Mining Xanthine Oxidase Inhibitors from an Edible Seaweed Pterocladiella capillacea by Using In Vitro Bioassays, Affinity Ultrafiltration LC-MS/MS, Metabolomics Tools, and In Silico Prediction

The prevalence of gout and the adverse effects of current synthetic anti-gout drugs call for new natural and effective xanthine oxidase (XOD) inhibitors to target this disease. Based on our previous finding that an edible seaweed Pterocladiella capillacea extract inhibits XOD, XOD-inhibitory and anti-inflammatory activities were used to evaluate the anti-gout potential of different P. capillacea extract fractions. Through affinity ultrafiltration coupled with liquid chromatography tandem mass spectrometry (LC-MS/MS), feature-based molecular networking (FBMN), and database mining of multiple natural products, the extract’s bioactive components were traced and annotated. Through molecular docking and ADMET analysis, the possibility and drug-likeness of the annotated XOD inhibitors were predicted. The results showed that fractions F4, F6, F4-2, and F4-3 exhibited strong XOD inhibition activity, among which F4-3 reached an inhibition ratio of 77.96% ± 4.91% to XOD at a concentration of 0.14 mg/mL. In addition, the P. capillacea extract and fractions also displayed anti-inflammatory activity. Affinity ultrafiltration LC-MS/MS analysis and molecular networking showed that out of the 20 annotated compounds, 8 compounds have been previously directly or indirectly reported from seaweeds, and 4 compounds have been reported to exhibit anti-gout activity. Molecular docking and ADMET showed that six seaweed-derived compounds can dock with the XOD activity pocket and follow the Lipinski drug-like rule. These results support the value of further investigating P. capillacea as part of the development of anti-gout drugs or related functional foods.


Introduction
Hyperuricemia is a chronic metabolic disease characterized by elevated serum uric acid levels due to long-term purine metabolic disorder or decreased uric acid excretion in the body [1,2].The long-term high concentration of serum uric acid leads to the crystallization of uric acid in joints and soft tissues in the body, causing damage to connective tissues and triggering gout [3,4].In acute gouty arthritis, the interaction between urate crystals and phagocytes (such as macrophages and infiltrating white blood cells) induces the secretion of various inflammatory mediators, including cytokines, chemokines, and interleukin, triggering typical inflammatory reactions [5,6].Therefore, the usual means of treating gout aim to inhibit the production of uric acid and promote the excretion of uric acid.Xanthine oxidase (XOD, EC 1.17.3.2) is the key enzyme involved in the metabolism of human purines into uric acid.It exists in the liver, intestine, serum, and lactating breast, catalyzing the gradual hydroxylation of hypoxanthine into xanthine, and then into uric acid [7].Its activity directly determines the rate of uric acid formation to a certain extent.Therefore, this target enzyme has received much attention in the treatment of or intervention in cases of hyperuricemia and gout.
Studies have shown that inhibiting the catalytic activity of xanthine oxidase can effectively reduce the production of uric acid, making this an important means to relieve and treat hyperuricemia and gout in the clinic [8].Allopurinol [9], febuxostat [10], and topiroxostat [11] are the most commonly used xanthine oxidase inhibitors in clinical practice for hyperuricemia and gout.However, these synthetic inhibitors are associated with strong side effects in clinical use, including varying degrees of liver damage, neurological adverse effects, and others.Allopurinol may even cause 'allopurinol hypersensitivity syndrome' [12,13].Therefore, it is of great significance to find new XOD inhibitors from natural sources, with strong activity and low toxicity, for the development of new antigout functional foods or drugs.Currently, natural xanthine oxidase inhibitors such as apigenin [14,15], quercetin [15,16], galangin [17,18], and myricetine [19,20] have been reported, most of which are found in terrestrial plants.However, their xanthine oxidase inhibition activity is lower than that of positive drugs, which may hinder their further research and development.
The ocean is a vast treasury of medicinal and edible biological resources, among which seaweeds have the advantage of a huge biomass and cultivability for sustainable development.Till now, there have been several reports on anti-gout active substances including XOD inhibitors derived from seaweeds.For instance, vine alkaloid isolated from Caulerpa prolifera has an irreversible XOD-inhibitory effect with an IC 50 value of 26.92 µM [21].The fucoidan from Laminaria japonica was found to completely reverse the negative changes induced by adenine in mice, restoring the activities of adenosine deaminase (ADA) and XOD in the liver to normal levels [22], which can effectively reduce the serum uric acid content and blood uric acid content of hyperuricemia mice and rats [23].The Enteromorpha prolifera polysaccharide significantly reduced serum uric acid (UA), serum blood urea nitrogen, and serum and hepatic XOD, and also improved histological parameters in hyperuricemic mice [24].These findings suggest that seaweeds are a meaningful avenue for XOD inhibitor exploration.
For highly efficient discovery of enzyme inhibitors, affinity ultrafiltration mass spectrometry (UF-LC/MS) is a powerful tool that has been increasingly used in screening of bioactive compounds from natural products.In the process of biological affinity ultrafiltration, the ligand-enzyme complexes are retained by the ultrafiltration membrane from the mixture, and then, the ligands released from the complex in the next step of treatment are further identified and quantified using high-performance liquid chromatography-mass spectrometry (HPLC-MS) analysis to achieve rapid identification of bioactive molecules from complex mixtures.Compared with the traditional separation-dependent procedure of active ingredients discovery from natural medicinal plants, UF-LC/MS greatly reduces the cost in terms of time, samples, and expensive reagents [25].Furthermore, several metabolomics tools have been developed to automate secondary metabolite identification such as MSDAIL [26], MSFINDER [27], Global Natural Products Social Molecular Networking (GNPS) [28], and its updated version, feature-based molecular networking (FBMN) [29,30].Previously, we presented a combined strategy named Bio-LCMS-GNPS to connect UF-LC/MS and GNPS, which provided a new approach to enzyme inhibitor discovery from terrestrial and marine bioresources [31].
Pterocladiella capillacea (S.G.Gmelin) belongs to the family of Pterocladiaceae, the order of Gelidiales, and the Phylum of Rhodophyta.It is mainly found in tropical and sub-tropical waters and partially inhabits temperate zones.As an edible seaweed plant, P. capillacea has been traditionally used as a source of jelly production in eastern Asian countries [32].Currently, reports have appeared on the antibacterial [33], antioxidant [34], and bacterial cell agglutination potential [35] of P. capillacea.In a preliminary study based on local seaweeds in Zhanjiang, China, the crude extracts of P. capillacea exhibited strong xanthine oxidase-inhibitory activity.To explore the value of this edible seaweed as anti-gout drugs or functional foods, we further evaluated the XOD-inhibitory and anti-inflammatory activities of its fractions, traced the possible XOD inhibitors via UF-LC-MS/MS, annotated them using metabolomics tools including MSDIAL, MSFINDER, and FBMN, and predicted the action mechanism and drug-likeness of the annotated XOD inhibitors via molecular docking and ADMET, as described in this paper.

Evaluation of XOD Inhibition Activity
The P. capillacea samples were collected along the seashore of Naozhou Island, Zhanjiang, China (see voucher specimens in Figure 1a).Its crude extract was prepared by anhydrous ethanol extraction using an air-dried seaweed sample.Then, the extract was fractioned by a silica gel column to produce twelve primary fractions F1-F12, as shown in the thin-layer chromatography (TLC) images (Figure 1b,c).The XOD inhibition activity screening indicated that fractions F4 and F6 ranked as the top two for activity (Figure 1f).Considering the relative simplicity of the fingerprint and the relatively higher amount of F4 compared with F6, this fraction was preferentially chosen for further study.The separation of F4 on Sephadex LH-20 gel column yielded four secondary fractions, F4-1 to F4-4 (Figure 1d,e).Among them, F4-2 and F4-3 displayed the top two highest XOD inhibition ratios (IRs: 66.47% ± 7.96% and 77.96% ± 4.91%, respectively) at the final concentration of 0.14 mg/mL (Figure 1g).At the same dose, the positive control, allopurinol showed an IR of 92.21% ± 3.38%.
Pterocladiella capillacea (S.G.Gmelin) belongs to the family of Pterocladiaceae, the order of Gelidiales, and the Phylum of Rhodophyta.It is mainly found in tropical and subtropical waters and partially inhabits temperate zones.As an edible seaweed plant, P. capillacea has been traditionally used as a source of jelly production in eastern Asian countries [32].Currently, reports have appeared on the antibacterial [33], antioxidant [34], and bacterial cell agglutination potential [35] of P. capillacea.In a preliminary study based on local seaweeds in Zhanjiang, China, the crude extracts of P. capillacea exhibited strong xanthine oxidase-inhibitory activity.To explore the value of this edible seaweed as anti-gout drugs or functional foods, we further evaluated the XOD-inhibitory and anti-inflammatory activities of its fractions, traced the possible XOD inhibitors via UF-LC-MS/MS, annotated them using metabolomics tools including MSDIAL, MSFINDER, and FBMN, and predicted the action mechanism and drug-likeness of the annotated XOD inhibitors via molecular docking and ADMET, as described in this paper.

Evaluation of XOD Inhibition Activity
The P. capillacea samples were collected along the seashore of Naozhou Island, Zhanjiang, China (see voucher specimens in Figure 1a).Its crude extract was prepared by anhydrous ethanol extraction using an air-dried seaweed sample.Then, the extract was fractioned by a silica gel column to produce twelve primary fractions F1-F12, as shown in the thin-layer chromatography (TLC) images (Figure 1b,c).The XOD inhibition activity screening indicated that fractions F4 and F6 ranked as the top two for activity (Figure 1f).Considering the relative simplicity of the fingerprint and the relatively higher amount of F4 compared with F6, this fraction was preferentially chosen for further study.The separation of F4 on Sephadex LH-20 gel column yielded four secondary fractions, F4-1 to F4-4 (Figure 1d,e).Among them, F4-2 and F4-3 displayed the top two highest XOD inhibition ratios (IRs: 66.47% ± 7.96% and 77.96% ± 4.91%, respectively) at the final concentration of 0.14 mg/mL (Figure 1g).At the same dose, the positive control, allopurinol showed an IR of 92.21% ± 3.38%.

Evaluation of Anti-Inflammatory Activity
Since xanthine oxidase catalyzes the production of urea from xanthine and hypoxanthine, generating a large amount of hydrogen peroxide and superoxide anion, it is possible that substances with xanthine oxidase-inhibitory activity may also exhibit antiinflammatory activity, and many studies have confirmed this hypothesis [36][37][38].Considering that inflammation can severely influence the progress of gout, the XOD-inhibitory samples F4 and F6, and the crude extracts found in the preliminary screening, were also evaluated for their anti-inflammatory activity.
As shown in Figure 2a, the NO production in the RAW264.7 cells of the control group (C) significantly increased compared with the blank group (B) (p < 0.001) under the stimulus of lipopolysaccharide (LPS), while samples F4 and F6, and the crude extract can significantly decrease the NO production at a dose of 20 µg/mL.Meanwhile, F4, F6, and the crude extract did not show toxicity to the cells at this dose (Figure 2b).This result was consistent with that of XOD inhibition activity, indicating that these samples also possess anti-inflammatory activity.respectively (sample numbers are marked below the start line).(f,g) The XOD inhibition rates of fractions F1-F12, F4-1-F4-4, crude extract, and allopurinol.

Evaluation of Anti-Inflammatory Activity
Since xanthine oxidase catalyzes the production of urea from xanthine and hypoxanthine, generating a large amount of hydrogen peroxide and superoxide anion, it is possible that substances with xanthine oxidase-inhibitory activity may also exhibit anti-inflammatory activity, and many studies have confirmed this hypothesis [36][37][38].Considering that inflammation can severely influence the progress of gout, the XOD-inhibitory samples F4 and F6, and the crude extracts found in the preliminary screening, were also evaluated for their anti-inflammatory activity.
As shown in Figure 2a, the NO production in the RAW264.7 cells of the control group (C) significantly increased compared with the blank group (B) (p < 0.001) under the stimulus of lipopolysaccharide (LPS), while samples F4 and F6, and the crude extract can significantly decrease the NO production at a dose of 20 µg/mL.Meanwhile, F4, F6, and the crude extract did not show toxicity to the cells at this dose (Figure 2b).This result was consistent with that of XOD inhibition activity, indicating that these samples also possess anti-inflammatory activity.The control group (C) was RAW264.7 cells stimulated with lipopolysaccharide (LPS).The sample groups (F4, F6, and crude extract) were RAW264.7 cells stimulated with LPS (dose: 1 µg/mL) and treated with the samples (dose: 20 µg/mL).(b) The blank group (B) was cells with viability of RAW 264.7 without treatment.The sample groups (F4, F6, and crude extract) were cells with viability of RAW 264.7 and treated with the samples (dose: 20 µg/mL).Data were expressed as mean ± SD (n = 3).**** p < 0.001, vs. control; ** p < 0.01, vs. control; * p < 0.05, vs. blank.

UF-LC-MS Screening of XOD Ligands in P. capillacea Extract
To trace the active molecules targeting XOD in active secondary fractions F4-2 and F4-3, affinity ultrafiltration-LC-MS/MS analyses were performed.
After the affinity ultrafiltration treatment, a preliminary HPLC analysis was conducted to check the capturing effect of the enzyme on the possible ligands contained in the samples.This was based on the hypothesis that compounds specifically bound to XOD should exhibit higher peaks in the process groups (P) incubated with XOD than in the corresponding blank groups (B) which were incubated with the inactivated enzyme.Indeed, larger peaks of the captured ligands were observed in the chromatograms of both F4-2 and F4-3 after affinity ultrafiltration treatment.

UF-LC-MS Screening of XOD Ligands in P. capillacea Extract
To trace the active molecules targeting XOD in active secondary fractions F4-2 and F4-3, affinity ultrafiltration-LC-MS/MS analyses were performed.
After the affinity ultrafiltration treatment, a preliminary HPLC analysis was conducted to check the capturing effect of the enzyme on the possible ligands contained in the samples.This was based on the hypothesis that compounds specifically bound to XOD should exhibit higher peaks in the process groups (P) incubated with XOD than in the corresponding blank groups (B) which were incubated with the inactivated enzyme.Indeed, larger peaks of the captured ligands were observed in the chromatograms of both F4-2 and F4-3 after affinity ultrafiltration treatment.

Comparison of Metabolite Profiles by LC-MS/MS and Multiple Database Mining
To further characterize the possible XOD ligands in F4-2 and F4-3, the above samples for groups 4-2(B), 4-2(P), 4-3(B), and 4-3(P) were analyzed using Orbitrap LC-MS/MS.The data were aligned via MS-DIAL to accurately localize the XOD ligands based on peak area comparison between different treatment groups, obtain their chromatographic and mass spectroscopic information (retention times, molecular weights, molecular formulae, and MS/MS spectra), and annotate them via FBMN, MSDIAL, and MSFINDER.In addition, the sources of compounds were manually screened by searching open and accessible natural product databases, including PubChem, Natural Product Dictionary (DNP), NPASS, LOTUS, COCONUT, and The Natural Products Atlas.As shown in Figure 4 and Tables 1  and 2, a total of 20 compounds from fractions F4-2 and F4-3 were identified as XOD ligands based on their much higher peak areas in the affinity ultrafiltration process groups ("P" groups) than in the corresponding enzyme inactivated blank groups ("B" groups).Compounds 1-11 were from F4-2, compounds 12-20 were from F4-3, and compounds 7-8 were from both (Figure 4).

Comparison of Metabolite Profiles by LC-MS/MS and Multiple Database Mining
To further characterize the possible XOD ligands in F4-2 and F4-3, the above samples for groups 4-2(B), 4-2(P), 4-3(B), and 4-3(P) were analyzed using Orbitrap LC-MS/MS.The data were aligned via MS-DIAL to accurately localize the XOD ligands based on peak area comparison between different treatment groups, obtain their chromatographic and mass spectroscopic information (retention times, molecular weights, molecular formulae, and MS/MS spectra), and annotate them via FBMN, MSDIAL, and MSFINDER.In addition, the sources of compounds were manually screened by searching open and accessible natural product databases, including PubChem, Natural Product Dictionary (DNP), NPASS, LOTUS, COCONUT, and The Natural Products Atlas.As shown in Figure 4 and Tables 1  and 2, a total of 20 compounds from fractions F4-2 and F4-3 were identified as XOD ligands based on their much higher peak areas in the affinity ultrafiltration process groups ("P" groups) than in the corresponding enzyme inactivated blank groups ("B" groups).Compounds 1-11 were from F4-2, compounds 12-20 were from F4-3, and compounds 7-8 were from both (Figure 4).Note: MW-MF searching: molecular weight (MW) and molecular formula (MF) searching in multiple natural product databases.Y: Yes; N: No.The structures of these compounds are presented in Figure 5. Note: MW-MF searching: molecular weight (MW) and molecular formula (MF) searching in multiple natural product databases.Y: Yes; N: No.The structures of these compounds are presented in Figure 6.For these ligands, the MSDIAL-MSFINDER-FBMN pipeline provided some annotated structures through MS/MS matching.However, manual searching in natural product databases showed that most of these structures were not from taxonomically close sources (Order of Gelidiales or Phylum of Rhodophyta).It was deduced that the MS/MS spectral libraries of MSDIAL, MSFINDER, and FBMN had not indexed a sufficient amount of records from this taxon.Thus, the molecular weight (MW) and molecular formula (MF) information of the annotations were used to search the natural product databases as well.And the taxonomical range was restricted to the Order of Gelidiales to obtain the most relevant hits.Finally, 15 compounds were annotated through MS/MS matching combined with MW-MF searching, 4 compounds were annotated through MW-MF searching, and 1 compound remained unknown (compound 16).The annotation details, including metabolite information, results, methods, biological sources, and reports on anti-gout or antiinflammatory related activities, are summarized in Tables 1 and 2 and Figures 5 and 6        Among the 20 annotated compounds (XOD ligands), 8 are natural products derived from seaweeds, including 4 from P. capillacea endophytic fungus, and 10 (compounds 1, 5, 7, 8, 11, 12, 15, 17, 18, and 19) have been reported for anti-gout-related or anti-inflammatory activities, providing scientific support for the XOD-inhibitory and anti-inflammatory activity of the extract and fractions of P. capillacea.
Based on MS/MS spectral similarity, the FBMN molecular networks were visualized to display the metabolites in fractions F4-2 (Figure 5

Molecular Docking and ADMET Analysis
Seven annotated compounds with seaweed-related origins but without reports of anti-gout-related activity were predicted to have an affinity for XOD using molecular docking and were then subjected to ADMET drug-likeness analysis.The Molybdopterin domain is the catalytic center of XOD [57], which is used for docking the treated ligand compound and calculate the minimum binding affinity.The lower the minimum binding affinity, the stronger the affinity between ligand and XOD, and the higher the binding stability.The molecular docking scoring results are shown in Table 3.For these ligands, the MSDIAL-MSFINDER-FBMN pipeline provided some annotated structures through MS/MS matching.However, manual searching in natural product databases showed that most of these structures were not from taxonomically close sources (Order of Gelidiales or Phylum of Rhodophyta).It was deduced that the MS/MS spectral libraries of MSDIAL, MSFINDER, and FBMN had not indexed a sufficient amount of records from this taxon.Thus, the molecular weight (MW) and molecular formula (MF) information of the annotations were used to search the natural product databases as well.And the taxonomical range was restricted to the Order of Gelidiales to obtain the most relevant hits.Finally, 15 compounds were annotated through MS/MS matching combined with MW-MF searching, 4 compounds were annotated through MW-MF searching, and 1 compound remained unknown (compound 16).The annotation details, including metabolite information, results, methods, biological sources, and reports on anti-gout or anti-inflammatory related activities, are summarized in Tables 1 and 2 and Figures 5 and 6.The MS/MS spectra of these compounds are provided in the (Supporting Information Figures S1-S21).

Molecular Docking and ADMET Analysis
Seven annotated compounds with seaweed-related origins but without reports of antigout-related activity were predicted to have an affinity for XOD using molecular docking and were then subjected to ADMET drug-likeness analysis.The Molybdopterin domain is the catalytic center of XOD [57], which is used for docking the treated ligand compound and calculate the minimum binding affinity.The lower the minimum binding affinity, the stronger the affinity between ligand and XOD, and the higher the binding stability.The molecular docking scoring results are shown in Table 3.The ranking of the minimum binding affinity shows that the XOD protein binds most stably to Chondroterpene B, followed by Chondroterpene E, Proximadiol, and Octadeca-2,4,6,8-tetraenoic acid.The docking diagram showed that Chondroterpene B enters deeply into the XOD active pocket and forms one hydrogen bond with residues SER-1080 and THR-1083 of the XOD protein and two hydrogen bonds with residue SER-1082, respectively (Figure 7a).Chondroterpene E enters deeply into the XOD active pocket and forms one hydrogen bond with residues SER-1080 and SER-1082 and forms two hydrogen bonds with residues GLN-1040 and THR-1083 each, respectively (Figure 7b).Proximadiol enters deeply into the XOD active pocket and forms a hydrogen bond with residue GLN-1194 of the XOD protein (Figure 7c).Octadeca-2,4,6,8-tetraenoic acid enters into the XOD active pocket and forms two hydrogen bonds with residues ARG-880 and THR-1010 each, respectively (Figure 7d).In addition, the docking results showed that all of the seven compounds except sphingosine enter the XOD activity pocket.
The ADMET properties of a drug are the absorption, distribution, metabolism, excretion, and toxicity of the drug in the human body, and these are key factors used to evaluate whether a compound can become a drug or not.To validate the drug-likeness of the compounds annotated in the active fractions of P. capillacea, they were subjected to online ADMET prediction, and the results are shown in Table 4.The ADMET properties of a drug are the absorption, distribution, metabolism, excretion, and toxicity of the drug in the human body, and these are key factors used to evaluate whether a compound can become a drug or not.To validate the drug-likeness of the compounds annotated in the active fractions of P. capillacea, they were subjected to online AD-MET prediction, and the results are shown in Table 4.Likewise, the seven annotated compounds, except the C 18 -sphingosine, showed suitable water solubility (−4 < LogS < 0.5), excellent absorption in the human small intestine (HIA < 0.3), and low effect on drug metabolism (CYP inhibitor).Octadeca-2,4,6,8-tetraenoic acid and Sphingosine have poor oral bioavailability (PPB > 90%) and may also cross the blood-brain barrier, while the other five compounds have low blood-brain barrier penetration (BBB > 0.7) and therefore do not cause side effects on the central nervous system.Moreover, Chondroterpene C, Chondroterpene H, Proximadiol, and Chondroterpene E may have improved safety characteristics, since they may not cause hepatotoxicity, liver damage, or skin sensitization (H-HT, DILI, Skin Sensitization < 0.3).Generally, all the seven compounds meet the five principles of Lipinski's rule for oral drugs [58].

Discussion
In the present study, two extract fractions (F4 and F6) from edible seaweed P. capillacea were discovered to have strong XOD-inhibitory and anti-inflammatory activities.From F4, with relatively simple components, two active subfractions, F4-2 and F4-3, were obtained and exhibited remarkable XOD-inhibitory activity.By using UF-LC-MS/MS, metabolomics tools, multiple natural product database mining, molecular docking, and ADMET analysis, 20 plausible XOD ligands in the active subfractions were localized and annotated, and 7 annotated seaweed-derived compounds were further predicted to have binding affinity with XOD and drug-likeness.Among them, Chondroterpene C, Chondroterpene E, and Proximadiol displayed strong XOD affinity and good drug-likeness properties.
Hyperuricemia and gout have become increasingly prevalent globally, while current clinical drugs like allopurinol have severe side effects including triggering 'allopurinol hypersensitivity syndrome' [13].Therefore, there is a need for new effective drugs and functional foods to cope with this issue.P. capillacea, a globally spread seaweed that is traditionally eaten in eastern Asia, has not been reported for its anti-gout activity.In this study, potent fractions and subfractions from P. capillacea have been found to inhibit XOD activity and inflammation.The potency of P. capillacea fractions and subfractions is generally comparable to the control allopurinol at the same dose.Furthermore, the fractions did not show cytotoxicity to macrophage cells.Since XOD is the key target in hyperuricemia and gout, and inflammation also plays an important role in the development of gout [36], P. capillacea, as an edible seaweed with huge biomass, may have great potential as a new source for the development of anti-hyperuricemia and anti-gout drugs or functional foods.
Preliminary UF-LC-MS/MS tracing, annotating, and predictive investigations located the active peaks and suggested 20 compounds as plausible XOD inhibitors in P. capillacea, including 8 compounds with seaweed and P. capillacea-related origins, and 9 compounds with XOD-inhibitory or anti-inflammatory reports.Furthermore, molecular docking of seven compounds with seaweed and seaweed-related origins and no reported anti-gout effects showed that six of them enter the XOD activity pocket and dock with XOD by forming hydrogen-bonding forces with amino acid residues.Three compounds, Chondroterpene C, Chondroterpene E, and Proximadiol, displayed strong XOD affinity and good drug-likeness properties.To some extent, the above results have provided reasonable explanations for the XOD-inhibitory and anti-inflammatory activity of P. capillacea extracts and useful clues for future studies on natural product research for anti-gout treatments using this seaweed.
UF-LC-MS/MS has been increasingly used in enzyme inhibitor discovery, where the LC-MS profile of an affinity-treated enzyme group is usually compared with that of the blank group without enzyme incubation to recognize the ligands specifically binding with the enzyme [66][67][68].In the blank group, the molecules absorbed by the ultrafiltration membrane are taken as non-specific binders.However, it is not easy to discriminate the molecules absorbed by the surface of the proteins using this method.In this study, the authors used heat-inactivated enzymes for the incubation treatment as a blank group instead of only ultrafiltration membrane, by which the small molecular ligands that specifically interact with the target enzyme may be more accurately recognized from complex natural product mixtures.Although LC-MS/MS can provide rich information about the samples, in this study, a preliminary HPLC-DAD analysis was performed for the ultrafiltration experimental samples before formal LC-MS/MS analysis.This measurement can not only verify the effect of ultrafiltration experiments, but also provide more spectrometric information on the active peaks, in addition to mass spectra for further natural product purification, since HPLC-DAD is a more frequently used approach in regular analytic tasks.
Bioactivity-coupled molecular networking analysis can rationalize natural product isolation and help in deduplication before time-consuming isolation tasks.Compared with our previously reported "Bio-LCMS-GNPS" strategy used in acetylcholinesterase inhibitors and antioxidants mining [31], improvements have been made in two aspects of the present study.Firstly, FBMN has been used to construct molecular networks instead of classical GNPS, and more metabolomics tools and multiple natural product databases are utilized to annotate ligands together.Since high-resolution mass spectrometry-based FBMN provides molecular formulae, more accurate peak area integration, and improved retention times for the features, and recognize different adduct ions of the same substance [69], it has an advantage in comparing peak areas of ligands between different groups and providing molecular formulae for library searching over classical GNPS.FBMN combined with MSDIAL, MSFINDER, and natural products databases, also provides more comprehensive annotation than GNPS individually.Secondly, following the annotation, in silico molecular docking and ADMET were performed to evaluate the possibility of annotated compounds as enzyme ligands and their drug-likeness, which is more reliable than mere metabolomics annotation.Therefore, this updated bioactivity-coupled mass spectrometric metabolomics pipeline is named "Bio-LCMS-Metabolomic-in silico Prediction".
The application of this pipeline to the discovery of XOD inhibitors in P. capillacea has not only provided plausible explanations for the bioactivity of samples, but has also located the XOD ligands in the samples.The feature information, including retention time, molecular weights, MS/MS spectra, and UV spectra, will guide future compound isolation and elucidation.As with all the MS metabolomics-based studies, the isolation, structural elucidation, and bioactivity study of isolated pure compounds is the final verification and "golden criterion" for the annotated results.This study and ongoing similar processes on fraction F6 will guide our in-depth chemistry and biology studies on the anti-gout constituents of P. capillacea for drugs and functional foods purposes.
In conclusion, the present study has revealed the great potential of the edible seaweed P. capillacea for developing anti-hyperuricemia and anti-gout drugs and functional foods.Furthermore, it has established a new bioactivity-coupled metabolomics investigating pipeline "Bio-LCMS-Metabolomic-in silico Prediction" for bioactive natural product discovery.
An Agilent 1200 high-performance liquid chromatography and a Thermo Orbitrap Fusion LUMOS Tribrid liquid chromatography-mass spectrometer (Orbitrap LC-MS/MS, Thermo Fisher Scientific, Waltham, MA, USA) were used to analyze the samples.A 96-well microplate reader (Bio-Tek Epoch 2, Bio Tek Instruments, Winooski, VT, USA) was used for spectrophotometric measurements.An Allegra X-30R high-speed centrifugator (Beckman Coulter, Brea, CA, USA) was used for ultrafiltration experiment.

Preparation of Crude Extracts of P. capillacea
The air-dried P. capillacea (3 kg) was crushed and soaked in anhydrous ethanol for 24 h with a ratio (g:mL) of 1:10, extracted three times, filtered, and concentrated to dryness below 40 • C with a rotary evaporator to obtain the crude extract, and then stored at −20 • C.

In Vitro Determination of XOD-Inhibitory Activity
The in vitro XOD-inhibitory activities were evaluated with UV transparent 96-well microwell plates using a modified method based on Andriana's report [70], in which the enzyme activity was reflected by the UV absorbance of uric acid derived from xanthine.

Cellular Anti-Inflammatory Activity Assay
The content of NO in cells and cell survival rate were determined in accordance with a previously reported method [71].In accordance with the manufacturer's protocol, the NO level of RAW264.7 cells treated with or without lipopolysaccharide and P. capillacea extracts (crude extract, F4 and F6) was evaluated using the Griess reagent system.The viability of RAW264.7 cells treated with or without P. capillacea extracts (crude extract, F4 and F6) was evaluated via CCK-8.RAW264.7 cells were seeded in a 24-well plate at a density of 5 × 10 4 cells/well.In all the test groups, the cells were treated with the extracts (crude extract, F4 and F6) of P. capillacea with a concentration of 20 µg/mL.Subsequently, they were activated with LPS (1 µg/mL) for 24 h, except for the blank group.Finally, NO production was estimated via spectrophotometry at 540 nm.
Cell viability assay: RAW264.7 cells were inoculated in 96-well plates at a density of 5 × 10 4 cells/well for 24 h.P. capillacea extract (crude extract, F4 and F6) was added to reach a final concentration of 20 µg/mL and incubated for 24 h.After 10 µL of CCK-8 solution was added to each well for 1 h, absorbance values were measured at 450 nm.4.6.UF-LC-MS/MS 4.6.1.Affinity Ultrafiltration Treatment First, 10 mg of sample (in 100 µL methanol) was applied onto a reverse SPE column, eluted with 1 mL pure water to remove salts, and then with 1 mL methanol to collect the eluent.This eluent solution was then filtrated through 0.22 µm membrane, dried completely, and finally dissolved in 20 µL of DMSO plus 80 µL of water to form the sample solution used for affinity ultrafiltration.For the process group (P), 50 µL of sample solution, 50 µL of 0.5 U/mL of XOD solution, and 100 µL of Tris-HCL buffer were added to a 1.5 mL centrifuge tube and incubated at 37 • C for 30 min.Afterwards, the unbound small molecules were removed by centrifugation at 11,200× g for 10 min and repeatedly washed using Tris-HCL buffer and centrifugation 3 times.Then, the ligands were released from the retained ligand-enzyme complexes on the ultrafiltration membrane using 100 µL of methanol-water (4:1) (centrifugation at 11,200× g for 10 min, repeated 3 times).The eluate was dried and dissolved in 1 mL of LC-MS pure methanol for HPLC and Orbitrap LC-MS/MS analyses.For the blank group (B), the difference was that the XOD solution was denatured at 100 • C for 30 min before use.For the crude sample group (C), the sample was prepared directly to 5 mg/mL after desalting, without ultrafiltration.

Annotation of the Bioactive Molecules Using Metabolomics Tools
The standard pipeline of feature-based molecular networking (FBMN) with MS-DIAL was performed by referring to a previous report [29] and following the instructions on the webpage of the GNPS platform (https://ccms-ucsd.github.io/GNPSDocumentation/featurebasedmolecularnetworking-with-ms-dial/, accessed on 17 December 2022).The parameters for clustering and compound matching were set as follows: minimum matching fragment of 6; minimum cluster size of 2; cosine threshold of 0.7; and search database range of the entire GNPS library.Data visualization was performed using Cytoscape 3.7.2software.In addition, the annotation of compounds was also performed via MS-DIAL database matching and MS-FINDER matching with all its indexed natural product databases [72,73].

Molecular Docking and ADMET Analysis
The sdf structures of ligands were downloaded from the PubChem database and were batch processed into pdbqt files suitable for molecular docking using openbabel 2.3.2.The crystal structure of XOD protein was obtained from the Protein Database Bank (PDB ID: 3nvw).Water molecules and guanine ligands were removed from the protein using Pymol 2.5.4 software and the protein was exported to pdb format.Then, the protein was hydrogenated and structurally optimized using Autodock vina 1.5.7, and finally exported to pdbqt format.Molecular docking of all ligands to XOD was performed using Autodock vina 1.5.7 and the docking results are visualized using Pymol 2.5.4.
ADMET analysis was performed online on the ADMETlab prediction website (https: //admetmesh.scbdd.com/,accessed on 2 April 2023) to predict the adsorption, distribution, metabolism, excretion, and toxicity characteristics of compounds by entering the SMILES codes of the compounds.

Figure 1 .
Figure 1.The TLC fingerprints and XOD-inhibitory activity of the crude extract and fractions of Pterocladiella capillacea.(a) The image of the seaweed material P. capillacea.(b-e) The TLC images of fractions F1-F12, F4-1-F4-4, and crude extract detected under 254 nm and 365 nm UV light,

Figure 3 .
Figure 3.The comparison of HPLC traces of the ultrafiltration membrane-retained samples after incubation with XOD or inactivated XOD ultrafiltration (the chromatograms were monitored under 290 nm).(a) The comparison of affinity ultrafiltration process group (4-2(P)) and inactivated enzyme blank group (4-2(B)) for sample F4-2.(b) The comparison of affinity ultrafiltration process group (4-3(P)) and inactivated enzyme blank group (4-3(B)) for sample F4-3.The numbers associated with arrows marked the main peaks.

Figure 3 .
Figure 3.The comparison of HPLC traces of the ultrafiltration membrane-retained samples after incubation with XOD or inactivated XOD ultrafiltration (the chromatograms were monitored under 290 nm).(a) The comparison of affinity ultrafiltration process group (4-2(P)) and inactivated enzyme blank group (4-2(B)) for sample F4-2.(b) The comparison of affinity ultrafiltration process group (4-3(P)) and inactivated enzyme blank group (4-3(B)) for sample F4-3.The numbers associated with arrows marked the main peaks.
. The MS/MS spectra of these compounds are provided in the Supporting Information (Figures S1-S21).

Figure 5 .
Figure 5.The FNMB molecular network for the fraction sample 4-2 based on positive ion MS/MS spectral similarity.Sub-figures (a)-(i) show the details of the amplified clusters including the annotated compounds 1-11 (C1-C11), respectively.The nodes display the measured average masses of the molecular ions with identical MS/MS spectra.The sizes of the nodes reflect the relative amount of the corresponding compounds.The different colors of sections in the nodes represent different sample groups, i.e., : 4-2(C) (Group1), 4-2(B) (Group 2), 4-2(P) (Group 3), and Blank (Group 4).

Figure 5 .
Figure 5.The FNMB molecular network for the fraction sample 4-2 based on positive ion MS/MS spectral similarity.Sub-figures (a-i) show the details of the amplified clusters including the annotated compounds 1-11 (C1-C11), respectively.The nodes display the measured average masses of the molecular ions with identical MS/MS spectra.The sizes of the nodes reflect the relative amount of the corresponding compounds.The different colors of sections in the nodes represent different sample groups, i.e.,

Figure 5 .
Figure 5.The FNMB molecular network for the fraction sample 4-2 based on positive ion MS/MS spectral similarity.Sub-figures (a)-(i) show the details of the amplified clusters including the annotated compounds 1-11 (C1-C11), respectively.The nodes display the measured average masses of the molecular ions with identical MS/MS spectra.The sizes of the nodes reflect the relative amount of the corresponding compounds.The different colors of sections in the nodes represent different sample groups, i.e., : 4-2(C) (Group1), 4-2(B) (Group 2), 4-2(P) (Group 3), and Blank (Group 4).

Figure 6 .
Figure 6.The FBMN molecular network for the fraction sample 4-3 based on positive ion MS/MS spectral similarity.Sub-figures (a)-(i) show the details of the amplified clusters including the annotated compounds 7, 8, and 12-20 (C7, C8, and C12-C20), re-spectively.The nodes display the measured average masses of the molecular ions with identical MS/MS spectra.The sizes of the nodes reflect the relative amount of the corresponding compounds.The different colors of sections in the

Figure 6 .
Figure 6.The FBMN molecular network for the fraction sample 4-3 based on positive ion MS/MS spectral similarity.Sub-figures (a-i) show the details of the amplified clusters including the annotated compounds 7, 8, and 12-20 (C7, C8, and C12-C20), re-spectively.The nodes display the measured average masses of the molecular ions with identical MS/MS spectra.The sizes of the nodes reflect the relative amount of the corresponding compounds.The different colors of sections in the nodes represent different sample groups, i.e., presents ligands 1-11) and F4-3 (Figure 6 presents ligands 7, 8, 12-20).

Table 1 .
The annotation of bioactive molecules in fraction F4-2 through UF-LC-MS/MS combined with metabolomics tools and bioresource searching.

Table 2 .
The annotation of bioactive molecules in fraction F4-3 through UF-LC-MS/MS combined with metabolomics tools and bioresource searching.

Table 3 .
The docking results of 7 annotated compounds with XOD crystal structure (PDB ID: 3nvw).

Table 3 .
The docking results of 7 annotated compounds with XOD crystal structure (PDB ID: 3nvw).

Table 4 .
The results of ADMET analysis of the annotated compounds.

Table 4 .
The results of ADMET analysis of the annotated compounds.