Human serum albumin-specific recognition of the natural herbal extract of Stryphnodendron polyphyllum through STD NMR, hyphenations and docking simulation studies

Abstract

Over the last two decades, new and more advanced strategies that help in the rapid screening and identification of new ligands for a specific macromolecule have become an important domain. From this viewpoint, the effectiveness of STD NMR, Tr-NOESY, and STD-TOCSY has been utilized to evaluate the binding potential of the natural extract of Stryphnodendron polyphyllum, used as a herbal medicine in Brazil, towards human serum albumin. Moreover, 1D-DOSY experiments have also been carried out for the discriminations of different molecular weight compounds present in this extract. Following the STD, Tr-NOESY, and TOCSY analysis, a hyphenated system comprising LC-SPE-NMR was utilized to see the complete structural assignments through 2D spectra. The combined results from NMR spectroscopy and separation methods provided myricetin-3-O-rhamnopyranoside (1), quercetin-3-O-glucopyranoside (2), quercetin-3-O-xylopyranoside (3), and quercetin-3-O-rhamnopyranoside (4) as the active site blockers. Moreover, epitope results and additional Tr-NOESY cross peaks suggested the presence of the flattened conformations of these ligands within the ligand–HSA complex through the edge protons. Similarly, STD competition studies with the ligand–HSA complex were demonstrated by varying the concentration of spy molecule that selectively binds with Sudlow's site II. Finally, docking simulations targeting both Sudlow sites (I and II) were performed, which interestingly mimic the STD competition results and showed that these compounds (1–4) are more prone towards binding site-1 inhibition. Therefore, we suggest that the sequence of techniques presented in this study can be considered as a simple and fast analytical tool for screening natural extracts to get better leads against any specific target.

---

Introduction

The recognition of ligands prior to separation from a folk medicinal extract is a big deal for new drug discovery, especially where the components are sensitive towards solvents, light or some isolation methods. Nowadays, herbal extracts are being produced by various pharmaceutical methods to fulfil the growing need for medication and nutritional supplements,¹ for dietary purposes as well as in the cosmetics industries.^2,3 There is great demand for drugs derived from natural products (herbal medicines) in the global market, regardless of country or region.⁴ Therefore, the recognition of a ligand of interest from a bioactive mixture is a key to new drug discovery, and more significantly, this emerging industry (herbal medicine) demands high standards of production with more economical and environmentally friendly new methods.

Since herbal extracts are sensitive towards environmental changes, NMR seems to be an ideal choice that offers the economically and environmentally healthy conditions required by the drug discovery process.⁵ Unlike other biochemical assays, NMR demands neither prior knowledge about the function of the protein and/or protein labeling nor target specific instrumental set-up.⁶ On the other hand, the screening techniques (mass spectrometry, circular dichroism, ultraviolet spectroscopy, etc.) used to analyze organic mixtures suffer from a number of drawbacks, like lack of information regarding proton exchange,⁷ unable to resolve overlaps in signals⁸ and demanding compounds with optical absorption.⁹ Although the aforementioned techniques are widely used in drug discovery, it is often not easy to develop a simple and rapid ligand screening method using these techniques because of differences in the molecular polarities, solubility, and presence of isomers, as well as the lack of optical absorption, etc. In addition, structure assignments of unknown compounds and their contribution towards the biological activity are also blurred.

Molecular recognition lies at the heart of all biological processes involving the human body since almost all body functions involve ligand–receptor and receptor–receptor type interactions. Therefore, ligand–protein interaction is an important primary domain for discovery of new drugs. Over the past two decades, the screening of natural product mixtures with receptors by using NMR as a detector has been extensively reported.¹⁰ Among the different NMR methods used for studying ligand–receptor interactions, STD NMR¹¹ has become one of the most commonly used methods for ligand recognition,^10,12 analyzing molecular kinetics and determining the dissociation constants¹³ of bound state ligands. Besides screening ligands, NMR has been demonstrated for fingerprinting the content of complex mixtures.^14,15 However, in all such studies the structural characterization remains a bottleneck,⁶ unless some separation methods are involved. However, the use of hyphenated methods like LC-UV-SPE-NMR can beautifully eliminate this characterization bottleneck, particularly where the NMR is unable to make distinctions between the signals due to overlap.¹⁶ Recently, Ferreira and co-workers¹⁷ supplied an excellent way to find leads from crude, through the significant results obtained by utilizing direct NMR techniques following the docking simulation. Thus, this development has made NMR arguably the most important tool for the recognition of ligands from complex natural products.

In the present study, we have employed STD NMR and Tr-NOESY experiments to study the binding of ligands from Stryphnodendron polyphyllum fully bloomed flower extract towards HSA protein. Human serum albumin is the most widely studied^18,19 and the most abundant protein in the body, in the range 30–50 g L⁻¹, constituting around 50% of total serum protein by concentration.²⁰ It is a well-known protein, responsible for the maintenance of blood pH²¹ and osmotic pressure¹⁸ and the transportation of small molecules; however, it possesses a great affinity for fatty acids.²² HSA consists of three structurally similar helical domains with a total of 585 amino acids, where each domain consists of two sub-domains, named A and B, connected by a flexible loop.²³ Six helices are present in the A subdomain while the B subdomain contains four helices, where a number of binding sites within the subdomains IIA and IIIA specific for Sudlow's site I and II, respectively, are present.²³

The screening of ligands from a natural product was conducted by using direct NMR applications along with hyphenated methods. To date, to the best of our knowledge, no single study that has reported the interactions of this highly bioactive plant towards a specific target, HSA protein. Here, we have selected an ethyl acetate extract of Stryphnodendron polyphyllum (the plant commonly known as barbatimão). In folk medicine, the extract of barbatimão is frequently used in Brazil for the treatment of leucorrhea, diarrhea, sepsis and inflammatory disorders, as well as for blood clotting and wound healing purposes.^24,25 The antibacterial and antioxidant activity of barbatimão extract is also reported.²⁶ In the near future, this approach may open a way to new drug discoveries based on data from ligand–protein interactions, and can provide further evidential proof of the usefulness of NMR.

Results and discussion

Drug discovery is the most important and perhaps the most challenging task to pharmacists and chemists. In most cases, the synthesis of the medicinal chemistry oriented bioactive compound is expensive and time-consuming, especially with a complex molecule possessing several equivalent reactive positions. Nevertheless, molecular recognition directly from a bioactive material may somehow facilitate this tough task.

Soon after the provision of the screening strategies, and to have an idea regarding the size of different components within this fully bloomed flower extract, an alternate approach to bio-fractionation, a 1D-DOSY experiment, was selected. It is well understood that the intensities of protons belonging to low molecular weight compounds are reduced or even completely wiped out by increasing the gradient strength.²⁷ Therefore, rather than utilizing bio-fractionation through the cut-off membrane, a series of 1D-DOSY analyses by varying gradient strength (5%, 20%, 40%, 80% and 95%) was employed to perform the fractionation based on molecular weight. Based on this series of 1D-DOSY spectral results, the assumption was made that there were at least two types of components present, as shown in Fig. 1. The changes in signal due to the decrease in gradient strength from 40% to 5% are indicated by arrows with an asterisk (*) and psi (Ψ), providing evidence of an almost similar component between δ = 7.9 and 12.4 ppm (Fig. 1). Spectral results corresponding to the gradient strength (20% and 80%) are not given because no recognizable changes were found in them. Therefore, after obtaining the realistic 1D-DOSY results, we assumed that there were smaller to medium size components (full spectra Fig. S-1 in ESI†), and the decision about whether the interaction originates from a smaller size or medium compound will be demonstrated later by STD NMR.

Fig. 1 Comparison of the 1D DOSY spectra of extract, performed by using big delta = 0.6 s, little delta = 1 ms, eddy current delays = 5 ms and with the following values of gradient strength: 95% 40% and 5%. Arrows with * and Ψ represent the small and medium size molecular weight compounds, respectively.

In the first step of this recognition process, screening was applied to 1 mg of fully bloomed flower extract of Stryphnodendron polyphyllum using STD NMR. First of all, a ¹H-NMR spectrum in 5 μM HSA (Fig. 2A) was recorded that represents a ligand–protein complex system and acts as a control while, in Fig. 2B suppression of the HOD signal was carried out to see the hidden anomeric protons under the large HOD signal. Given complete numbering was done by 2D NMR data obtained (not shown here) thereby, separation methods. The STD NMR clearly showed (Fig. 3) that the aromatic, as well as the sugar region (2.8–5.5 ppm) of the spectra, is involved in major binding. Similarly, further evidence of binding was also observed in the line broadening and change in relaxation rate as a result of protein addition.

Fig. 2 The ¹H-NMR spectra ((A) normal proton and (B) with HOD signal suppression) of the ethyl acetate fraction of the fully bloomed flower extract of Stryphnodendron polyphyllum in a 50 μM solution of HSA protein acquired with a Bruker 600 MHz AVANCE III spectrometer with a cryogenic TCI probe at 298 K. Signals are represented with empty and filled squares for compound quercetin-3-O-glucopyranoside (2) and quercetin-3-O-xylopyranoside (3), respectively, and similarly, with empty and filled circles for compound myricetin-3-rhmanopyranoside (1) and quercetin-3-O-rhamnopyranoside (4), respectively. Complete nomenclature numbering became possible after the evaluation of 1D and 2D spectra through a separation method.

Fig. 3 ¹H-NMR spectrum with HOD signal suppression (A), STD NMR spectrum (B) and the magnified region (δ = 3.0–8.2 ppm) of the STD spectrum in a rectangle (C). All experiments were performed using 1 mg of extract in a 50 μM protein solution and in a 3 mm NMR tube on a Bruker 600 MHz spectrometer. To avoid the crowding as a result of numbering, sometimes, a single arrow with varying colors is shown, where navy blue, sky blue, red and black represent the protons belonging to compounds 1, 2, 3 and 4, respectively. An asterisk represents the signals from the protein and some impurities. More detail is provided in the experimental conditions.

Fig. 3 presents the proton NMR spectrum as a reference at the top, the STD spectrum in the middle and the bottom spectrum is the magnified region (chemical shift δ = 3.0 to 8.2 ppm) of the STD spectrum in the dotted rectangle, for more clarity. The observations that come out from the figure are: some signals of impurities are present in the region δ = 2–2.5 ppm, background signals from protein, and almost all signals showed some level of enhancement—an indication of binding to HSA protein. Over the last decade, it has been shown that the signals in the STD spectrum represent the engagement in binding with some binding residues of the receptor.^5,17 Conversely, the absence of signals in the STD spectrum shows that there is no binding/affinity towards the respective receptor.^5,11–17

The amplification gradient is the result of a saturation received by the particular proton from the receptor, and it also determines the vicinity of the ligand to the binding cavity. Thus, the intensity of the signal depends on two factors: (A) the length of saturation time, and (B) the number of protons.^13,17 The STD build-up stack was obtained by performing experiments with varying saturation times (1 s, 2 s, 3 s, 4 s, and 5 s), and the outcome stack model of the spectra is shown in Fig. S-2 ESI.† Group epitope mapping calculations were made by comparing the individual proton integrals, and normalizing others with respect to this, which was earmarked for 100%. After evaluating a series of experiments performed at different saturation times, a saturation time of ca. 3 s was discovered to be most efficient where all signals have excellent resolution and intensity. Possible recognizable compounds with their respective amplification factors, after comparing the data from STD NMR, Tr-NOESY, TOCSY, and the separation method, are shown in Fig. 4, where the sugar moieties of all compounds are represented separately. Intriguingly, the interaction footprints from these molecules revealed that the protons at position H-6 (6.26 ppm) and H-8 (6.44 ppm) showed singlets. Following this, the second largest signals were obtained from the sugar region.

Fig. 4 Structures and relative amplification factors (AF) in percentage with absolute numbering for the recognizable compounds, are shown. Where structure (1) with substituent A represents myricetin-3-O-rhmanopyranoside, (2) with B is quercetin-3-O-glucopyranoside, (3) with C is quercetin-3-O-xylopyranoside and (4) with D is quercetin-3-O-rhamnopyranoside. For more clarity of the amplification factors, all sugars are presented separately.

A 100% STD effect was obtained from the H-6 protons of all the recognizable compounds, showing a major involvement in the interaction. However, the H-8 protons of these compounds contributed a 94% STD effect. Among the other aromatic signals, position H-2′/H-6′ of myricetin-3-O-rhamnoside (1) was taller with a 73% STD effect, while the same protons from quercetin-3-O-rhamnoside (2) provided between 46% and 50% STD effects. So, the protons H-2′/H-6′ from compound 1 engaged in stronger binding as compared to the same protons from compound 4. However, the protons H-6′ and H-2′ from compounds 2 and 3 revealed different amplification factors of 13% and 27%, and 27% and 39%, respectively. Conversely, the proton at position H-5′ remained prominent with 63%, 39% and 61% STD effects for the compound 2, 3 and 4 respectively. From these aromatic epitope results, it is clear that the only sides (edges) of all the recognizable compounds (1–4) were towards the binding cavities of the protein, thereby receiving more saturation transfer, and hence the larger interaction.

Likewise, the epitope results remarkably showed that both rhamnose units could provide better interaction compared to the glucose and xylose units. In addition, all the axial protons from these molecules provided greater amplification and obviously good interaction towards the HSA protein. Among the equatorial protons; H-3′′ from compound 1 and 4 could provide noticeable STD effects of 82% and 77%, respectively. The second highest STD effect was found from the sugar moiety of compound 1, H-4′′ and H-5′′ provided well-nigh effects of 39% and 43%, representing its moderate interactions. Similar protons from compound 4 could provide almost the same effect of 43%. By combining the results of compounds 1 and 4 for position H-4′′ and H-5′′, the interacting behavior towards HSA was found to be similar. However, the signal for H-4′′ from compounds 2 and 3 also provided STD effects of 41% and 36%, respectively. H-5′′ for 2 provided 41% STD, and both protons (H-5a′′ and H-5b′′) belonging to 3 showed a bit higher binding with 57% and 25% STD effects, respectively. Interestingly, the anomeric protons belonging to all compounds showed a negligible STD effect between 1–5%. From this epitope mapping, it can be concluded that the parts of the sugars that are towards the flavones (aglycone) skeleton showed less binding than the part far from it. Thus, these epitope results revealed that both edges are in intimate contact with the binding sites.

Furthermore, Tr-NOESY experiments were carried out to identify the bound conformations of these STD-exposed interacting molecules within the HSA–ligands complex. Tr-NOESY is the best-known experiment for the evaluation of the bound conformations for the ligand–receptor complexes at equilibrium, where the bound ligand's geometric information is transferred to unbound.^5,17,28,29 In NOE, the major distinguishing parameter between the bound and unbound ligands is the correlation time (τ_c); small molecules possess short τ_c and a longer relaxation rate. However, in the bound state this shorter correlation time (τ_c) converts to the longer—a characteristic of large mass molecules, resulting in a strong NOE. Similarly, the Tr-NOESY build-up rate in solution was also much faster when compared to NOESY.^17,28 In 2D NOESY without the addition of protein, all cross peaks had positive signs when compared to the diagonal (Fig. 5A). However, in the sugar region some signals also revealed negative signs, which might be due to the larger size or an uncorrected phase. Nevertheless, in Tr-NOESY all peaks represented the same significance to the diagonal and, more importantly, the buildup rate was much faster when matched to NOESY without protein, giving evidence of protein binding (see Fig. 5B). However, the mixing time for both NOE experiments was kept the same at 400 ms. In the next step, the bound conformations of these compounds within the ligands–receptor complex were studied.

Fig. 5 Two-dimensional NOESY spectrum without HSA protein (A) and Tr-NOESY (B) with Bruker standard phase sensitive pulse sequence for the extract in D₂O–methanol-d₄ (85 : 15 v/v) and protein in PBS buffer solution with pH 7.4. For the Tr-NOESY, the pulse sequence was modified by adding the 3500 Hz spinlock for the removal of background protein signals before the first π/2 pulse. Both spectra were acquired by applying a 400 ms mixing time after evaluating a numbers of different mixing time spectra on the Bruker 600 MHz NMR spectrometer with a cryoprobe at 298 K. In NOESY, the cross peaks observed were opposite to the diagonal while for the Tr-NOESY spectrum both had the same sign. The signals are represented by the blue and red squares for compounds 2 and 3, and blue and red circles for compounds 1 and 4, respectively.

Protons H-6′′, H-5′′, H-4′′ and H-3′′ of compound 1 and 4 were not present in the NOESY experiment; however, in the Tr-NOESY experiment these were very prominent (Fig. 5B). In fact, the flexibility of these protons has been reduced in the bound conformation. Thus, the short correlation time of these protons was no longer existed in the bound state, thereby interacting with binding cavities and hence affording the strong negative NOE results. Similarly, in the aromatic region, the strong signals from protons H-6 and H-8 were also absent, but they came into sight due to binding to the protein in the Tr-NOESY experiment. Interestingly, some protons from the glycone part of compound 2 and 3, like H-2′′ and H-3′′, also showed strong negative NOEs in the Tr-NOESY spectrum. It is worth noting that the signals from the anomeric protons were present in both spectra (Fig. 5A and B). The Tr-NOESY experiments mimic the epitope mapping STD results by providing similar results. Therefore, using the combined evidence from both the STD and Tr-NOESY experiments, we argue that the portions of the small molecules (1, 2, 3 and 4) that are engaged in strong binding are the edges of each molecule. Therefore, by taking compound 1 as a reference, the bound conformations have been suggested through 3D models in Fig. 6A–C, which show a flattened type existence in the binding cavity.

Fig. 6 The suggested 3D conformations obtained after the STD amplification factor calculations and Tr-NOESY cross peak information for the bound state of myricetin-3-O-rhamnopyranoside in a solution of the HSA-myricetin-3-O-rhamnopyranoside complex. In all conformations ((A), (B), and (C)), the structure is in the extended form. The numbering in (B) and the green protons in (C) represent the strong bindings.

For further clarification about the structures and their bindings, the same sample (with protein) was considered for 2D TOCSY and STD-TOCSY experiments. The 2D TOCSY experiment is well comprehended for complex structure identification and has the added advantage of determining the ligand protons proximity to the neighboring binding sites of the receptor.^11,30 Signal intensities were the primary source for STD-TOCSY establishments; by comparing the off-resonance to the STD spectrum, more intense signals were considered to indicate more intimate contact towards protein binding sites, while the fewer intense signals were more solvent exposed. Intact protons that were present in the STD NMR spectra experienced the same saturation, resulting in strong cross peaks. Interestingly, the aromatic region (Fig. 7A and B) as well as the glycone part (Fig. S-3 ESI†) protons, showed strong cross peaks that could be attributed to the closest approach to binding sites in this ligands–protein complex.

Fig. 7 Two-dimensional standard TOCSY (A) and STD-TOCSY spectrum (B) acquired using a 50 μM solution of HSA in 1 mg extract solution in methanol-d₄ and D₂O (15 : 85 v/v). MLEV-17 sequences were employed to acquire these experiments by adding a mixing time of 60 ms at 298 K with a 600 MHz spectrometer. In STD-TOCSY, the parameters for the saturation of the protein were kept the same as were used earlier for the STD experiments. The magnified region (δ = 6–7.8 ppm) of both the standard TOCSY and the STD-TOCSY are shown and signals are represented by the blue and red squares for compounds 2 and 3, and blue and red circles for compounds 1 and 4, respectively.

Among the aromatic protons of the recognized compounds, H-6′ and H-5a′′ of compounds 2 and 3 (small blue and red squares, respectively) were prominent, and they could provide evidence of binding in STD-TOCSY. Similarly, H-2′ from compounds 1 and 4 (small blue and red circles, respectively) and protons H-2′/6′ from compound 1 provided excellent agreement with the STD spectrum and hence interactions (Fig. 6B). Intriguingly, signals from H-6 and H-8 positions from these compounds (1, 2, 3 and 4) mimic exactly the STD and Tr-NOESY spectral results of having strong cross peaks, and thereby strong contact with binding sites. Thus, 2D STD-TOCSY again provided the strongest evidence of flattened conformations during the bindings of the recognizable ligands. In addition, the methyl signals of a rhamnose sugar from compounds 1 and 4 provided the same results (Fig. S-3 ESI†). In the STD-TOCSY results, some signals from the 2D TOCSY spectrum were missing, which might be due to a lower degree of saturation received from the protein because of the large distance from the binding site or being present in some shallow site.

At this stage, on the premise of STD NMR, Tr-NOESY and STD-TOCSY experiments, the structural skeleton was almost clarified and evidence of the presence of a flavonoid type skeleton with attached sugars was provided, but no guarantee about the absolute structural assignments was provided yet. However, there was a substantial need for some separation method that could lead to a full characterization and a certainty about the number of components involved in their binding. Thus, the obvious choice was LC-SPE-NMR, which not only enables characterization of the structure, but is also helpful in providing the exact number of components involved as well.^31,32 To accomplish this purpose, 3 mg of the extract was dissolved in 1 mL of methanol–water (85 : 15 v/v). Afterwards, a separation method with shorter run-time and high resolution was observed (Fig. S-4 ESI†). Using the gradient solvent with varying concentrations and with the injection of 15 μL of sample with a flow rate of 1 mL min⁻¹ led to a chromatogram (see Fig. S-5 ESI†) comprising 13 peaks (4 major and 9 minor). After 30 times of trapping the four major compounds on the general phase cartridge of SPE system, each peak was eluted to the NMR probe using 230 μL of CD₃OD. All of the trapped compounds provided very clear 2D NMR spectra (COSY, HSQC, and HMBC) after matching up this spectral data with earlier reported results,^33–35 which led to the conclusion that myricetin-3-O-rhamnopyranoside, quercetin-3-O-glucopyranoside, quercetin-3-O-xylopyranoside and quercetin-3-O-rhamnopyranoside are compounds 1, 2, 3 and 4, respectively.

STD NMR titration studies

After getting the information related to the bound conformations of these interacting ligands towards the HSA through STD NMR and Tr-NOESY, the following step was conducted to observe the binding sites for each ligand. In this perspective, a competitive ligand with a well-known and fully described binding site (a so-called spy molecule) was necessary to make a comparison with the unknown molecule at the corresponding site. To assess this goal, we applied STD NMR competition studies based on the signal intensities. STD competition studies can confirm whether this spy and unknown ligand compete for the same binding site on the protein or not. The addition of a spy molecule in a solution containing a ligand can either reduce or completely abolish the signals of that ligand if the spy is a better binder to the specific site. Conversely, if this ligand itself binds tighter than the spy, it results in the reduction or complete elimination of the spy signal in the STD experiment. Despite these possibilities, if both compounds have different binding sites, none of the signals will be affected.

Thus, in STD NMR competition studies, L-tryptophan is used as a reference (spy molecule) that preferably binds to Sudlow's site II of human serum,³⁶ and has been extensively applied to site-selective competition studies.^37,38 In this study, the unknown amounts of the four natural products trapped through the SPE system were further utilized by making 1 : 100 and 1 : 350 molar solutions of protein to L-tryptophan ratio for each single STD experiment after obtaining 1 and 2D spectra. All these competition experiments were performed by using a fixed concentration of HSA (50 μM) for each ligand. By close observation of the spectra, it could be presumed that the STD intensities of these ligands protons remained the same, even though, when increasing the molar excess of the spy molecule (Fig. 8 and S-6–S-8 ESI†). These outcomes of the competition study strongly suggest that no ligand targets the selected site of the HSA, subsequently binding with some other sites. Due to lack of sufficient quantities (from the separation system), further STD competition experiments for the hunt of Sudlow's site I were not performed with other spy molecules that selectively bind to the respective locus. Therefore, rather than exploring the site selectivity of these ligands towards Sudlow's site I through STD competition experiments, further observations were made through docking simulation and molecular dynamics studies.

Fig. 8 STD competition experiments showing only the aromatic region (δ = 5.7–8.1 ppm) recorded with a Bruker 600 MHz spectrometer at 298 K. (A) STD spectrum of the HSA–quercetin-3-O-glucoside (2) complex (HSA 50 μM); (B) STD spectrum of the complex system with HSA vs. spy molecule (1 : 100 molar excess) and compound 2; and (C) STD spectrum of the complex system with HSA vs. spy molecule (1 : 350 molar excess) and compound 2. In both spectra (B) and (C), the signals of 2 are not affected by increasing the concentration of the spy molecule. Signals from the spy molecule (L-tryptophan) are represented by S, whereas those from quercetin-3-O-glucoside are un-numbered.

Docking simulation

With a specific end goal to investigate the binding sites for the recognizable ligands, the next question was whether the ligands will prefer Sudlow's I or II for their best-fit accommodations. Both sites (Sudlow's I and II) were targeted here for docking simulation and compounds were analyzed individually within both binding sites for their binding potentials. Molecular docking simulation is a computational strategy used to tie up the ligand with the targeted protein in order to spot their interaction patterns, binding affinities, and mechanism of action. Here, to uncover the binding and inhibitory capability of these sugars containing the flavonoid skeleton towards the HSA, the selected amino acid residues are given in the Experimental section. Each compound was individually analyzed for Sudlow's site I and II.

With respect to site-1, compound 1 docked within the binding groove of human serum albumin with active residues. Arg257, the most interacting residue of the binding site, interacted via a strong hydrogen bond of 2.08 Å that linked the NH of Arg257 to the hydroxyl group of the ligand's sugar. Another quite weak hydrogen bond of 3.31 Å was also observed in between the Arg222 NH and the OH of the ligand's resorcinol, as shown in Fig. 9. Similarly, compound 2 (Fig. S-9a ESI†), displayed more intense interacting capability with three strong hydrogen bonds and the highest binding free energy of −7.3211 kcal mol⁻¹ (Table 1). These bonds are mediated by the carboxylic oxygen of Glu153, the guanidine NH of Arg257 and the guanidine NH of Arg218 towards the hydroxyl hydrogen, the hydroxyl oxygen of the sugar and the hydroxyl oxygen of the resorcinol moiety at two extremes of the ligand molecule.

Fig. 9 3D bound conformations of the myricetin-3-O-rhamnoside, generated by Molecular Operating Environment (MOE) 2011.10 within the HSA–myriscetin-3-O-rhamnoside complex for Sudlow's site I (A) and II (B). The PDB file for the HSA complex with amino camptothecin was taken from the protein data bank with ID: 4L8U. The best-fit accommodation of this structure after 30 different pose evaluations provided almost similar results as deduced earlier through STD AF and Tr-NOESY studies, i.e. a flattened extended structure is involved in binding.

Table 1 MOE docking and molecular dynamic results for the compounds (1–4) within binding site-1 of human serum albumin with their free binding energies

Site I
Compounds	S	E-score1	E-refine	E-score2
1	−6.969235	−15.22145	−17.91924	−6.969235
2	−7.32114	−12.14727	−27.00223	−7.32114
3	−6.6642	−12.1306	−25.2395	−6.6642
4	−6.940517	−13.40245	−23.50109	−6.940517

---

However, compound 3 also showed comparable results with the same number of hydrogen bonds supporting its residency within the binding site. Arg257, Glu153 and Ser192 via their NH, OCO and OH were found to be involved in hydrogen bonding with the hydroxyl oxygen of the ligand's hexose sugar, as depicted in Fig. S-10a ESI.† The distances of interaction that enable the compound to be selected as effective are 2.6 Å, 2.21 Å, and 2.31 Å, respectively. Moreover, a hydrophobic interaction mediated from the CH of Glu292 captured the resonating resorcinol of compound 3, resulting in a CH⋯π interaction. Conversely, compound 4 behaved quite differently with different interacting residues within the same binding pocket and with good binding energy of −6.9405 kcal mol⁻¹ (Table 1). The compound's catechol OH interacted with the carboxylic oxygen of Asp451 with a binding length of 2.03 Å, however, a bond of 2.14 Å was observed in between the hydroxyl hydrogen of the hexose sugar and the hydroxyl oxygen of Ser192. Beside this, Ala291 also played a supportive role in ligand binding via hydrophobic interaction, established via its CH towards the resorcinol ring of the molecule (displayed in Fig. S-11a ESI†).

On the other hand, when the same compounds were screened against site II, a quite different binding pattern was exposed, as shown in Table 2 with different conformations. Compound 1 (Fig. 9B) placed itself at the edge of the defined binding site and interacted with the outside residues. Fig. 9B shows the compound placement with an active site (purple) and outside (gray) residues. It interacted strongly through hydrogen bonds of 1.88 Å, 1.99 Å, and 2.36 Å that originated from Asn429, Lys519 and Arg428 towards the OH of pyrogallol, the ketonic oxygen of the centered ring and the OH of the hexose sugar moieties of compound 1. Furthermore, the active site Lys190 and Leu115 also interacted via hydrogen bonding of 2.23 Å and 3.23 Å with the ligand to support its binding at the boundary of the protein's binding pocket. The orientation of compound 2 (Fig. S-9b ESI†) within binding pocket 2 ensured the correct placement of the molecule surrounding its hot spot residues. Arg117 and Asp183, two important active site residues, formed hydrogen bonds of 1.92 Å and 2.22 Å with the hexose sugar oxygen and the catecholic OH groups of the compound, depicted in Fig. S-9b and S-10b ESI.† Compound 3 and 4 presented in Fig. S-10b ESI† and Fig. S-11b,† respectively, also docked within the binding groove catching the protein's hot spot residues. Arg114, Arg117, and Arg186 were found to be the main players in compound 3's binding through their guanidine NH. Agr114 mediated two strong hydrogen bonds of 2.38 Å and 2.27 Å towards the ether-linked oxygen of the hexose sugar. Moreover, Arg117 and Arg186 radiated a single (2.45 Å) and three hydrogen bonds (1.99 Å, 2.73 Å and 2.78 Å) in the direction of the sugar's hydroxyl oxygen and the carbonyl oxygen of the resorcinol and the centered ring, respectively. Compound 3 with stronger bonding and higher binding free energy (−6.9428 kcal mol⁻¹, Table 2) is considered to have more potential for this site of HSA, ensuring its greater blocking ability for site-2. Moreover, compound 4 seized Arg114 (1.92 Å) through hydrogen bonding with a binding interaction of 1.92 Å with a total free energy as shown in Table 2.

Table 2 MOE docking and molecular dynamic results for the compounds (1–4) within binding site-2 of human serum albumin with their free binding energies

Site II
Compounds	S	E-score1	E-refine	E-score2
1	−6.483576	−12.29182	−20.07261	−6.483576
2	−6.516641	−11.81531	−8.509714	−6.516641
3	−6.9428	−13.7778	−14.4524	−6.9428
4	−6.61349	−11.05528	−17.81524	−6.61349

---

Conclusions

In this study, direct NMR spectroscopic methods are suggested as a tool for the rapid and conclusive characterization of ligand recognition from the natural extract to pursue new leads. Here, the combined use of STD-NMR, transfer NOESY experiments, and molecular docking furnished details on the molecular recognition of the sugar-containing flavonoid ligands, specifically, it delivered details of bound conformations of binding ligands in the complex. Moreover, similar STD NMR experiments were performed on ligand–HSA complexes and the spy molecule, which were designed to observe the competition for Sudlow's site II. It is worth noting that the combined results from the STD competition experiments and the docking protocol are in good agreement and have empowered us to conclude that these compounds do not compete for the same binding site as L-tryptophan does. From in-depth analysis of the compounds' orientations, affinities, interaction patterns and binding energies of two sites of human serum albumin, it can be concluded that the sugar-attached compounds are more prone towards binding site-1 inhibition. A number of hydrogen bonds with an active site residues ensured their inhibitory potency and confirmed the stability of the protein–ligand complexes. Compounds 2 and 3 were found to be most active for site-1 and site-2, respectively, from analysis of their docking results. Similarly, the epitope mapping results were also in close agreement with the bound conformation results of molecular docking, and revealed a collective statement that these structures remained more flattened instead of twisting in the complex and showed a greater affinity from their edges. Therefore, the structural information gleaned from studying this ligand–protein interaction may help to develop new drug designs and better ligands/inhibitors for HSA protein. Moreover, this approach can also be extended to get the metabolomic fingerprinting of herbal extracts available in the market under some trade names for their quality premises and efficacy.

Experimental section

Sample preparation

For the NMR experiments, 1 mg of fully bloomed flower extract from Stryphnodendron polyphyllum was dissolved in D₂O and CD₃OD, (85 : 15% v/v ratio) while the HSA protein (CAS number: 0070024907, essentially fatty acid free, ≥97% pure, purchased from Sigma-Aldrich, Brazil and used without further purification) solution was prepared in sodium phosphate buffer (pH 7.4) in D₂O. A freshly prepared solution of extract and protein (50 μM) in a 3 mm NMR tube was used for STD NMR and Tr-NOESY experiments; however, for the TOCSY and STD-TOCSY the same solution was used the next day. For the competition STD titration experiments, the solid phase extraction (SPE) trapped amount was further utilized, after obtaining 1D and 2D NMR spectra. Due to the small amount of eluted sample from the SPE cartridges, only two concentrations (between 5–10 mM) of the reporter molecule (L-tryptophan) were added to perform the competition experiment for each compound (1–4).

NMR spectroscopy

All nuclear magnetic resonance spectra (¹H NMR) were recorded on a Bruker Avance III 600 MHz (hydrogen nucleus) spectrometer equipped with a cryoprobe TCI (triple resonance ¹³C/¹⁵N/¹H) and pulse field gradients along the z direction. Data acquisition and processing were performed with the Bruker default software Topspin 3.0 version installed on the machine.

STD NMR

All STD experiments were recorded on a Bruker 600 MHz instrument with a 50 μM solution of HSA protein in pH 7.4 phosphate buffer in D₂O and 1 mg of barbatimão flower extract in D₂O–CD₃OD (85 : 15 v/v). Acquisitions were performed between 296–300 K using the Bruker standard STD pulse sequence with HOD signal suppression, and by using the Gaussian shaped soft pulse train (50 ms, separated by 2 ms delays between the pulses) for selective protein irradiation, and an alternation between on and off resonances. Irradiation of the protein resonances was made possible by putting the on-resonance frequency at −0.50 ppm, while the off-resonance frequency was applied at 35.0 ppm away where no NMR resonances of the ligand or protein were present. For the STD build-up and to calculate the corresponding epitope mapping effect, a series of STD spectra were acquired with different saturation times (1 s, 2 s, 3 s, 4 s and 5 s) and with a recovery delay of 3 s to avoid incomplete relaxation to thermal equilibrium. Investigation of the saturation time-dependence STD NMR experiments with 1–5 s showed that the saturation time of 3 s led to the efficient transfer of saturation from the protein to the ligand protons. Similarly, for the competition experiments the saturation time as well as the relaxation delay were set to 3 s to allow complete relaxation of the system. As STD NMR is a pseudo 2D technique, to obtain the 1D STD difference NMR spectrum subtraction of spectra (on-resonance from the off-resonance) was performed through phase cycling.

NOESY and Tr-NOESY studies

2D-NOESY and Tr-NOESY experiments were recorded using the Bruker phase sensitive pulse sequence. A little modification to the standard pulse sequence was performed by adding a spin lock filter of approximately 3.5 kHz after the first 90 degree pulse to remove the background protein signals to get a two-dimensional Tr-NOESY spectrum. 2D-Tr-NOESY experiments were performed with the same sample that was used earlier for STD NMR experiments by using the times of 200, 400, 500 and 600 ms at 298 K in a 3 mm NMR tube. Two-dimensional Tr-NOESY spectra were observed to determine the most favorable mixing time for Tr-NOE effects, which was found to be 400 ms with a relaxation delay of 3 s to get the fully thermodynamic relaxation between scans. After optimization of all parameters necessary for efficient Tr-NOE effect transfer, the Tr-NOESY experiment was acquired by using the States-TPPI method, with 4K points in t₂ and 256 increments in t₁ with a total of 24 scans. Before processing, a line broadening function of 0.30 Hz was added in both dimensions (F₁ and F₂) with a squared cosine window function.

2D-TOCSY and STD-TOCSY studies

Two-dimensional total correlation spectroscopy (TOCSY) and STD-TOCSY spectra were acquired with the same protein and ligand solution that was used for the 2D-Tr-NOESY experiments. The standard MLEV-17 (Malcolm Levitt's composite-pulse decoupling sequence) with a mixing time of 60 ms and a relaxation delay of 3 s was applied. The spectral width of the standard and the STD-TOCSY experiment was kept same at 7211.539 Hz in both dimensions (F₁ and F₂). To obtain the TOCSY and STD-TOCSY spectra, 64 transients in the t₂ dimension and 256 and 384 interlaced mode increments for the on and of resonance in t₁ were added for TOCSY and STD-TOCSY, respectively. The pre-saturation frequencies for the protein (on-resonance) and away from any signal (off-resonance) were added to −0.5 ppm and 35 ppm, respectively. The saturation of the protein was achieved by the cascade of 60 Gaussian shaped soft pulses of 50 ms duration, each with a delay of 200 ms in between a total saturation time of 3 s. Prior to processing the Fourier-transform, both 2D spectra were multiplied with a 90° shifted squared sine bell function in both (F₁ and F₂) dimensions and zero-filled with 4K and 1K data points in t₂ and t₁, respectively.

HPLC and solid phase extraction system conditions

3 mg of the fully bloomed flower extract was dissolved in methanol (HPLC grade, TEDIA USA) and water (ultrapure water; MilliQ) at a ratio of 85 : 15 v/v. Afterwards, the corresponding solution was filtered through a Millipore filter (Tedia PVDF) to provide a dust free sample. For the whole process of this separation method, the Agilent 1200 series of HPLC system was utilized by connecting a Prontosil C18 column (250 × 2.0 mm i.d, particle size 4 mm), sampling 15 μL with a 1 mL min⁻¹ flow rate and a gradient solvent of methanol and water. The complete information on the gradient system is provided in Fig. S-4 (ESI†). Moreover, a line of 5 DAD wavelengths (210, 254, 280, 330 and 360 nm) was employed for the complete investigation of the sample and 0.05% trifluoroacetic acid was added to get signal attenuation. The resulting chromatogram is shown in Fig. S-5 (ESI†) with 13 peaks.

To increase the concentration of the existing compounds for NMR experiments, a Prospect 2 SPE system (Netherlands) with 210 nm DAD detection wavelength was utilized. Furthermore, the dilution before the trapping was performed through a makeup pump with pure water at a flow rate of 3 mL min⁻¹ then each compound was trapped on a HySphere resin GP cartridge (10 × 2 mm). Thus, each compound was trapped 30 times within the same cartridge to ensure sufficient concentration for 2D spectral data. Before the elution, each compound was dried with N₂ for 35 minutes and then eluted into the NMR probe by adding deuterated methanol-d₄ as the solvent for the NMR experiments.

Docking and molecular dynamic studies

Molecular Operating Environment (MOE) 2012.10 was used for docking simulation, binding energy calculations, and affinity and interaction evaluation. Four sugar containing small organic compounds were used to find their binding affinities at the binding sites of human serum albumin. Sketching of the compounds was done using ChemDraw followed by their conversion to 3D format, atom type, and stereochemistry correction, charge application and minimization using MOE 2012.10. Human serum albumin protein contains two distinct binding sites with amino acids; Tyr150, Glu153, Lys195, Lys199, Arg218, Arg222, His242, Arg257, Leu284, Ser287, Ala291 and Arg114, Leu115, Val116, Arg117, Ile142, Tyr138, His146, Tyr161, Asp183, Leu185, Arg186, Lys190, Ser193 for site-1 and site-2, respectively. Based on the resolution, PDB-4L8U was selected as a target protein from the protein data bank. Protein preparation includes its correction using the auto-correction module in MOE, followed by its protonation and minimization with charge application. A grid map surrounding the active site was generated using active site residues through the surface and maps option in MOE to specify the binding cavity. Default MOE docking parameters with Triangle Matcher (algorithm) and London dG (scoring function) were used and results were refined by the additional rescoring function GBVI/WSA dG with preservation of the 30 top-ranked poses of each compound. The top-ranked conformation of each compound was extracted to determine their specific interactions and orientations inside the binding pocket.

Acknowledgements

The authors of this manuscript acknowledge the generous financial support for these studies from The World Academy of Sciences (TWAS) Trieste, Italy and Conselho Nacional de Desenvolvimento Cientfíco e Tecnológico (CNPq) Brazil.

References

1. A. G. Schauss, X. Wu, R. L. Prior, B. Ou, D. Patel, D. Huang and J. P. Kababick, J. Agric. Food Chem., 2006, 54, 8598–8603.
2. N. G. Masih and B. S. Singh in Phytochemical Screening of Some Plants Used in Herbal Based Cosmetic Preparations, ed. L. D. Khemani, M. M. Srivastava and S. Srivastava, Springer, Berlin, Heidelberg, 2012, pp. 111–112.
3. A. S. Bouldin, M. C. Smith, D. D. Garner, S. L. Szeinbach, D. A. Frate and E. M. Croom, Soc. Sci. Med., 1999, 49, 279–289.
4. W. P. Jones, C. Young-Won and A. D. Kinghorn, Curr. Drug Targets, 2006, 7, 247–264.
5. B. Meyer and T. Peters, Angew. Chem., Int. Ed., 2003, 42, 864–890.
6. M. Politi, M. I. Chávez, F. J. Cañada and J. Jiménez-Barbero, Eur. J. Org. Chem., 2005, 2005, 1392–1396.
7. H. J. Maple, R. A. Garlish, L. Rigau-Roca, J. Porter, I. Whitcombe, C. E. Prosser, J. Kennedy, A. J. Henry, R. J. Taylor, M. P. Crump and J. Crosby, J. Med. Chem., 2011, 55, 837–851.
8. J. Jaumot and R. Gargallo, Curr. Pharm. Des., 2012, 18, 1900–1916.
9. D. C. Carter and J. X. Ho, in Advances in Protein Chemistry, ed. J. T. E. F. M. R. C. B.Anfinsen and S. E.David, Academic Press, 1994, vol. 45, pp. 153–203.
10. L. P. Calle, F. J. Canada and J. Jimenez-Barbero, Nat. Prod. Rep., 2011, 28, 1118–1125.
11. M. Mayer and B. Meyer, Angew. Chem., Int. Ed., 1999, 38, 1784–1788.
12. R. P. Venkitakrishnan, O. Benard, M. Max and J. L. Markley, Methods Mol. Biol., 2012, 914, 47–63.
13. S. A. K. Tanoli, N. Tanoli, S. Usmani, H. Zaheer Ul and A. Ferreira, Cent. Eur. J. Chem., 2014, 12, 332–340.
14. A. R. Bilia, M. C. Bergonzi, G. Mazzi and F. F. Vincieri, J. Agric. Food Chem., 2001, 49, 2115–2124.
15. A. E. Taggi, J. Meinwald and F. C. Schroeder, J. Am. Chem. Soc., 2004, 126, 10364–10369.
16. J. W. Jaroszewski, Planta Med., 2005, 71, 795–802.
17. S. A. K. Tanoli, N. U. Tanoli, T. M. Bondancia, S. Usmani, R. Kerssebaum, A. G. Ferreira, J. B. Fernandes and Z. Ul-Haq, Analyst, 2013, 138, 5137–5145.
18. D. Singh, S. Bharti, S. Agarwal, R. Roy and K. Misra, J. Mol. Model., 2014, 20, 1–9.
19. M. D. Shortridge, D. S. Hage, G. S. Harbison and R. Powers, J. Comb. Chem., 2008, 10, 948–958.
20. P. Putignano, G. A. Kaltsas, M. Korbonits, P. J. Jenkins, J. P. Monson, G. M. Besser and A. B. Grossman, J. Clin. Endocrinol. Metab., 2000, 85, 3309–3312.
21. C. Bertucci, M. Pistolozzi, G. Felix and U. H. Danielson, J. Sep. Sci., 2009, 32, 1625–1631.
22. A. Bhattacharya, T. Grune and S. Curry, J. Mol. Biol., 2000, 303, 721–732.
23. M. Fasano, S. Curry, E. Terreno, M. Galliano, G. Fanali, P. Narciso, S. Notari and P. Ascenzi, IUBMB Life, 2005, 57, 787–796.
24. G. C. Lopes, A. C. C. Sanches, C. V. Nakamura, B. P. Dias Filho, L. Hernandes and J. C. P. d. Mello, J. Ethnopharmacol., 2005, 99, 265–272.
25. M. P. Corrêa and I. B. d. D. Florestal, Dicionário das plantas úteis do Brasil e das exóticas cultivadas, Imprensa Nacional, Rio de Janeiro, 1921.
26. G. C. Lopes, C. V. Nakamura, B. P. Dias Filho and J. C. P. Mello, Rev. Bras. Farmacogn., 2003, 13, 24–27.
27. M. Politi, J. Alvaro-Blanco, P. Groves, A. Prieto, J. A. Leal, F. J. Cañada and J. Jiménez-Barbero, Eur. J. Org. Chem., 2006, 2006, 2067–2073.
28. U. L. Guenther and C. Ludwig, Front. Biosci., Landmark Ed., 2009, 14, 4565–4574.
29. D. Potenza, F. Vasile, L. Belvisi, M. Civera and E. M. V. Araldi, ChemBioChem, 2011, 12, 695–699.
30. T. Haselhorst, H. Blanchard, M. Frank, M. J. Kraschnefski, M. J. Kiefel, A. J. Szyczew, J. C. Dyason, F. Fleming, G. Holloway, B. S. Coulson and M. von Itzstein, Glycobiology, 2007, 17, 68–81.
31. J. W. Jaroszewski, Planta Med., 2005, 71, 691–700.
32. A. Marston and K. Hostettmann, Planta Med., 2009, 75, 672–682.
33. S. A. Kawashty, S. A. M. Mosharrafa, M. El-Gibali and N. A. M. Saleh, Biochem. Syst. Ecol., 2000, 28, 915–917.
34. M. Kalegari, M. D. Miguel, J. d. F. G. Dias, A. L. L. Lordello, C. P. d. Lima, C. M. S. Miyazaki, S. M. W. Zanin, M. C. d. S. Verdam and O. G. Miguel, Braz. J. Pharm. Sci., 2011, 47, 635–642.
35. E. D. Rodrigues, D. B. da Silva, D. C. R. de Oliveira and G. V. J. da Silva, Magn. Reson. Chem., 2009, 47, 1095–1100.
36. M. L. Conrad, A. C. Moser and D. S. Hage, J. Sep. Sci., 2009, 32, 1145–1155.
37. L. Unione, S. Galante, D. Diaz, F. J. Canada and J. Jimenez-Barbero, MedChemComm, 2014, 5, 1280–1289.
38. K. J. Fehske, U. Schläfer, U. Wollert and W. E. Müller, Mol. Pharmacol., 1982, 21, 387–393.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c5ra01457c

---