters' order is (1,2,4a,10b). Next, specific optical rotations of the two configurations (4aS,3S,10bR) and (4aR,3R,10bS) were computed. Since a pair of enantiomers will have opposite optical activity, the experimental optical rotation was expected to match with only one of the two. According to Table 3, the configuration (4aS,3S,10bR) is levorotatory, and the configuration (4aR,3R,10bS) is dextrorotatory. Apart from the sign, there is a noticeable difference between the magnitudes of the experimental and the computed OR of the configuration (4aS,3S,10bR), which could be due to impurities in the sample (see 1 H spectrum in Supplementary Materials). However, since the purity of the sample is approximately 90%, it can be assumed that compound 1 determines the overall sign of the sample. Therefore, the absolute configuration of compound 1 is defined as (4aS,3S,10bR). The NMR spectra were initially recorded in CD 3 OD and showed the general characteristics of lycorine-type alkaloids [27]. In the downfield region, two singlets corresponding to the para-oriented aromatic protons appeared at δ H greater than 6.0 ppm (H-10/6.88 ppm and H-7/6.76 ppm), together with a broad singlet at δ H 5-6 ppm (H-3/5.59 ppm), which represented the olefinic proton attached to C-3. Furthermore, similar to many lycorine-type alkaloids, compounds 2 and 3 also showed two doublets corresponding to an AB system of the benzylic methylene at C-6 (H-6α and H-6β, J = 14.1 Hz) [27]. Taking into consideration the fact that compounds 2 and 3 do not have a methylenedioxy moiety as discussed above, and the fact that a single methoxy-group (δ H 3.84 and δ C 56.5) is present, compounds 2 and 3 appear to have one hydroxy-and one methoxy-group in positions 8 and 9, instead of the methylenedioxy moiety. The position of the methoxy-group was deduced from a NOESY cross-peak between δ H 3.84 and H-7 (δ H 6.76) and an HMBC correlation between δ H 3.84 and δ C 147.8 ( Figure 2). Another two key HMBC correlations were between H-7 and C-6 and between H-10 and C-10b. Furthermore, the COSY spectrum showed strong cross-peaks between three couples of protons: H-2 and H-1, H-1 and H-10b, and H-10b and H-4a ( Figure 2). Taking these data together and by comparison with reported NMR data of lycorine-type alkaloids, it was found that the planar structures of compounds 2 and 3 were identical to the structure of pseudolycorine. This was confirmed by the molecular formula C 16 H 19 NO 4 obtained by HRMS-based elemental analysis. However, several hydrogens of compounds 2 and 3 at the upfield region (2.50-4.00 ppm) were more deshielded than those of pseudolycorine ( Figure S19), which could be indicative for a differing stereochemistry. DP4+ probability was examined thereafter for compounds 2 and 3 in order to obtain proof for their relative configuration. Nonetheless, no decisive probability (90-100%) was observed. Although the SRRS configuration had the highest probability (61%), the dihedral angle between H-10b and H-4a of this configuration can only produce a J H-10b/H-4a of 5.52 Hz (Table 2). Therefore, using the experimental J H-10b/H-4a of 11.0 Hz as a constraint, the SRRS configuration was ruled out, and based on the calculated J-couplings, SRRR, SRSS, SSRR or SSSS could be the possible configuration (Table 2). Since SSSS is the known configuration of pseudolycorine, only three possibilities remained: SRRR, SRSS and SSRR. Although these three configurations possessed a DP4+ probability of 0% (Table 2), the use of J H-10b/H-4a as a restriction in this case is of more reliable since it is well-known that a transconjunction between two fused rings of lycorine-type skeleton exhibits a typical J-coupling of 10-11 Hz. 27 The DP4+ probability, however, only takes into account experimental and calculated chemical shifts as inputs, and the coupling constants are not incorporated in the stereochemistry determination by DP4+ [30]. In addition, the accuracy of DP4+ probability can be affected by many reasons and unsuccessful examples were also reported by the authors of this method [29].

Compound
Further analysis of NOE interactions also inferred that compounds 2 and 3 are diastereomers of pseudolycorine. As can be seen in Figure 2, spatial proximities were observed for H-1, H-2 and H-10b in the NOESY spectrum, indicating that these three hydrogens are on the same side of the plane. H-1 and H-2 appeared as two broad singlets with no noticeable splitting, suggesting a cis-orientation. In addition, H-10b and H-4a shared a J-coupling of 11.1 Hz, typical for the trans-orientation between these two hydrogens reported for the lycorine-type scaffold as mentioned above [27]. Therefore, combining NOE correlations and J-coupling information, it can be deduced that H-1, H-2 and H-10b were on the same side and were on the opposite side of H-4a. Among the three remaining possibilities as discussed above, the SRSS is the sole configuration that can show all these NOESY correlations. Hence, the relative configuration of compounds 2 and 3 was established as (1S,2R,4aS,10bS)pseudolycorine, or its enantiomer. The structures of compounds 2 and 3 were elucidated with the configuration of one chiral center (at position H-2) differing from the configuration reported for pseudolycorine, and thus, correspond to 2-epi-pseudolycorine. To be noticed, C-4 was not observed in CD 3 OD, whereas it could be observed in (CD 3 ) 2 SO in both HMBC and 13 C-NMR spectra. Table 4 shows NMR data of compounds 2 and 3 in CD 3 OD and (CD 3 ) 2 SO. As reported in Table 3, the experimental optical rotation of 2-epi-pseudolycorine (compounds 2 and 3) was only −8.0 and was significantly different from the computed values, ranging from 227.9 to 230.6, if dextrorotatory, and from −230.4 to −227.7, if levorotatory. Based on these data, it was not possible to deduce the absolute configuration of compounds 2 and 3. At this point, it was hypothesized that the sample consisted of a racemic mixture. The fact that the measured value is not 0, but −8, could be caused by the presence of some impurities and/or the measurement or weighing errors due to the small amount of material. In order to determine whether the sample consisted of a racemate, an enantioseparation was performed, and the obtained HPLC chromatogram of the mixture of compounds 2 and 3 indeed showed two peaks with almost equal peak areas ( Figure S20). Hence, the mixture of compounds 2 and 3 was confirmed to be a racemate. Compounds 2 and 3, occurring as an enantiomeric mixture, were reported for the first time and the name 2-epi-pseudolycorine was adopted. Vibrational circular dichroism (VCD) spectrum of compounds 2 and 3 in (CD 3 ) 2 SO was devoid of any signal, confirming the existence of a racemic mixture ( Figure S23). Finally, the assignment of (−)-2-epi-pseudolycorine (2) to the configuration (1S, 2R, 4aS, 10bS) and (+)-2-epi-pseudolycorine (3) to the configuration (1R, 2S, 4aR, 10bR) was made according to their simulated optical rotations (see Table 3). Cytotoxicity   Compounds 1, 4, 5, 7, 8, and 11 (O-demethyl-norlycoramine, 6α-hydroxyhippeastidine, 6β-hydroxyhippeastidine, 2-epi-lycorine, zephyranthine, and 9-O-demethyl-7-O-methyllycorenine, respectively) exhibited weak inhibition of SARS-CoV-2 (EC 50 = 40-77 µM), at non-cytotoxic concentrations (CC 50 > 100 µM) (see Figure 4) . Compounds 1, 4 [33]. Although these antiviral assays also run in Vero-E6 cells, the RT-PCR read-out was after 18-24 h, whereas we used a high-content imaging read-out after 5 days, which might explain the increased cytotoxicity in our assay. Compounds 2, 3, and 9, ((±)-2-epi-pseudolycorine and ungeremine, respectively) lacked antiviral activity, as well as cytotoxicity. Detailed results can be found as Supplementary Materials.

Anti-SARS-CoV-2 Activity and
Preliminary structure-activity relationship (SAR) was examined for six lycorine-type alkaloids isolated in this study ( Figure 5). Using lycorine as the standing stone, 2-epilycorine only has one stereocenter at C-2 that differs from lycorine, but this entirely inverted its activity as well as its cytotoxicity compared to lycorine with an EC 50 of 54 µM at noncytotoxic concentration (CC 50 > 100 µM). This suggested that ligand-host interaction occurs at ring C in the structure of 2-epi-lycorine, and that the chirality of C-2 plays an important role in improving the potency and removing the cytotoxicity on the Vero-E6 cell line. Furthermore, an enhancement of activity was observed for zephyranthine (EC 50 = 39 µM, CC 50 > 100 µM) compared to 2-epi-lycorine implying that the double bond at C-3 and C-4 is not necessary for the activity. The disappearance of the double bond increases the flexibility of ring C, and thus could favorize the formation of interactions between two hydroxyl groups and residues at the binding site. On the other hand, the opening of the dioxymethylene moiety at C-8 and C-9 resulted in the loss of the antiviral activity of (±)-2-epi-pseudolycorine in comparison with 2-epi-lycorine. Looking closely at their structures, oxygen attached to C-9 can solely be a hydrogen-bonding acceptor when the dioxymethylene moiety remains, but 9-OH in the case of (±)-2-epi-pseudolycorine is a phenolic hydroxyl which could participate in an ion interaction and/or hydrogen bonding as either a hydrogen-bonding acceptor or a hydrogen-bonding donor. Therefore, 9-OH might generate unfavorable interactions with residues at the active site, which in turn cancels the antiviral activity. This suggested that ring A, especially substituent at C-9, might be involved in the pharmacophore for the antiviral activity. Another observation is that aromatization of rings B and C as can be seen in the structure of ungeremine eliminated cytotoxicity on the Vero-E6 cell line, but did not improve its anti-SARS-CoV-2 potency. Our findings are in agreement with previously reported SAR studies about anticancer activity and the inhibitory effects of lycorine-type alkaloids against dengue and hepatitis C viruses [34][35][36][37]. In general, ring C with substitution on C-1 and C-2 is of key importance for the activities and selectivity; and next comes to the role of substitution on ring A. However, with the limited structures in the current study, the effects of chirality at C-1 and different types of substituents are not covered. This remains an interesting question to be addressed in the future.
TLC was performed on pre-coated silica gel F254 plates (Merck, Darmstadt, Germany), and the spots were observed under UV light (254 and 366 nm). After subsequently spraying with the Dragendorff and NaNO 2 10% reagents, TLC plates were observed under visible light. Dragendorff reagent was prepared by combining a mixture A and a solution B. Mixture A was prepared by suspending 0.85 g of bismuth subnitrate (Bi 5 O(OH) 9 (NO 3 ) 4 ) in 40 mL of water and 10 mL of glacial acetic acid. Solution B was prepared by dissolving 8 g of potassium iodide (KI) in 20 mL of water.
The following instruments were used for fractionation, isolation and structure elucidation. For flash chromatography, a Reveleris iES system from Grace (Columbia, MD, USA) using the Reveleris Navigator TM v.1 software and commercially packed Claricep flash columns containing 40 g irregular deactivated silica gel (Agela Technologies, Wilmington, DE, USA). For analytical purposes, an Agilent 1200 series HPLC-DAD system (Agilent Technologies, Santa Clara, CA, USA) equipped with OpenLAB v.A.01.05 software was used, together with a Phenomenex Kinetex EVO C18 (250 × 4.6 mm, 5 µm) column. For purification, a preparative HPLC system coupled with DAD and MS detectors was used. All compartments of this system were supplied by Waters (Milford, MA, USA), including Mass-Lynx v.4.1 software. Separation was carried out on a semi-preparative Phenomenex Kinetex C18 column (250 × 10 mm, 5 µm). NMR spectra were recorded on two different NMR instruments (400 MHz): an Avance Nanobay III and a DRX-400 system (Bruker BioSpin, Rheinstetten, Germany). NMR data processing was performed with TopSpin v.4.0.6 from Bruker. Finally, accurate mass measurements were conducted for all isolated compounds using a Xevo G2-XS QToF mass spectrometer (Waters, Milford, MA, USA) coupled to an Acquity UPLC system and an Acquity HSS T3 UPLC column (100 × 2.1 mm, 1.8 µm) was used. Optical rotation were recorded on a Jasco P-2000 spectropolarimeter (Easton, MD, USA) at 589 nm (20 • C). The system was equipped with Spectra Manager TM software.

Plant Material
The fresh bulbs of Hymenocallis littoralis (Jacq.) Salisb. were collected in April 2017 in Dubréka (Guinea-Conakry) and provided by Prof. A. M. Baldé (University Gamal Abdel Nasser of Conakry, Guinea). The voucher specimen (No. 2aHK2) is kept at the Institute for Research and Development of Medicinal and Food Plants of Guinea, Dubréka. After cleaning and lyophilization, the total weight of the plant material was 3.3 kg.

Extraction and Isolation
Freeze-dried, powdered bulbs were macerated and percolated with approximately 30 L of 80% methanol. The extract was filtered and dried under reduced pressure. Then, liquid-liquid partition was applied in order to obtain subfractions. In short, the crude extract was suspended in water and was acidified with 10% HCl to a pH < 3 before performing liquid-liquid partition with DCM (I). Next, the pH of the acidified phase was increased to a pH ≥ 9 by adding NH 4 OH (25%), followed by a subsequent liquid-liquid partition with DCM (II), EtOAc and n-BuOH. In this way, four fractions (DCM (I), DCM (II), EtOAc and n-BuOH) were obtained and were examined by TLC with the mobile phase DCM-MeOH-NH 4 OH (90:10:1). Observation of the TLC plates after spraying with the Dragendorff + NaNO 2 10% reagents indicated that all fractions contained alkaloids, and the majority of alkaloids were present in the DCM (I) fraction.
The DCM (I) (3.1 g) and DCM (II) extracts (3.0 g) were fractionated by flash chromatography. After TLC investigation, silica gel was chosen as the suitable stationary phase. The percentage of NH 4 OH was fixed at 0.1% in the solvent systems. Other parameters were set as follows: (1)-Detectors: ELSD and UV (254 and 270 nm), (2)-Flow rate: 25 mL/min. The step-wise gradient used for fractionating the DCM (I) and DCM (II) extracts was: 0-5 min (0% MeOH), 5-15 min (5% MeOH), 15-30 min (10% MeOH), 30-40 min (15% MeOH), 40-45 min (20% MeOH), 45-50 min (25% MeOH), 55-60 min (70% MeOH), and the gradient was held when a peak was eluted. As for the fractionation of the EtOAc extract (2.2 g), column chromatography was used with a sample/stationary phase (silica gel) ratio of 1/50. Isocratic elution was applied, using DCM-MeOH-NH 4 OH (90/10/1, v/v) as the solvent system, by virtue of the result of TLC analysis. Finally, MeOH was used to elute any remaining compounds of interest from the column. The n-BuOH fraction (4.6 g) was also fractionated by column chromatography on a silica gel column. On the basis of TLC investigation, a three-step gradient was chosen to elute compounds from the column: With regard to the recrystallization, in the current work, impure crystals were first collected and were subsequently rinsed with DCM, DCM-MeOH (99:1, v/v) and DCM-MeOH (95:5, v/v) to remove impurities. Next, the crystals were dissolved in a minimal volume of DCM-MeOH (50:50, v/v). Finally, the solution obtained was left to dry in open air, resulting in the formation of solid crystals of the desired constituent. In this way, 1.0 g and 18.0 mg were purified of compounds 6 and 10, respectively. Compound 6 was found to be the most abundant alkaloid and was present in the DCM (II) and EtOAc extracts, and compound 10 was only purified from the EtOAc extract.
Compounds 1 (5.8 mg), 7 (1.5 mg), and 9 (1.5) were purified by silica gel column chromatography. Compound 1 was isolated from the n-BuOH extract and compounds 7 and 9 from the DCM (II) extract. The solvent systems used were deduced from TLC analysis.

Computational Details
Conformational analysis was performed by PCMODEL (version 10.0) using the Monte Carlo algorithm and the MMFF94 force field. All conformers within an energy window of 10 kcal.mol −1 were selected. Specific optical rotations were simulated for the sodium D-line wavelength (589.3 nm), as experimental optical rotations were measured at the same wavelength. The obtained conformers were first geometrically optimized in the gas phase. After dereplication, Boltzmann weighting was applied using the sum of electronic and thermal free energies at 298.15 K, and only conformers with energies within an energy window of 2.5 kcal.mol −1 from the global minimum were considered as the contributing ones.

Cells and Virus
VeroE6-eGFP cells (provided by Dr. K. Andries, J&JPRD, Beerse, Belgium) were maintained in DMEM (Gibco cat no. 41965-039) supplemented with heat-inactivated 10% FBS and 500 µg/mL Geneticin (Gibco cat no. 10131-0275) and kept in a humidified 5% CO 2 incubator at 37 • C. For the production of virus stocks and during virus experiments, the cells were maintained in an assay medium. The VeroE6-eGFP assay medium consisted of DMEM supplemented with 2% FBS.
The SARS-CoV-2 isolate used in this study was the BetaCov/Belgium/GHB-03021/2020 (EPI ISL407976|2020-02-03), which was isolated from a Belgian patient returning from Wuhan in February 2020. The isolate was passaged seven times on Vero-E6 cells, which introduced two series of amino acid deletions in the spike protein [42]. The infectious content of the virus stock was determined by titration on Vero-E6 cells. The SARS-CoV-2related work was conducted in the high-containment BSL3+ facilities of the KU Leuven Rega Institute (3CAPS) under licenses AMV 30112018 SBB 219 2018 0892 and AMV 23102017 SBB 219 2017 0589 according to institutional guidelines.

Antiviral Assay
The SARS-CoV-2 antiviral assay in Vero-E6 cells is derived from a previously established SARS-CoV-1 assay [43]. Stock solutions of the various compounds in DMSO (10 mM) were prepared. On day -1, the test compounds were serially diluted in assay medium and VeroE6-eGFP cells were plated corresponding to a final density of 25,000 cells per well in black 96-well plates (Greiner Bio-One, Vilvoorde, Belgium; Catalog 655090). The plates were incubated overnight (37 • C, 5% CO 2 and 95% relative humidity). On day 0, the cells with compound were infected with SARS-CoV-2 (at 20 CCID 50 per well). The plates were incubated in a humidified incubator at 37 • C and 5% CO 2 . At 4 days p.i., the wells were examined for eGFP expression using an argon laser-scanning microscope. The microscope settings were excitation at 488 nm and emission at 510 nm, and the fluorescence images of the wells were converted into signal values. The antiviral activity was expressed as EC 50 defined as the concentration of compound achieving 50% inhibition of the virus-reduced eGFP signals as compared to the untreated virus-infected control cells. The cytotoxicity of the compounds for VeroE6 cells in the absence of virus was evaluated in a standard MTS-assay as described previously [44].
In vitro anti-SARS-CoV-2 screening revealed weak anti-SARS-CoV-2 activity of Odemethyl-norlycoramine, 2-epi-lycorine, 6α-hydroxyhippeastidine, 6β-hydroxyhippeastidine, zephyranthine, and 9-O-demethyl-7-O-methyllycorenine (EC 50 = 40-77 µM) at non-cytotoxic concentrations (CC 50 > 100 µM). Lycorine and pancratistatin exhibited cytotoxicity with a CC 50 of 1.2 µM and 0.13 µM, respectively. Investigation of the structure-activity relationship for the lycorine-type alkaloids in this study suggested a crucial role of ring C: the meaningful ligand-host interactions appeared to be related to ring C; the spatial flexibility of ring C was required to exhibit anti-SARS-CoV-2 activity; the stereochemistry of C-2 on ring C might determine the activity and cytotoxicity. In view of the limited scope of this study, the chirality effect of C-2 was explored, but that of C-1 remains unknown. Therefore, further research on a bigger library of lycorine-type analogs with a wide chemical diversity is necessary. Besides, the potential of galanthamine-type and crinine-type scaffolds as anti-SARS-CoV-2 agents remains a compelling research topic.