Platanosides from Platanus × acerifolia: New molecules, SAR, and target validation of a strong lead for drug-resistant bacterial infections and the associated sepsis

Abstract

Three undescribed (1–3) and nine known (4–12) platanosides were isolated and characterized from a bioactive extract of the May leaves of Platanus × acerifolia that initially showed inhibition against Staphylococcus aureus Targeted compound mining was guided by an LC-MS/MS-based molecular ion networking (MoIN) strategy combined with conventional isolation procedures from a unique geographic location. The novel structures were mainly determined by 2D NMR and computational (NMR/ECD calculations) methods. Compound 1 is a rare acylated kaempferol rhamnoside possessing a truxinate unit. 6 (Z,E-platanoside) and 7 (E,E-platanoside) were confirmed to have remarkable inhibitory effects against both methicillin-resistant S. aureus (MIC: ≤ 16 μg/mL) and glycopeptide-resistant Enterococcus faecium (MIC: ≤ 1 μg/mL). These platanosides were subjected to docking analyses against FabI (enoyl-ACP reductase) and PBP1/2 (penicillin binding protein), both of which are pivotal enzymes governing bacterial growth but not found in the human host. The results showed that 6 and 7 displayed superior binding affinities towards FabI and PBP2. Moreover, surface plasmon resonance studies on the interaction of 1/7 and FabI revealed that 7 has a higher affinity (K_D = 1.72 μM), which further supports the above in vitro data and is thus expected to be a novel anti-antibacterial drug lead.

1. Introduction

Sepsis or septicaemia is a life-threatening immune-system response to a bacterial, viral, parasitic, or fungal infection. An estimated 50 million cases of sepsis occur each year, leading to about 11 million deaths globally [1]. Thus, sepsis is considered the leading cause of preventable death in the world by the Global Sepsis Alliance. Staphylococcus aureus causes an estimated 50 % of sepsis and is thus a primary concern regarding drug resistance [2]. S. aureus toxins lead to tissue necrosis [3] much like acetaminophen (paracetamol, APAP) induced liver toxicity. Both result in an increase in procalcitonin (PCT), which is used as a measure of both bacteria-induced sepsis and liver damage by APAP overdose and is thus recognized as one of the most critical bacterial-causing diseases in humans [4]. The treatment choices for S. aureus are usually penicillin and vancomycin. However, serious side effects and drug resistance limit their utility [5,6].

Natural products (NPs) and their derivatives have historically contributed significantly to pharmacotherapy, especially for infectious diseases. When compared to the synthetic classes, NPs demonstrated highly impressive structural diversity and complexity. In fact, since the golden age of antibiotics in the mid-20th century, NPs have always been serving as powerful therapeutics against pathogenic bacteria [7]. For instance, over the time frame from 1981 to 2019, of the 162 approved anti-bacterial agents, 89, or 54.9 %, are NPs or their structural analogues [8]. However, new classes of antibiotics have been rare, and the lack of research and resources have resulted in a two-decade-long antibiotic development void [9]. Thus, the significance of this new class could not be greater.

Identifying new molecules early in the workflow using Molecular Ion Network (MoIN) minimizes time, effort, and the cost of large screening campaigns [10,11]. The visualization of networks using Cytoscape enables the direct observation of similarities and differences between two or more samples in which similar entities within the network are clustered together. In contrast, disparate or unique entities are grouped separately. With the use of this approach, platanosides (PTSs), as well as their analogs, were found in the early workflow of purification [12].

PTSs belong to a unique class of flavonoids, which generally consist of the components of the kaempferol flavone rhamnose glycoside moiety with p-coumaroyl group(s) attached to the carbohydrate unit. The leaves of the Oriental plane tree (P. orientalis, chinar) and American sycamore (P. occidentalis, American plane tree) have been documented to be rich sources of flavonoids (incl. PTSs) with interesting antibacterial activity against methicillin-resistant S. aureus (MRSA) [13–19]. Such flavonoids were often encountered from the buds [20–28] of Platanus × acerifolia Willd (PaW), but only a few have been previously reported from its leaves [29,30]. Most recently, PTSs possessing significant S. aureus inhibitory activities have been unveiled from the January (Winter) PaW leaves [31].

In this study, the EtOAc-soluble fraction of the 90 % MeOH extract of the PaW leaves collected in the Spring (May 2017) showed antibacterial activity against S. aureus. Subsequently, an in-depth investigation of this fraction was carried out using various chromatographic techniques due to its strong potential to inhibit bacterial growth of S. aureus, resulting in the isolation and characterization of three undescribed flavonol glycosides (1–3) and nine known acylated kaempferol rhamnosides (4–12) (Figs. 1, 2). Furthermore, the antibacterial activities against both MRSA and glycopeptide-resistant Enterococcus faecium (GREfm) of the isolates were evaluated, showing the MICs of 6 (Z, E-platanoside) and 7 (E, E-platanoside) were 4–16 μg/mL (MRSA) and 0.5–1 μg/mL (GREfm).

Fig. 1.

MoIN of the mentioned bioactive fraction. left: Organized landscape of complete network; right: the MoIN cluster with signals at m/z 579, 595, and 765 for new PTSs and at m/z 763, 725, 617, and 601 for known PTSs.

Fig. 2.

The structures of compounds of 1–12.

2. Results and discussion

2.1. Isolation and structural elucidation

Isolation of the unreported molecules was facilitated under the guidance of UPLC-HRMS/MS (positive mode)-based molecular ion networking (MoIN) as enabled through the Cytoscape platform (https://cytoscape.org). This analysis revealed three unique signals at m/z 765.2, 595.1, and 579.2 with different retention times (Fig. 1). Subsequently, target purification of these m/z 579, 595, 765 metabolites afforded compounds 1, 2, and 3.

Compound 1 was obtained as a yellow amorphous powder, and its molecular formula was established as C₃₉H₃₄O₁₅ based on the protonated molecule at m/z 743.1975 [M + H]⁺ detected by HRESIMS (calcd. 743.1970) and ¹³C NMR data (Table S1). The UV spectrum (Fig. S1) exhibited two characteristic absorptions at 269 and 340 nm, indicating the presence of a 3-OH substituted flavanol skeleton [32,33], which was corroborated by the absorption bands at 3412 (br, hydroxy), 1718 (ester carbonyl), 1654 (conjugated carbonyl), and 1515 (aromatic ring) cm⁻¹ in its IR spectrum (Fig. S2) [34]. The ¹H NMR spectrum (Table S1) revealed an AAʹBBʹ spin system due to the aromatic proton signals at δ_H 7.82 (2H, br d, J = 8.3 Hz, H-2ʹ, H-6ʹ) and 6.97 (2H, br d, J = 8.3 Hz, H-3ʹ, H-5ʹ), along with two meta-coupled doublets at δ_H 6.47 (1H, br s, H-8) and 6.27 (1H, br s, H-6), as well as three hydroxyl groups at δ_H 12.63 (1H, s), 11.02 (1H, s), and 10.32 (1H, s), characteristic resonance for 5-OH, 7-OH, and 4ʹ-OH, respectively. Undoubtedly, the data above delineated a kaempferol backbone [35]. Meanwhile, the ¹H–¹H COSY motif (Fig. 3) of H-6″/H-5″/H-4″/H-3″/H-2″/H-1″ in conjunction with the ¹³C (Table S1) and HSQC NMR analysis (Fig. S5) revealed the existence of an α-rhamnose moiety [13,35], which was found to locate at C-3 due to the key HMBC correlation from H-1″ (δ_H 5.38) to C-3 (δ_C 135.07) (Fig. 3). Besides, four methine protons at δ_H 4.03 (1H, dd, J = 9.8, 6.0 Hz, H-7a), 4.17 (1H, dd, J = 9.8, 8.8 Hz, H-7b), 3.72 (1H, dd, J = 9.5, 6.0 Hz, H-8a), 3.89 (1H, dd, J = 9.5, 8.8 Hz, H-8b) were observed and the correlations of H-7a/H-7b/H-8b/H-8a/H-7a in ¹H–¹H COSY experiment (Fig. 3) confirmed the presence of a cyclobutane ring [34]. In addition, two 4-hydroxyphenyl groups were constructed owing to the observation of two sets of well-resolved proton signals: δ_H 6.83 (2H, br d, J = 8.0 Hz, H-2a, H-6a), 6.87 (2H, br d, J = 7.8 Hz, H-2b, H-6b), 6.52 (2H, br d, J = 8.0 Hz, H-3a, H-5a) and 6.54 (2H, br d, J = 7.8 Hz, H-3b, H-5b), together with δ_H 9.15 (1H, s) for 4a-OH and 9.17 (1H, s) for 4b-OH. The HMBC correlations (Fig. 3) from H-2a/6a to C-7a (δ_C 44.80), H-2b/6b to C-7b (δ_C 45.18), H-7a to C-9a (δ_C 173.53), and H-7b to C-9b (δ_C 175.24) suggested that the attachment of the two 4-hydroxyphenyl groups as vicinal arrangement in the truxinyl (head-to-head). Furthermore, the HMBC correlation between H-3″ (δ_H 5.03) and C-9a (δ_C 173.53) indicated the attachment of truxinate unit at C-3″ of l-rhamnose.

Fig. 3.

Key 2D-NMR correlations of compound 1.

The relative configuration of the cyclobutane ring was determined via the NOESY analysis (Fig. 3). The NOE correlations of H-7a/H-7b, H-8a/H-8b, H-8a/H-2a (H-6a), and H-8b/H-2b (H-6b) established cis-configurations between H-7a and H-7b, as well as between H-8a and H-8b, but a trans-geometry between H-7a and H-8a, which is typically conserved in a β-truxinate structure. In fact, the NMR data of the cyclobutane protons (Table S1) were in good agreement with those of biginkgoside A, a β-truxinic-type flavonol glycoside cyclodimers isolated from Ginkgo biloba leaves in our group [34]. Accordingly, the most probable stereoisomer is either (7aR*,7bS*,8aS*,8bR*)-1A or (7aS*,7bR*,8aR*,8bS*)-1B. In order to finalize the relative configuration of 1, the ¹³C NMR calculations for both 1A and 1B were performed by means of GIAO at the mPW1PW91/6–31 g (d, p) level [36,37]. The results obtained through DP4 + probability analyses predicted 1A to match well with the experimental data of 1 with 100.00 % probability (Fig. 4, Tables S2 and S3). The absolute configuration of 1 was then established by comparing the experimental and calculated electronic circular dichroism (ECD) spectra of 1A. The negative Cotton effects (CEs) around 227 nm and 318 nm, along with the positive CE at 252 nm in the experimental ECD spectrum, are congruent with those in the calculated one but antipodal to the isomer [(7aS,7bR,8aR,8bS)- 1] (Fig. 5, Table S4). Thus, the structure of compound 1 was finally ascertained as (7aR,7bS,8aS,8bR)-kaempferol-3-O-α-L-(3″-β-truxinate)-rhamnopyranoside. Naturally occurring compounds possessing a truxinate unit have been discussed in a preceding work [34]. The plausible biosythetic ptathway of 1 was most likely derived from a precursor (i.e., 8) and a p-coumaric acid via a [2 + 2]-cycloaddition reaction [34].

Fig. 4.

NMR calculations and DP4+ analysis outcomes of stereoisomers 1A and 1B using the means of GIAO at mPW1PW91/6–31g (d, p) level.

Fig. 5.

Experimental and calculated ECD spectra of 1 in MeCN.

Compound 2 was isolated as a pale yellow powder, and its positive-mode HRESIMS data showed a sodium adduct ion at m/z 601.1322 [M + Na]⁺ (calcd. 601.1316), which in conjunction with ¹³C NMR data (Table S5) corresponded to the molecular formula C₃₀H₂₆O₁₂. The absorption bands at 253 and 313 nm in the UV spectrum, and 3442 (hydroxy), 1655 (conjugated carbonyl), and 1511 (aromatic ring) cm⁻¹ in the IR spectrum were characteristic for a flavonol structure [34]. A comparison of the NMR data of 2 (Table S5) with those of 1 (Table S1) displayed close structural similarity. Unlike compound 1, the typical signals of the β-truxinate group attached at C-3″ were absent in 2. Instead, the presence of a (Z)-p-coumaroyl moiety was evidenced from the ¹H NMR data [δ_H 7.80 (2H, br d, J = 8.4 Hz, H-2ʹ″, 6ʹ″), 6.93 (1H, d, J = 12.8 Hz, H-7ʹ″), 6.82 (2H, br d, J = 8.4 Hz, H-3ʹ″, 5ʹ″), and 5.93 (1H, d, J = 12.8 Hz, H-8ʹ″)], the correlations between H-7ʹ″/H-8ʹ″, H-3ʹ″/H-2ʹ″ and H-6ʹ″/H-5ʹ″ in ¹H–¹H COSY spectrum and the HMBC correlations from H-7ʹ″ to C-2ʹ″ (6ʹ″) (δ_C 134.83) and from H-8ʹ″ to C-1ʹ″ (δ_C 128.51) of p-coumaroyl moiety. Furthermore, the coumaroyl was assigned to the 3″-position of the 3-O-rhamnosyl moiety since the rhamnosyl H-3″ signal was shifted downfield to δ_H 5.12 (1H, br d, J = 10.1 Hz) and C-3″ downfield at δ_C 75.74. According to all the NMR data, including the obvious HMBC correlations (Fig. S15), compound 2 showed great resemblance to another compound previously isolated from the buds of P. orientalis, known as kaempferol 3-O-α-L-(3″-E-p-coumaroyl)-rhamnopyranoside [27]. Therefore, compound 2 was determined as kaempferol 3-O-α-L-(3″-Z-p-coumaroyl)-rhamnopyranoside.

Compound 3 was obtained as a yellow powder. The molecular formula was recorded to be C₃₀H₂₆O₁₃ from the molecular ion peak [M + Na]⁺ at m/z 617.1273 (calcd. 617.1266) in the positive-mode HRESIMS data, which was in accordance with ¹³C NMR data (Table S5). The UV (λ_max 269, 315 nm) and IR (ν_max 3442, 1655, 1510 cm⁻¹) spectra were alike to that of compound 2, corresponding to a flavonol structure. The NMR spectra of compounds 3 and 2 were also similar (Table S5), though a main difference resided in an additional hydroxyl group of B-ring in 3. Specifically, a 1,3,4-trisubstituted aromatic ring at δ_H 7.40 (1H, br s, H-2ʹ), 7.37 (1H, br d, J = 8.3 Hz, H-6ʹ), and 6.97 (1H, d, J = 8.3 Hz, H-5ʹ) indicated the presence of a 3′,4′-dihydroxy B ring system in quercetin aglycone [38]. Another two differences were observed when comparing the ¹H NMR data of compounds 3 and 2: H-2″ was shifted downfield (Δ + 1.10 ppm), while H-3″ was shifted upfield (Δ − 1.14 ppm) in 3, indicating that the (Z)-coumaroyl moiety was located at C-2″ position. Likewise, compound 3 structurally bears a close resemblance to 2″-(E)-p-coumaroyl quercitrin with the only exception for the configuration of p-coumaroyl at C-2″ position [39]. Thus, 3 was identified as quercetin-3-O-α-L-(2″-Z-p-coumaroyl)-rhamnopyranoside.

2.2. Antibacterial activity of isolated compounds

Inspired by the fact that flavonoids are effective resources for designing antimicrobial drugs and the results of the preliminary antibacterial bioassay, the isolates 1–12 were initially evaluated for their inhibition effects against Newman S. aureus, with vancomycin-HCl (Van. HCl) was used as a positive control. The results (Table S6) demonstrated compounds 6 and 7 as active ingredients with MIC values of 4–8 μg/mL. As shown in Table 1, 6 exhibited a greater potent inhibitory effect against MRSA (MIC: 4 μg/mL) than the positive control (MIC: 8 μg/mL). Besides, both 6 and 7 significantly inhibited the growth of GREfm with MIC values of 0.5 and 1 μg/mL, respectively. The structure–activity relationship (SAR) of the related compounds showed that the flavonoid moiety connected to the p-coumaroyl group through the sugar and the hydroxy groups at the positions of C-5, C-7, and C-4′ were vital for the antibacterial activity, which is consistent with the previous reports [13,31]. Additionally, the configuration around the double bonds of the p-coumaroyl group plays a decisive role. Specifically, the Z-configuration of the p-coumaroyl unit in the R₂-position showed enhanced activity compared with the E-configuration and vice versa for R₃-position.

Table 1.

In vitro antibacterial activity of the isolated PTSs.

Compound	MIC (μg/mL)
	Mu50	Van (B)
6	4c	0.5c
7	16a	1b
Van. HCl	8b	256a

Mu50: methicillin-resistant S. aureus; Van (B): glycopeptide-resistant E. faecium. Different superscript letters (i.e., a-c) indicate significant differences among the compounds in each column at p < 0.05.

2.3. In-silico molecular docking study

In this context, the compounds derived from the present study were subjected to docking analyses against FabI (enoyl-ACP reductase) and PBP1/2 (penicillin binding protein), both of which are pivotal enzymes governing bacterial growth. Importantly, neither of these enzymes exhibit a substantial degree of conservation in the human host [40,41]. As a result, these enzymes could function as viable anti-bacterial targets, with a high likelihood of compounds effectively targeting these enzymes across various species.

FabI serves as a catalyst in the biosynthesis of fatty acids, essential components within the bacterial cell wall structure. The inhibition of FabI’s activity leads to the impediment of bacterial growth [42]. Conversely, PBPs are integral to the synthesis of peptidoglycan, a principal constituent of the bacterial cell wall. PBPs bind to β-lactam antibiotics, such as penicillin, thereby disabling their capacity to synthesize peptidoglycan. Mutations and the amplification of PBPs are orchestrated as defensive mechanisms by bacteria to mitigate the inhibitory effects exerted by β-lactam antibiotics [43,44]. Notably, mutations in the penicillin binding site of PBPs reduce their binding affinity to penicillin, leading to antibacterial resistance. This phenomenon, however, can be circumvented through the use of non-β-lactam antibacterial compounds that bind either to the penicillin binding site or an allosteric binding site. Such compounds have the potential to counteract the resistance mechanism. In essence, investigating the interaction of the studied compounds with FabI and PBPs sheds light on their prospective antibacterial activities.

The molecular docking outcomes exhibited a consistent pattern, wherein larger molecules demonstrated notably improved results. For instance, both di- and mono-coumaroyl derivatives prominently displayed superior binding affinities towards bacterial FabI enzymes (S. aureus, A. baumannii, and P. aureginosa) (Fig. 6), featuring a binding affinity spectrum ranging from −9.1 to −12.3 kcal/mol. Furthermore, this compound set also exhibited robust binding affinities to PBP enzymes, encompassing both the penicillin and allosteric binding sites. Equally noteworthy, the novel compounds 1–3 demonstrated impressive binding affinities towards FabI (−9.7 to −11.4 kcal/mol) (Fig. 6) and PBP2 (−8.6 to −9.5 kcal/mol). These outcomes suggest the dual potential of the new compounds as antibacterial agents, thereby regulating bacterial growth.

Fig. 6.

The binding of compounds 1 (blue), 2 (pink), and 3 (green) to (a) S. aureus FabI; (b) binding domain of S. aureus FabI; (c) A. baumannii FabI; (d) binding domain of A. baumannii FabI; (e) P. aeruginosa FabI; (f) binding domain of P. aeruginosa FabI.

Although compounds 6 (Z,E-platanoside) and 7 (E,E-platanoside) both displayed strong binding affinities to the FabI and PBP enzymes of S. aureus, a discernible difference emerged. 6 (FabI: −11 kcal/mol; PBP2: −9.2 kcal/mol; PBP2 allosteric site: −9.9 kcal/mol) demonstrated superior affinity compared to compound 7 (FabI: −10.6 kcal/mol; PBP2: −7.1 kcal/mol; PBP2 allosteric site: −9.3 kcal/mol) (Fig. 7). This observation corresponded with the established activity trend for 6 and 7 in the antimicrobial assay (Table 1). Furthermore, despite 1 demonstrating a higher in-silico binding affinity compared to 7, the binding position of 7 resides deeper within the binding pocket. Conversely, 1 binds near the opening of the binding pocket, suggesting a greater likelihood of easier dissociation for 1 as opposed to 7. Subsequently, surface plasmon resonance (SPR, an essential technique for hit validation) [45–47] studies on the interaction of the selected platanosides (e.g., 1 and 7, both samples are available) and FabI (a pivotal enzyme governing bacterial growth) revealed 7 has a higher affinity (K_D = 1.72 μM) than that of 1 (K_D = 5.38 μM) (Fig. 8), aligning with the inhibitory activity observed in antibacterial assays.

Fig. 7.

The binding of compounds 6 (orange) and 7 (purple) to (a) S. aureus FabI and (b) S. aureus PBP2.

Fig. 8.

SPR-based binding assay of 1 (a) and 7 (b) with FabI. The compounds were prepared at concentrations of 0.39 μM, 0.78 μM, 1.56 μM, 3.13 μM, 6.25 μM, and 12.5 μM, respectively. K_D is the equilibrium dissociation constant, a basic parameter to evaluate the binding property of the compound-receptor.

A pronounced theme illuminated by the docking outcomes underscores the compelling efficacy of the larger di- and mono-coumaroyl derivatives. These sizable compounds exhibited notable binding affinities across a diverse array of enzymes such as bacterial FabI and PBP1/2. For instance, the molecular docking results demonstrated that larger compounds like 6 and 7 displayed notably strong binding affinities to these enzymes despite marked variations in dimensions and amino acid residues within the binding domains (Table S7).

A comprehensive examination of the binding orientation of the di-coumaroyl compound, exemplified by 6 (Z,E-platanoside), to these enzymes, unveiled that the compound’s structural scaffold possessed favorable attributes that facilitated its fitting into the binding domains (Fig. 9). Compound 6, characterized by di-coumaroyl structure, comprised a highly flexible configuration encompassing a flavonol, a glycoside, and two coumaroyl moieties coupled with multiple aromatic rings, carbonyl, and hydroxyl functionalities. The outcomes indicated that the linkages among these structural components are rotatable, while the two coumaroyl appendages can be adjusted and folded accordingly to conform to the binding domains of diverse enzymes. The carbonyl and hydroxyl functionalities demonstrated a propensity to establish multiple hydrogen bonding interactions with specific amino acid residues, whereas the aromatic rings in the flavonol and coumaroyl initiated π-π, π-σ, π-alkyl, and π-sulfur interactions, effectively anchoring 6 within the binding domains.

Fig. 9.

Binding poses and interactions of 6 to FabI, PBP1, and PBP2.

In summary, the molecular docking analysis revealed that larger compounds, particularly di- and mono-coumaroyl derivatives, including the novel compounds 1–3, exhibited robust binding affinities towards the anti-inflammation and antibacterial targets. The observed binding affinities, along with their consistent correspondence to the experimental antimicrobial activity of compounds 6 (Z,E-platanoside) and 7 (E,E-platanoside), emphasize the promising potential of these compounds in addressing the intricate interplay between inflammation and bacterial infections.

3. Conclusions

In conclusion, the current investigation confirmed how plants belonging to the Platanus genus represent a reliable source of flavonoids to study the SAR and supply the platanosides. Herein, we only conducted the studies on PaW leaves. Given the favored antimicrobial activity, the EtOAc-soluble fraction of the 90 % MeOH extract of Spring PaW leaves was submitted for further isolation. As a result, three previously undescribed flavonol glycosides (1–3) were obtained. Interestingly, compound 1 is a rare naturally occurring flavonol glycoside possessing a β-truxinate motif [33]. Furthermore, platanosides 6 and 7 exhibited potent inhibitory effects against both MRSA and GREfm. Additional studies are warranted to clarify their antibacterial mechanisms, as well as the possibility of exploring the use of these platanosides in infectious diseases. The effectiveness of the platanosides against drug-resistant bacteria and organ damage to chemical insults from acetaminophen, a common cause of liver damage leading to liver transplant, makes these molecules a highly promising set of drug leads for the control of sepsis. The in-silico molecular docking results provide compelling evidence supporting the antibacterial potential of the compounds isolated in this study, including the novel compounds 1–3. The alignment between the demonstrated binding affinities and the established antimicrobial efficacy, as observed in platanosides 6 and 7, underscores the considerable promise embedded within these compounds, effectively poised to navigate the intricate interplay linking inflammation and bacterial infections.

4. Experimental

4.1. General experimental procedures

The optical rotations were measured using an Autopol IV automatic polarimeter. UV absorptions and IR spectra were recorded on a Hitachi U-2900E double beam spectrophotometer and a Thermo Scientific Nicolet Is5 FTIR spectrometer, respectively. NMR spectra were recorded by a Bruker Avance III 400, a Bruker Avance DRX-500, and/or a Bruker Avance 600 MHz spectrometer. Chemical shifts were expressed in δ (ppm) and referenced to the residual solvent signals. ESIMS were measured on a Waters UPLC H-Class-SQD or an Agilent 1100 series mass spectrometer. HRESIMS were recorded on a Bruker Daltonics micro TOF-QII or an AB 5600 + Q TOF spectrometer. Semi-preparative HPLC was performed on a Waters e2695 system with a Waters 2998 PDA and a Waters 2424 ELSD; ODS column (SunFire, 5 μm, 150 × 4.6 mm; SunFire, 5 μm, 250 × 10 mm; Cosmosil, 5 μm, 250 × 10 mm). Flash chromatography was performed on an EZ-L100-P200 (Lisure Science Ltd., Suzhou, China). Column chromatography (CC) was performed using silica gel (100–200 mesh, Kang-Bi-Nuo Silysia Chemical Ltd., Yantai, China), MCI (CHP20P, 75–150 μm, Mitsubishi Chemical Industries, Tokyo, Japan), and Sephadex LH-20 (GE Healthcare Bio-Sciences AB, Uppsala, Sweden). Silica gel-precoated plates (HSGF254, Kang-Bi-Nuo Silysia Chemical Ltd., Yantai, China) were used for TLC detection. Spots were visualized using UV light (254 and/or 365 nm) and by spraying with 10 % H₂SO₄-EtOH.

4.2. Plant materials

The leaves of PaW were collected from Zhangjiang area, Shanghai, China in May 2017 (spring). The plant samples were authenticated by Prof. Ze-Xin Jin (Taizhou University). The voucher specimens [Nos. 20170501, 20171001 (green leaves), 20171002 (fallen leaves) and 20180101] were conserved at the Herbarium of the School of Pharmaceutical Sciences, Taizhou University, China.

4.3. Bacterial strains

Newman Staphylococcus aureus, a methicillin-sensitive strain, and Mu50, a Healthcare-associated methicillin-resistant S. aureus (MRSA) strain isolated in Japan, as well as Van (B), a van-B phenotype glycopeptide-resistant Enterococcus faecium (GREfm) strain isolated in China were kindly supplied by Prof. Lefu Lan (Shanghai Institute of Materia Medica, Chinese Academy of Sciences, Shanghai, China).

4.4. Isolation and purification

The air-dried May leaves were grinded using an electric mill. 35 kg of dry powder were extracted four times using 105 L of 90 % MeOH. The obtained 90 % MeOH extract (5.6 kg) was then suspended in water and extracted with petroleum ether (PE), ethyl acetate (EtOAc) and normal butanol (n-BuOH), successively. All the fractions were concentrated by rotatory evaporator till dryness yielding PE (700 g), EtOAc (150 g), and n-BuOH (400 g) extracts, which were kept at room temperature for the phytochemical analysis.

The EtOAc fraction (150 g) was loaded to column chromatography over silica gel with a gradient-elution system of EtOAc/MeOH (neat EtOAc → 97.5:2.5 → 95:5 → 90:10 → 80:20, v/v) to yield five main fractions (Fr. 1–Fr. 5). Fr. 1 (22 g) was fractionated by an MCI column (from 30 % MeOH in H₂O to 100 % MeOH) to afford five sub-fractions Fr. 1A–Fr. 1D. Fr. 1B (6 g) was then successively purified using MCI column chromatography (CC) with gradient mobile phase of MeOH/H₂O (from 30 % to 100 %, v/v), Sephadex LH-20 CC, and semi-preparative HPLC [SunFire; flow rate, 3.0 mL/min]. Afterwards, compound 3 (2.2 mg, t_R = 27.5 min) was obtained using semi-preparative HPLC [MeCN/H₂O (containing 0.05 % TFA, v/v) 35:65, v/v], while compounds 1 (20.0 mg, t_R = 18.7 min), 2 (2.3 mg, t_R = 16.5 min), and 11 (22.5 mg, t_R = 14.7 min) were generated by semi-preparative HPLC [MeOH/H₂O (containing 0.05 % TFA, v/v) 60:40, v/v]. Fr. 1C (1 g) was loaded over Sephadex LH-20 CC with MeOH as eluent, and further separated by semi-preparative HPLC [SunFire, MeOH/H₂O (containing 0.05 % TFA, v/v) 60:40, v/v] to give 12 (12.0 mg, t_R = 14.3 min). In a similar way, compounds 4 (35.2 mg, t_R = 24.5 min), 5 (4.1 mg, t_R = 27.1 min), 6 (25.6 mg, t_R = 25.7 min) and 7 (90.3 mg, t_R = 28.9 min) were finally purified by semi-preparative HPLC [1.5 g, MeOH/H₂O (containing 0.05 % TFA, v/v) 70:30, v/v] from Fr. 1D (1.5 g) while compounds 8 (13.6 mg, t_R = 21.8 min), 9 (6.7 mg, t_R = 19.1 min), and 10 (2.9 mg, t_R = 20.5 min) ere garnered from Fr. 2 (2 g) under the same HPLC condition.

4.4.1. Kaempferol- 3-O-α-L-(3″-β-truxinate)-rhamnopyranoside (1)

Yellow amorphous powder; ${\left[\alpha \right]}_{\text{D}}^{20}-133$ (c 0.1, MeOH); UV (MeOH) λ_max (log ε) 271 (3.02), 341 (3.75) nm; ECD (c 1.37 × 10⁻³ M, MeCN) λmax (Δ ε) 227 (−9.7), 252 (+5.16), 301 (−4.22) nm; IR (film) νmax 3500, 1732, 1655, 1620, 1510, 1385, 1177, 1142, and 1082 cm⁻¹; ¹H NMR and ¹³C NMR data (in DMSO‑d₆), see Table S1; HR-ESIMS m/z 743.1975 [M + H]⁺ (calcd. for C₃₉H₃₄O₁₅, 743.1970, Δ = 0.6 ppm).

4.4.2. Kaempferol-3-O-α-L-(3″-Z-p-coumaroyl)-rhamnopyranside (2)

Pale yellow powder; ${\left[\alpha \right]}_{\text{D}}^{20}-39$ (c 0.1, MeOH); UV (MeOH) λ_max (log ε) 253 (2.88), 313 (3.92) nm; IR (film) ν_max 3442, 2925, 1682, 1655, 1615, 1511, 1382, and 1052 cm⁻¹; ¹H NMR and ¹³C NMR data (in CD₃OD), see Table S5; HR-ESIMS m/z 601.1322 [M + Na]⁺ (calcd. for C30H26O12, 601.1316, Δ = 0.9 ppm).

4.4.3. Quercetin-3-O-α-L-(2″-Z-p-coumaroyl)-rhamnopyranside (3)

Yellow powder; ${\left[\alpha \right]}_{\text{D}}^{20}+9$ (c 0.1, MeOH); UV (MeOH) λ_max (log ε) 269 (2.95), 315 (3.88) nm; IR (film) ν_max 3442, 2925, 2850, 1677, 1655, 1615,1510, 1382, and 1045 cm⁻¹; ¹H NMR and ¹³C NMR data (in CD₃OD), see Table S5; HR-ESIMS m/z 617.1273 [M + Na]⁺ (calcd. for C30H26O13, 617.1266, Δ = 1.1 ppm).

4.5. Computational calculations of ¹³C NMR and ECD data

The procedures of NMR/ECD calculations referred to the preceding works [37]. Conformational searches of selected compounds were performed by the Spartan 14 program (Wavefunction, Inc.) in the MMFF force field, with a threshold of 3.0 kJ/mol compared with the minimum energy conformer. The conformers were initially optimized at the b3lyp/6–31g(d,p) in gas. NMR calculations on the optimized conformations and TMS were performed using the means of gauge-independent atomic orbital (GIAO) at the mPW1PW91/6–31g(d,p) level, and the calculated shielding constants were directly submitted to statistical analyses with DP4+ probability by weighing the Boltzmann distribution rate [36,37]. The time-dependent density functional theory (TD-DFT) ECD calculations were conducted in MeCN at the b3lyp/6–31+g(d,p) level for all selected conformers of compounds [37]. Rotatory strengths for a total of 30 excited states were calculated. The ECD spectra were generated using SpecDis version 1.6 (University of Würzburg, Würzburg) from dipole-length rotational strengths by applying Gaussian band shapes with a σ of 0.3 eV² [48].

4.6. In vitro antibacterial susceptibility assays

The minimum inhibitory concentration (MIC) was evaluated by 96-well micro-dilution method based on previous reports [31,49]. In brief, overnight cultures of the bacteria were suspended into the Cation-adjusted Mueller-Hinton II broth (CAMHB) reaching a bacterial suspension of 0.5 McFarland turbidity [(1 ~ 2) × 10⁸ CFU/mL)]. Compounds were dissolved with DMSO to 2.0 mg/mL as stock solutions. All the samples were diluted with CAMHB to 64 μg/mL as the initial concentration. Further two-fold serial dilutions were performed to make their final concentrations in 96-well plates ranging from 64 μg/mL to 0.125 μg/mL. 100 μL of bacterial dilution (5 × 10⁵ CFU/mL) was distributed in each well, as well as negative controls, growth controls (containing culture broth with equal amounts of DMSO, without compounds) and positive controls [containing culture broth plus vancomycin hydrochloride (Van. HCl)]. The 96-well plates were incubated at 37 °C for 16 h and the MIC values of the tested compounds were defined as the lowest concentration to completely inhibit bacterial growth. All the tests were performed in triplicate.

4.7. Statistical analysis

The data were analyzed for their variances by one-way ANOVA using Graphpad prism (version 9). Differences were assessed by Tukey’s multiple comparison statistical analysis and determined to be significant at p < 0.05.

4.8. Molecular docking study

The ligand structures were optimized using the MM2 energy-minimized function in the Chem3D Ultra version 16.0. The crystal structures of the receptor proteins (Table S8) were obtained from Protein Data Bank [50,51]. AutoDockTools version 1.5.6 were used to prepare the receptor proteins and ligands for the molecular docking experiment. AutoDock Vina programme was used to perform the docking and calculate the binding affinity [52,53]. And lastly, the results were processed and analysed using the BIOVIA, Dassault Systèmes, Discovery Studio Visualizer, v21.1.0.20298, San Diego: Dassault Systèmes, 2023.