B-type Proanthocyanidins with Dentin Biomodification Activity from Cocoa (Theobroma cacao)

Abstract

To enable translational studies, a scalable preparative isolation scheme was developed for underivatized cocoa (Theobroma cacao) proanthocyanidins (PACs), affording six all-B-type oligomeric PACs including a new tetramer 4. Their structures including absolute configuration were unambiguously established by comprehensive spectroscopic and chemical methods. Evaluation of the PACs’ dentin biomodification properties employed dynamic mechanical and infrared spectroscopic analyses in dentin bioassay models. PAC treatment enhanced the biomechanical strength of dentin by 5- to 15-fold compared to untreated dentin. Among the PAC agents, the pentamer, cinnamtannin A3 (6), led to the highest complex modulus value of 131 MPa, whereas the “branched” tetramer, 4, showed the lowest, yet still significant, bioactivity. This study of specifically singly-linked medium-length oligomeric PACs indicates that the linkage site is paramount in determining the potency of these PACs as dentin biomodifiers.

Graphical Abstract

INTRODUCTION

Proanthocyanidins (PACs), also known as condensed tannins (CTs), are oligomeric and polymeric flavan-3-ols, joined by A-type (2→O→7 or 2→O→5) and/or B-type (4→8 or 4→6) linkages and represent end products of the flavonoid biosynthetic pathway.^1,2 PACs have attracted considerable research interests in the fields of human and animal nutrition, for their diverse health-promoting benefits,^3–5 as new biomaterials such as biosensors,⁶ and as dentin biomodification agents,⁷ for their ability to crosslink with biomacromolecules. Understanding the bioactivities of PACs requires in-depth investigation of their chemical complexity, including the relative and absolute stereochemistry of the monomers, their degree of polymerization (DP), exact interflavan linkage (IFL) types, and their degree of galloylation.^8–10

Our dental natural products research has identified a series of plant-derived PACs as potential dentin biomodifiers. Promising agents include PACs isolated from pine (Pinus massoniana) and cinnamon (Cinnamomum verum) barks, ranging from dimers to hexamers featuring mixed A- and B-type IFLs, as well as B-type dimers and trimers from grape seed extract (GSE), all with potent activities in enhancing the mechanical properties of dentin.^11–14 The recent implementation of graphical structural descriptors, termed PAC Block ARrays (PACBAR),¹⁵ has facilitated these interdisciplinary studies, particularly with regard to the exploration of structure activity relationships (SARs) of PACs with complex A- and B-type linkage patterns and their impact on dentin biomodification activities.¹⁶

B-type PACs are found abundantly in food sources such as grape seeds, cocoa, apple, and green tea, which are generally valued as dietary nutrients for their health/resilience promoting benefits.^3,17 However, biological evaluation of trimeric and higher oligomeric B-type PACs has almost exclusive been using crude or sometimes PAC-enriched fractions,^18,19 as rigorously characterized pure compounds are hardly accessible. One of the underlying reasons might be that B-type PACs pose more challenges in purification and structural analysis than their A-type counterparts. Due to their multiple phenolic groups, the commonly used separation methods for PACs, which are primarily based on molecular weight such as Sephadex LH or Toyopearl HW, tend to interfere strongly via surface affinity interactions (e.g., intermolecular hydrogen bonding, hydrophobic, and π-π interactions).^20,21 Compared to A-type PACs containing additional ether bonds that confer conformational stability upon the molecule, the rotational constraints of the singleton linkages in B-type PACs lead to atropisomerism. This explains the common occurrence of multiple peaks or poor peak shapes (“humps”) in LC profile as well as peak broadening in NMR spectra acquired at ambient temperatures,^14,22,23 both reflecting the presence of multiple atropisomeric species. This study taking advantage of advanced separation and spectroscopy tools developed over the last few decades, by employing countercurrent separation (CCS) techniques, modern diol phases for HPLC, tandem high-resolution mass spectrometry (HRMS), low temperature NMR, to purify and characterize single, underivatized (native) B-type PACs at both the analytical and preparative scale. Moreover, it has been demonstrated that quantum-mechanics-driven iterative and functionalized spin analysis of the ¹H NMR spectra (HiFSA) yields highly precise spin parameters that enable the determination of the absolute configuration of flavan-3-ol monomers in PACs.^24,25

Cocoa beans are the seeds of Theobroma cacao, used for chocolate production, and flavan-3-ols as well as PACs are their main phenolic components.²⁶ In a preliminary study, the trimeric and tetrameric PAC enriched fractions of cocoa showed higher biomodification activity than the crude extract, whereas both significantly increased the biomechanical property of dentin.²⁷ B-type PACs from dimers up to pentamers have been reported as the characteristic metabolites from cocoa beans, and their underivatized structures were elucidated by low-temperature NMR, albeit without determination of their absolute configurations.^28–31 It is known that the processing conditions of cocoa products, such as fermentation, roasting, and drying can lead to structural and concentration variation of PACs, including epimerization.^31–33 Therefore, the structural characterization of PACs including the rigorous absolute stereochemistry determination is imperative to produce reproducible results in cross-disciplinary research.

In order to explore the structures and dentin biomodification activity relationships of oligomeric PACs with B-type linkages, cocoa was selected for isolation efforts targeting medium-DP oligomers. The present study led to the isolation and full characterization of six oligomeric PACs with DP from 3~5 from cocoa. The chemical structures and absolute configuration of the new tetramer, 4, as well as of five previously isolated PACs (1-3, 5, 6) were fully established via extensive NMR and ECD spectroscopic analysis, as well as through acid-catalyzed furanolysis. Dentin biomodification induced by B-type PACs were evaluated by dynamic mechanical analysis (DMA) and Fourier-transform infrared (FTIR) spectroscopy.

MATERIALS AND METHODS

General Experimental Procedures.

ECD spectra were recorded on a JASCO-815 spectrometer with a 0.10 cm quartz cuvette at sample concentrations of 0.10 mg/mL in MeOH. NMR spectra were acquired on a JEOL 400 MHz or 600 MHz spectrometers equipped with a Royalprobe™. High-resolution electrospray ionization (ESI) mass spectra were acquired by a Bruker Impact II quadrupole time of flight (qTOF) mass spectrometer. Column chromatography (CC) was performed on Sephadex LH-20 gel (Pharmacia, Uppsala, Sweden) and TOYOPEARL HW-40F (TOSOH Bioscience LLC, PA, USA). Centrifugal partition chromatography (CPC) was carried out on an SCPE-250 (Gilson, Inc., Middleton, WI, USA). TLC was with SIL G/UV254 (Macherey-Nagel, Inc., Bethlehem, PA, USA), visualized under UV light (254 nm) and then sprayed with vanillin-sulfuric acid reagent followed by heating. Semi-preparative HPLC was performed on a Shimadzu HPLC equipped with a binary solvent delivery pump LC-20AB, UV-Vis detector SPD-20A and a vacuum degasser, using a YMC-Pack ODS-AQ (250 × 10 mm I.D., 5 μm, 120 Å) column. All solvents used were obtained from Fisher Scientific (Fair Lawn, NJ, USA) or Sigma-Aldrich (St. Louis, MO, USA). NMR solvents were purchased from Cambridge Isotope Laboratories (Tewksbury, MA, USA). The NMR data was processed with MestReNova (ver. 14.2.0–26256, Mestrelab Research, S.L., Compostela, Spain). Quantum-mechanical analysis of the ¹H NMR spectra (HifSA) were carried out using the Cosmic Truth (CT; ctb.nmrsolutions.fi) software tools.

Plant Materials.

The extract (100 g) of fermented and unroasted seeds of Theobroma cacao was prepared by extraction with 30:70 acetone-water, followed by lyophilization. The material (Lot CMIE-7LJJKF) was provided by Barry Callebaut, Lebbeke-Wieze, Belgium in 2014 and was stored at −20 °C until used and stability confirmed by HPLC.

Diol Phase HPLC Profiling and UHPLC-MS analysis.

A sample for diol-phase HPLC analysis was prepared at 20 mg/mL, by directly dissolving the T. cacao extract in MeOH. ACN/HOAc (98:2) was solvent A, and MeOH/H₂O/HOAc (95:3:2) was solvent B. HPLC analysis was performed according to a method previously reported,³⁴ with further optimization of the total analysis time (0–5 min, 7% solvent B; 5–30 min, 7 to 30% solvent B; 30–32 min, 30 to 100% solvent B; 32–40 min, 100% solvent B) and using a Develosil™ HPLC 100 Diol (5 μm, 4.6 × 250 mm) column. The PDA chromatograms were extracted at 280 nm. A sample for UHPLC-MS analysis was prepared at 1 mg/mL in MeOH and the separation was performed on a CORTECS^® C18 (2.7 μm, 3.0 × 100 mm) column using gradient elution with an ACN/H₂O/FA solvent system changing from (5:95:0.01) to (30:70:0.01) in 8 min.

Optimization of Countercurrent Solvent System.

The shake flask method was performed to determine the K_U/L values of constituents in the PAC-rich cocoa extract in different solvent systems. The values were calculated as K_U/L = A_U/A_L, where A_U and A_L were the peak areas (UV 280 nm) of the target analytes in the upper and lower phase, respectively.³⁵ Briefly, ~1 mg of the extract was added to 5 mL of pre-equilibrated two-phase solvent system. The mixture was vigorously shaken and left to stand at room temperature until the phases were fully separated. Equal aliquots of each phase were then transferred to separate vials using a pipette, evaporated to dryness, each residue was dissolved in 1 mL of MeOH, and analyzed by HPLC with UV detection at 280 nm.

Extraction and Isolation.

Cocoa extract (30 g) was dissolved in 70% aqueous acetone (400 mL) overnight, filtered, and evaporated in vacuo, then freeze-dried to afford the PAC-enriched fraction (20 g). Divided into five aliquots of 4 g each, this fraction was subjected to separation on CPC using EtOAc/n-BuOH/MeOH/H₂O (6:0.1:1:5, v/v) in ascending mode at a flow rate of 10 mL/min and 3000 rpm. The CPC fractions were pooled into 6 subfractions, A-F, according to their TLC profiles. Fraction C (1.0 g) was further purified by Sephadex LH-20 CC using gradient elution from 70% to 100% aqueous MeOH and finally 70% aqueous acetone, affording four subfractions C1-C4. Subfraction C2 contained the major trimer 1 (procyanidin C1, 434.0 mg). Subfraction C3 was subjected to reserve-phase semi-prep HPLC (ACN/H₂O/FA, 18:82:0.1, 2.6 mL/min) to afford 2 (8.3 mg). Fraction D (542.0 mg) underwent Sephadex LH-20 CC using gradient elution from 90% to 100% aqueous MeOH and final washing with 70% aqueous acetone, giving four subfractions D1-D4. Subfraction D3 consisted of the purified major tetramer 3 (cinnamtannin A2, 300.4 mg). Subfraction D4 (55.0 mg) was further separated by semi-prep HPLC (ACN/H₂O/FA, 14:86:0.1, 3 mL/min) to yield 4 (4.7 mg) and 5 (12.8 mg). Fraction E (157.0 mg) was combined with subfraction D4 (680.0 mg) and separated by HW-40 CC in gradient elution mode from 70% to 100% aqueous MeOH plus a final 70% aqueous acetone elution, yielding three subfractions E1-E3. Subfraction E2 consisted of the purified major pentamer 6 (cinnamtannin A3, 146.0 mg). The purities (%, w/w) of the isolates 1–6, all fully soluble in MeOH, were calculated as the sums of observed rotamers of each isolates, determined as 95.4%, 93.7%, 94.5%, 90.2%, 98.0%, and 99.4%, respectively, by the 100% qHNMR method (Figures S40–S46).

Furanolysis.

To confirm the absolute configuration of the isolated PACs, furanolysis was performed.³⁶ The reference standard, epicatechin-(4→2″)-menthofuran, was prepared as previously described.³⁷ Similarly, 1~2 mg samples of each of the PACs 1–6 were prepared as 10 mM solutions in MeOH and then treated with 40 mM menthofuran. All reactions were performed in 0.1 N HCl in MeOH solution at 40 °C for one hour. Reaction mixtures were neutralized with 40 mM NaOAc and extracted with EtOAc. The reaction products were identified by LCMS and chiral phase HPLC analysis. The chiral phase HPLC analysis was performed using a Chiralpak^® IA (5 μm, 4.6 mm × 250 mm), elution with an isocratic solvent system of hexane/isopropanol/TFA (65:35:0.1, 1 mL/min).

Procyanidin C1 (micro-PACBAR: EC-8EC-8EC) (1):

Brownish amorphous powder; ECD (MeOH) λ_max (Δɛ) 218.5 (50.02), 238 (22.89), 278 (−1.93) nm; HRMS (ESI) m/z [M + H]⁺ Calcd for C₄₅H₃₉O₁₈ 867.2131; Found 867.2165. ¹H and ¹³C NMR data see Table 1.

Table 1.

NMR data of trimers 1 and 2 in methanol-d₄.

				1a,c		2b,c
Unit	Ring	No.	δ C	δH, J (Hz)	δ C	δH, J (Hz)
I	C	2	76.82*	5.0841 d (−0.70)	77.34	4.8919 ddd (1.07, −0.87, −0.71)
		3	73.49	3.9880 d (2.12)	73.09	3.8684 dd (2.10, 1.07)
		4	36.98	4.7022 d (2.12)	37.42	4.5864 d (2.10)
	A	5	158.00		158.12
		6	96.12	5.9914 d (2.34)	95.89	5.9745 d (1.16)
		7	157.70		159.18
		8	96.03	6.0198 d (2.34)	96.23	5.9704 d (1.16)
		9	154.84		159.29
		10	101.76		99.92
	B	1’	132.53*		132.51
		2’	114.95	6.9118 dd (1.84, −0.70)	115.31	6.9164 dd (2.04, −0.71)
		3’	145.29d		145.62d
		4’	145.81d		145.68d
		5’	115.85*	6.7591 d (8.21)	115.64	6.7061 d (8.26)
		6’	119.10	6.7056 d (8.21,1.84)*	119.52	6.6797 ddd (8.26, 2.04, −0.87)
II	F	2	76.82*	5.2524 d (0.66)	76.97	4.9460 ddd (1.46, −0.81, −0.68)
		3	72.99	3.9569 dd (1.95, 0.66)	73.39	3.8752 dd (1.46, 1.23)
		4	37.16	4.7097 d (1.95)	37.27	4.5373 d (1.23)
	D	5	156.40		156.59
		6	96.98	5.9067 s	108.23
		7	156.90		156.28
		8	106.95		96.23	6.1774 s
		9	154.54		155.67
		10	102.18		101.62
	E	1’	132.53*		132.28
		2’	114.86	7.0331 d (1.92)	115.02	6.8640 dd (2.00, −0.68)
		3’	145.30d		145.82d
		4’	145.89d		145.57d
		5’	115.85*	6.7446 d (8.27)	115.78	6.7332 d (8.11)
		6’	118.60	6.7106 dd (8.27, 1.92)*	119.07	6.6565 ddd (8.11, 2.00, −0.81)
III	I	2	79.47	5.0020 d (1.47)	79.28	4.9083 ddd (1.06, −0.84, −0.83)
		3	66.68	4.3248 ddd (4.40, 2.98, 1.47)	66.80	4.2409 ddd (4.35, 2.71, 1.06)
		4	29.93	2.9627 dd (−16.55, 4.40), 2.8221 dd (−16.55, 2.98)	29.77	2.8949 dd (−17.04, 4.35), 2.7802 dd (−17.0, 2.71)
	G	5	156.50		156.60
		6	97.28	5.9431 s	96.89	5.9085 s
		7	156.58		156.60
		8	107.56		106.55
		9	154.53		154.42
		10	100.09		100.05
	H	1’	131.91		132.10
		2’	115.07	7.1300 d (1.98)	114.92	7.1130 dd (2.00, −0.83)
		3’	145.57d		145.41d
		4’	145.61d		145.84d
		5’	115.88*	6.6882 d (8.17)	115.91	6.7090 d (8.37)
		6’	118.88	6.9093 dd (8.17, 1.97)	118.52	6.7898 ddd (8.37, 2.00, −0.84)

^aRecorded on 400/100 MHz at 255 K, resp.

^bRecorded on 600/150 MHz at 260 K, resp.

^cThe δ_H and J values were generated via HifSA.

^dInterchangeable assignment.

^*Overlapped signals.

Epicatechin-(4β→6)-epicatechin-(4β→8)-epicatechin; (micro-PACBAR: EC-6EC-8EC) (2):

Brownish amorphous powder; ECD (MeOH) λ_max (Δɛ) 215.5 (48.37), 239.5 (46.09), 274 (−1.26) nm; HRMS (ESI) m/z [M + Na]⁺ Calcd for C₄₅H₃₈NaO₁₈ 889.1950; Found 889.1947. ¹H and ¹³C NMR data see Table 1.

Cinnamtannin A2 (micro-PACBAR: EC-8EC-8EC-8EC) (3):

Brownish amorphous powder; ECD (MeOH) λ_max (Δɛ) 219 (40.71), 238 (20.93), 280.5 (−2.43) nm; HRMS (ESI) m/z [M + H]⁺ Calcd for C₆₀H₅₁O₂₄ 1155.2765; Found 1155.2800. ¹H and ¹³C NMR data see Table 2 and 3.

Table 2.

¹³C NMR data of the tetramers 3–5 and pentamer 6 in methanol-d₄.

Unit	Ring	No.	3a	4a	5a	6b
I	C	2	76.82*	77.66	77.37	76.78
		3	73.50	73.33	73.18	73.44
		4	37.08	37.47	37.42*	37.28
	A	5	158.00	158.16*	159.04c	156.91
		6	96.07	96.47*	95.86c	97.29*
		7	158.20	160.26	159.14c	157.89*
		8	95.98	96.97	96.18c	95.99
		9	157.90	159.32	158.12	154.91
		10	101.80	98.29	100.14*	101.75
	B	1’	132.54	132.14	132.60	132.52*
		2’	114.99	115.06	115.32	114.92
		3’	145.70c	145.70*	145.45–145.60	145.17–145.92
		4’	145.40c	145.70*	145.45–145.60	145.17–145.92
		5’	115.84*	114.96	115.67–116.00	115.80–116.00
		6’	119.10	119.61	119.59	119.03
II	F	2	76.66	76.85	76.81	76.69*
		3	73.00	73.04	73.26	72.81*
		4	37.41	37.47	37.42*	37.49*
	D	5	156.40	155.43	156.76	156.38*
		6	97.03	108.34	108.43	97.03
		7	157.00	154.16	156.55*	158.00
		8	107.11	107.16*	96.34	107.12
		9	154.80	152.89	155.57	154.91
		10	102.15	103.68	101.54	102.41
	E	1’	132.61	132.37	132.36	132.52*
		2’	114.84	114.95	115.05	114.92*
		3’	145.70c	145.49*	145.45–145.60	145.17–145.92
		4’	145.20c	145.70*	145.45–145.60	145.17–145.92
		5’	115.84*	115.89*	115.67–116.00	115.80–116.00
		6’	118.68*	118.71	119.09	118.74*
III	I	2	76.82*	79.34	76.60	76.78
		3	72.84	67.08	73.01	72.81*
		4	37.23	29.83	37.01	37.49*
	G	5	156.40*	156.72	156.50*	156.38*
		6	97.30	97.14	96.84	96.06
		7	156.90	156.49	156.99	157.89*
		8	107.62	107.16 *	106.06	107.47
		9	155.00	154.51	154.84	154.58
		10	102.35	99.89	102.36	102.32
	H	1’	132.49	131.9	132.50	132.52*
		2’	114.97	114.97	114.78	114.92*
		3’	145.90c	145.50*	145.21c	145.17–145.92
		4’	145.40c	145.70*	145.77c	145.17–145.92
		5’	115.84*	115.89*	115.67–116.00	115.80–116.00
		6’	118.68*	118.67	118.34	118.74*
IV	L	2	79.48	77.16	79.49	76.69*
		3	66.69	73.72	66.87	72.98
		4	29.98	37.77	29.92	37.09
	J	5	156.60	158.17	156.50*	156.38*
		6	97.20	96.47*	97.32	97.29*
		7	156.70	159.59	156.56	158.14
		8	107.60	96.13	107.59	107.30
		9	154.60	159.48	154.50	155.03
		10	100.09	99.29	100.14*	102.27
	K	1’	131.99	132.58	132.01	132.52*
		2’	115.09	115.01	115.10	114.82
		3’	145.61c	145.44*	145.56c	145.17–145.92
		4’	146.00c	145.49*	145.88c	145.17–145.92
		5’	115.84*	115.89*	115.67–116.00	115.80–116.00
		6’	118.88	119.09	118.92	118.64
V	O	2				79.44
		3				66.72
		4				29.93
	M	5				156.60
		6				97.29*
		7				156.71
		8				107.58
		9				154.60
		10				100.04
	N	1’				132.03
		2’				115.12
		3’				145.17–145.92
		4’				145.17–145.92
		5’				115.80–116.00
		6’				118.83

^aRecorded on 150 MHz at 260 K.

^bRecorded on 100 MHz at 255 K.

^cInterchangeable assignment.

^*Overlapped signals.

Table 3.

¹H NMR (δ_H in ppm, J in Hz) data of the tetramers 3 and 4 in methanol-d₄.^a

Unit	Ring	No.	3	4
I	C	2	5.1095 dd (0.66, −0.68)	4.9324 d (0.89)*
		3	4.0245 dd (1.57, 0.66)	3.9790 dd (2.00, 0.89)
		4	4.7441 d (1.57)	4.8273 d (2.00)*
	A	6	6.0112 d (2.16)	6.0823 d (2.39)
		8	6.0433 d (2.16)	6.1223 d (2.39)
	B	2’	7.1176 dd (2.00, −0.68)	6.9478 d (2.39)
		5’	6.7208 d (8.17)b	6.7112 d (8.20)b
		6’	6.8002 dd (8.17, 2.00)	6.5954 dd (8.20, 2.39)
II	F	2	5.2851 s	5.1715 d (0.93)
		3	4.1080 s	3.9403 dd (2.00, 0.93)
		4	4.7834 brs	4.5987 d (2.00)
	D	6	5.9568 s	-
	E	2’	7.0591 dd (2.00)	7.0078 d (2.01)
		5’	6.7550 d (7.95)b	6.7873 d (8.20)b
		6’	6.7759 dd (7.95, 2.00)b	6.7021 dd (8.20, 2.01)b
III	I	2	5.3108 s	4.9327 ddd (1.93, 1.22, −0.80)*
		3	4.0094 s	4.2278 dd (2.71, 1.22)
		4	4.7362 brs	2.9239 dd (−16.92 4.91) 2.7705 ddd (−16.92, 2.71, 1.93)
	G	6	5.9723 s	5.9527 s
	H	2’	6.9332 d (2.00)	7.1038 d (2.00)
		5’	6.7590 d (8.10)b	6.7631 d (8.20)b
		6’	6.7302 dd (8.10, 2.00)b	6.7986 dd (8.20, 2.00, −0.80)
IV	L	2	5.0195 d (1.00)	4.8292 d (0.97)*
		3	4.3415 dd (4.53, 3.26)	3.8388 dd (2.00, 0.97)
		4	2.9762 dd (17.00, 4.53) 2.8357 ddd (17.00, 3.26, 1.00)	4.4543 d (2.00)
	J	6	5.9617 s	5.9800 d (2.20)*
		8	-	5.9820 d (2.20)*
	K	2’	7.1512 d (2.00)	6.9234 d (2.06)
		5’	6.7793 d (8.10)b	6.7631 d (8.10)b
		6’	6.9284 dd (8.10, 2.00)	6.6363 ddd (8.10, 2.06, −0.80)

^aRecorded on 600 MHz at 260 K. The δ_H and J values were generated via HifSA.

^bThe assignments of the severely overlapped region were predicted by HifSA calculations.

^*Overlapped signals.

Epicatechin-(4β→8)-[epicatechin-(4β→6)-]epicatechin-(4β→8)-epicatechin (micro-PACBAR: EC-8EC(6-EC)-8EC) (4):

Brownish amorphous powder; ECD (MeOH) λ_max (Δɛ) 220.5 (61.29), 237.5 (58.00), 275 (−0.34) nm; HRMS (ESI) m/z [M + H]⁺ Calcd for C₆₀H₅₁O₂₄ 1155.2765; Found 1155.2749. ¹H and ¹³C NMR data see Table 2 and 3.

Epicatechin-(4β→6)-epicatechin-(4β→8)-epicatechin-(4β→8)-epicatechin (micro-PACBAR: EC-6EC-8EC-8EC) (5):

Brownish amorphous powder; ECD (MeOH) λ_max (Δɛ) 215.5 (60.24), 237.5 (42.12), 273 (−1.05) nm; HRMS (ESI) m/z [M + H]⁺ Calcd for C₆₀H₅₁O₂₄ 1155.2765; Found 1155.2758. ¹H and ¹³C NMR data see Table 2 and 4.

Table 4.

¹H NMR (δ_H in ppm, J in Hz) data of tetramer 5 and pentamer 6 in methanol-d₄.^a

Unit	Ring	No.	5b	6c
I	C	2	4.9159 dd (1.80, −0.68)	5.1136 s
		3	3.8506 dd (2.50, 1.80)	4.0340 d (1.29)
		4	4.5704 d (2.50)*	4.7541 d (1.29)
	A	6	5.9488 d (2.31)*	6.0129 d (2.05)*
		8	5.9485 d (2.31)*	6.0444 d (2.05)
	B	2’	6.9021 d (2.03)	7.1309 d (2.25)*
		5’	6.7327 d (8.17)d	6.9281 d (8.20)d
		6’	6.6765 ddd (8.17, 2.03, −0.68)	6.7489 d (8.20, 2.25)
II	F	2	4.9528 d (2.00)	5.3324 d (2.16)
		3	4.0108 dd (2.50, 2.00)	4.1445 d (2.16, 1.04)
		4	4.5704 d (2.50)*	4.8088 d (1.04)
	D	6	-	5.9759 s
		8	6.1928 s	-
	E	2’	6.8912 d (2.10)	6.9386 d (1.92)
		5’	6.7118 d (8.24)d	6.7829 d (8.10)d
		6’	6.6947 dd (8.24, 2.10)d	6.8319 dd (8.10, 1.92)
III	I	2	5.1568 d (2.20)	5.3167 s
		3	3.8924 dd (2.20, 2.00)	4.0155 d (1.81)
		4	4.6165 d (2.00)	4.7956 d (1.81)
	G	6	5.9185 s	6.0078 s
	H	2’	7.0208 d (1.80)	7.0678 d (1.88)
		5’	6.7197 d (8.24)	6.7631 d (8.25)d
		6’	6.5825 dd (8.24, 1.80)	6.7298 dd (8.25, 1.88)d
IV	L	2	4.9445 dd (2.80, −1.38)	5.2968 d (0.83)
		3	4.2737 ddd (4.50, 3.50, 2.80)	4.1313 dd (2.33, 0.83)
		4	2.9237 dd (−16.35, 4.50) 2.7779 dd (−16.35, 3.50)	4.7350 d (2.33)
	J	6	5.9137 s	5.9612 s*
	K	2’	7.0685 d (2.05)	7.1313 d (2.45)*
		5’	6.7197 d (8.20)d	6.7537 d (8.53)d
		6’	6.8510 dd (8.20, 2.05, −1.38)	6.7832 dd (8.53, 2.45)d
V	M	2		5.0218 d (1.09)
		3		4.3417 ddd (4.02, 3.22, 1.09)
		4		2.9777 dd (−16.95, 4.02) 2.8356 dd (−16.95, 3.22)
	O	6		5.9616 s*
	N	2’		7.1518 d (2.00)
		5’		6.7222 d (7.63)
		6’		6.9283 dd (7.63, 2.00)d

^aThe δ_H and J values were generated via HifSA.

^bRecorded on 600 MHz at 260 K.

^cRecorded on 400 MHz at 255 K.

^dThe assignments of the severely overlapped region were predicted by HifSA calculations.

^*Overlapped signals.

Cinnamtannin A3 (micro-PACBAR: EC-8EC-8EC-8EC-8EC) (6):

Brownish amorphous powder; ECD (MeOH) λ_max (Δɛ) 219 (105.42), 238 (54.60), 279 (−4.87) nm; HRMS (ESI) m/z [M + H]⁺ Calcd for C₇₅H₆₃O₃₀ 1443.3399; Found 1443.3431. ¹H and ¹³C NMR data see Table 2 and 4.

Dynamic Mechanical Analysis (DMA) of Dentin.

Dentin matrix specimens were subjected to a strain sweep to assess viscoelastic properties, as published elsewhere.^37,38 Briefly, dentin specimens (0.5×1.7×7 mm) were demineralized and treated with compounds 1–6 (1% w/v in 20 mM HEPES) for 1 h (n = 3) followed by rinsing. Viscoelastic properties (complex modulus - E*; damping capacity - tanδ) of the dentin matrix were obtained following a strain sweep (amplitude: 1 to 100 μm at 1 Hz), performed before and after treatment with PACs. The E* represents the elastic (storage modulus, E′) and viscous (loss modulus, E″) properties of a material, whereas the damping capacity is calculated by the ratio of both components (tanδ = E″/E′).³⁹ Results obtained before and after treatment were used to calculate the fold variation of E*. The intra-group equality of variances was determined by Levene’s test and found to meet the assumption of homogeneous distribution for E* (p = 0.174), but not for tan δ (p = 0.016). Data were statistically analyzed using one-way ANOVA and Tukey and Games-Howell post-hoc test, respectively (α = 0.05, SPSS v.25, SPSS).

Biochemical Analysis of the Dentin Extracellular Matrix.

Fourier-transform infrared spectroscopy (FTIR, Nicolet 6700, Thermo Fisher Scientific) was applied for the biochemical characterization of the functional groups of the dentin extracellular matrix (ECM), as previously described.^16,38 Briefly, spectra were collected with an attenuated total reflectance (ATR) apparatus from 4,500 to 600 cm⁻¹. An accumulation of 128 scans per sample (n = 3) was collected at a resolution of 4 cm⁻¹ and analyzed with OMNIC Spectra (Thermo Fisher Scientific). The spectra were normalized by mean values of all data sets after baseline subtraction using Origin Pro 8 (OriginLab, Northampton, MA) software. The main bands of type I collagen were assessed: amide I (C=O stretching mode at 1,630 cm⁻¹), amide II (N-H bending and C-N stretching modes at 1,550 cm⁻¹), amide III (C-N stretching modes and N-H bending at 1,240 cm⁻¹), and wagging vibrations of CH₂ groups at 1,450 cm⁻¹. Additionally, the spectra of compounds 1–6 were collected using the same settings, as previously described.³⁷ The conformational modifications in collagen were quantified by fitting the areas and intensities of the assigned bands and using indices calculated between amide bands (II and III) and the CH₂ scissoring band (amide II/CH₂ and amide III/CH₂). Data were statistically analyzed to evaluate the effects of each compound to dentin collagen, using one-way ANOVA and Tukey’s post-hoc test (α = 0.05, SPSS v.25, SPSS). The equality of variances intra-groups was determined by the Levene test and found to meet the assumption of homogeneous distribution for E* (p = 0.174), not for tan δ (p = 0.016).

RESULTS AND DISCUSSION

Chromatographic Separation of DP Clusters.

Diol phase HPLC analysis showed the relative content of extractable PACs from T. cacao, which eluted as defined clusters in the order of DP (Figure 1A). The LC-MS metabolomic profile of cocoa extract revealed the presence of primarily B-type PACs with DP values ranging from 1 to 5 (Figure 1B). Higher-DP PACs were hardly ionizable and, thus, not observed under the chosen conditions.

Figure 1.

(A) Diol HPLC (280 nm) and (B) UHPLC-MS (ESI+) profiles of the investigated cocoa extract. The numbers in blue present the respective DP.

Centrifugal partition chromatography (CPC), countercurrent chromatography (CCC), and gel permeation chromatography are efficient methods in cocoa fractionation and isolation of oligomeric PACs.^30,40,41 In the present study, an optimized CPC solvent system for cocoa PAC separation was selected via the shake flask method, which is capable of yielding partition coefficient (K) values for target analytes.^35,42 Four solvent systems were selected, and the K values of major peaks are presented in Table S1. As a result, the EtOAc - n-BuOH- MeOH - H₂O (6/0.1/1/5) solvent system was selected as the K values of the target analytes were in the optimal range of 0.5~2 in this system. Employing reported CPC separation parameters,³⁷ the cocoa extract could be separated into six fractions (Figure 2). Fractions A and B mainly contained monomers and dimers, whereas fraction F was the extrusion fraction containing polymeric PACs. The major trimeric (1), tetrameric (3), and pentameric (6) PACs were obtained by further purification of fractions C to E via a single step of Sephadex LH-20 or Toyopearl HW-40 chromatographic column (CC), which provided good overall yield (0.5~1.4%, w/w of extract) and purity (94.5~99.4% by 100% qHNMR method, Figure S40, S42, S46) for the isolates. Therefore, this two-step preparative approach can serve as an efficient and feasible protocol for the scaled-up preparation of B-type oligomeric PACs. This approach is amenable to the 100+ mg scale, allowing for subsequent bioactivity evaluation and translational studies.

Figure 2.

Diol-HPLC (280 nm) analysis of the CPC fractions.

Structural Elucidation of PACs.

Compounds 1 and 2 are a pair of isomers that, based on HRESIMS data (Figure S33–S34), share the molecular formula of C₄₅H₃₈O₁₈. Three carbon resonances (CH/CH₂) in the low frequency (29~38 ppm) and six oxygenated methines around 65~80 ppm from the ¹³C NMR spectra, which were ascribed as C-4 and C-2/C-3 of each (epi)catechin unit, respectively, suggested both 1 and 2 being B-type trimers. PAC 1 was identified as the major trimer procyanidin C1 (Figure 3) known from cocoa via comparison with reference material by TLC and ¹H NMR spectrum (Figure S1). Comparing the ¹³C NMR spectra of 1 and 2, the most significant difference was one quaternary carbon resonance (δ_C 108.23, C-6 or C-8) in 2 being shifted to higher frequency compared with 1 (δ_C 106.95 and 107.56, II/III-C-8). This resonance was assigned as II-C-6 of 2 based on HMBC correlations from I-H-4 to II-C-5/6/7 and II-H-8 to II-C-6/7/9/10. Thus, the two IFLs in 2 were established as 4→6 and 4→8, respectively, and its NMR data (Table 1) was consistent with that of a compound isolated from apples.²²

Figure 3.

Classical chemical structures along with the graphical and micro PACBAR representations of the PAC oligomers 1–6 from cocoa (*new compound).

When the structures of the trimers, 1 and 2, were established previously by NMR spectroscopic analysis, only their relative configurations were reported.²² Herein, their absolute configurations were determined for the first time, using ECD spectroscopy and furanolysis. The positive Cotton effects (CEs) at 220~240 nm and negative CEs at 260~280 nm were observed in both trimers 1 and 2 (Figure 5), indicating the β orientation of the IFLs and α orientation of the B-rings.^11,43 The furanolysis including the reference compounds were prepared as described before.³⁷ The cleavage products of both 1 and 2 were determined as epicatechin and epicatechin-(4→2″)-menthofuran by LCMS and chiral HPLC analysis (Figure 6 and S46). Therefore, the absolute configurations and overall structures of the trimers 1 and 2 were determined as epicatechin-(4β→8)-epicatechin-(4β→8)-epicatechin and epicatechin-(4β→6)-epicatechin-(4β→8)-epicatechin (Figure 3), respectively, which in micro PACBAR nomenclature can be written as EC-8EC-8EC and EC-6EC-8EC, respectively.¹⁵

Figure 5.

ECD spectra of PACs 1–6 (left), with a magnified view of the 260–300 nm low abundance range (right).

Figure 6.

Chiral HPLC (280 nm) analysis of the furanolysis products of PACs, 1–6, and the reference standards.

Based on their HRESIMS spectra (Figures S35–S37), compounds 3–5 also possessed one identical molecular formula, C₆₀H₅₀O₂₄. Compound 3 represents the major tetramer from cocoa and was identified as cinnamtannin A2 by comparing its NMR spectra (Table 2 and 3) with reference data.^22,30 This conclusion was further confirmed by ECD spectroscopy (Figure 5) and furnanolysis (Figure 6 and S39).

Compound 4 showed the typical ¹H NMR pattern of PACs exhibiting: the AMX spin systems of 1,3,4-trisubstituted aromatic B-rings at high frequencies (δ_H 6.50~7.10); doublet-type signals from AB spin system systems or singlet signal from A-ring aryl hydrogens (δ_H 5.95~6.13, H-6/H-8); AMX resonances between 3.80 and 5.20 ppm arising from the methine hydrogens (H-2/H-3/H-4) of the heterocyclic C-rings; and the methylene hydrogens H₂-4 of the terminal (epi)catechin unit at low frequencies (Figure S15). Comparing the ¹³C NMR spectra of 3 and 4 (Table 2), one quaternary carbon resonance at δ_C 108.34 was shifted to higher frequency in 4 compared to 3 (δ_C 107.11~107.62), indicating that the interflavan linkage sites were at either C-6 or C-8 of the lower units. Detailed analysis of the 2D NMR spectra of 4 revealed that a “branched” (4→6) unit originated from the middle unit of this tetramer. The observation of ROESY correlations from I-H-4 to II-H-2′/6′ and II-H-4 to III-H-2′/6′ confirmed the 4→8 IFLs between units I and II as well as units II and III. The HMBC correlations from IV-H-4 to C-5/6/7, and from I-H-4 to C-7/8/9 of unit II, located the branched unit as being linked in position II-C-6 (Figure 4). The small coupling constants (³J = 0.8~2.0 Hz) for H-2/3 in each unit suggested the cis-configuration of H-2 and H-3 (Table 3).^11,43 The presence of the carbons C-2 of units I, II, and IV at relatively low frequencies (δ_C 77.66, 76.85, 77.16, resp.) indicated that all IFLs were in trans position relative to C-2, based on the γ-gauche effect.¹²

Figure 4.

Key 2D NMR correlations of 4 (red – HMBC, blue – COSY, green – ROESY).

The absolute configuration was further established by ECD spectroscopy and furanolysis. The positive CE at 220~240 nm, and negative CE at 260~280 nm, indicated the β orientation of the IFLs at C-4 and the α orientation of the B-rings in 4 (Figure 5).^11,43 The furanolysis cleavage products were identified as epicatechin and epicatechin-(4→2″)-menthofuran, respectively, by HPLC on chiral phase column and via LC-MS (Figure 6 and S39). Therefore, tetramer 4 was determined as epicatechin-(4β→8)-[epicatechin-(4β→6)-]epicatechin-(4β→8)-epicatechin as shown in Figure 3, equivalent to the PACBAR code of EC-8EC(6-EC)-8EC.

Being an isomer of 3, tetramer 5 showed a highly similar NMR profile. Yet, small but significant chemical shift differences were observed for the three quaternary carbons at δ_C 106.06~108.43 in 5 vs. δ_C 107.11~107.62 in 3. This could be ascribed to structural differences affecting C-6 or C-8 of the lower units. Comprehensive analysis of the 2D NMR spectra of 5 found a 4→6 IFL between units I and II. This conclusion was based on the HMBC correlations between I-H-4/H-3 with II-C-6, as well as the ROESY cross peaks between II-H-4 with III-H-2′/6′, and III-H-4 with IV-H-2′/6′. The relative configuration of 5 was established by analyzing the coupling constants of H-2 and H-3, as well as the differential carbon chemical shifts of C-2 based on the γ-gauche effect. As such, compound 5 was determined to likely be identical with a tetramer identified previously from apple (Malus pumila).²³ However, while the ¹³C NMR data was generally consistent with that provided in the reference, the assignments of II-H-2′ and IV-H-2′ were determined to be incorrect as reported²³ and were, thus, revised according to the ROESY correlations from H-2/H-3 to H-2′ of each unit (Table 4). Finally, the absolute configuration of 5 was determined by ECD spectroscopic analysis (Figure 5) coupled with furanolysis (Figure 6 and S39). Overall, the absolute structure of tetramer 5 was established as epicatechin-(4β→6)-epicatechin-(4β→8)-epicatechin-(4β→8)-epicatechin as shown in Figure 3, and encoded in the micro-PACBAR system as EC-6EC-8EC-8EC.

Compound 6 was isolated from the most polar fraction obtained from the CPC fractionation. The molecule exhibited a molecular ion at m/z 1443.3431 in the HRESIMS spectrum (Figure S38). Moreover, in its ¹³C NMR spectrum, one methylene (δ_C 29.93) and four methine groups (δ_C 37.09~37.49) were observed at low frequencies, suggesting that 6 was an all-B-type pentamer. Upon systematic assignment of the hydrogen and carbon resonances, the IFLs were all determined to be (4→8), which was particularly confirmed by the ROESY correlations from the H-4 of the upper to H-2′/6′ of the adjacent lower units.⁴⁴ The constitutional monomers were determined to be epicatechin from the couplings ³J_H2,H3 < 2.2 Hz (Table 4). Moreover, the IFLs were established as being trans configured relative to C-2 based on the γ-gauche effect, which made the C-2 carbon (δ_C76.69~76.78) of units I-IV resonate at relatively low frequencies (Table 2). The structure of this molecule was previously described as cinnamtannin A3 from apple²³ and unroasted cocoa beans³⁰, and the NMR and ECD data were consistent with those reported previously (Figure 5).

Biomechanical and Biochemical Analysis of the B-type PAC-biomodified Dentin Matrix.

Dynamic mechanical analysis (DMA) was used to determine the viscoelastic properties of the dentin matrix, by measuring the complex modulus (E*) and damping capacity (tan δ).³⁸ The biomodification of the dentin matrix mediated by B-type PACs showed a 5- to 15-fold increase in E* (p < 0.001, Table 5). The B-type pentamer 6 exhibited the highest E* value among the treated groups, whereas tetramer 4, featuring a relatively rare branched (4→6) IFL, yielded lower E* statistically similar to the untreated dentin, yet with still a 5-fold increase. Meanwhile, the B-type PACs biomodification did not affect the damping capacity (tan δ) of the dentin matrix.

Table 5.

Summary of the biomechanical and chemical characteristics of dentin matrices biomodified by PACs 1–6.^*

	Biomechanical analysis		Biochemical analysis
Compounds	Complex Modulus (E*)	Damping capacity (Tan δ)	Amide II/CH2 (A1550/A1450)	Amide III/CH2 (A1240/A1450)
1	93.2 ± 14.5 AB	0.29 ± 0.02 AB	6.34 ± 0.44 A	1.37 ± 0.14 AB
2	93.0 ± 22.6 AB	0.25 ± 0.02 AB	5.98 ± 0.13 AB	1.32 ± 0.06 B
3	77.8 ± 21.7 AB	0.25 ± 0.03 AB	5.38 ± 0.32 AB	1.28 ± 0.09 AB
4	44.0 ± 21.5 BC	0.28 ± 0.02 A	4.98 ± 0.39 BC	1.25 ± 0.04 B
5	86.6 ± 19.9 AB	0.25 ± 0.01 A	5.30 ± 0.66 ABC	1.28 ± 0.17 AB
6	130.9 ± 32.6 A	0.20 ± 0.01 B	4.38 ± 0.27 C	1.00 ± 0.28 AB
Control	9.7 ± 3.1 C	0.20 ± 0.04 AB	6.36 ± 0.04 A	1.66 ± 0.07 A

^*Identical letters indicate a lack of statistically significant differences among the groups in each column (p < 0.05).

Fourier-transform infrared (FTIR) spectroscopy has been established for the characterization of the structure of collagens and collagen-based materials.^16,45 Figure S47 shows the FTIR spectra of PAC-treated vs. unmodified (control) collagen of dentin dentin extracellular matrix (ECM). New peaks, shifts, and changes of absorbance intensity of bands were observed between 1150~1000 cm⁻¹ in the fingerprint region of the spectra of PAC-treated dentin, which indicated the structural modification of dentin collagen. The additional peaks shown in the PAC-treated dentin matrix (Figure 7A) correspond to the characteristic functional groups of PACs (Figure 7B). In the region of 1200~900 cm⁻¹, the absorptions around 1100 were assigned to the C-O stretching of the PAC C-ring,⁴⁶ bands at 1150 cm⁻¹ were attributed to the symmetric stretching of C-O-C bonds in PACs,¹⁶ and the shoulder peaks around 1120~1115 cm⁻¹ were assigned tentatively to the C–H in plane deformation resonances.⁴⁷ The observation of these characteristic bands from PAC-treated dentin matrix indicated that PACs retained its chemical integrity during the treatment.

Figure 7.

(A) Biochemical characterization of type I collagen in the dentin matrix using FTIR spectroscopy (fingerprint region from 1170 to 950 cm⁻¹). Arrows point to the exclusive peaks in PAC-treated samples (1120, 1100, 1050 cm⁻¹), as well as to major peaks in the control sample (1080 and 1030 cm⁻¹). (B) IR spectrum of compounds 1–6.

The ratios of amide II and amide III to CH₂ band were selected to quantify structural modifications of dentin collagen. The amide II/CH₂ ratios significantly decreased in tetramer 4-, and pentamer 6-treated dentin matrix when compared to control (p < 0.001). This suggested the formation of PAC-collagen cross-linking (N-H hydrogen bonding) and changes of the collagen secondary structure. Significant differences were observed when comparing dentin matrices treated with trimer 2 and tetramer 4 vs the untreated matrix (control) (p = 0.007), where a decrease in amide III/CH₂ was evident. In previous studies, the 4→8 vs 4→6 IFLs in complex, mixed A/B-type PACs were found to have limited effect on the biomechanical properties of dentin matrices, and the most potent PACs, in terms of DP, were the trimers and tetramers.^16,37

The current biomechanical and biochemical observations add new perspectives to our previous understanding of the SARs of oligomeric PACs interaction with dentin. Mainly, they provide the new insight that branching (i..e, an additional 4→6 IFL in a middle unit) of B-type PACs may limit the mechanical enhancement of dentin matrix despite showing similar structural changes on the collagen compared to their unbranched 4→8 IFLs analogues. Another new conclusion is that the DP has little or no influence on biomechanical properties among B-type trimers and pentamers as both showed similar E* values, most likely because they are already in the range of DP with optimum biomechanical performance.¹⁶ Some biological studies have reported that A-type PAC fractions were more effective and stable than B-type PAC fractions.^48,49 Trimeric PACs from red wine (Vitis vinifera) with 4→8 linkages showed stronger skin-whitening effects than 4→6 linkages PACs in vitro assay.⁵⁰ Although the underlying mechanisms are still unclear, the present results further consolidate our proposal that DP, interflavan linkage types, and linkage sites are structural features that collectively determine which PACs are most suitable with regard to size and hydrophilic-hydrophobic properties, unfolding the most robust interaction with biomacromolecules such as dentin collagen. Studied with the same agent concentrations and dentin models, the overall B-type PACs activities in the current study were still somewhat lower than those observed with AB-mixed type PACs.³⁷ Nevertheless, the higher DP B-type PACs described here have suitable dentin biomodification properties and remain viable candidates for future translation studies.

B-type PACs are exclusive metabolites of higher plants and a widely studied class of natural products in the area of biomedicine and food chemistry, due to their ubiquity in human diets. However, their absolute and sometimes even their relative stereochemistry and the associated structural implications have often been neglected. In summary, the current study describes a scale-up preparative isolation of B-type PACs from cocoa that enabled comprehensive structural determination including the establishment of absolute configuration via ECD and acid-catalyzed depolymerization, at a scale that can support translational studies.

Furthermore, the isolation of B-type oligomeric PACs including a new tetramer 4 expanded the available library of dentin biomodification agents for SAR studies. The dental bioassays revealed the new observations that, among B-type oligomeric PACs, the linkage sites (C-6 vs C-8) can influence the biomechanical effect of dentin collagen. Meanwhile, trimers and pentamer with identical linkage sites showed similar effects and fall within the optimal DP range (3~5) of PACs with dentin bioactivity. The results further support the translational potential of specific medium-DP PACs from cocoa in restorative dental applications.