The methanol (MeOH) extract of I. lacunosa seeds was partitioned between hexane and 95% MeOH. The 95% MeOH-soluble fr. was chromatographed on Diaion HP20, Sephadex LH-20, and octadecyl silica (ODS) columns, and HPLC using ODS furnished five compounds (1–5) as amorphous powders.

The HR-positive-ion and HR-negative-ion electrospray ionization time-of-flight mass spectrometry (ESI-TOF-MS) data for 1 (lacunosin I) showed [M + Na]+ and [M–H]– ion peaks at m/z 1621.7816 and 1597.7852, respectively, indicating a molecular formula of C72H126O38. The 1H-NMR spectrum of 1 revealed signals due to seven anomeric protons [δ 6.47 (1H, d, J = 1.5 Hz), 6.36 (1H, d, J = 7.5 Hz), 5.96 (1H, d, J = 8.0 Hz), 5.84 (1H, d. J = 7.5 Hz), 5.12 (1H, d, J = 7.5 Hz), 4.87 (1H, d, J = 7.5 Hz), 4.80 (1H, d, J = 7.5 Hz)], one methoxy group [δ 3.64 (3H, s)], two sets of non-equivalent methylene protons [δ 2.73 (1H, dd, J = 7.5, 15.5 Hz), 2.71 (1H, dd, J = 5.0, 15.5 Hz); δ 2.62 (1H, ddd, J = 8.0, 9.0, 15.5 Hz), 2.51 (1H, ddd, J = 7.0, 9.0, 15.5 Hz)] adjacent to the carbonyl groups, seven secondary methyl groups [δ 1.97 (3H, d, J = 6.0 Hz), 1.83 (3H, d, J = 6.0 Hz), 1.63 (3H, d, J = 6.0 Hz), 1.60 (3H, d, J = 6.0 Hz), 1.47 (3H, d, J = 6.0 Hz), 1.45 (3H, d, J = 7.5 Hz), 1.35 (3H, d, J = 7.5 Hz)], and two primary methyl groups [δ 0.97 (3H, t, J = 7.0 Hz), 0.93 (3H, t, J = 7.0 Hz)], together with a signal presumably due to one H-2 [δ 2.81 (1H, dq, J = 7.5, 7.5 Hz)] of the niloyl residue. The 13C-NMR spectrum shows signals corresponding to three carboxyl (δ 174.8, 173.7, 172.9) and seven anomeric carbons (δ 106.2, 103.5, 102.7, 102.1, 101.7, 101.2, 96.8). The 1H- and 13C-NMR signals were assigned based on the 1H-1H COSY, 1H-1H total correlation spectroscopy (TOCSY), HMQC, HMBC, and NOESY spectra (Tables 1 and 2). The coupling constant of the 1H-NMR signals attributed to the anomeric and methine protons, along with the C–H coupling constant (1JC-1–H-1) of the anomeric carbon signals [12] observed in the 13C-NMR spectrum, indicated that the monosaccharide units of 1 consisted of 1 mol each of fucopyranose with β in 4C1 conformation and rhamnopyranose with α in 1C4 conformation, 2 mol of glucopyranose with β in 4C1 conformation, and 3 mol of quinovopyranose with β in 4C1 conformation. The absolute configurations of the monosaccharide components quinovose, glucose, fucose, and rhamnose of the crude resin glycoside fraction of I. lacunosa seeds were previously identified as d, d, d, and l, respectively, based on the HPLC analysis of their thiocarbamoyl-thiazolidine derivatives [10]. In contrast, the organic acid component of 1 was identified as 1 mol of nilic acid. The absolute configuration of the nilic acid component of the crude resin glycoside fraction of I. lacunosa seeds was previously confirmed to be 2R,3R through specific rotation measurement and HPLC analyses with a chiral stationary phase of its p-bromophenacyl ester [10]. In the HMBC spectrum of 1, important correlations are observed between the signals of H-1 of the first glucosyl residue (Glc) and C-2 of the first quinovosyl residue (Qui); H-2 of Glc and C-1 of the first rhamnosyl residue (Rha); H-1 of the second glucosyl residue (Glc') and C-3 of Rha; H-1 of the fucosyl residue (Fuc) and C-4 of Rha; and H-1 of the second quinovosyl residue (Qui') and C-2 of Glc' (Fig. 1). Furthermore, key cross-peaks in the NOESY spectrum of 1 are detected between the signals of H-1 Glc and H-2 of Qui; H-1 of Rha and H-2 of Glc; H-1 of Glc' and H-3 of Rha; H-1 of Fuc and H-4 of Rha; and H-1 of Qui' and H-2 of Glc' (Fig. 1). Thus, 1 is composed of the sugar chain of β-d-quinovopyranosyl-(1 → 2)-O-β-d-glucopyranosyl-(1 → 3)-[O-β-d-fucopyranosyl-(1 → 4)]-O-α-l-rhamnopyranosyl-(1 → 2)-O-β-d-glucopyranosyl-(1 → 2)-β-D-quinovopyranose. In a previous study, we demonstrated that the chemical shift of the methylene carbon (β-carbon) at the β-position of the terminal methyl group of the aglycone moiety (Agl) in the resin glycoside, as observed in the 13C-NMR spectrum, served as a reliable indicator of the position of the hydroxyl group of the aglycone [13]. Specifically, the chemical shift of the β-carbon varied significantly depending on the number of methylene groups between the oxymethine and terminal methyl groups of Agl. The chemical shift at the β-carbon in pyridine-d5 was approximately δc 32.5 when there were four methylene groups, approximately δc 27.5 when there were three methylene groups, and approximately δc 37.5 when there were two methylene groups. Based on the chemical shift (δ 37.7) of the β-carbon of the terminal methyl group observed in the 13C-NMR spectrum, the 1H-1H COSY correlations between the signals of H2-2 (δ 2.73, 2.71) of Agl and H-3 (δ 4.42) of Agl, the HMBC correlation between the signals of the methoxy protons and C-1 (δ 172.9) of Agl; the aglycone of 1 was identified as methyl ipurolate. The absolute configuration of the ipurolic acid component in the crude resin glycoside fraction of I. lacunosa seeds is the 3S,11S, which was previously determined from the 1H-NMR spectrum of the (+)-α-methoxy-α-trifluoromethylphenylacetic acid (MTPA) derivative of its methyl ester [10]. In addition, glycosylation shifts [14, 15] at C-10 (–3.8 or –3.9 ppm), C-11 (+ 10.2 ppm), and C-12 (–3.0 ppm) of the methyl ipurolate moiety (Ipa) were detected compared with the 13C-NMR data for methyl ipurolate in the literature [16]. A correlation in the HMBC spectrum of 1 was observed between the signals of H-1 of Qui and C-11 of Ipa. These data suggest that 1 is composed of multifidinic acid C [17], which was previously reported to be the glycosidic acid component of the crude resin glycoside fraction from I. lacunosa seeds [10]. Furthermore, from the unassigned NMR signals corresponding to one primary methyl group, a β-d-quinovopyranosyl residue, an oxymethine carbon, a β-carbon (δ 37.4), and a carboxyl carbon, in conjunction with the molecular formula of 1, it was suggested that 1 was composed of quamoclinic acid B [18], a glycosidic acid component of the crude resin glycoside fraction of I. lacunosa seeds [10]. The absolute configuration of the aglycone in quamoclinic acid B (7-hydroxydecanoic acid) was previously determined through 1H-NMR analysis of the (+)-MTPA derivative of its methyl ester, which was obtained during our studies on the components of the crude resin glycoside fraction [10]. The presence of quamoclinic acid B as a component of 1 is supported by the following evidence. A correlation was observed in the HMBC spectrum of 1 between the signals of H-1 of the third quinovosyl residue (Qui'') and C-7 of the first 7S-hydroxydecanoyl residue [Hda; Agl of the first quamoclinic acid B residue]. The negative-ion ESI-TOF-MS/MS of [M–H]– ion showed a fragment ion peak at m/z 1281.5956 [M–(quamoclinic acid B unit)]–. Based on these findings, it can be inferred that 1 has a structure analogous to that of QM-11 (6) [19]. Compound 6 is composed of 1 mol each of 2R,3R-nilic acid and quamoclinic acid B, which are linked via ester bonds to the sugar moiety of the methyl ester of quamoclinic acid C [18]. Quamoclinic acid C is an epimer of multifidinic acid C, in which the Qui and Fuc are substituted with β-D-fucopyranosyl and β-D-quinovopyranosyl residues, respectively (Fig. 2). Comparison of the 1H-NMR spectra of 1 and multifidinic acid C methyl ester (7) [10] indicated remarkable downfield shifts (∆δ = δ1–δ7) of the signals corresponding to H-2 (∆δ = 1.10) of Rha and H-4 (∆δ = 1.65) of Fuc due to acylation. These data suggested that the ester linkages were located at C-2 of Rha and C-4 of Fuc. In addition, the HMBC spectrum of 1 showed key cross-peaks between the signals of H-2 of Rha and C-1 of Hda, and H-4 of Qui and C-1 of the niloyl residue (Nla) (Fig. 1). Accordingly, the structure of 1 was defined as methyl 3S,11S-ipurolate 11-O-β-d-quinovopyranosyl-(1 → 2)-O-β-d-glucopyranosyl-(1 → 3)-[O-(4-O-2R,3R-niloyl)-β-d-fucopyranosyl-(1 → 4)]-O-(2-O-7S-hydroxydecanoyl 7-O-β-d-quinovopyranoside)-α-l-rhamnopyranosyl-(1 → 2)-O-β-d-glucopyranosyl-(1 → 2)-β-d-quinovopyranoside, which is an isomer of 6, in which the glycosidic acid component of 6 was substituted by multifidinic acid C (Fig. 2). The NMR assignments of Ipa and Hda were performed with according to previously reported data [10, 20].

Key HMBC and NOESY correlations observed for 1–5 and key 1H-1H COSY correlations observed for 5 (600 MHz, in pyridine-d5)

Using ESI-TOF-MS, the molecular formula of 2 (lacunosin II) was determined to be C77H134O39 by HR-positive-ion and HR-negative-ion modes. The 1H- and 13C-NMR spectra of 2 were similar to those of 1, except for additional signals due to the presence of 1 mol of the 2-methylbutyryl residue (Mba) (Tables 1 and 2). The absolute configuration of the 2-methylbutyric acid component in the crude resin glycoside fraction derived from I. lacunosa seeds was determined to be S, following the same procedure as those used for nilic acid [10]. The NMR signals were assigned using the same 2D-NMR spectroscopy used for 1. A comparison of the 1H-NMR signals corresponding to the sugar moieties of 2 and 1 revealed a downfield shift (1.61 ppm) in the signal corresponding to H-4 of Qui' in 2. The other signals, including H-2 of Rha and H-4 of Fuc, which are indicative of acylation shifts, showed chemical shifts almost identical to those of 1. In addition, 2 exhibits significant correlations in the HMBC spectrum between the signals of H-1 of Qui and C-11 of Ipa; H-1 of Glc and C-2 of Qui; H-3 of Rha and C-1 of Glc'; H-4 of Rha and C-1 of Fuc; H-1 of Qui' and C-2 of Glc'; H-1 of Qui'' and C-7 of Hda; the methoxy protons and C-1 of Ipa; H-2 of Rha and C-1 of Hda; H-4 of Fuc and C-1 of Nla; and H-4 of Qui' and C-1 of Mba (Fig. 1). No HMBC correlations were observed to confirm the sugar linkage between Glc and Rha. However, significant correlations were identified in the NOESY spectrum of 2, including cross-peaks between the signals of H-1 of Qui and H-11 of Ipa; H-1 of Glc and H-2 of Qui; H-1 of Rha and H-2 of Glc; H-1 of Glc' and H-3 of Rha; H-1 of Fuc and H-4 of Rha; and H-1 of Qui' and H-2 of Glc' (Fig. 1). Moreover, although no HMBC correlation was detected between H-2 of Rha and C-1 of Hda, the observed acylation-induced shift in the signals suggested that Hda was esterified at C-2 of Rha. Consequently, 2 was concluded to be a homolog of 1 in which Mba was bonded to C-4 of Qui' (Fig. 2).

HR-positive-ion and HR-negative-ion ESI-TOF–MS data for 3 (lacunosin III) revealed that the molecular formula of 3 is C83H142O43. The 1H- and 13C-NMR signals of 3, assigned using the same 2D-NMR spectroscopic techniques as 1, were similar to those of 2, especially those of Agl, which were nearly identical. However, additional signals corresponding to 1 mol each of the tigloyl residue and β-quinovopyranosyl residue with 4C1 conformation were observed, along with the disappearance of signals corresponding to Mba (Tables 1 and 2). Tiglic acid was identified as an organic acid component in the crude resin glycoside fraction of I. lacunosa seeds; however, its geometrical isomer, angelic acid, was not detected. By comparing the 13C-NMR signals of the sugar moieties between 3 and 2, downfield shifts (∆δ = δ3–δ2) were observed at the signals attributed to C-3 (∆δ = 7.8) and C-5 (∆δ = 0.9) of Qui', and an upfield shift was observed at that attributed to C-4 (∆δ = − 1.9) of Qui' [14, 15]. The other signals are nearly identical. In addition, acylation shifts were observed for the same signals as those observed for 2. In the HMBC spectrum of 3, important cross-peaks were detected between the signals of H-1 of Qui and C-11 of Ipa; H-1 of Glc and C-2 of Qui; H-4 of Rha and C-1 of Fuc; H-1 of Qui' and C-2 of Glc'; H-1 of the fourth quinovosyl residue (Qui''') and C-3 of Qui'; H-4 of Fuc and C-1 of Nla; H-4 of Qui' and C-1 of the tigloyl residue (Tig); H-1 of Qui'' and C-7 of Hda; and methoxy protons and C-1 of Ipa (Fig. 1). No valid HMBC correlation was detected between H-2 of Rha and C-1 of Hda, nor was any correlation detected to confirm the sugar linkages at the anomeric positions of Rha and Glc'. However, the ester linkage of Hda was determined based on the observed acylation shifts, leading to the conclusion that the carboxyl group of Hda is bound to C-2 of Rha. Furthermore, the sugar linkages at the anomeric positions of Rha and Glc' were elucidated by following the NOESY correlations, which indicated that Rha is attached to C-2 of Glc and Glc' is attached to C-3 of Rha. These correlations included the following cross-peaks: between H-1 of Qui and H-11 of Ipa; H-1 of Glc and H-2 of Qui; H-1 of Rha and H-2 of Glc; H-1 of Glc' and H-3 of Rha; H-1 of Fuc and H-4 of Rha; H-1 of Qui' and H-2 of Glc'; and H-1 of Qui''' and H-3 of Qui' (Fig. 1). Thus, 3 was identified as methyl 3S,11S-ipurolate 11-O-β-d-quinovopyranosyl-(1 → 3)-O-(4-O-tigloyl)-β-d-quinovopyranosyl-(1 → 2)-O-β-d-glucopyranosyl-(1 → 3)-[-O-(4-O-2R,3R-niloyl)-β-d-fucopyranosyl-(1 → 4)]-O-(2-O-7S-hydroxydecanoyl 7-O-β-d-quinovopyranoside)-α-l-rhamnopyranosyl-(1 → 2)-O-β-d-glucopyranosyl-(1 → 2)-β-d-quinovopyranosyide (Fig. 2). Compound 3 contained a new glycosidic acid, an isomer of lacunosinic acid H [10], in which the terminal rhamnosyl residue was substituted with Qui'''.

The molecular formula of 4 (lacunosin IV) was determined to be C94H164O47 using the HR-positive-ion and HR-negative-ion modes in ESI-TOF–MS. The 1H- and 13C-NMR spectral data of 4, assigned using 2D-NMR spectroscopy, were similar to those of 2. In particular, the data for Ipa were nearly identical. However, additional signals attributed to 1 mol each of α-rhamnopyranosyl and quamoclinic acid B residues are observed, along with the absence of signals corresponding to 1 mol of niloyl residue in 4 (Tables 1 and 2). A comparison of the 13C-NMR signals of 4 and 2 revealed a downfield shift in the signal attributed to C-3 (4.7 ppm) of Fuc in 4. In contrast, the signals corresponding to the other sugar moieties in 2 were nearly identical to those observed in 4. The HMBC spectrum of 4 reveals significant correlations, including H-1 of Glc with C-2 of Qui; H-2 of Glc with C-1 of Rha; H-3 of Rha with C-1 of Glc'; H-4 of Rha with C-1 of Fuc; H-1 of Qui' with C-2 of Glc'; H-3 of Fuc with C-1 of Rha', C-1 of Qui'', or C-1 of the fifth quinovosyl residue (Qui''''); H-1 of Qui'' with C-7 of Hda; and H-1 of Qui'''' with C-7 of the second 7-hydroxydecanoyl residue (Hda'; Agl of the second quamoclinic acid B residue); and methoxy protons with C-1 of Ipa (Fig. 1). Although no HMBC correlations suitable for determining the sugar linkages at the anomeric positions of Qui and Rha' were observed, the NOESY spectrum of 4 revealed notable cross-peaks, including those between the signals of H-1 of Qui and H-11 of Ipa; H-1 of Glc and H-2 of Qui; H-1 of Rha and H-2 of Glc; and H-1 of Fuc and H-4 of Rha. In addition, a prominent cross-peak was observed between H-1 of the second rhamnosyl residue (Rha') and H-3 of Fuc (Fig. 1). These data indicate that 4 contains 1 mol of 2-methylbutyric acid, 2 mol of quamoclinic acid B, and 1 mol of new glycosidic acid methyl ester, identified as methyl 3S,11S-ipurolate 11-O-β-d-quinovopyranosyl-(1 → 2)-O-β-d-glucopyranosyl-(1 → 3)-[O-α-l-rhamnopyranosyl-(1 → 3)-O-β-d-fucopyranosyl-(1 → 4)]-O-α-l-rhamnopyranosyl-(1 → 2)-O-β-d-glucopyranosyl-(1 → 2)-β-d-quinovopyranoside. This new glycosidic acid methyl ester was found to be a homolog of the lacunosinic acid D methyl ester (8) [10], in which the aglycone methyl 3S,11S-dihydroxyhexadecanoate was substituted with methyl 3S,11S-ipurolate. Comparing the 1H-NMR spectral data corresponding to sugar moieties of 4 and 8 [10], downfield shifts are observed for the signals corresponding to H-2 (1.06 ppm) of Rha, H-4 (1.54 ppm) of Fuc, and H-4 (1.67 ppm) of Qui' in 4. In addition, key HMBC correlations were observed between the signals of H-4 of Fuc and C-1 of Hda'; and H-4 of Qui' and C-1 of Mba (Fig. 1). These data indicate that Hda, Hda', and Mba are attached to C-2 of Rha, C-4 of Fuc, and C-4 of Qui', respectively. Consequently, 4 was concluded to be methyl 3S,11S-ipurolate 11-O-(4-O-2S-methylbutyryl)-β-d-quinovopyranosyl-(1 → 2)-O-β-d-glucopyranosyl-(1 → 3)-[O-α-l-rhamnopyranosyl-(1 → 3)-O-(4-O-7S-hydroxydecanoyl 7-O-β-d-quinovopyranoside)-β-d-fucopyranosyl-(1 → 4)]-O-(2-O-7S-hydroxydecanoyl 7-O-β-d-quinovopyranoside)-α-l-rhamnopyranosyl-(1 → 2)-O-β-d-glucopyranosyl-(1 → 2)-β-d-quinovopyranoside (Fig. 2).

The HR-negative-ion ESI-TOF-MS and NMR data revealed that the molecular formula of 5 (lacunosinic acid I) was C34H62O18. The 1H-NMR spectrum of 5 showed signals corresponding to three anomeric protons, two secondary methyl groups, two non-equivalent methylene protons adjacent to the carbonyl group, and one primary methyl group. The 13C-NMR spectrum shows signals attributed to one carboxyl carbon and three anomeric carbons, along with 30 aliphatic carbons. NMR assignments were performed using the same 2D-NMR spectroscopies as those employed for 1 (Table 3). These data suggest that 5 consists of 1 mol of α-l-rhamnopyranose with 1C4 conformation and 2 mol of β-d-glucopyranose with 4C1 conformation as monosaccharides, and dihydroxyhexadecanoic acid as the aglycone. The planar structure of the aglycone was determined from the correlations observed in the 2D-NMRspectra of 5. In the 1H-1H COSY spectrum of 5, significant correlations were observed sequentially from the H2-2 signals of Agl to H2-3 of Agl/H-4 (δ 4.94) of Agl (Fig. 1). In addition, HMBC correlations were observed between the signals of H2-2 of Agl and C-4 (δ 81.0) of Agl, indicating the presence of a hydroxyl group at C-4 (Fig. 1). The chemical shift of the signal attributed to the β-carbon of Agl in 5 was δ 32.4, suggesting that the remaining hydroxyl group was located at C-11. Thus, the aglycone of 5 is identified as 4,11-dihydroxyhexadecanoic acid, which is a novel compound. Comparison of the 13C-NMR data of 5 with those of the corresponding methyl pyranosides [21] revealed glycosylation shifts [14, 15] at the C-2 (+ 4.9 ppm) of the third glucosyl residue (Glc'') and C-2 (+ 4.5 ppm) of Glc. In the HMBC spectrum of 5, key cross-peaks were detected between the signals of H-1 of Glc'' and C-11 of Agl; H-1 of Glc and C-2 of Glc''; and H-1 of Rha and C-2 of Glc (Fig. 1). In addition, the NOESY spectrum provided important correlations between the signals of H-1 of Glc and H-2 of Glc''; and H-1 of Rha and H-2 of Glc (Fig. 1). The absolute configurations of the monosaccharide components of the crude resin glycoside fraction obtained from I. lacunosa seeds have already been determined [10]. However, that of the aglycone in 5 could not be determined because of the yield of 5. Therefore, 5 was identified as 4,11-dihydroxyhexadecanoic acid 11-O-α-l-rhamnopyranosyl-(1 → 2)-O-β-d-glucopyranosyl-(1 → 2)-O-β-d-glucopyranoside.