Myrtucommulones and Related Acylphloroglucinols from Myrtaceae as a Promising Source of Multitarget SARS-CoV-2 Cycle Inhibitors

The LABEXTRACT plant extract bank, featuring diverse members of the Myrtaceae family from Brazilian hot spot regions, provides a promising avenue for bioprospection. Given the pivotal roles of the Spike protein and 3CLpro and PLpro proteases in SARS-CoV-2 infection, this study delves into the correlations between the Myrtaceae species from the Atlantic Forest and these targets, as well as an antiviral activity through both in vitro and in silico analyses. The results uncovered notable inhibitory effects, with Eugenia prasina and E. mosenii standing out, while E. mosenii proved to be multitarget, presenting inhibition values above 72% in the three targets analyzed. All extracts inhibited viral replication in Calu-3 cells (EC50 was lower than 8.3 µg·mL−1). Chemometric analyses, through LC-MS/MS, encompassing prediction models and molecular networking, identified potential active compounds, such as myrtucommulones, described in the literature for their antiviral activity. Docking analyses showed that one undescribed myrtucommulone (m/z 841 [M − H]−) had a higher fitness score when interacting with the targets of this study, including ACE2, Spike, PLpro and 3CLpro of SARS-CoV-2. Also, the study concludes that Myrtaceae extracts, particularly from E. mosenii and E. prasina, exhibit promising inhibitory effects against crucial stages in SARS-CoV-2 infection. Compounds like myrtucommulones emerge as potential anti-SARS-CoV-2 agents, warranting further exploration.


Introduction
Brazil holds areas of high natural wealth in terms of plant biodiversity, including two out of the thirty-four hotspots spread across the world: Cerrado and the Atlantic Forest.These ecoregions remain highly threatened, subject to rapid urbanization and high levels of resource exploitation [1].
The Myrtaceae family is commonly found in many of these biodiversity hotspots, such as southwestern Australia, as well as in the Brazilian Cerrado and Atlantic Forest [2], where Myrtaceae stands out as one of the most diverse families of woody plants [1].The LABEX-TRACT from the Federal University of Rio de Janeiro is a plant extract bank that holds a representative number of samples, containing species from the Atlantic Forest and the Amazon regions, which represent an outstanding source of potentially pharmacologically active compounds, as well as new chemical entities.
Several studies, by using both in silico and in vitro approaches, have demonstrated the relevance of the Brazilian flora in tackling SARS-CoV-2 infection and, therefore, the COVID-19 pandemic.Previous studies from our group have shown the capacity of flavonoids isolated from Siparuna cristata (Poepp.& Endl.) A. DC. to inhibit SARS-CoV-2 in vitro replication [3], as well as the capability of glycosylated phenylethanoid glycosides to act as 3-Quimiotripsin-like protease (3CL pro ) and Papain-like protease (PL pro ) inhibitors, of which forsythoside and verbascoside (abundant phenylpropanoids in Brazilian plants) were the most promising multitarget compounds [4].The potential of Ampelozizyphus amazonicus Ducke bark extracts against SARS-CoV-2, which is a plant traditionally used by Brazilian quilombola Amazonian riverine communities to treat COVID, has been demonstrated by their capacity to inhibit the interaction between the SARS-CoV-2 Spike protein and angiotensin-converting enzyme 2 (ACE2) receptor and, therefore, reduce viral replication in Calu-3 cells.This effect was attributed to phenolic compounds found in these extracts [5].
SARS-CoV-2's Spike protein, 3CL pro , and PL pro are significant viral targets for developing multitarget compounds to combat the virus [6].The Spike protein, responsible for viral-host cell fusion, is crucial for the virus's entry into the host cell [7].By targeting this protein, the interaction with the host's ACE2 receptor can be prevented, thereby inhibiting the entry of the virus [8].3CL pro , a homodimeric protease, is vital for the viral replication process.Its unique structure, comprising three domains and an active site with a catalytic dyad (HIS 41 and CYS 145), makes it an ideal target for antiviral drugs [9][10][11].PL pro , on the other hand, is a monomeric protease with a distinct active site formed by a catalytic triad (CYS 111, HIS 272, and ASP 286).Its enzymatic activity is integral for viral maturation and replication [12][13][14].
The Myrtaceae family has economic importance, as its species are commonly used as condiments, for essential oil extraction, and for wood and fruit production [15].Moreover, their valuable therapeutic properties, such as antibacterial, antifungal, antiviral, antioxidant, anti-inflammatory, cytotoxic, and antiproliferative activities, have stimulated research efforts to identify bioactive compounds in their extracts.These extracts are known to be rich in terpenes, polyphenols, and other unique acylphloroglucinol compounds named myrtucommulones [2,[16][17][18].These phloroglucinol derivatives can be chemically divided into two main subclasses: oligomeric acylphloroglucinols and phloroglucinol-terpene adducts [2], with a very large structural diversity.Among these compounds, the myrtucommulones are related to various biological activities.Several have shown significant activities against Epstein-Barr virus (EBV), herpes simplex virus (HSV), respiratory syncytial virus (RSV), and human immunodeficiency virus (HIV), and more recently, docking analysis have suggested that they can be effective against SARS-CoV-2 [19][20][21][22].
In this context, the objective of this work was to investigate through chemometric analyses correlations between the bioactive compounds present in extracts of the Myrtaceae family on the entry and replication mechanisms associated with the Spike protein, ACE2, and 3CL pro and PL pro proteases of SARS-CoV-2 in host cells through in vitro and in silico analysis.The species studied included Eugenia brasiliensis Lam., Eugenia prasina O. Berg., Eugenia mosenii (Kausel) Sobral, Myrcia strigipes Mart., and Myrcia splendens Sw.DC., all from Atlantic Forest and available in the LABEXTRACT.

Myrtaceae Extracts Inhibit RBD:ACE2 Complex Formation and SARS-CoV-2 Proteases Activity In Vitro
The ability of the extracts to inhibit the formation of RBD:ACE2 complex was assessed by Lumit™ kit (Promega), a bioluminescence-based kit suitable for the trial of potential inhibitory samples [23].At 250 µg•mL −1 , only the E. brasiliensis (12%) extract was not able to inhibit RBD:ACE2 interaction above 70% (Table 1).The extract of E. prasina stood out as the most active, showing an inhibitory rate of 96%, followed by E. mosenii extract (84%).
Table 1.Myrtaceae extracts and their effect on the entry and replication mechanisms associated with the Spike protein (250 µg•mL −1 ) in host cells and inhibition of 3CL pro and PL pro proteases (100 µg•mL −1 ) of SARS-CoV-2.The FRET-based assay was employed to measure SARS-CoV-2 3CL pro and PL pro proteolytic activities by using the fluorescence resonance energy transfer (FRET) approach.This assay has been applied to investigate potential enzyme activity inhibitors from both synthetic and natural sources [24].

Myrtaceae Extracts Inhibit SARS-CoV-2 Spike-Pseudotyped VSV Entry into VERO ACE2 Cells
Since most Myrtaceae extracts showed good inhibition of the RBD:ACE2 complex formation, a SARS-CoV-2 Spike-pseudotyped VSV that contains in the envelope the SARS-CoV-2 Spike protein and a VSV replication machinery with a luciferase reporter gene was used to measure the potential of Myrtaceae extracts to block SARS-CoV-2 Spikemediated cell entry.For this assay, Vero E6 cells that overexpress ACE2 (Vero E6/ACE2) were used.When the Spike pseudotyped viruses are incubated with cells containing the ACE2, it is possible for them to recognize the receptor and attach and enter the cells through Pharmaceuticals 2024, 17, 436 4 of 27 the binding between the Spike protein and the receptors on the cell surface.It had been previously shown using RBD:ACE2 complex formation assays that Myrtaceae extracts present the ability to inhibit the protein-protein interaction, suggesting that substances in the extracts could act as blockers of virus cell entry.In follow-up neutralization assays, using Spike-pseudotyped VSV, extracts were able to significantly block receptor-mediated cell entry (Figure 1A) at the highest tested concentration (250 µg•mL −1 ).The extract from E. mosenii seemed to be the most active, inhibiting cell entry up to 80% even at the lowest tested concentration (25 µg•mL −1 ) and presenting the lowest EC 50 value (<25 µg•mL −1 ) (Table 2).The extracts only induced cell toxicity in concentrations above 500 µg•mL −1 (Figure 1B).
formation, a SARS-CoV-2 Spike-pseudotyped VSV that contains in the envelope the SARS-CoV-2 Spike protein and a VSV replication machinery with a luciferase reporter gene was used to measure the potential of Myrtaceae extracts to block SARS-CoV-2 Spikemediated cell entry.For this assay, Vero E6 cells that overexpress ACE2 (Vero E6/ACE2) were used.When the Spike pseudotyped viruses are incubated with cells containing the ACE2, it is possible for them to recognize the receptor and attach and enter the cells through the binding between the Spike protein and the receptors on the cell surface.It had been previously shown using RBD:ACE2 complex formation assays that Myrtaceae extracts present the ability to inhibit the protein-protein interaction, suggesting that substances in the extracts could act as blockers of virus cell entry.In follow-up neutralization assays, using Spike-pseudotyped VSV, extracts were able to significantly block receptor-mediated cell entry (Figure 1A) at the highest tested concentration (250 µg•mL −1 ).The extract from E. mosenii seemed to be the most active, inhibiting cell entry up to 80% even at the lowest tested concentration (25 µg•mL −1 ) and presenting the lowest EC50 value (<25 µg•mL −1 ) (Table 2).The extracts only induced cell toxicity in concentrations above 500 µg•mL −1 (Figure 1B).

Chemometric Analyses Using Partial Least Squares Regression Prediction Model
To predict the potential active compounds that inhibit all the Spike:ACE2 interactions (Lumit™ kit) and r3CL pro and rPL pro proteolytic activities (FRET), the 15 ethanol extracts from the five species of the Myrtaceae family underwent liquid chromatography coupled to tandem mass spectrometry (LC-MS/MS) analysis using two ionization sources in both positive and negative modes (LC-ESI/APCI-(+/−)-MS/MS). Chemometric analyses were performed alongside binary values of activities in the three target models (excluding SARS-CoV-2 Spike-pseudotyped VSV entry assay).A Partial Least Squares (PLS) regression analysis model was constructed to correlate the LC-MS data with the inhibitory activity of the extracts.
The PLS regression used the LC-MS dataset of the 15 samples as independent variables (X), encompassing a total of 170 parameters (m/z ions), with the dependent variable (Y) being the inhibition activities against the three target models.This algorithm combines principal component analysis (PCA) and linear regression, seeking latent variables that maximize the correlation between predictors (m/z ions) and biological activity.Based on the PLS analyses, patterns in the sample distribution were identified, allowing classification into two groups based on average activity values across the three target models: more active samples (≥50% inhibition; active) and less active samples (<50% inhibition; non-active).
All four chemical datasets (ESI+/− and APCI+/−) for each plant extract were modeled, and the data obtained from negative electrospray ionization (ESI) showed the highest explained variance, R² and Q² values, and the lowest Root Mean Square Error (RMSE) in the four constructed models.Therefore, data from this ionization mode were selected to proceed with prediction analyses.At the outset, three models were developed to individually evaluate the Spike:ACE2, r3CL pro , and rPL pro targets.Within these models, our goal was to assess the influence of ions obtained through LC-MS/MS analyses in explaining the inhibitory activity response.Consequently, this approach enabled us to identify ions exhibiting the strongest correlation with the target under investigation, thereby providing insights into their potential roles in the conducted experimental assays.
In the investigation model of Spike:ACE2 interaction inhibitors, the explained variance was 99.7% utilizing five factors, which is higher than the values observed in the models for investigating r3CL pro inhibition (81% with six factors) and rPL pro inhibition (85% with four factors).Similar trends were observed in terms of accuracy and reproducibility parameters: R2/Q2 values of 0.95/0.91,0.77/0.65,and 0.85/0.74for Spike:ACE2, r3CL pro , and rPL pro , respectively (Supplementary Figures S1-S3).In all cases, the Root Mean Square Error of Calibration (RMSEC) was below 0.2, and the Root Mean Square Error of Cross-Validation (RMSECV) was under 0.3.These parameters collectively indicate a robust fit for the constructed models.
When examining the one-target models that were created, it becomes evident that there are certain ions displaying contributions exceeding the established threshold, and some ions are similar across all three evaluated targets.As a result, the decision was made to develop a fourth model, a multitarget model.This choice was based on the premise that molecules capable of affecting more than one target hold greater promise in the quest for compounds with anti-SARS-CoV-2 potential.To construct this model, the same data matrix and normalization techniques employed in the one-target models were used, with the sole distinction lying in the assessed response.The response (Y) for this model was formulated following this logic: samples exibiting activity in all three targets were categorized as multitarget active (1, Active), whereas samples demonstrating activity in only one or two targets were designated as one-target active (−1, Non-active) (Figure 2A,B).matrix and normalization techniques employed in the one-target models were used, with the sole distinction lying in the assessed response.The response (Y) for this model was formulated following this logic: samples exibiting activity in all three targets were categorized as multitarget active (1, Active), whereas samples demonstrating activity in only one or two targets were designated as one-target active (−1, Non-active) (Figure 2A,B).The multitarget PLS model achieved an explained variance of 98.1% during calibration, employing seven factors and 96.9% during internal validation.The model presented an R² of 0.98 and a Q² of 0.96, indicating a reliable fit.Additionally, it had low RMSE values, with RMSECV (validation error) at 0.15, higher than RMSEC (calibration error) at 0.11.An expected RMSECV > RMSEC relationship confirmed the representativeness of the calibration data and the absence of overfitting in the model.
In the score plot of the PLS model (Figure 2A), the separation between the two distinct groups is evident (active and non-active).The X-loadings weights plot (Figure 2B) reveals the ions that make the most positive contribution in explaining the desired activity, specifically, ions that are found exclusively and/or in greater abundance in extracts with higher activity.This  The multitarget PLS model achieved an explained variance of 98.1% during calibration, employing seven factors and 96.9% during internal validation.The model presented an R² of 0.98 and a Q² of 0.96, indicating a reliable fit.Additionally, it had low RMSE values, with RMSECV (validation error) at 0.15, higher than RMSEC (calibration error) at 0.11.An expected RMSECV > RMSEC relationship confirmed the representativeness of the calibration data and the absence of overfitting in the model.
In the score plot of the PLS model (Figure 2A), the separation between the two distinct groups is evident (active and non-active).The X-loadings weights plot (Figure 2B) reveals the ions that make the most positive contribution in explaining the desired activity, specifically, ions that are found exclusively and/or in greater abundance in extracts with higher activity.This process identified eight ions with m/z values of [M − H] − 431.1, 439.4,445.1, 453.4,473.2, 633.3, 661.5, and 841.3.

Annotated Compounds for the Predicted Ions in the Multitarget Model Using Molecular Networking Analyses
Molecular networks are now widely used in the natural products chemistry area since the creation of the open-source Global Natural Products Social Molecular Networking (GNPS) online platform [25].This tool allows the visualization of tandem mass spectrometry data (MS 2 ) to highlight structural similarities between metabolites of complex matrices and to help in the annotation of the detected metabolites [26].
The clusters formed can be considered "molecular families" since the compounds should share key chemical characteristics for the node's connection.The predicted ions in the chemometric analysis formed a network whose compounds did not match the GNPS library search.However, through this concept of molecular networks, they could belong to the same class of compounds.The ions that most contribute to the anti-SARS-CoV-2 activity were grouped into Network 1 (Figure 3).
The clusters formed can be considered "molecular families" since the compounds should share key chemical characteristics for the node's connection.The predicted ions in the chemometric analysis formed a network whose compounds did not match the GNPS library search.However, through this concept of molecular networks, they could belong to the same class of compounds.The ions that most contribute to the anti-SARS-CoV-2 activity were grouped into Network 1 (Figure 3).8).It is interesting to note that these ions were more prevalent in E. mosenii, which was the most active and multitarget species among the analyzed anti-SARS-CoV-2 targets.However, the species E. prasina also presents compounds contained in Networks 1 and 2, which could partially explain the separation of these two species from the others as being the most active in the multitarget PLS model (Figure 2).
The parent masses and MS 2 fragments of these compounds strongly suggest they are myrtucommulones, a class of oligomeric acylphloroglucinols commonly found in the Myrtaceae species (Figure 4).These compounds have attracted the attention due to their diverse molecular architectures that, in general, are composed of an acylphloroglucinol nucleus attached to one or more syncarpic acid residues via an iso-butyl or iso-pentyl bridge(s) [2,17].8).It is interesting to note that these ions were more prevalent in E. mosenii, which was the most active and multitarget species among the analyzed anti-SARS-CoV-2 targets.However, the species E. prasina also presents compounds contained in Networks 1 and 2, which could partially explain the separation of these two species from the others as being the most active in the multitarget PLS model (Figure 2).
The parent masses and MS 2 fragments of these compounds strongly suggest they are myrtucommulones, a class of oligomeric acylphloroglucinols commonly found in the Myrtaceae species (Figure 4).These compounds have attracted the attention due to their diverse molecular architectures that, in general, are composed of an acylphloroglucinol nucleus attached to one or more syncarpic acid residues via an iso-butyl or iso-pentyl bridge(s) [2,17].Assignments were made based on the presence of characteristic ions from the portions of syncarpic acid and acylphloroglucinol, resulting from the cleavage of these molecules on both sides of the iso-butyl or iso-pentyl bridge (Figure 4 and Table 3).This mechanism was previously described for β-triketone heterodimers linked by an iso-pentyl moiety found in the acetonitrile extract of Eucalyptus gregsoniana, a species from the Myrtaceae family [27].Of the eight predicted ions, two compounds were annotated based Assignments were made based on the presence of characteristic ions from the portions of syncarpic acid and acylphloroglucinol, resulting from the cleavage of these molecules on both sides of the iso-butyl or iso-pentyl bridge (Figure 4 and Table 3).This mechanism was previously described for β-triketone heterodimers linked by an iso-pentyl moiety found in the acetonitrile extract of Eucalyptus gregsoniana, a species from the Myrtaceae family [27].Of the eight predicted ions, two compounds were annotated based on previously described fragmentation patterns [17], and three were proposed as undescribed compounds (Table 3).The remaining ions m/z [439.4]− and m/z [661.5]− have the same fragmentation pattern and are present in Network 1 of the myrtucommulone class (Figure 3).The ion at m/z of [453.4]− exhibits a comparable fragmentation pattern and is found within Network 2 of other acylphloroglucinol derivatives.It displays the mass corresponding to the acylphloroglucinol monomer in the MS/MS spectrum (Supplementary Figure S4).Compounds 1 (semimyrtucommulone) and 2 (nor-semimyrtucommulone) were previously described in Myrtus communis and have acylphloglucinol and syncarpic acid residues linked by an iso-butyl bridge [17], as well as compound 3, not yet reported in the literature.These structures are classified as dimeric type myrtucommulones [2].
The fragmentations follow the aforementioned patterns (Figure 5), which means that cleavage on one side of the iso-butyl bridge creates the pairs of fragments at m/z  and 3, respectively.These ions are characteristic of the syncarpic acid residue (Supplementary Figures S5-S7).
Compounds 4 and 5 present the same fragmentation pattern but are characteristic of trimeric and tetrameric molecules, respectively, meaning that they contain more than two residues of acylphloroglucinol and/or syncarpic acid (Figure 5, Supplementary Figures S8 and S9).Similarly to compounds 3, 4 and 5, several unknown compounds with a similar fragmentation pattern were observed in Network 1, suggesting that new compounds of this class may be present in E. mosenii and E. prasina.
Phloroglucinols and their derivatives, in particular myrtucommulones, display diverse molecular architectures and a wide range of biological profiles, such as antimicrobial, antiviral, anti-inflammatory, antiproliferative, and cytotoxic properties [2].
Compounds 4 and 5 present the same fragmentation pattern but are characteristic of trimeric and tetrameric molecules, respectively, meaning that they contain more than two residues of acylphloroglucinol and/or syncarpic acid (Figure 5, Supplementary Figures S8  and S9).Similarly to compounds 3, 4 and 5, several unknown compounds with a similar fragmentation pattern were observed in Network 1, suggesting that new compounds of this class may be present in E. mosenii and E. prasina.Phloroglucinols and their derivatives, in particular myrtucommulones, display diverse molecular architectures and a wide range of biological profiles, such as antimicrobial, antiviral, anti-inflammatory, antiproliferative, and cytotoxic properties [2].
The antiviral activity of the hydroalcoholic leaf extracts of Myrtus communis L. was investigated by Moradi et al. [19] against herpes simplex virus-1 (HSV-1), who observed an inhibitory effect on the virus both before and after entering the cell.In the review study The antiviral activity of the hydroalcoholic leaf extracts of Myrtus communis L. was investigated by Moradi et al. [19] against herpes simplex virus-1 (HSV-1), who observed an inhibitory effect on the virus both before and after entering the cell.In the review study conducted by Nicoletti et al. [17], myrtucommulones extracted from species of the genus Kunzea and Melaleuca also showed moderate effects against HSV-1.
Myrtucommulones and related acylphloroglucinols from Myrtaceae were evaluated for their in vitro activity against the respiratory syncytial virus (RSV), showing promising effects compared with the ribavirin control.Studies suggested that the antiviral activity of these compounds might be influenced by the side chains of the phloroglucinol moiety [20,21].Phloroglucinols have also been shown to be effective against other viruses, such as human immunodeficiency virus (HIV), herpes or enterovirus, and preliminary docking analysis data suggest that they can be effective against SARS-CoV-2 [22].In addition, macrocarpals A-E (phloroglucinol dialdehyde diterpene derivatives) were isolated from Eucalyptus globulus Labill.(Myrtaceae) showed pronounced anti-HIV activity [2].
Liquid chromatography coupled with tandem mass spectrometry (UHPLC-MS/MS) analysis of the five Myrtaceae ethanol extracts showed chemical differences in their profiles (Figure 6), evidencing the more significant presence of myrtucommulones in the multitarget extracts (E.prasina and E. mosenii).Through the Feature Based Molecular Network (FBMN) workflow, other molecular networks were constructed (Supplementary Figure S4).For the unpredicted ions, there was a match in the GNPS library for 2 ′ ,6 ′ -dihydroxy-4 ′ -methoxydihydrochalcone (m/z 271.1 [M − H] − ) in E. brasiliensis (Supplementary Figure S10) [30] and myricitrin for E. prasina [31], as well as the presence of quinic and ellagic acids for M. splendens [32], confirmed by their fragmentation profiles.In addition, M. splendens and M. strigipes show chemical similarities in the range of 24 to 28 min, with evidence of the presence of tannins already reported in the literature for species of the Myrtaceae family [32].

Figure S4
).For the unpredicted ions, there was a match in the GNPS library for 2′,6′dihydroxy-4′-methoxydihydrochalcone (m/z 271.1 [M − H] − ) in E. brasiliensis (Supplementary Figure S10) [30] and myricitrin for E. prasina [31], as well as the presence of quinic and ellagic acids for M. splendens [32], confirmed by their fragmentation profiles.In addition, M. splendens and M. strigipes show chemical similarities in the range of 24 to 28 min, with evidence of the presence of tannins already reported in the literature for species of the Myrtaceae family [32].

Myrtaceae Extracts Inhibit SARS-CoV-2 Replication in Infected Lung Cells
Despite Vero E6 cells being recognized as an important lineage for SARS-CoV-2 replication [33], they may not adequately mimic the infection event in human cells [34,35].Therefore, in a second phase, we conducted experimental assays with the human pneumocyte lineage, widely recognized as the most effective model for in vitro infection and screening of anti-coronavirus drugs due to its high permissibility to SARS-CoV-2 through a transmembrane serine protease 2 (TMPRSS2)-dependent mechanism viral entry [36][37][38].
Firstly, Calu-3 cells were infected at a multiplicity of infection (MOI) of 0.01, and the extracts were evaluated for their antiviral activity at a single concentration (50 µg•mL −1 ) for 24 h of treatment, aiming to select the plant species with the best antiviral performance.However, the data indicated that all extracts exhibited high potential as inhibitors of SARS-CoV-2 replication, reducing viral infection by more than 80% (Figure 7).
Firstly, Calu-3 cells were infected at a multiplicity of infection (MOI) of 0.01, and the extracts were evaluated for their antiviral activity at a single concentration (50 µg•mL −1 ) for 24 h of treatment, aiming to select the plant species with the best antiviral performance.However, the data indicated that all extracts exhibited high potential as inhibitors of SARS-CoV-2 replication, reducing viral infection by more than 80% (Figure 7).Based on the initial screening, we further investigated whether the anti-SARS-CoV-2 activity of the studied species would remain effective in Calu-3 cells that were infected and exposed to different extract concentrations (curve: 0.781, 1.563, 3.125, 6.25, 12.5, 25, 50, and 100 µg•mL −1 ) for 24 h.In vitro analysis demonstrated that the five Myrtaceae extracts were able to inhibit approximately 100% of SARS-CoV-2 replication when treated with a high tested concentration (100 µg•mL −1 ).Remarkably, it was observed that the high antiviral effect was maintained in treatments until 12.5 µg•mL −1 (100-60%) (Figure 8).We highlight that all extracts presented lower EC50 values in Calu-3 cells, with a concentration below 8.15 µg•mL −1 (Table 4).These data indicate that the concentration required to obtain     Considering that the extracts were toxic to Vero E6/ACE2 cells only at concentrations above 500 µg•mL −1 , we aimed to assess the viability of Calu-3, non-infected and treated with the intermediate concentration of 200 µg•mL −1 for 72 h, using the colorimetric assay that detects lactate dehydrogenase (LDH) enzyme.Overall, all five plant species showed only low or moderate cytotoxicity.The results demonstrated that the best cell viability indices were observed in treatments with extracts from E. prasina (92%), E. brasiliensis (82%), and E. mosenii (74%), while M. strigipes was responsible for causing the greatest cytotoxic effect in a pulmonary cell model (40%) (Figure 9).The concentration used in cytotoxicity assays was up to 24 times higher than the maximum EC 50 obtained in antiviral assays.The viability data were obtained in comparison with control cells treated with vehicle (DMSO) in the same proportion.Plant extracts and plant-derived secondary metabolites have broad-spectrum antimicrobial activity and can exhibit properties against SARS-CoV-2 replication [5,[39][40][41].Here, chemometric analyses suggest that myrtucommulones, more prevalent in the multitarget species E. mosenii and E. prasina, may be the main contributor to the strong inhibitory virus effect exhibited by these extracts.Furthermore, our antiviral assays with Calu-3 cells revealed that other species within this family also demonstrated potent effects against the coronavirus, raising the hypothesis that the action of these biocompounds may occur on other proteases that are also targets of the infection or even that antiviral efficacy may vary in different cell lineages.
Interestingly, studies have revealed that extracts from the species Scutellaria barbata D. Don (SBD), belonging to the Lamiaceae family, inhibited the action of the viral proteases CLpro and TMPRSS2, indicating that SBD blocked the entry of SARS-CoV-2 pseudoparticles into the Calu-3 cells by affecting TMPRSS2 activity [42].Molecular docking analyses revealed that the bioactive compound gingerol, extracted from Zingiber officinale (Zingiberaceae family), prevents the entry and fusion of coronavirus by blocking the TMPRSS2 receptor [43,44].Flavonoids such as kaempferol, luteolin, sulforaphane, quercetin, and cryptotanshinone, found in different plant families, also exhibit inhibitory Plant extracts and plant-derived secondary metabolites have broad-spectrum antimicrobial activity and can exhibit properties against SARS-CoV-2 replication [5,[39][40][41].Here, chemometric analyses suggest that myrtucommulones, more prevalent in the multitarget species E. mosenii and E. prasina, may be the main contributor to the strong inhibitory virus effect exhibited by these extracts.Furthermore, our antiviral assays with Calu-3 cells revealed that other species within this family also demonstrated potent effects against the coronavirus, raising the hypothesis that the action of these biocompounds may occur on other proteases that are also targets of the infection or even that antiviral efficacy may vary in different cell lineages.
Interestingly, studies have revealed that extracts from the species Scutellaria barbata D. Don (SBD), belonging to the Lamiaceae family, inhibited the action of the viral proteases CLpro and TMPRSS2, indicating that SBD blocked the entry of SARS-CoV-2 pseudoparticles into the Calu-3 cells by affecting TMPRSS2 activity [42].Molecular docking analyses revealed that the bioactive compound gingerol, extracted from Zingiber officinale (Zingiberaceae family), prevents the entry and fusion of coronavirus by blocking the TMPRSS2 receptor [43,44].Flavonoids such as kaempferol, luteolin, sulforaphane, quercetin, and cryptotanshinone, found in different plant families, also exhibit inhibitory properties on the same serine protease, in addition to acting as anti-inflammatory agents in the treatment of this respiratory disease [45].In summary, these findings elucidate the several active compounds found in medicinal plants and open ways for future research that encompasses the mechanisms of action of acylphloroglucinol compounds and their immunomodulatory effects in COVID-19, as well as the role of the Myrtaceae family on other therapeutic targets, such as the transmembrane serine protease 2, required for SARS-CoV-2 invasion into pneumocytes cells.
In view of the promising results obtained from the in vitro assays and the prediction studies, we decided to perform molecular docking analyses for compounds 1 to 5 to characterize their behavior at the binding sites (3CL pro , PL pro , and Spike) and to elucidate the existing molecular interactions.

Molecular Docking of ACE2 and Spike Protein
The comprehensive molecular docking analysis with the SARS-CoV-2 Spike protein and ACE2 has provided us with a deeper understanding of the potential binding affinities of various ligands.The fitness score, a reflection of binding strength, was used as the primary metric for evaluation.
Collectively, these scores underline the potential interactions and affinities of the ligands with both the Spike protein and ACE2 (Table 5).Our decision to prioritize the interaction analyses for ligand compound 5 was driven by its superior docking scores, suggesting a promising interaction profile.

Interactions of the Ligand Compound 5 with ACE2
Our investigation into the molecular interactions of ACE2 has unveiled additional specific interactions, emphasizing the complexity of ACE2's molecular associations.
Starting with hydrophobic interactions, the residue 321 PRO displayed an interaction at 3.98 Å.The residue 376 HIS showed an interaction at 3.91 Å.The residue 479 PHE presented a hydrophobic interaction with 3.86 Å. Remarkably, the residue 485 TYR displayed two distinct interactions at distances of 3.39 Å and 3.67 Å, respectively.
In the realm of hydrogen bonds, the residue 322 THR formed a bond with 1.64 Å, an angle of 2.43 Lastly, our data also highlighted π-cation interactions.Specifically, the residue 489 ARG showcased a π-cation interaction with 4.36 Å and a deviation of 1.98 Å.

Interactions of the Ligand Compound 5 with the Spike Protein
In our in-depth analysis of molecular interactions between the ligand and the SARS-CoV-2 Spike protein, specific interactions have been identified, highlighting the possible intricate association between them.
Amongst the hydrophobic interactions, the Spike protein residue 340 GLU showed an interaction at 3.32 Å.The residue 341 VAL exhibited an interaction at 3.93 Å.The residue 344 ALA displayed a hydrophobic interaction with 3.66 Å.Furthermore, the residue 346A ARG formed an interaction at 3.35 Å.The residue 347 PHE also demonstrated a hydrophobic interaction with 3.48 Å.The residue 354 ASN presented an interaction with 3.57 Å. Impressively, the residue 358 ILE showcased two distinct interactions: the first with 3.86 Å and the second at a shorter distance of 3.04 Å.
Regarding hydrogen bonds, the Spike protein residue 340 GLU formed a bond with 1.57 Å, an angle of 2.53 • , and a dihedral angle of 167.25 While our comprehensive investigation has elucidated a multitude of interactions between the ligand and both the SARS-CoV-2 Spike protein and ACE2, it is noteworthy to mention that the interactions identified are not among those described in known hotspots.Given the significance of hotspot interactions in biological function and therapeutic targeting, this observation suggests that these proteins may not be the most promising candidates for further molecular dynamic studies.Future work will prioritize ligands and protein targets that exhibit interactions within or near known functional hotspots.
Figure 10.Interaction of the PL pro -compound 5 complex in the BL2loop.The protein is colored in pale cyan, the residues of BL2loop are colored in green (residues G266, N267, Y268, Q269, C270, and G271), and the ligand compound 5 is colored in orange (PDBid: 7JRN).
Figure 10.Interaction of the PL pro -compound 5 complex in the BL2loop.The protein is colored in pale cyan, the residues of BL2loop are colored in green (residues G266, N267, Y268, Q269, C270, and G271), and the ligand compound 5 is colored in orange (PDBid: 7JRN).
As seen with PL pro , the docking analyses showed that compound 5 also obtained the highest score with 3CL pro , equal to 65.01, among the substances analyzed.Compound 4 had the second-best score (59.99), followed by compound 3 (58.43),nor-semimyrtucommulone (2) (55.55), and semimyrtucommulone (1) (50.14), in that order.The evaluation of the receptor-ligand interaction for compound 5 revealed an H-bond between O7 (Supplementary Figure S11) and CYS145 (2.32 Å), representative of the HIS41-CYS145 catalytic dyad [10].OH29 from the same position forms an H-bond with ARG188 (2.39 Å), and OH23 forms a bond at 2.52 Å with GLN189, located in the S2 region of the active groove [46].
Considering that 3CL pro has a described allosteric binding region between the homodimer chains [10], the possible inhibition of the enzyme at this site was also tested.Again, compound 5 stood out from the other compounds, with a score of 63.61. Figure 11 shows the interaction of compound 5 in the region between the protease chains.A chemical interaction was predicted with OH29 of the ligand (Supplementary Figure S11), which formed a 1.71 Å H-bond with Leu287 of the A chain.The ranking of the other compounds continues with compound 4 leading with a 63.37 score, then semimyrtucommulone (1) (56.61), then compound 3 (54.02),and lastly the nor-semimyrtucommulone (2) (53.41).As seen with PL pro , the docking analyses showed that compound 5 also obtained the highest score with 3CL pro , equal to 65.01, among the substances analyzed.Compound 4 had the second-best score (59.99), followed by compound 3 (58.43),norsemimyrtucommulone (2) (55.55), and semimyrtucommulone (1) (50.14), in that order.The evaluation of the receptor-ligand interaction for compound 5 revealed an H-bond between O7 (Supplementary Figure S11) and CYS145 (2.32 Å), representative of the HIS41-CYS145 catalytic dyad [10].OH29 from the same position forms an H-bond with ARG188 (2.39 Å), and OH23 forms a bond at 2.52 Å with GLN189, located in the S2 region of the active groove [46].
Considering that 3CL pro has a described allosteric binding region between the homodimer chains [10], the possible inhibition of the enzyme at this site was also tested.Again, compound 5 stood out from the other compounds, with a score of 63.61. Figure 11 shows the interaction of compound 5 in the region between the protease chains.A chemical interaction was predicted with OH29 of the ligand (Supplementary Figure S11), which formed a 1.71 Å H-bond with Leu287 of the A chain.The ranking of the other compounds continues with compound 4 leading with a 63.37 score, then semimyrtucommulone (1) (56.61), then compound 3 (54.02),and lastly the norsemimyrtucommulone (2) (53.41).To further investigate the results obtained, MD were performed, and the results can be viewed in the Supplementary Data (Movies S1-S3).These simulations were used to evaluate the dynamics of the interaction between the PL pro -compound 5 (Movie S1) and 3CL pro active site-compound 5 (Movie S2) and 3CL pro allosteric groove-compound 5 (Movie S3) complexes.Our results for PL pro showed that in the first 50 nanoseconds of our simulation, we observed that the average number of hydrogen bonds formed between PL pro and compound 5 was typically around 3. However, during certain moments within To further investigate the results obtained, MD were performed, and the results can be viewed in the Supplementary Data (Movies S1-S3).These simulations were used to evaluate the dynamics of the interaction between the PL pro -compound 5 (Movie S1) and 3CL pro active site-compound 5 (Movie S2) and 3CL pro allosteric groove-compound 5 (Movie S3) complexes.Our results for PL pro showed that in the first 50 nanoseconds of our simulation, we observed that the average number of hydrogen bonds formed between PL pro and compound 5 was typically around 3. However, during certain moments within these nanoseconds, this number increased, reaching up to 7 hydrogen bonds (Figure 12A).After 50 ns, the average became 2 h-bonds formed until the last 10 ns.In this mark, no h-bonds were predicted (Figure 12A).Ligand RMSD has a 3 staged profile, with an average of 1.8 Å during the first 50 ns, 2.5 Å from 50 ns to approximately 175 ns, and in the last of the 25 ns reaching an RMSD higher than 10 Å (Figure 12B).This high RMSD and the lack of predicted h-bonds at the end of the simulation allude to the exit of the ligand from the binding site.This is demonstrated in the RMSD of the BL2loop, which maintains an average of 1.8 Å until approximately 175 ns when the average climbs to approximately 4.5 Å, showing that this region has moved (Figure 12C).The RMSF graph (Figure 12D) shows how the BL2loop residues (residues 266, 267, 268, 269, 270, 271) highly fluctuated during the simulation (approximately 5 Å, with some residues reaching 6 Å).Furthermore, the MD results for complex 3CL pro -compound 5 targeted to the active site of the protease did not show any hydrogen bonds with residues of interest (HIS41 and CYS145).However, an average of 2 h-bond was reported during the 200 ns of simulation, with greater frequency up to 100 ns (Figure 13A).In this period, the residues with the highest occupancy were GLN189 and ARG188, confirming the result obtained with Maestro.Moreover, the ligand RMSD (Figure 13B) showed instability with a plateau starting at 110 ns, maintaining an average of 7 Å.This can be justified due to the larger size of the ligand and the fact that the active site of the protease is shallow [47], so when it reaches this plateau there is a smaller area of the ligand that can remain stable in the interaction with the target.In contrast to the instability of the ligand, the catalytic cavity remained relatively stable during the 200 ns (Figure 13C), maintaining a mean RMSD of 1.5 Å.Our results indicate that although the ligand remained partially in the groove, it Furthermore, the MD results for complex 3CL pro -compound 5 targeted to the active site of the protease did not show any hydrogen bonds with residues of interest (HIS41 and CYS145).However, an average of 2 h-bond was reported during the 200 ns of simulation, with greater frequency up to 100 ns (Figure 13A).In this period, the residues with the highest occupancy were GLN189 and ARG188, confirming the result obtained with Maestro.Moreover, the ligand RMSD (Figure 13B) showed instability with a plateau starting at 110 ns, maintaining an average of 7 Å.This can be justified due to the larger size of the ligand and the fact that the active site of the protease is shallow [47], so when it reaches this plateau there is a smaller area of the ligand that can remain stable in the interaction with the target.In contrast to the instability of the ligand, the catalytic cavity remained relatively stable during the 200 ns (Figure 13C), maintaining a mean RMSD of 1.5 Å.Our results indicate that although the ligand remained partially in the groove, it did not generate stability in the protein, given the increase in ligand RMSD (Figure 13B) and the increase in variability assessed by RMSF. Figure 13D shows high movement in the region formed by the residues of interest (HIS41, CYS145, GLU166, GLN189, and ARG188).The results of the simulation in aqueous solution showed greater stability of the ligand in the allosteric site than when compared to the active site.The compound 5 ligand did not show any h-bond with LEU287.However, GLY170 stood out.A 95% occupancy and an average of 4 h-bonds with GLY170 and compound 5 were predicted over the entire simulated time (Figure 14A).In contrast to the ligand RMSD targeted to the active site, the ligand RMSD in the predicted allosteric pocket showed greater stability (Figure 14B), with an average RMSD of less than 1 Å over the 200 ns.While the cavity was more unstable, the RMSD varied between 0.5 Å and 2.5 Å over 200 ns (Figure 14C).One hypothesis is that this is due to greater perturbation of the residues for the ligand to fit into the cavity.This can also be assessed by the flexibility of the residues in the RMSF plot of the protein (Figure 14D).The results of the simulation in aqueous solution showed greater stability of the ligand in the allosteric site than when compared to the active site.The compound 5 ligand did not show any h-bond with LEU287.However, GLY170 stood out.A 95% occupancy and an average of 4 h-bonds with GLY170 and compound 5 were predicted over the entire simulated time (Figure 14A).In contrast to the ligand RMSD targeted to the active site, the ligand RMSD in the predicted allosteric pocket showed greater stability (Figure 14B), with an average RMSD of less than 1 Å over the 200 ns.While the cavity was more unstable, the RMSD varied between 0.5 Å and 2.5 Å over 200 ns (Figure 14C).One hypothesis is that this is due to greater perturbation of the residues for the ligand to fit into the cavity.This can also be assessed by the flexibility of the residues in the RMSF plot of the protein (Figure 14D).Moving forward to the analysis of the pharmacokinetic properties, the Lipinski rule of 5 criteria predicts that compound 3, semimyrtucommulone (1), and norsemimyrtucommulone (2) are promising drugs.Furthermore, Table 6 shows that their lipophilicity satisfies membrane permeability according to the LogP value.The Caco-2 assessment of the same three compounds indicates satisfactory absorption criteria, which are presented in Supplementary Table S1.While compounds 4 and 5 are metabolized through their properties as inhibitors of cytochromes CYP1A2, CYP2C19, CYP2D6, and CYP3A4.Moving forward to the analysis of the pharmacokinetic properties, the Lipinski rule of 5 criteria predicts that compound 3, semimyrtucommulone (1), and nor-semimyrtucommulone (2) are promising drugs.Furthermore, Table 6 shows that their lipophilicity satisfies membrane permeability according to the LogP value.The Caco-2 assessment of the same three compounds indicates satisfactory absorption criteria, which are presented in Supplementary Table S1.While compounds 4 and 5 are metabolized through their properties as inhibitors of cytochromes CYP1A2, CYP2C19, CYP2D6, and CYP3A4.The high content of Plasma Protein Binding (PPB) restricts the bioavailability of all five compounds.Moreover, all compounds must be capable of crossing the Blood-Brain Barrier (BBB) (Supplementary Table S1).The tested molecules are expected to cause hepatotoxicity in humans (H-HT) and drug-induced liver injury (DILI).Nevertheless, they do not demonstrate genotoxicity, as concluded by the Ames test.The compounds do not have carcinogenic or respiratory toxicity properties (see Table 7), except for semimyrtucommulone 1, which shows potential for both.In addition, since the in silico prediction indicated the presence of a PAINS warning, it is possible that this higher promiscuity in compounds 4 and 5 can interfere with the in vitro testing and, therefore, generate false-positive results.In brief, compound 5 shows greater promise in interacting with ACE2-Spike, PL pro , and 3CL pro targets of SARS-CoV-2 based on in silico analysis.Nonetheless, its physicochemical features suggest that structural modifications to the molecule may be required.Such changes may also enhance the receptor-ligand complex interaction, rendering compound 5 an encouraging multitarget drug candidate.

Chemicals, Materials and Plant Extracts
E. brasiliensis (RB00658589/AL1707) and M. strigipes (RB01071480/TROVO526) were collected in the Floresta da Tijuca (State of Rio de Janeiro), E. mosenii (RB00776077/AL1901) and M. splendens (RB00662950/AL1721) were collected in the Parque Nacional de Itatiaia (State of Rio de Janeiro), and E. prasina (RB00831331/PARDOC2233) was collected in Parque Nacional da Serra dos Órgãos (PARNASO-State of Rio de Janeiro).The first code in parentheses refers to the voucher specimen number deposited at Jardim Botanico do Rio de Janeiro herbarium, and the second represents the original reference code in the extract bank.The leaves were dried in a ventilated oven and grounded in a hammer mill.Extracts of the leaves were obtained by maceration with ethanol: water (9:1), filtrated, and the solvent removed by rotary evaporation under reduced pressure.These extracts belong to the Federal University of Rio de Janeiro plant extract bank (LABEXTRACT, Centro de Ciências da Saúde, Universidade Federal do Rio de Janeiro).This work was authorized by the Directing Council of Genetic Heritage (Conselho de Gestão do Patrimônio Genético, CGEN) with bioprospecting purposes by the authorization ACADBDD.
ii. LDH assay: Calu-3 (1.5 × 10 4 cells/well) was maintained in a 96-well plate (Jet Biofil, Guangzhou, China) with E. mosenii, E. prasina, M. splendens, M. strigipes and E. brasiliensis, at a concentration of 200 µg•mL −1 for 72 h at 37 • C, 5% CO 2 .The supernatants were collected to evaluate cell viability using the CytoTox 96 ® Non-Radioactive colorimetric kit (Promega, Madison, USA) according to the manufacturer's instructions.Lysis solution 1× (Promega, Madison, USA) was used to lyse Calu-3 cells, and purified lactate dehydrogenase (LDH) was used as a positive control.The reaction product was analyzed at 490 nm using a 96-well plate reader (Loccus, São Paulo, Brazil).All the extracts were resuspended in 100% DMSO for the in vitro tests.The DMSO final concentrations do not exceed 1% (v/v) in the experiments, thereby not affecting cell growth.
The mass spectrometer (MS), equipped with an electrospray ionization (ESI) and atmospheric pressure chemical ionization (APCI) sources and an ion trap analyzer (with 1.000 of resolution), was operated in positive-and negative-ion modes.High-purity nitrogen (N2) was used as the sheath gas (20 arbitrary units) and auxiliary gas (10 arbitrary units).High-purity helium (He) was used as the collision gas.Full scan data acquisition (mass range: m/z 100-1000) and data-dependent acquisition (topN = 3) were performed.The normalized collision energy of the collision-induced dissociation (CID) cell was set at 35 eV.All analyses were carried out for three different preparation samples of each of the five species of the Myrtaceae family.

Data Processing, Statistical Analysis and Molecular Networking
Graphs of biological data were generated using the GraphPad Prism 8.0 software.The results RBD:ACE2 interaction and r3CL pro and rPL pro inhibition were determined using linear regression.For SARS-CoV-2 Spike-pseudotyped VSV neutralization assay, the EC 50 values were determined with the nonlinear regression of log (inhibitor) vs. normalized response.Values correspond to the best curve generated based on R 2 values ≥ 0.9.All experiments were realized with three technical replicates.
The data obtained from LC-MS analyses were converted to the mzML format using the parameters: subset, msLevel 1-2 and polarity positive or negative, as well as Peak Picking, vendor algorithm (msLevel 1-2) in the Proteowizard-MSconvert version 3.02 tool.Then, the data were processed in MZmine v.2.53.The mass detection varied according to the intensity and quality of the signals obtained in each analysis method (ESI or APCI [+/−]).The mass chromatogram construction and wavelet deconvolution were carried out using the ADAP Chromatogram Builder algorithm.Isotopes were eliminated through the Isotopic Peaks Grouper, with the most intense isotope selected as the representative, and alignment was performed using the join aligner.The table list was saved to an ASCII text file.
Process chemical data and biological data were then exported to UnscramblerX software version 10.4.The study aimed to predict potential active compounds inhibiting Spike:ACE2 interaction and proteolytic activities of r3CLpro and rPLpro.Fifteen ethanol extracts from Myrtaceae family species underwent LC-MS/MS analysis in positive and negative modes.Chemometric analyses were performed, and a Partial Least Squares Regression model correlated LC-MS data with inhibitory activities.The PLS regression identified sample patterns, classifying them into two groups based on average activity values: active (≥50% inhibition) and non-active (<50% inhibition) samples.
The MZmine results were uploaded to the open-source Global Natural Products Social (GNPS) online platform, along with the open-source converted files (.mzML) and metadata table.These files were used to build the molecular networks and to dereplicate the present compounds.Along with GNPS annotation, a custom database was also used.The parameters used were those recommended in the GNPS documentation for unity resolution data.The GNPS-generated data were downloaded and processed in Cytoscape 3.9.1 to obtain the results of the molecular networking analysis.

Anti-SARS-CoV-2 Assay in Calu-3 Cells
Monolayers of Calu-3 cells were infected with SARS-CoV-2 at a multiplicity of infection (MOI) of 0.01 for 1 h at 37 • C and 5% CO 2 .The supernatants were removed, and the cells were treated with a concentration curve (0.781, 1.563, 3.125, 6.25, 12.5, 25, 50 and 100 µg•mL −1 ) of Myrtaceae extracts for 24 h.Then, the supernatant was collected and used for the virus titration assay.The concentration of the extract required to obtain 50% of maximum effective virus inhibition (EC 50 ) was assessed compared with control cells (infected and untreated cells) [49,50].

Statistical Analysis
Cytotoxicity and antiviral assays on Calu-3 cells were performed with four technical replicates per assay.The generated data were analyzed using GraphPad Prism 10.0 software (GraphPad Software, La Jolla, CA, USA).Differences between the compounds were analyzed using one-way ANOVA with post-testing of Dunnett, and the EC 50 values were determined using nonlinear regression of Log(inhibitor) vs. normalized response.The equations used to fit the best curve were generated based on R 2 values 0.81 to 0.92.
For PL pro , the binding site was defined with a centre in the coordinates of the original ligand GRL0617 present in the crystal with radius of 10 Å.The radius simulation for 3CL pro was established with 10 Å from HIS 41 in target docking and 35 Å for blind docking, which in the last case, includes the union of the two chains.For the RBD, a blind docking was realized, with the simulation radius of 70 Å with centre in LEU 455.Finally, docking with ACE2 followed the same methodology described for Spike.However, in the latter case, the radius of 80 A was determined from GLN598 establishing the simulation region.
The interaction analysis between amino acid residues and substances was performed using Maestro (Schrödinger Release 2023-3: Maestro, Schrödinger, LLC, New York, NY, USA, 2023) and Protein-Ligand Interaction Profiler.
The PL pro complex preparation was performed using the MCPB.pytool [57] and GAMESS-US [58] to properly simulate the zinc atom present in the protein and its effects on protein motility.
Force fields used were ff19SB, gaff2, and OPC for waters.Solvation was made in a truncated Octahedron box with a 12.0 Å distance from the box edge.Salt molarity was calculated using the method described by Machado [59] to neutralize the system and be at 0.15 M.
Relaxation was made with 1000 cycles of Steepest descent.Temperature was set to 310 K. Final simulations were for 200 ns with MonteCarlo Barostat.SHAKE was used for restraining hydrogen bonds.Time step was defined to 0.001.

Conclusions
The study's findings indicate that among the five Myrtaceae species examined, two (E.prasina and E. mosenii) demonstrated noteworthy potential as multitarget agents against SARS-CoV-2 targets (Spike:ACE2, PLpro, and 3CLpro).Moreover, all five species of the Myrtaceae family exhibited high inhibitory capacity against SARS-CoV-2 B.1 lineage replication in the human lung epithelial cells model.Notably, ions identified as myrtucommulones were correlated with the observed activity.In addition, the established workflow employs molecular network databases and predictive chemometric models through tandem mass spectrometry to accelerate the dereplication of molecules before subsequent compound isolation.This abbreviates the lengthy process of bioprospecting using bio-guided fractionation.The comprehensive investigation seamlessly integrates biological and chemical analyses, significantly advancing our understanding of the antiviral properties of Myrtaceae.This study represents the first documentation of such activity for these compound types, encouraging further isolation efforts and testing on the models employed.

Figure 2 .
Figure 2. Scores (A), X-loadings weight (B) for the PLS model based on the LC-MS data obtained in the electrospray ionization in negative mode of the 15 samples of five species of Myrtaceae family, and the inhibition values of the extracts on the RBD:ACE2 interaction and the activity of the r3CL pro and rPL pro enzymes of SARS-CoV-2.The dashed line indicates the cutoff for contributions exceeding 0.05 (loadings weights).

Figure 2 .
Figure 2. Scores (A), X-loadings weight (B) for the PLS model based on the LC-MS data obtained in the electrospray ionization in negative mode of the 15 samples of five species of Myrtaceae family, and the inhibition values of the extracts on the RBD:ACE2 interaction and the activity of the r3CL pro and rPL pro enzymes of SARS-CoV-2.The dashed line indicates the cutoff for contributions exceeding 0.05 (loadings weights).

Figure 3 .
Figure 3. Network 1 showing the oligomeric acylphloroglucinols type compounds, ions that most contribute to the anti-SARS-CoV-2 activity.The molecular networks 1 and 2 (Figure 3 and Supplementary Figure S4) display the nodes of the predicted ions with m/z [445.1]− (1), m/z [431.1]− (2), m/z [473.2]− (3), m/z [633.3]− (4), m/z [841.3]− (5), m/z [439.4]− (6), m/z [661.5]− (7), and m/z [453.4]− (8).It is interesting to note that these ions were more prevalent in E. mosenii, which was the most active and multitarget species among the analyzed anti-SARS-CoV-2 targets.However, the species E. prasina also presents compounds contained in Networks 1 and 2, which could partially explain the separation of these two species from the others as being the most active in the multitarget PLS model (Figure2).The parent masses and MS 2 fragments of these compounds strongly suggest they are myrtucommulones, a class of oligomeric acylphloroglucinols commonly found in the Myrtaceae species (Figure4).These compounds have attracted the attention due to their diverse molecular architectures that, in general, are composed of an acylphloroglucinol nucleus attached to one or more syncarpic acid residues via an iso-butyl or iso-pentyl bridge(s)[2,17].

Figure 3 .
Figure 3. Network 1 showing the oligomeric acylphloroglucinols type compounds, ions that most contribute to the anti-SARS-CoV-2 activity.The molecular networks 1 and 2 (Figure 3 and Supplementary Figure S4) display the nodes of the predicted ions with m/z [445.1]− (1), m/z [431.1]− (2), m/z [473.2]− (3), m/z [633.3]− (4), m/z [841.3]− (5), m/z [439.4]− (6), m/z [661.5]− (7), and m/z [453.4]− (8).It is interesting to note that these ions were more prevalent in E. mosenii, which was the most active and multitarget species among the analyzed anti-SARS-CoV-2 targets.However, the species E. prasina also presents compounds contained in Networks 1 and 2, which could partially explain the separation of these two species from the others as being the most active in the multitarget PLS model (Figure2).The parent masses and MS 2 fragments of these compounds strongly suggest they are myrtucommulones, a class of oligomeric acylphloroglucinols commonly found in the Myrtaceae species (Figure4).These compounds have attracted the attention due to their diverse molecular architectures that, in general, are composed of an acylphloroglucinol nucleus attached to one or more syncarpic acid residues via an iso-butyl or iso-pentyl bridge(s)[2,17].

Figure 6 .
Figure 6.Aligned chromatograms of Myrtaceae ethanol extracts by UHPLC-ESI(-)3D-IT-MS/MS highlighting the predicted ions.The recognition that Myrtaceae extracts act as blockers of the ACE2:RBD interaction in Vero E6/ACE2 cells infected with Spike-pseudotyped VSV and as inhibitors of the SARS-CoV-2 r3CLpro and rPLpro proteolytic activities led us to investigate the Myrtaceae extracts as inhibitors of the SARS-CoV-2 replication in infected Calu-3 cells, a model of human lung epithelial cells (human pneumocytes type II model).

Figure 6 .
Figure 6.Aligned chromatograms of Myrtaceae ethanol extracts by UHPLC-ESI(-)3D-IT-MS/MS highlighting the predicted ions.The recognition that Myrtaceae extracts act as blockers of the ACE2:RBD interaction in Vero E6/ACE2 cells infected with Spike-pseudotyped VSV and as inhibitors of the SARS-CoV-2 r3CLpro and rPLpro proteolytic activities led us to investigate the Myrtaceae extracts as inhibitors of the SARS-CoV-2 replication in infected Calu-3 cells, a model of human lung epithelial cells (human pneumocytes type II model).

Figure 7 .
Figure 7. Myrtaceae extracts inhibit virus replication in Calu-3 cells.Infected cells with SARS-CoV-2 at MOI (multiplicity of infection) 0.01 were treated with five different plant extracts at 50 µg•mL −1for 24 h at 37 °C and 5% CO2.After this period, the supernatants were harvested, and the virus titles were accessed by plaque-forming units (PFU/mL) assay.The percentage of inhibition was obtained by comparison with the infected/untreated group.*** p ≤ 0.001 in relation to infected control (n = 4).

Figure 7 .
Figure 7. Myrtaceae extracts inhibit virus replication in Calu-3 cells.Infected cells with SARS-CoV-2 at MOI (multiplicity of infection) 0.01 were treated with five different plant extracts at 50 µg•mL −1 for 24 h at 37 • C and 5% CO 2 .After this period, the supernatants were harvested, and the virus titles were accessed by plaque-forming units (PFU/mL) assay.The percentage of inhibition was obtained by comparison with the infected/untreated group.*** p ≤ 0.001 in relation to infected control (n = 4).Pharmaceuticals 2024, 17, x FOR PEER REVIEW 12 of 29

Figure 11 .
Figure11.Interaction of the 3CL pro -compound 5 complex in the allosteric site.The A-chain of the homodimer is colored in grayish green, and the B-chain residues that interacted in the cavity are colored in magenta.A-Leu287 is represented by dark green sticks.Compound 5 is shown in orange within the alternative interaction groove (PDBid: 6XQT).

Figure 11 .
Figure11.Interaction of the 3CL pro -compound 5 complex in the allosteric site.The A-chain of the homodimer is colored in grayish green, and the B-chain residues that interacted in the cavity are colored in magenta.A-Leu287 is represented by dark green sticks.Compound 5 is shown in orange within the alternative interaction groove (PDBid: 6XQT).

29 Figure 12 .
Figure 12. Results of the molecular dynamics of the PL pro -compound 5 complex in 200 ns.(A) Predicted h-bonds between the BL2loop residues and compound 5. (B) Calculated RMSD of the ligand compound 5 for the duration of the simulation.(C) Calculated RMSD of the BL2loop for the duration of the simulation.(D) Calculated RMSF for the residues of the PL pro .

Figure 12 .
Figure 12. Results of the molecular dynamics of the PL pro -compound 5 complex in 200 ns.(A) Predicted h-bonds between the BL2loop residues and compound 5. (B) Calculated RMSD of the ligand compound 5 for the duration of the simulation.(C) Calculated RMSD of the BL2loop for the duration of the simulation.(D) Calculated RMSF for the residues of the PL pro .

Pharmaceuticals 2024 , 29 Figure 13 .
Figure 13.Results of the molecular dynamics of the 3CL pro -compound 5 complex in 200 ns.(A) Predicted h-bonds between compound 5 and the residues CYS145, HIS41, GLU166, ARG188, and GLN189.(B) Calculated RMSD of the ligand compound 5 for the duration of the simulation.(C) Calculated RMSD of the catalytic site for the duration of the simulation.(D) Calculated RMSF for the residues of the 3CL pro .

Figure 13 .
Figure 13.Results of the molecular dynamics of the 3CL pro -compound 5 complex in 200 ns.(A) Predicted h-bonds between compound 5 and the residues CYS145, HIS41, GLU166, ARG188, and GLN189.(B) Calculated RMSD of the ligand compound 5 for the duration of the simulation.(C) Calculated RMSD of the catalytic site for the duration of the simulation.(D) Calculated RMSF for the residues of the 3CL pro .

Figure 14 .
Figure 14.Results of the molecular dynamics of the 3CL pro -compound 5 complex in 200 ns.(A) Predicted h-bonds between compound 5 and the residues GLY170.(B) Calculated RMSD of the ligand compound 5 for the duration of the simulation.(C) Calculated RMSD of the allosteric site for the duration of the simulation.(D) Calculated RMSF for the residues of the 3CL pro .

Figure 14 .
Figure 14.Results of the molecular dynamics of the 3CL pro -compound 5 complex in 200 ns.(A Predicted h-bonds between compound 5 and the residues GLY170.(B) Calculated RMSD of the ligand compound 5 for the duration of the simulation.(C) Calculated RMSD of the allosteric site for the duration of the simulation.(D) Calculated RMSF for the residues of the 3CL pro .

Table 2 .
EC 50 values for the five Myrtaceae extracts in Spike-pseudotyped VSV assay.

Table 3 .
LC-MS/MS data of annotated and proposed compounds (undescribed) in the molecular network (MS/MS spectra in Supplementary FiguresS5-S9).

Table 5 .
Fitness of top-ranked individuals from GOLD docking.

Table 6 .
In silico prediction of physicochemical properties of compounds 1-5.
MW: Molecular Weight; nHA: number of hydrogen bond acceptors; nHD: number of hydrogen bond donors; TPSA: topological polar surface area; nRot: number of rotatable bonds; LogP: logarithm of the n-octanol/water distribution coefficient.

Table 6 .
In silico prediction of physicochemical properties of compounds 1-5.

Table 7 .
Toxicity prediction.Interpret the colors so that green means the substance is not toxic, yellow means it may be toxic, and red means toxic.