Polyphenol fatty acid esters as serine protease inhibitors: a quantum-chemical QSAR analysis

Abstract

We investigated the ability of polyphenol fatty acid esters to inhibit the activity of serine proteases trypsin, thrombin, elastase and urokinase. Potent protease inhibition in micromolar range was displayed by rutin and rutin derivatives esterified with medium and long chain, mono- and polyunsaturated fatty acids (1e–m), followed by phloridzin and esculin esters with medium and long fatty acid chain length (2a–d, 3a–d), while unmodified compounds showed only little or no effect. QSAR study of the compounds tested provided the most significant parameters for individual inhibition activities, i.e. number of hydrogen bond donors for urokinase, molecular volume for thrombin, and solvation energy for elastase. According to the statistical analysis, the action of elastase inhibitors is opposed to those of urokinase and thrombin. Cluster analysis showed two groups of compounds: original polyphenols together with rutin esters with short fatty acid chain length and rutin esters with long fatty acid chain length.

Keywords::

• Rutin
• phloridzin
• esculin
• serine proteases
• quantum-chemical calculations

Introduction

Serine proteases belong to the most studied class of proteolytic enzymes, and are thus a primary target for drug discovery^1,2. Serine protease enzymes of interest include trypsin–like enzymes, such as trypsin, kallikrein, plasmin, thrombin and tryptase; chymotrypsin–like enzymes, including chymotrypsin, cathepsin G and chymase; elastase–like enzymes, including neutrophil elastase and elastase; and carboxypeptidase-like enzymes³. These enzymes are involved in a diverse array of biological functions, such as digestion, immune response, blood pressure regulation, blood clotting, cardiovascular function, hormone processing, angiogenesis, bone remodelling and ovulation^3,4. However, under pathological conditions they are responsible for the onset of many diseases, including cancer, cardiovascular disorders, inflammation, coagulation diseases, multiple sclerosis, Alzheimer’s disease, diabetes mellitus, atherosclerosis, and pancreatitis^3,5–9. Appropriate drug therapy can comprise the inhibition of a particular serine protease implicated in the pathology and/or symptomatology of a disease. Hence, substantial interest concerns the identification of serine protease inhibitors, which possess high selectivity for specific serine proteases³.

Natural polyphenolic compounds, such as flavonoids, chalcones and coumarins, have been recognized to have various beneficial bioactivities on human health, making them an attractive target to be used in prevention and therapy^7,10. Current knowledge suggests that polyphenolic compounds have the ability of selective inhibition of a wide range of enzymes including pathological proteases^11–14. Strong inhibitory effects of polyphenols, including flavonoids, on various serine proteinases have been reported^11,15,16.

Recently, enzymatic modifications of flavonoids have been reported. There is evidence that selective flavonoid acylation provides a useful tool for improvement of existing biological properties, including antioxidant, antimicrobial and anticarcinogenic activity^17–20, and it may introduce various beneficial properties to the novel compounds, such as penetration through the cell membrane²¹. Lin et al.²² observed increased 5α-reductase inhibition after acylation of (-)-epigallocatechins. Salem et al.²³ showed that the acylation of isorhamnetin-3-O-glucoside enhanced its capacity to inhibit xanthine oxidase. Our recent investigations indicate that lipophilic flavonoid derivatives are strong inhibitors of transport enzymes such as sarcoplasmic reticulum Ca²⁺-ATPase and plasma membrane Ca²⁺-ATPase^24,25.

Based on the recent findings that acylation of polyphenols may contribute to the improvement of enzyme inhibition, our goal of this work was to test rutin, phloridzin and esculin fatty acid esters, to examine the influence of the selective modification of the basic polyphenolic skeleton on serine protease inhibition and to assess the significance of individual molecular properties for the resulting inhibition activities using QSAR study.

Experimental protocols

Material

Rutin (1), phloridzin (2) and esculin (3) were purchased from Sigma-Aldrich Chemical Co. The individual rutin, phloridzin and esculin derivatives (1a–m, 2a–d and 3a–d) were synthesized at the Food Research Institute, Biocentre Modra.

Lipase B from Candida antarctica immobilized on acrylic resin (EC 3.1.1.3, 10 000 U/g), trypsin from bovine pancreas (EC 3.4.21.4, 12 900 U/mg), thrombin from bovine plasma (EC 3.4.21.5, 2 500 NIH U/mg), elastase from porcine pancreas (EC 3.4.21.36, 840 U/mg), Nαbenzoyl-D,L-arginine p-nitroanilide hydrochloride, N-glycine-arginine p-nitroanilide dihydrochloride, Nα-benzoyl-phenylalanyl-valyl-arginine p-nitroanilide hydrochloride, N-succinyl-L-alanyl-L-alanyl-L-alanine p-nitroanilide, quercetin (4), 4-guanidobenzoic acid hydrochloride (5), capric, caprylic, caproic, lauric, palmitic, oleic, linoleic, linolenic, arachidonic and erucic acid were purchased from Sigma-Aldrich Chemical Co. 2-Methylbutan-2-ol, trifluoroacetic, butyric, myristic and stearic acid were obtained from Fluka Chemie GmbH. Urokinase 500 000 HS from human urine (EC 3.4.21.73, 500 000 IU/mg) was from Medac GmbH. Molecular sieves 4Å (10–20 mesh) and Silica gel 60 (230–400 mesh) were supplied by Merck. All solvents and other reagents were of analytical, spectrometric or HPLC grade.

General procedure for the synthesis of rutin, phloridzin and esculin esters

Rutin, phloridzin and esculin esters (1a–m, 2a–d and 3a–d) were synthesized via lipase–catalyzed esterification of glycosylated polyphenols (rutin 1, phloridzin 2 and esculin 3) with different acylating agents (fatty acids of C4–C22) in 2-methylbutan-2-ol at 60°C for 168 h according to.17,26 Molar ratio polyphenol:fatty acid was adjusted to 1:5. All reactants were previously dried in a vacuum dryer for 24 h at 80°C and the solvent with molecular sieves (100 g/l) for at least 7 days. To initiate direct esterification reaction, immobilized lipase from Candida antarctica (0.25 g) was added to the reaction mixture. The water content in the solvent was kept lower than 2% with molecular sieves (150 g/l). The extent of the reaction was monitored by HPLC analysis. The reaction was terminated by removing the enzyme and molecular sieves by filtration.

After the solvent was evaporated under vacuum, the reaction mixture was subjected to column chromatography on silica gel. The separation was performed with ethyl acetate/methanol 7:3 (v/v) to obtain respective fractions, which were analyzed by HPLC for rutin, esculin and phloridzin ester content. Subsequently, the solvent was evaporated and esters were re–crystallized from water. Final purity of the obtained products was determined using HPLC analysis.

Analytical methods

HPLC analysis was done using Purospher STAR RP-18e column (250 mm × 4.0 mm, 5 μm) with a UV detector. Single components were separated using a gradient of 10% methanol containing 0.1% trifluoroacetic acid (A) and methanol (B) at flow rate of 1 ml/min: 0 min (A:B 25/75), 20 min (100/0), 25 min (25/75). Rutin, phloridzin and esculin and their esters were detected at 280 nm and 360 nm. The elution was performed at 30°C. Sample injection volume was 20 μl. The conversion yields were calculated from the ratio between the molar concentration of rutin, phloridzin and esculin esters and the initial molar concentration of rutin, phloridzin and esculin. Calibration curve for rutin, phloridzin and esculin was obtained using standards in methanol.

The chemical structure of the purified esters of rutin, phloridzin and esculin with laurate and palmitate (1e, 1g, 2a, 2c, 3a and 3c) was determined by ¹H and ¹³C NMR analysis in DMSO-d6 using Varian Unity-Inova 600 MHz spectrometer.

¹H chemical shifts for rutin-4′′′-O-laurate (DMSO-d6): δ (ppm) 7.50 (H2′ and H6′), 6.83 (H5′), 6.34 (H8), 6.17 (H6), 5.42 (H1′′), 4.64 (H4′′′ acylated), 4.45 (H1′′′), 3.68–3.14 (9H rhamnoglucosyl), 0.75 (CH3 rhamnosyl).

¹³C chemical shifts for rutin-4′′′-O-laurate (DMSO-d6): δ (ppm) 177.34 (C4), 164.34 (C7), 161.22 (C9), 156.45 (C2), 156.39 (C5), 148.37 (C3′), 144.80 (C4′), 133.05 (C3), 121.43 (C6′), 121.06 (C1′), 116.14 (C5′), 115.12 (C2′), 103.75 (C10), 100.91 (C1′′), 100.47 (C1′′′), 98.69 (C6), 93.45 (C8), 76.48 (C3′′), 75.47 (C5′′), 74.04 (C2′′), 73.30 (C4′′′), 70.36 (C2′′′), 69.55 (C4′′), 68.09 (C3′′′), 66.79 (C6′′), 65.71 (C5′′′), 17.05 (CH3 rhamnose).

¹H chemical shifts for rutin-4′′′-O-palmitate (DMSO-d₆): δ (ppm) 7.50 (H_2′ and H_6′), 6.84 (H_5′), 6.35 (H₆), 6.18 (H₈), 5.43 (H_1′′), 4.64 (H_{4′′′ acylated}), 4.45 (H_1′′′), 3.68–3.16 (9H rhamnoglucosyl), 0.75 (CH₃ rhamnosyl).

¹³C NMR chemical shifts for rutin-4′′′-O-palmitate (DMSO-d₆): δ 177.41 (C4), 164.21 (C7), 161.27 (C9), 156.52 (C2), 156.42 (C5), 148.38 (C3′), 144.83 (C4′), 133.10 (C3), 121.45 (C6′), 121.11 (C1′), 116.20 (C5′), 115.15 (C2′), 103.86 (C10), 100.93 (C1′′), 100.49 (C1′′′), 98.68 (C6), 93.46 (C8), 76.51 (C3′′), 75.51 (C5′′), 74.08 (C2′′), 73.94 (C4′′′), 70.40 (C2′′′), 69.56 (C4′′), 68.14 (C3′′′), 66.77 (C6′′), 65.74 (C5′′′), 17.08 (CH₃ rhamnose).

¹H chemical shifts for phloridzin-6′′′-O-laurate (DMSO-d6): δ (ppm) 7.02 (H2′′), 6.63 (H3′′′), 6.10 (H3′), 5.94 (H5′), 4.97 (H1′′′), 4.26 (H6′′′ acylated), 3.56 (H5′′′), 3.40 (H₂), 3.30 (H2′′′ and H3′′′), 3.15 (H4′′′), 2.77 (H₃), 2.26 (H-α), 1.46 (H-β), 1.30–0.84 (aliphatic chain), 0.75 (CH3 rhamnosyl).

¹³C NMR chemical shifts for phloridzin-6′′′-O-laurate (DMSO-d6): δ 204.59 (C1), 172.80 (C=O), 165.31 (C6′), 164.43 (C4′), 160.56 (C2′), 155.25 (C4′′), 131.43 (C1′′), 129.08 (C2′′), 114.91 (C3′′), 105.12 (C1′), 100.54 (C1′′′), 96.87 (C5′), 94.48 (C3′), 76.36 (C3′′′), 73.84 (C5′′′), 73.06 (C2′′′), 69.76 (C4′′′), 63.01 (C6′′′), 44.94 (C2), 33.36 (C-α), 28.96 (C3), 31.27–13.93 (aliphatic chain), 24.34 (C-β).

¹H chemical shifts for phloridzin-6′′′-O-palmitate (DMSO-d₆): δ (ppm) 7.02 (H2′′), 6.64 (H3′′′), 6.11 (H3′), 5.95 (H5′), 4.98 (H1′′′), 4.26 (H6′′′ acylated), 3.57 (H5′′′), 3.41 (H₂), 3.31 (H2′′′ and H3′′′), 3.15 (H4′′′), 2.78 (H₃), 2.26 (H-α), 1.46 (H-β), 1.32–0.83 (aliphatic chain), 0.75 (CH3 rhamnosyl).

¹³C NMR chemical shifts for phloridzin-6′′′-O-palmitate (DMSO-d₆): δ 204.64 (C1), 172.85 (C=O), 165.37 (C6′), 164.49 (C4′), 160.59 (C2′), 155.31 (C4′′), 131.46 (C1′′), 129.13 (C2′′), 114.95 (C3′′), 105.16 (C1′), 100.58 (C1′′′), 96.91 (C5′), 94.52 (C3′), 76.40 (C3′′′), 73.88 (C5′′′), 73.09 (C2′′′), 69.80 (C4′′′), 63.04 (C6′′′), 44.99 (C2), 33.42 (C-α), 28.99 (C3), 31.34–13.96 (aliphatic chain), 24.37 (C-β).

¹H chemical shifts for esculin-6′-O-laurate (DMSO-d6): δ (ppm) 7.84 (H3), 7.30 (H5), 6.82 (H8), 6.22 (H4), 4.85 (H1′), 4.33 (H6′ acylated), 3.63 (H5′), 3.31 (H2′ and H3′), 3.19 (H4′), 2.26 (H-α), 1.44 (H-β), 1.30–0.85 (aliphatic chain).

¹³C NMR chemical shifts for esculin-6′-O-laurate (DMSO-d6): δ 172.74 (C=O), 160.40 (C2), 151.42 (C7), 144.26 (C4), 142.26 (C6), 114.09 (C5), 111.98 (C3), 103.16 (C8), 101.38 (C1′), 75.74 (C3′), 73.86 (C5′), 73.07 (C2′), 69.92 (C4′), 63.23 (C6′), 33.47 (C-α), 24.39 (C-β), 31.25–13.92 (aliphatic chain).

¹H chemical shifts for esculin-6′-O-palmitate (DMSO-d6): δ (ppm) 7.85 (H3), 7.31 (H5), 6.83 (H8), 6.23 (H4), 4.85 (H1′), 4.33 (H6′ acylated), 3.64 (H5′), 3.32 (H2′ and H3′), 3.19 (H4′), 2.27 (H-α), 1.44 (H-β), 1.30–0.88 (aliphatic chain).

¹³C NMR chemical shifts for esculin-6′-O-palmitate (DMSO-d6): δ 172.77 (C=O), 160.45 (C2), 151.44 (C7), 144.31 (C4), 142.30 (C6), 114.12 (C5), 111.99 (C3), 103.18 (C8), 101.41 (C1′), 75.77 (C3′), 73.89 (C5′), 73.11 (C2′), 69.95 (C4′), 63.25 (C6′), 33.49 (C-α), 24.40 (C-β), 31.26–13.95 (aliphatic chain).

Enzyme assay and inhibitory studies

For determination of trypsin, thrombin, urokinase and elastase inhibitory activity by polyphenolic compounds, we used photometric methods with chromogenic substrates Nα-benzoyl-D,L-arginine p-nitroanilide hydrochloride for trypsin, N-glycine-arginine p-nitro anilide dihydrochloride for urokinase, Nα-benzoyl-phenylalanyl-valyl-arginine pnitroanilide hydrochloride for thrombin and N-succinyl-L-alanyl-L-alanyl-L-alanine p-nitroanilide for elastase according to the methods described earlier^27–30.

The rate of hydrolysis of the substrates (0.3 mM) by trypsin (18 U/ml), thrombin (12.5 NIH U/ml), urokinase (62500 IU/ml) and elastase (8.4 U/ml) was monitored at 410 nm in Tris-HCl buffer (50 mM, pH 8.0) without the presence of Ca²⁺ or other activators, at 37°C, pH 7.6. Data scanning time was set to the 1st and 61st minute after the reaction start.

Rutin, esculin, phloridzin and their esters were initially solubilized at a concentration of 10 mM in dimethyl sulphoxide (DMSO) and subsequently incorporated in the reaction mixture to final concentrations in the range of 1–50 μM. Samples were run in triplicate in 96-well plates by MRX microplate reader. Each experiment was performed in triplicate. The inhibitory activity was expressed as the concentration required for 50% activity inhibition (IC₅₀). Data were expressed as means of the percentage of control ± S.D. (standard deviation of the mean) of the indicated number of observations. Statistical comparison between groups was performed using the one-way ANOVA followed by Tukey test. Differences among means were considered significant when p < 0.05.

QSAR analysis

Low energy conformations (starting geometries) of the compounds studied were obtained by Monte Carlo equilibrium conformer search (MMFF94) and subsequent optimization in DFT B3LYP 6-31G* method, all performed in programme SPARTAN’08 (Wavefunction Inc., USA). The following parameters were obtained for final conformers: hydration energy E_aq, energy of the highest occupied molecular orbital E_HOMO, energy of the lowest unoccupied molecular orbital E_LUMO, dipole d, molecular volume V, molecular surface S, polar surface area PSA, number of hydrogen bond donors HBD and number of hydrogen bond acceptors HBA. The values of octanol-water partition coefficients AlogP were calculated by Dragon programme (http://www.talete.mi.it). Statistical evaluation of the results was performed by Statistica 7.1 Software (StatSoft Inc., USA). The pair correlation matrix was calculated for all values and subsequently the best linear regressions were chosen for multiple linear regression calculations. To check the reliability of the regression equations, the compounds were divided into two groups: test compounds (2d and 4) and prediction set (the rest of the compounds). The compounds 2d and 4 were chosen as representatives of both extremes, ester of flavonoid glycoside and flavonoid aglycone. The best regression equations recalculated for the prediction set were further used for the prediction calculations of two test compounds 2d and 4. For cluster analysis of the compounds studied, the programme Statistica 7.1 and the Ward method with Manhattan distances were used³¹. Molecular lipophilic MLP and molecular electrostatic potentials were obtained using VEGA programme (http://www.ddl.unimi.it).32 Molecular weight MW of the compounds was calculated using Molinspiration software based on fragmental methods (http://www.molinspiration.com).

Results and discussion

Polyphenol fatty acid ester synthesis

We chose glycosylated polyphenolic compounds with proven health beneficial effects to study their inhibitory action on serine proteases³³. Chemical structures of the initial polyphenols with arrow-marked acylation sites are shown in Figure 1. The whole set of selectively acylated derivatives of rutin, phloridzin and esculin (1a–m, 2a–d, and 3a–d) with appropriate conversion yields and common physicochemical constants is listed in Table 1. To date, different procedures on the preparation and structural identification of some rutin, phloridzin and esculin fatty acid esters were previously described^{17,19,20,26,34}. We observed that the highest conversion (> 50%) of polyphenols to appropriate esters was reached when short and medium chain fatty acids (C4–C12) were introduced on the acyl acceptor molecule. As the carbon number of fatty acid grew (from C12 to C18), the conversion yield of esters gradually decreased. A lower conversion degree was observed in rutin derivatives due to the absence of the primary hydroxyl group, which is known to be preferentially esterified. The fatty acid chain length was recognized to play an important role in acylation. The effect of fatty acid chain length on enzymatic acylation of rutin, phloridzin and esculin was previously published with similar conversion yields reached³⁴.

Figure 1.  Chemical structures of the initial polyphenols (rutin 1, phloridzin 2 and esculin 3) with depicted positions of acylation.

Table 1.  Physicochemical parameters and conversion yields of rutin, phloridzin and esculin esters.

	Ester	Abbreviation	MW^a	Mp^b (°C)	RT^c (min)	Conversion (%)
1a	Rutin butyrate	R–4	680.61	155–158	5.34	61.8
1b	Rutin capronate	R–6	708.60	161–163	7.10	57.7
1c	Rutin caprylate	R–8	736.72	159–161	8.90	55.2
1d	Rutin caproate	R–10	764.77	151–154	9.42	57.9
1e	Rutin laurate	R–12	792.83	143–145	12.53	55.1
1f	Rutin myristate	R–14	820.88	132–135	14.89	57.3
1g	Rutin palmitate	R–16	848.94	142–144	17.40	47.0
1h	Rutin stearate	R–18	876.99	138–141	19.82	34.2
1i	Rutin oleate	R–18:1	874.97	151–154	16.38	26.6
1j	Rutin linoleate	R–18:2	872.96	155–158	17.16	26.9
1k	Rutin linolenate	R–18:3	870.94	138–144	17.90	28.3
1l	Rutin arachidonate	R–20:4	896.98	145–147	18.59	25.1
1m	Rutin erucate	R–22:1	931.08	172–175	20.93	15.7
2a	Phloridzin laurate	P–12	618.72	110–112	14.53	70.1
2b	Phloridzin myristate	P–14	646.77	116–118	17.23	54.3
2c	Phloridzin palmitate	P–16	674.83	122–124	19.72	55.8
2d	Phloridzin stearate	P–18	702.88	103–106	21.96	39.2
3a	Esculin laurate	E–12	522.59	160–163	15.58	95.4
3b	Esculin myristate	E–14	550.64	152–154	17.84	94.7
3c	Esculin palmitate	E–16	578.70	148–151	19.98	79.0
3d	Esculin stearate	E–18	606.75	129–132	21.88	71.6

Reaction conditions: 50 mM polyphenol, 250 mM fatty acid, 10 g/l Candida antarctica lipase, 150 g/l molecular sieve 4Å, in 2–methylbutan–2–ol, at 60°C, 168 h.

^aMolecular weight (MW) calculated using Molinspiration program.

^bMelting point (Mp) determined on a Kofler melting point apparatus, uncorrected.

^cRetention time (R_T) determined by HPLC analysis.

Serine protease inhibitory activity

To study inhibitory action of the acylated polyphenols, we focused on a group of serine proteases including digestive proteases trypsin and pancreatic elastase, blood coagulation protease thrombin and fibrinolytic protease urokinase. The inhibitory activities of the initial polyphenolic compounds 1, 2 and 3, their lipophilic derivatives 1a–m, 2a–d and 3a–d, as well as standards (4 and 5) on trypsin, thrombin, urokinase and elastase activity, in terms of the concentrations required to inhibit activity levels by 50% (IC₅₀), are summarized in Table 2. Quercetin was used as a natural polyphenolic reference compound due to its potent inhibitory effect on serine proteases and the synthetic standard 4-guanidinobenzoate was involved. Rutin and rutin derivatives esterified with medium and long chain fatty acids (1f–m) proved to be most effective serine protease inhibitors with most intensive inhibitory effect on thrombin, followed by urokinase and trypsin. Phloridzin esters (2a–d) showed some inhibitory effects on urokinase (IC₅₀ ~ 33 μM) and thrombin (IC₅₀ = 65–75 μM). Esculin derivatives (3a–d) were found to be effective inhibitors of elastase (IC₅₀ = 20–60 μM). No significant inhibitory properties were observed in the initial compounds, except in rutin.

Table 2.  Inhibitory activities of rutin, phloridzin, esculin and their lipophilic esters on trypsin, thrombin, urokinase and elastase expressed as concentrations required for 50% activity inhibition (IC₅₀) (SD < 5%, n = 4, R > 0,95) and subsistent logarithmic values (–logIC₅₀ +6)

	Compound	IC50 (μM)/ pIC50 (M)
		Trypsin	Thrombin	Urokinase	Elastase
1	Rutin	40/ 4.40	25/ 4.60	6/ 5.22	100/ 4.00
1a	Rutin butyrate	>500/ 3.29	22/ 4.66	8/ 5.10	>500/ 3.29
1b	Rutin capronate	>500/ 3.29	32/ 4.49	14/ 4.85	>500/ 3.29
1c	Rutin caprylate	>500/ 3.29	25/ 4.60	32/ 4.49	>500/ 3.29
1d	Rutin caproate	>500/ 3.29	32/ 4.49	20/ 4.70	>500/ 3.29
1e	Rutin laurate	90/ 4.05	14/ 4.85	50/ 4.30	>500/ 3.29
1f	Rutin myristate	130/ 3.89	10/ 5.00	15/ 4.82	>500/ 3.29
1g	Rutin palmitate	40/ 4.40	3/ 5.52	15/ 4.82	>500/ 3.29
1h	Rutin stearate	40/ 4.40	8/ 5.10	15/ 4.82	>500/ 3.29
1i	Rutin oleate	60/ 4.22	9/ 5.05	25/ 4.60	>500/ 3.29
1j	Rutin linoleate	45/ 4.35	5/ 5.30	30/ 4.52	>500/ 3.29
1k	Rutin linolenate	50/ 4.30	7/ 5.15	20/ 4.70	>500/ 3.29
1l	Rutin arachidonate	75/ 4.12	5/ 5.30	9/ 5.05	>500/ 3.29
1m	Rutin erucate	10/ 5.00	10/ 5.00	30/ 4.52	>500/ 3.29
2	Phloridzin	90/ 4.05	320/ 3.49	80/ 4.10	100/ 4.00
2a	Phloridzin laurate	55/ 4.26	70/ 4.15	9/ 5.05	>500/ 3.29
2b	Phloridzin myristate	65/ 4.19	75/ 4.12	33/ 4.48	>500/ 3.29
2c	Phloridzin almitate	60/ 4.22	75/ 4.12	18/ 4.74	>500/ 3.29
2d	Phloridzin stearate	215/ 3.67	65/ 4.19	13/ 4.89	>500/ 3.29
3	Esculin	150/ 3.82	250/ 3.60	25/ 4.60	100/ 4.00
3a	Esculin laurate	350/ 3.46	60/ 4.22	100/ 4.00	20/ 4.70
3b	Esculin myristate	120/ 3.92	54/ 4.27	100/ 4.00	25/ 4.60
3c	Esculin palmitate	210/ 3.68	54/ 4.27	100/ 4.00	30/ 4.52
3d	Esculin stearate	50/ 4.30	52/ 4.28	89/ 4.05	60/ 4.22
4	Quercetin	10/ 5.00	35/ 4.46	20/ 4.70	50/ 4.30
5	4–Guanidinobenzoate	450/ 3.35	400/ 3.40	20/ 4.70	500/ 3.30

The inhibitory activity of unmodified coumarin 3 was considerably increased after lipophilization. It seems that the shorter the fatty acid introduced on the coumarin skeleton, the more potent the inhibitory action gained by the compound. The inhibitory effect of esculin esters was comparable to that of quercetin. This could be explained by the fact that elastase has a preference for small, aliphatic side chains within the inhibitor ring structure^35,36.

Jedinak et al.³⁷ found that flavonoid glycosylation decreases inhibitory activity on thrombin. Our study shows that acylation of polyphenol glycosides improves serine protease inhibitory activity. We suggest that increased inhibitory properties of the acylated polyphenols may be connected with enhanced hydrophobicity of the compounds. Probably the acyl introduced on the polyphenolic skeleton would provide an interaction with hydrophobic region of serine enzymes, and would thus provide a better inhibitory activity. However, non-specific interactions of the longer acyl chains are also possible, and may result in a decrease in protease activity through aggregation induced by the hydrophobic interactions. We previously showed that lipophilic rutin derivatives were potent inhibitors of both sarcoplasmic reticulum Ca²⁺-ATPase and plasma membrane Ca²⁺-ATPase and that this inhibition was associated with conformational alterations in the enzyme structure^24,25. Lin and co-authors²² showed that acylation of (-)-epigallocatechins considerably increased 5α-reductase inhibition. They suggested that the long-chain fatty acid introduced on the flavonoid skeleton interacted with hydrophobic pocket of the enzyme. Other groups observed that the acylation of flavonoids such as isorhamnetin-3-O-glucoside and mesquitol led to the improvement of xanthine oxidase inhibitory activity^23,37.

Current knowledge suggests that a general mechanism might be involved in the enhanced inhibitory activity of the acylated polyphenols on structurally diverse classes of enzymes, which seems to be donated by the medium to long fatty acid chains.

QSAR analysis

The values of molecular descriptors were calculated for optimal geometries of the compounds studied. For illustration, the geometries of rutin and rutin stearate together with their molecular lipophilic and electrostatic potentials are shown in Figure 2.

Figure 2.  Geometry of rutin and rutin stearate together with molecular lipophilic potential MLP (upper part) and molecular electrostatic potential MEP (lower). Potentials were calculated by VEGA program.

In order to obtain a reasonable range of values for statistical calculations, the inhibitory activities were expressed as (–logIC₅₀ +6) (Table 2). Our QSAR study started with calculation of molecular parameters for the prediction set and test compounds (Table 3). For QSAR analysis purposes, the values of the inhibitory activities <3.3 were set to 3.29. For all values in Table 2 and 3, the total correlation matrix was obtained. Enzyme related correlations of the simple linear regressions are summarized in Table 4 (molecular surface S was omitted due to its linear dependence on molecular volume V). The best simple correlations of calculated data with serine proteinase inhibitory activity were obtained for polar surface area (Figure 3).

Table 3.  Values of octanol–water partition coefficient AlogP for pH = 7.0, hydration energy E_aq, energy of the highest occupied molecular orbital E_HOMO, energy of the lowest unoccupied molecular orbital E_LUMO, dipole d, molecular volume V, molecular surface S, polar surface area PSA, number of hydrogen bond donors HBD and number of hydrogen bond acceptors HBA.

	AlogP	Eaq (kJ/mol)	E HOMO (kJ/mol)	E LUMO (kJ/mol)	d (Debye)	S (Å^2)	V (Å^3)	PSA (Å^2)	HBD	HBA
1	–1.45	–119.61	–547.40	–161.05	3.71	554.22	532.94	221.05	10	16
1a	0.06	–134.07	–539.46	–149.37	3.84	629.28	609.70	213.81	9	16
1b	0.97	–131.79	–540.16	–149.45	3.70	669.95	646.63	213.94	9	16
1c	1.88	–130.80	–540.21	–149.44	3.63	709.98	683.50	213.85	9	16
1d	2.79	–132.39	–544.29	–173.50	5.71	746.01	720.11	217.02	9	16
1e	3.71	–145.58	–544.55	–175.56	6.02	784.51	756.77	216.66	9	16
1f	4.62	–139.41	–542.49	–148.42	5.72	834.73	795.00	222.68	9	16
1g	5.53	–88.45	–555.95	–166.67	4.43	927.75	888.27	216.47	9	16
1h	6.44	–138.55	–566.18	–192.59	6.24	865.63	829.67	211.23	9	16
1i	6.00	–152.82	–568.72	–195.59	6.28	905.64	866.44	211.14	9	16
1j	5.55	–167.35	–565.10	–206.20	4.26	897.21	862.71	208.55	9	16
1k	5.11	–194.95	–565.41	–207.44	4.38	887.55	858.23	208.49	9	16
1l	5.58	–198.67	–562.41	–203.53	4.33	887.66	854.47	207.62	9	16
1m	7.82	–122.02	–533.09	–161.89	5.88	990.57	937.77	216.66	9	16
2	0.75	–150.05	–585.92	–148.69	3.34	412.60	402.57	141.79	7	10
2a	5.90	–127.40	–553.28	–155.02	4.28	655.76	627.90	140.63	6	10
2b	6.82	–136.57	–568.27	–81.16	6.40	697.19	666.51	145.53	6	10
2c	7.73	–127.29	–569.04	–140.41	3.06	747.42	702.76	146.13	6	10
2d	–0.54	–122.91	–575.72	–128.04	3.33	781.60	739.34	146.02	6	10
3	5.90	–63.26	–564.31	–146.35	10.02	318.22	299.11	120.87	5	8
3a	5.53	–48.23	–564.53	–148.39	9.61	557.15	524.01	118.45	4	8
3b	6.44	–49.40	–589.33	–170.77	6.40	606.38	562.20	126.19	4	8
3c	7.35	–46.09	–590.66	–171.60	6.30	647.60	599.09	126.62	4	8
3d	1.50	–44.64	–593.22	–174.78	6.02	686.31	635.82	126.15	4	8
4	8.64	–68.32	–528.82	–177.79	0.21	277.41	265.39	106.46	5	7

^aThe values of octanol–water partition coefficients AlogP according to Ghose and Crippen were calculated by Dragon program. All other parameters were calculated by the program SPARTAN’08.

Table 4.  Values of correlation coefficients obtained for simple linear regression for experimental inhibition activities for trypsin (TRP), urokinase (URK), thrombin (TRB) and elastase (ELS) and independent variables AlogP, Eaq, E_HOMO, E_LUMO, d, V, PSA, HBD and HBA.

	TRP	URK	TRB	ELS
TRP	1.00	0.08	0.29	–0.01
URK	0.08	1.00	0.36	–0.63
TRB	0.29	0.36	1.00	–0.49
ELS	–0.01	–0.63	–0.49	1.00
AlogP	0.49	–0.22	0.17	0.11
Eaq	–0.11	–0.42	–0.46	0.79
EHOMO	0.03	0.56	0.33	–0.43
ELUMO	–0.31	0.02	–0.61	–0.01
d	–0.23	–0.41	–0.12	0.28
V	0.13	0.23	0.77	–0.68
PSA	–0.06	0.51	0.75	–0.71
HBD	0.03	0.60	0.68	–0.74
HBA	–0.05	0.52	0.76	–0.73

For all regressions p < 0.05, n = 25.

Figure 3.  Correlation of a descriptor polar surface area PSA with enzyme inhibitory activities (p < 0.05, n = 25).

Inhibitory activity toward urokinase was significantly influenced by number of hydrogen bond donors HBD, energy of the highest occupied molecular orbital E_HOMO and polar surface area PSA. This indicates that electron-donor properties of the inhibitor may play a role in the mechanism of inhibition exerted on this enzyme by the compounds studied. The best regression equation obtained for the prediction set was For the test compounds, we obtained URK = 5.02 (4.89 measured) for 2d and URK = 5.04 (4.70 measured) for 4.

1

The activities toward elastase showed interesting results when compared with urokinase and thrombin (Table 4), indicating opposite trends with the respective correlation coefficients R = –0.63 and R = –0.49. This fact could be interpreted as being in favour of selectivity of the compounds studied and deserves further examination since drug selectivity is obviously a positive property of a designed compound.

Thrombin activity showed the best correlations with V, HBA, PSA (polar surface area) and acceptable correlations with several other descriptors. The best regression equation obtained for triple combination was The values predicted for the test compounds were TRB = 4.45 (4.19 measured) for 2d and TRB = 3.56 (4.46 measured) for 4. The predicted value for quercetin 4 showed a large deviation from the experimental one. This is caused probably by the peripheral values of descriptors, mainly the value of V, for this compound. However, the predicted value for compound 2d revealed a deviation near the SEE value. Contrary to the prediction power, only regression equation (2) for thrombin achieved values of the regression coefficients in the range of p -level limit < 0.05.

2

The elastase related activities provided a significant simple correlation with hydration energy E_aq. Along with the opposite character of ELS activities when compared with URK and TRB, this indicates the importance of the desolvation process for the elastase inhibitors. The best regression equation obtained was The values predicted for the test compounds were ELS = 3.19 (3.29 measured) for 2d and ELS = 4.14 (4.30 measured) for 4, which are also within the range of SEE. According to the statistical significance of coefficients in equation (3), the regression for elastase reached the least satisfactory values from the entire calculated multiple regressions caused probably by insufficient variation of data. The analogous equations, where only rutin and its derivatives were taken into account, were also calculated for all enzymes. According to both correlation coefficient and statistical significance, the equations obtained were worse than for the heterogeneous set of compounds.

3

In the case of trypsin, no significant correlation was found, suggesting that some other parameters should be taken into account. Values of partition coefficient AlogP provided no significant correlation with the inhibition activities studied and will be further used in the examination of the transport properties of the compounds presented. All structural indications obtained from regression equations, given as values of correlation coefficients, are summarized in Table 4.

Based on the values of molecular weight (Table 1), number of hydrogen bond donors and acceptors (Table 3), it is evident that majority of the compounds tested in this study do not obey “Lipinski’s rule of five” and the question arises whether they, or related derivatives, can be expected to become prospective drugs. According to the authors Kinoshita et al.³⁹, who collected more then 7000 flavonoids with known or expected biological activities and tested them for “Lipinski’s rule of five”, all criteria were met only by 43% of the compounds. The values of Lipinski alert index LAI calculated by Dragon programme for the compounds studied were equal to 0 only for phloridzin, esculin, esculin derivatives and quercetin. Concerning the fact that “Lipinski’s rule of five” is not strongly valid for natural compounds with higher number of oxygen atoms and hence hydrogen bond donors and acceptors⁴⁰, these polyphenolic compounds might be further considered for drug discovery process.

In the dendrogram obtained by cluster analysis (Figure 4), two clusters appear: the first cluster contains rutin esters with long fatty acid chain length and the second one contains original polyphenols together with rutin esters with short fatty acid chain length. The first cluster is further divided into several subclusters: (1h–1m) and (1d–1g), where the subgroup (1h–1k) represents the most similar members, followed by the subgroups (1d, 1e), (1f, 1g) and (1l, 1m). This indicates that chain length is the deciding factor concerning the similarity of the compounds, while the presence of double bonds is a less significant factor. The second cluster contains small subcluster of the original polyphenols phloridzin, esculin and quercetin (2, 3 and 4) and a huge subcluster consisting of phloridzin esters (2a–2d), rutin short chain esters (1a–1c) and the subgroup containing the esculin esters together with rutin (1, 3a–3d).

Figure 4.  Results of the cluster analysis of the polyphenolic compounds studied according to the Ward method with Manhattan distances.

Based on QSAR analysis we can conclude that the inhibition potential of individual polyphenols was not influenced by substitution with fatty acids up to C10. Majority of the compounds acylated with long fatty acid chains contributed to higher inhibition activities. An exception was seen in elastase related activities, where the ester substitution provided an enhancement of the inhibitory activities only for the group of esculin derivatives. This is probably connected with the fact that only ELS activities provided a significant simple correlation with E_aq, which in absolute values was distinctly lower than for other compounds.

Statistical analysis obtained provided a more detailed view to the classification of the compounds under study and their mechanism of action, which will be the basis of our future study focused on molecular modelling.

Conclusions

We suggest that selective modification of polyphenols via lipophilization with fatty acids may represent a new approach to the production of potent low toxic serine protease inhibitors. Most potent serine protease inhibition was observed in rutin derivatives esterified with medium to long, mono- and polyunsaturated fatty acids. Recent findings indicate that these polyphenol-based acylated derivatives may be involved in a specific mechanism of enzyme activity inhibition.

Acknowledgments

We are thankful to Ing. Ladislav Soltes, Dr.Sc. for scientific advice and critical comments.

Declaration of interest

The work was supported by The Agency of the Ministry of Education, Science, Research and Sport of the Slovak Republic for the Structural Funds of EU, OP R&D of ERDF in the frame of the Project, Evaluation of natural substances and their selection for prevention and treatment of lifestyle diseases (ITMS 26240220040), by the Slovak Research and Development Agency LPP-0251-07 and by Grants VEGA 2/0030/11 and VEGA 2/0038/11. The authors report no declarations of interest.