Experimental and Computational Study to Reveal the Potential of Non-Polar Constituents from Hizikia fusiformis as Dual Protein Tyrosine Phosphatase 1B and α-Glucosidase Inhibitors

Hizikia fusiformis (Harvey) Okamura is an edible marine alga that has been widely used in Korea, China, and Japan as a rich source of dietary fiber and essential minerals. In our previous study, we observed that the methanol extract of H. fusiformis and its non-polar fractions showed potent protein tyrosine phosphatase 1B (PTP1B) and α-glucosidase inhibition. Therefore, the aim of the present study was to identify the active ingredient in the methanol extract of H. fusiformis. We isolated a new glycerol fatty acid (13) and 20 known compounds including 9 fatty acids (1–3, 7–12), mixture of 24R and 24S-saringosterol (4), fucosterol (5), mixture of 24R,28R and 24S,28R-epoxy-24-ethylcholesterol (6), cedrusin (14), 1-(4-hydroxy-3-methoxyphenyl)-2-[2-hydroxy -4-(3-hydroxypropyl)phenoxy]-1,3-propanediol (15), benzyl alcohol alloside (16), madhusic acid A (17), glycyrrhizin (18), glycyrrhizin-6’-methyl ester (19), apo-9′-fucoxanthinone (20) and tyramine (21) from the non-polar fraction of H. fusiformis. New glycerol fatty acid 13 was identified as 2-(7′- (2″-hydroxy-3″-((5Z,8Z,11Z)-icosatrienoyloxy)propoxy)-7′-oxoheptanoyl)oxymethylpropenoic acid by spectroscopic analysis using NMR, IR, and HR-ESI-MS. We investigated the effect of the 21 isolated compounds and metabolites (22 and 23) of 18 against the inhibition of PTP1B and α-glucosidase enzymes. All fatty acids showed potent PTP1B inhibition at low concentrations. In particular, new compound 13 and fucosterol epoxide (6) showed noncompetitive inhibitory activity against PTP1B. Metabolites of glycyrrhizin, 22 and 23, exhibited competitive inhibition against PTP1B. These findings suggest that H. fusiformis, a widely consumed seafood, may be effective as a dietary supplement for the management of diabetes through the inhibition of PTP1B.


Introduction
Diabetes mellitus (DM) is a serious chronic disease and an important public health problem. DM occurs when the pancreas does not produce enough insulin or when the body cannot effectively use insulin. In 2014, 422 million adults worldwide had DM and the prevalence of DM has been rising steadily for the past three decades [1]. Several underlying mechanisms contribute to the pathogenesis Compound 13 was obtained as a yellow syrup, and the HR-ESI-MS showed a pseudo molecular ion peak at m/z 607.3820 [M + H] + (calculated for C34H55O9, 607.3846), confirming a molecular formula of C34H54O9. The 1 H-and 13 C-NMR spectra for 13 indicated the presence of diacylglycerol, aliphatic chain with three double bonds, alkane dicarboxylic acid, and 2-methylpropenoic acid, strongly suggesting a glycerol FA derivative.
Similarly, typical absorptions for acylglycerol and FA with aliphatic chains were detected in the FT-IR data: 3705.55-3680. 48 , and one carbonyl carbon at δ C 173.2 in the 13 C-NMR spectra and a methyl signal at δ H 0.88, overlapping methylene protons between δ H 1.25 and 2.31, and six olefinic protons (δ H 5.36) in the 1 H-NMR spectra explain the presence of eicosatrienoic acid. The 1 H-NMR spectrum showed two methylene groups lying between three double bonds of eicosatrienoic acid at δ H 2.79 (2H), which could be assigned to H-7 and H-10 . The HMBC correlations of eicosatrienoic acid were also observed from H-5 to C4 , from H6 to C-7 , H-8 to C-7 , H-9 to C-10 , H-11 to C-10 , H-12 to C-13 , and from H-20 to C-18 and C-19 . The geometry of the three double bonds in this FA moiety was presumed to be cis-form based on the 13 C-NMR spectrum (δ C 25.8, 27.3, 27.4). The signals of carbons next to a double bond usually appear at δ C 27 to 28 in a cis-configuration, whereas those of a trans-configuration appear at δ C 32 to 33 [22,23]. 13 C-NMR spectrum (δC 25.8, 27.3, 27.4). The signals of carbons next to a double bond usually appear at δC 27 to 28 in a cis-configuration, whereas those of a trans-configuration appear at δC 32 to 33 [22,23].

Enzyme Kinetic Analysis of Active Compounds with PTP1B
Compounds 6, 13, 22 and 23 were subjected to enzyme kinetic study, since these compounds demonstrated potent activity against PTP1B. According to the Lineweaver-Burk plot and secondary plot of y-intercept (Table 1 and Figure 3), compounds 22 and 23 showed general competitive type inhibition against PTP1B, whereas compounds 6 and 13 showed inhibition in a non-competitive manner. The binding constant of inhibitor with enzyme-substrate complex (K iu ) and free enzyme (K ic ) was determined using the secondary plot of 1/V max,app (Y-intercept) and K m,app /V max,app (slope) of the respective linear regression of Lineweaver-Burk plot, respectively. As shown in Figure 3, K ic values for the inhibition of PTP1B were 3.17 and 10.17 µM for 22 and 23, respectively, and K iu values for inhibition of PTP1B by 6 and 13 were 24.43 and 4.13 µM, respectively. control acarbose (IC50 = 158.41 ± 1.05 μM). However, compounds 18 and 19 showed no activity under the tested concentration. Interestingly, unsaturated FAs C20:4 (Δ 7,9,11,13 ) (9) and C17:3 (Δ 8,11,14 ) (12) showed potent inhibition against α-glucosidase with IC50 values of 34.85 ± 2.39 and 43.90 ± 0.77 μM, respectively. In addition, neolignan 14 and trace amine 21 also showed moderate inhibition with IC50 values of 133.84 ± 3.86 and 273.23 ± 5.65 μM, respectively. In contrast, other compounds exhibited weak or no activity against α-glucosidase inhibition.

Molecular Docking Simulation in PTP1B Inhibition
Due to the potent inhibitory activity of 5, 6, 13, 22, and 23 against PTP1B, we conducted computational docking analysis using these compounds to evaluate binding affinities and aspects. Sterols 5 and 6 and compound 13 are well docked into the allosteric pocket of PTP1B (α3, α6, and α7 helices), whereas triterpenoids 22 and 23 are docked into the catalytic site ( Figure 4). Because 6 is mixture of 24R,28R and 24S,28R-epoxy-24-ethylcholesterol (6a and 6b), we also compared the binding aspect between the two isomers. Compound A (catalytic inhibitor) and compound B (allosteric inhibitor) were used as positive controls to verify the docking protocol.
As shown in Figure 4; Figure 5, best fitted models of 5, 6a, and 6b interacted with Glu200 in the α3 helix via H-bond and surrounded by hydrophobic residues in α3 (Phe196, Asn193, and Leu192) and α6 (Glu276 and Phe280) helices of enzyme with negative B-scores of −8.10, −7.90, and −8.66 kcal/mol, respectively. Interestingly, one difference was observed between the 5-PTP1B complex and the 6a/6b-PTP1B complex. Both 6a and 6b interacted with Pro188 residue via a hydrophobic bond ( Figure 5B,C), but the aliphatic side chain of 5 did not reach near Pro188 ( Figure 5A). Docking examination showed that 13 interacted with the allosteric site of the enzyme by positioning the long aliphatic chain toward the center of α3 and α6 helices of the enzyme, whereas the methacrylic acid moiety of 13 was located at the edge of the α3 helix and interacted with Asn193 and Lys197 via H-bond interactions ( Figure 5D). Although 13 showed strong potency against PTP1B inhibition in vitro, its binding affinity was poor due to the long aliphatic chain. However, four tight H-bond interactions between compound 13 and PTP1B residues including Tyr153, Lys150, Lys197, and Asn193 may play key roles in PTP1B inactivation.

Molecular Docking Simulation in PTP1B Inhibition
Due to the potent inhibitory activity of 5, 6, 13, 22, and 23 against PTP1B, we conducted computational docking analysis using these compounds to evaluate binding affinities and aspects. Sterols 5 and 6 and compound 13 are well docked into the allosteric pocket of PTP1B (α3, α6, and α7 helices), whereas triterpenoids 22 and 23 are docked into the catalytic site ( Figure 4). Because 6 is mixture of 24R,28R and 24S,28R-epoxy-24-ethylcholesterol (6a and 6b), we also compared the binding aspect between the two isomers. Compound A (catalytic inhibitor) and compound B (allosteric inhibitor) were used as positive controls to verify the docking protocol. As shown in Figure 4; Figure 5, best fitted models of 5, 6a, and 6b interacted with Glu200 in the α3 helix via H-bond and surrounded by hydrophobic residues in α3 (Phe196, Asn193, and Leu192) and α6 (Glu276 and Phe280) helices of enzyme with negative B-scores of −8.10, −7.90, and −8.66 kcal/mol, respectively. Interestingly, one difference was observed between the 5-PTP1B complex and the 6a/6b-PTP1B complex. Both 6a and 6b interacted with Pro188 residue via a hydrophobic bond (Figures 5B and 5C), but the aliphatic side chain of 5 did not reach near Pro188 ( Figure 5A). Docking examination showed that 13 interacted with the allosteric site of the enzyme by positioning the long aliphatic chain toward the center of α3 and α6 helices of the enzyme, whereas the methacrylic acid moiety of 13 was located at the edge of the α3 helix and interacted with Asn193 and Lys197 via H-bond interactions ( Figure 5D). Although 13 showed strong potency against PTP1B inhibition in vitro, its binding affinity was poor due to the long aliphatic chain. However, four tight H-bond interactions between compound 13 and PTP1B residues including Tyr153, Lys150, Lys197, and Asn193 may play key roles in PTP1B inactivation. In contrast to sterols and compound 13, the best docked models of compounds 22 and 23 were placed into the catalytic site of PTP1B. As shown in Figure 4C, binding orientations of 22 and 23 were slightly different. The PTP1B-22 complex had a negative B-score (Table 2) of −9.09 kcal/mol with two H-bonds with Lys116 and Lys 120 as well as a salt-bridge interaction with Lys120 residue. Hydrophobic interactions between 22 and Phe182, Gly183, Arg221, Glu115, Thr263, Asp265, and Lys120 residues were also observed ( Figure 5E). However, the PTP1B-23 complex had a B-score of -8.90 kcal/mol with two H-bonds with Gly183 and Asp48 residues and a salt-bridge interaction between carboxyl moiety of 23 and Lys116. As shown in Figure 5F, 23 was surrounded by Tyr46, Val49, Ala217, Phe182, and Gln262 residues via hydrophobic interaction.  In contrast to sterols and compound 13, the best docked models of compounds 22 and 23 were placed into the catalytic site of PTP1B. As shown in Figure 4C, binding orientations of 22 and 23 were slightly different. The PTP1B-22 complex had a negative B-score (Table 2) of −9.09 kcal/mol with two H-bonds with Lys116 and Lys 120 as well as a salt-bridge interaction with Lys120 residue. Hydrophobic interactions between 22 and Phe182, Gly183, Arg221, Glu115, Thr263, Asp265, and Lys120 residues were also observed ( Figure 5E). However, the PTP1B-23 complex had a B-score of -8.90 kcal/mol with two H-bonds with Gly183 and Asp48 residues and a salt-bridge interaction between carboxyl moiety of 23 and Lys116. As shown in Figure 5F, 23 was surrounded by Tyr46, Val49, Ala217, Phe182, and Gln262 residues via hydrophobic interaction.

Discussion
Growing evidence has linked PTP1B with insulin resistance, T2DM, and obesity. Numerous studies have revealed that PTP1B negatively controls leptin and insulin signaling pathways [12]. Therefore, a considerable effort has been expended on generating small molecule inhibitors of PTP1B to promote the insulin signaling pathway in insulin resistant states. By following the conventional method of producing inhibitors that target the catalytic site of an enzyme, many selective and reversible PTP1B inhibitors were discovered [35]. However, these small molecule inhibitors, which often possessed phospho-Tyr mimetic moieties, were highly charged and lacked oral bioavailability, showing limitations in their potential for drug development. Therefore, the development of an allosteric inhibitor is urgently needed to develop orally bioavailable inhibitors of PTP1B [36]. We previously demonstrated that non-polar fractions of H. fusiformis methanol extract showed potent PTP1B and α-glucosidase inhibition [15]. Various non-polar components such as 24-ketocholesterol, fucosterol, 24,28-epoxyfucosterol, fucoxanthin, and saringosterol have been isolated from this seaweed [14,37]. However, the systematic extraction and isolation of compounds from H. fusiformis as well as the mechanisms of PTP1B and α-glucosidase inhibition through detailed enzyme kinetics and molecular docking simulation have not been reported. In this study, we isolated one new and 20 known compounds from the non-polar fraction of H. fusiformis methanol extract and evaluated the PTP1B and α-glucosidase inhibitory activity of the isolated compounds. Enzyme assay results revealed that unsaturated and saturated FAs, sterols, and triterpenoid glycosides showed good inhibitory activity against PTP1B. Shibata et al. reported that unsaturated FAs at 10 µM drastically inhibited PTP1B, whereas saturated FAs showed moderate inhibition [38]. Interestingly, in rat adipocytes, long-time treatment of saturated free FAs inhibited insulin-stimulated glucose uptake, but short-time treatment enhanced glucose transport [39]. Similarly, in our results, unsaturated FAs showed significantly strong PTP1B inhibitory activity with IC 50 values in the range of 4.86-16.43 µM. In contrast, among saturated FAs, palmitic acid (2) showed moderate activity with an IC 50 value of 49.39 ± 1.39 µM. In addition, C17:3 (∆ 8,11,14 ) (7) and the new compound 13 showed notable inhibition among the isolated 22 compounds. Together, our results and the previously reported data suggest that FAs could be an important factor responsible for T2DM.
A previous study showed that fucosterol (5) from Pelvetia siliquosa possessed anti-diabetic activity in streptozotocin-induced Sprague-Dawley rats [40]. Another report demonstrated that 5 is a non-competitive PTP1B inhibitor in vitro and improved insulin resistance by inhibition of PTP1B and stimulation of insulin signaling pathway in insulin-resistant HepG2 cells [41]. However, information on the biological activity of fucosterol epoxide (6) is limited. As shown in Table 1, 5 and its epoxide (6) showed PTP1B inhibitory activities. Interestingly, 6 showed 3 times stronger activity than 5. Enzyme kinetic analysis using Lineweaver-Burk plot and its secondary plot and computational docking analysis demonstrated that 5, 6, and 13 are non-competitive inhibitors and well docked into the allosteric pocket placed~20 Å away from the catalytic site of PTP1B [42]. Best fitted models of 5 and 6 interacted with Glu200 in the α3 helix via H-bond and surrounded by hydrophobic residues in α3 and α6 helices of enzymes such as Phe280, Phe196, Leu192, and Ala189. However, the lack of interaction between compound 5 and Pro188 explains its lower PTP1B inhibitory potency compared to 6.
PTP1B enzyme exists in two conformations: open and closed forms. In the open form, the WPD loop, which contains Trp179-Asp181 residues, is beside the catalytic site to form an open-binding pocket, which is accessible for the substrate. In the closed-form, the WPD loop covers the substrate-binding site of the enzyme, forming a catalytically competent state. For the WPD loop to close, Pro188-Phe191-Leu192 residues must move to accommodate Trp179 [42]. However, this movement is blocked by compound 6 directly via hydrophobic interaction. Thus, the allosteric inhibitor 6 could prevent the movement of the WPD loop and maintain the loop in an open (inactive) form. In the case of 13, this compound also hydrophobically interacted with Pro188 residue with four H-bond interactions with Tyr153, Lys150, Lys197 and the key allosteric site residue Asn193. These interactions may play critical roles in PTP1B inactivation in the PTP1B-13 complex.
We also found that triterpenoid glycosides 19 and 18 are effective and moderate PTP1B inhibitors, respectively. Compound 19, which is a 6 -methyl ester of 18, showed 2.2 times stronger PTP1B inhibition than compound 18. In addition, 18α and 18β-glycyrrhetinic acids (22 and 23), metabolites of 18 and 19, are stronger PTP1B inhibitors compared with 18 and 19. Although the PTP1B inhibitory activities of 22 and 23 were previously described by Na et al. [43], the inhibitory mechanisms and structure-activity relationships have not been reported. In our enzyme kinetic and computational study, triterpenoids 22 and 23 showed competitive inhibition activity against the PTP1B enzyme and were strongly fitted into the catalytic site of the enzyme. Due to the different configuration (α and β) of the hydrogen atom at C-18 position, binding aspect was slightly changed. The carboxyl moiety of 22 and Lys120, Lys116, Tyr46 and Ser216 residues interacted via hydrogen bonds including salt bridge and conventional H-bonds, respectively. These interactions may contribute to the strong PTP1B inhibitory activity of 22.
Regarding α-glucosidase inhibitory activity, 9 showed notable inhibitory activity among the FAs. However, we could not define the correlation among α-glucosidase inhibitory activity, unsaturation, and number of carbon atoms. In addition, sterols and triterpenoid glycosides did not show any inhibition against α-glucosidase under the tested concentrations, but triterpenoids 22 and 23 exhibited similar effect with the positive control, acarbose.
This study has four important findings: (i) the isolation and structure identification of compounds from H. fusiformis, (ii) the identification of FAs as PTP1B and α-glucosidase inhibitors, (iii) the demonstration that sterols derived from H. fusiformis function as PTP1B inhibitors, and (iv) the demonstration that glycyrrhizin and its metabolites function as PTP1B and α-glucosidase inhibitors. Notably, glycyrrhizin (18) is metabolized by β-d-glucuronidase or intestinal flora to glycyrrhetinic acid [44,45]. Therefore, the in vivo anti-diabetic activity of 18 may be attributed to the PTP1B and α-glucosidase inhibitory activity of its metabolite, glycyrrhetinic acid.
In conclusion, the in vitro experimental and in silico computational results from this study confirmed that compounds isolated from H. fusiformis exhibit potent PTP1B and α-glucosidase inhibitory activity. Among the isolated compounds, FAs and triterpenoid derivatives showed potent inhibitory activity against both enzymes. However, sterols did not show any inhibition activity against α-glucosidase. Taken together, these results suggest that constituents of H. fusiformis could be used as promising anti-diabetic materials to delay the absorption of glucose via inhibition of α-glucosidase enzyme in the digestive organs and to enhance the insulin signaling pathway via inhibition of the PTP1B enzyme in insulin-sensitive organs.

General Experimental Procedures
The specific rotations were operated on a JASCO DIP-370 digital polarimeter. The 1 H-and 13 C-NMR spectra were recorded in methanol-d 4 and chloroform-d on a JEOL JNM ECP-400 spectrometer (Tokyo, Japan) at 400 MHz and 100 MHz, respectively. The infrared (IR) spectra were measured on a Mattson Polaris FT/IR-300E spectrophotometer. Mass spectra were recorded using a Quattro II mass spectrometer. Column chromatography was conducted using Diaion HP-20, Sephadex LH-20 (20-

Plant Material
Seaweed H. fusiformis was purchased from Wando, Republic of Korea. A whole plant voucher specimen was registered and deposited at the Department of Food and Life Science, Pukyong National University, Busan, South Korea (Professor Jae Sue Choi).

In Vitro α-Glucosidase Inhibitory Activity Assay
Enzyme inhibition studies were carried out spectrophotometrically in a 96-well micro-plate reader using a procedure reported by Li et al. [47]. Acarbose was used as a positive control.

In Vitro PTP1B Inhibitory Activity Assay
The inhibitory activity of isolated compounds against truncated form of human recombinant PTP1B was evaluated using pNPP as a substrate [48]. The amount of p-nitrophenyl produced after enzymatic dephosphorylation of pNPP was estimated by measuring the absorbance at 405 nm using a micro-plate spectrophotometer (Molecular Devices, Sunnyvale, CA, USA). Ursolic acid was used as a positive control.

Kinetic Parameters of Active Compounds towards PTP1B Inhibition
The inhibition constant (K i ) and inhibition mode for the inhibition of PTP1B was calculated by the Lineweaver-Burk plot and its secondary plot of the slope and the y-intercept of compounds [49,50]. The kinetic parameters were obtained over various concentrations of substrate (0 to 2 mM) and inhibitors (0, 4.7, 23.3, and 116.6 µM for compound 6; 0, 2.5, 5, and 10 µM for compound 13; 0, 5, 10, 20, and 40 µM for compounds 22 and 23). Graphs were generated using SigmaPlot 12.0 (Systat Software Inc., San Jose, CA, USA).

Statistical Analysis
All experiments were carried out in triplicate and repeated on three separate days. All data are expressed as the mean ± standard deviation (SD) (n = 3).