The MeOH extracts of the fruits of G. xanthochymus and G. subelliptica were subjected to several purification steps. Six new polyprenylated acylphloroglucinols, xanthochymusones N and O (1 and 2), (–)-garciyunnanin L (3), garsubelones C–E (4–6), and two known analogues were obtained (Fig. 1). Known compounds were identified as garsubelone B (7) [17] and lupulone B (8) [18] by comparison of the physical and chemical data to the reported literature. Compound 3 were proven to be enantiomer of ( +)-garciyunnanin L [19] via NMR, optical rotation, and ECD data.

Xanthochymusone N (1) was isolated as a yellow gum. Its molecular formula, C38H52O8, was established by analysis of 13C NMR (Table 1) and HRESIMS data (m/z 635.3586, [M – H]−, cacld for 635.3584). The UV spectrum revealed the presence of conjugated groups with maximum absorptions at 232, 293, and 327 nm. The FTIR spectrum displayed absorption bands due to hydroxy (3400 cm−1), carbonyl (1724 cm−1), and olefinic (1621 and 1467 cm−1) functionalities. The 1H NMR spectrum (Table 1) exhibited signals of two singlet aromatic protons (δH 7.40 and 6.78, s), olefinic protons (δH 4.92 and 4.00), and nine singlet methyl groups (δH 0.52 ~ 1.65). The 13C and DEPT-NMR data revealed a total of 38 carbon resonances (Table 1), including a nonconjugated carbonyl (δC 209.2), a group of enol carbons at δC 122.4 and 164.7, six quaternary carbons (δC 75.8, 73.2, 72.8, 60.9, 54.1, and 49.5), two methines (δC 43.4 and 41.6). The above signals, in combination with a number of PPAPs that have been isolated from this plant, [13, 20,21,22,23,24,25,26] indicated that compound 1 was a PPAP derivative. Comparison of its 1D NMR data with those of garcinialone [27], a type B PPAP with 6/6/6/6/6 fused ring system, indicated that they were structurally similar. The core carbon system of 1 was confirmed by the correlations from δH 0.99 and 0.92 (Me-37 and Me-38) to δC 60.9 (C-1), 43.4 (C-7), and 49.5 (C-8), from δH 2.82 and 2.37 (H2-10) to C-1 and δC 164.7 (C-2) and 209.2 (C-9), from a hydroxyl proton at δH 5.44 (4-OH) to δC 122.4 (C-3), 75.8 (C-4), and 54.1 (C-5), from δH 1.97 and 1.36 (H2-22) to C-5, C-9, and δC 38.5 (C-6), and from δH 1.59 (Me-26) to δC 41.6 (C-23), 73.8 (C-24), and 51.6 (C-25) in the HMBC spectrum, in combination with the correlations of δH 1.04 (H-7)/2.32 and 1.44 (H2-6), and δH 2.33 (H-23)/H2-22 in the 1H–1H COSY spectrum (Fig. 2). An oxidized isoprenyl group linked to C-23 was further determined via HBMC correlations of δH 1.37 and 1.39 (Me-30 and Me-31) with δC 72.8 (C-29) and 40.1 (C-28), coupled with the proton spin system of H-23/δH 1.83 and 1.34 (H2-27)/δH 2.07 and 1.73 (H2-28) in the 1H–1H COSY spectrum. Other parts of the planer structure of 1 were confirmed to be identical to those of garcinialone by detailed analysis of the 2D NMR spectroscopic data (Fig. 2). The relative configuration of 1 could be defined by using a combination of 1H–1H coupling constants, conformational analysis, NOE experiments, as well as biosynthetic consideration. As described in the literature, these rare PPAPs could be biosynthetically generated from type B PPAPs with a bicyclo[3.3.1]nonane-2,4,9-trione core carrying a lavandulyl substituent by Alder-ene reaction [19, 27]. As the cyclization took place from the less hindered α-phase, the resulting β-axial hydroxyl group (4-OH) was formed. In the 1H spectrum of 1 (measured in CDCl3), the 3 J coupling constants of both H-6a (δH 2.32, t, J = 13.6 Hz) and H-22b (δH 1.36, t, J = 13.6 Hz) were 13.6 Hz. So, H-6a, H-7, H-22b, and H-23 were axial while the C-7 prenyl group and C-23 substituent were equatorial. Then, the NOE contacts of 4-OH with H-6ax, H-22ax, and Me-26 indicated that the relative configuration of C-7 and C-23, and C-24 (Fig. 3). The NOE correlation of Me-26 (δH 1.33) with H-27a (δH 1.85) measured in CD3OD also confirmed the relative configuration of C-23 and C-24. The above deduction suggested that B ring adopted boat conformation while C ring adopted chair conformations.

Key 1H–1H COSY and HMBC correlations of compounds 1 and 4–6

Simplified and configuration optimized molecular model of 1. Black arrows, coupling constants; blue arrows, NOE correlations

It is worth noting that the chemical shift of C-7 (δC 43.4) and the difference in chemical shifts of the two H-6 protons (0.88 ppm) and the configurational assignment of C-7 substituent (endo) is against the Grossman-Jacobs rule of determination of the C-7 configuration [3, 15]. The fact may be ascribable to the conformational change of B ring (from chair to boat) after the Alder-ene reaction (Fig. 3). So, the Grossman-Jacobs rule is no longer applicable to determination of the C-7 configuration of this kind of PPAPs.

Xanthochymusone O (2) was assigned the molecular formula C38H48O6 by analysis of its 13C NMR and HREIMS data. The 1H and 13C NMR data of 2 (Table 1) resembled those of garciyunnanin L [19]. The structure of 2 was shown to be a double-bond regioisomer of garciyunnanin L (Δ29,30 in 2 and Δ28,29 in garciyunnanin L). The 2D NMR data showed that the other structural features of 2 were identical to those of garciyunnanin L (Fig. 2).

The 1H and 13C NMR spectra of 3 are identical to those of ( +)-garciyunnanin L [19]. Nevertheless, the opposite optical rotations of 3 and ( +)-garciyunnanin L ([α]D = –21 and + 28.5, respectively) and opposite experimental CD curves indicate that these compounds are enantiomeric. Considering that compounds 1–3 are co-isolates of G. xanthochymus and their absolute configurations should be consistent with those of xanthochymusones reported previously [14].

Garsubelone C (4) has NMR spectra nearly identical to garsubelone B (7) [17], except the isopropyl group in garsubelone B was replaced by a sec-butyl group (δC 49.5, C-11; 17.8, C-12; 26.7, C-13; and 11.7, C-13a) (Table 2), so we assign to garsubelone C the structure 4. The HMBC correlation of Me-12 (δH 1.02, d, J = 6.5 Hz) with C-10 (δC 208.6) and the correlations of Me-12/δH 1.78 (H-11)/1.95 and 1.32 (H2-13)/δH 0.81 (Me-13a) in the 1H–1H COSY spectrum confirmed the assignment (Fig. 2). Since the absolute configuration of garsubelone B was undoubtfully determined as 1R,5S,7R on the basis of X-ray crystallographic data and garsubelones B and C are isolated from the same plant (G. subelliptica) [17], the absolute configurations of C-1, C-5, and C-7 for compound 4 should be the same as those of garsubelone B.

The molecular formula of garsubelone D (5) was determined as C30H42O4 by analysis of its 13C NMR (Table 3) and HRESIMS data (m/z 489.2978, [M + Na]+), 68 mass units more than that of lupulone B (8) [18]. On the basis of analysis of its 1D (Table 3) and 2D NMR data, compound 5 was assigned to possess the same backbone as lupulone B. The structural novelty of 5 involved the presence of a geranyl side chain rather than a prenyl group. The geranyl group linked to C-5 was confirmed by the 1H–1H COSY correlations of H2-21/H-22 and H2-25/H2-26/H-27, in combination with the HMBC correlations of Me-24 (δH 1.75) with C-22 (δC 123.0) and C-25 (δC 40.9) and of both Me-29 (δH 1.66) and Me-30 (δH 1.56) with C-27 (δC 125.3) (Fig. 2). The NOE contact of Me-24/H2-21 in the NOESY spectrum defined the E configuration of the corresponding ene.

Garsubelone E (6) was assigned the molecular formula C30H42O4 by analysis of its 13C NMR and HREIMS data. The 1H and 13C NMR data of 6 (Table 3) resembled those of xanthochymusone K [16]. Instead of a hydroxymethyl in xanthochymusone K, a methyl carbon at δC 18.3 (C-25) appeared in 6, suggesting that 6 possessed a complete prenyl group. This suggestion was further supported by the correlations of Me-24 (δH 1.58) and Me-25 (δH 1.56) with δC 119.3 (C-22) and 136.0 (C-23), as well as 1H–1H COSY correlations of δH 2.68 and 2.51 (H2-21)/δH 4.79 (H-22). The 2D NMR data showed that the other structural features 6 were identical to those of xanthochymusone K (Fig. 2).

The previous assignments of some PAPs were controversial and have been proved to be incorrect in this study. Garsubelones B and C have same planar structures and relative configuration to those of hyperselancins A and B (Scheme 1) [28], respectively, however, their reported 1H and 13C NMR spectroscopic data are different (Table 2). The chemical shift of C-7 (δC 47.7) and the difference in chemical shifts of the two H-6 protons (0.05 ppm) of both hyperselancins A and B indicates that the C-7 substituents of the two compounds should be endo according to the Grossman-Jacobs rule [3, 15]. The original authors also referred to the Grossman-Jacobs rule, but unfortunately they seem to have confused the concept of endo/exo and α/β [28]. Hence, we revise the relative configuration of C-7 of hyperselancins A and B (Scheme 1).

Structural revision of hyperselancins A and B and garcinielliptone F

Garsubelone B were found to share identical 1H and 13C NMR data to garcinielliptone F (Table 4) [29, 30], a PPAP resulting from hydrolysis of the enol ether of garsubelone B, suggesting structural reassignment of garcinielliptone F (Scheme 1). The original authors may have misinterpreted the mass spectrometry data. Interestingly, we detected signals corresponding to another set of non-dominant compound in the NMR spectra of garsubelone B/C and structurally similar compounds: hyperselancins A and B [28], androforin A [31], and hypercohin I [32]. Analysis of the 2D NMR data of garsubelone B/C indicated that these signals correspond to positional isomers of the pyran ring (C-2–O–C-16) in this class of compounds (Scheme 2). Based on the HPLC analysis results of garsubelone B (Fig. S1), it can be inferred that this class of compounds exhibits slow tautomerism, with the dominant isomer having the pyran ring positioned at C-4–O–C-16 (Scheme 2).

Transformation between 7 and 7a

Garcinielliptone G was reported along with garcinielliptone F and was later found to induce apoptosis in acute leukemia cells [29, 30, 33]. However, the structure of garcinielliptone G was also misassigned due to the misuse of MS data. In 2021, hyperscabin D with the same skeleton was reported [34], and its 1H and 13C NMR data (particularly the signals near the pyran ring) showed excellent agreement with those of garcinielliptone G (Table 5). The IR spectrum of hyperscabin D did not show absorption peak corresponding to hydroxyl groups, nor were any signals of active hydrogen observed in the 1H spectrum [34]. The above evidence supports the correct assignment of hyperscabin D and our structural revision of garcinielliptone G (Scheme 3).

Structural revision of garcinielliptone G

In 2023, Quan et al. reported two polyprenylated acylphloroglucinols, garxanthochins A and B (Scheme 4), from G. xanthochymus [35]. However, the 1H and 13C NMR data of garxanthochin A are identical to those of xanthochymusone K reported by our group (Table 6) [16]. The chemical shifts of C-25 (δC 61.8) and C-28 (δC 81.5) does not support the linkage of C-4–O–C-25 but support the linkage of C-4–O–C-28. In the 1H–1H COSY spectrum, a broad singlet at δH 2.22 (hydroxyl hydrogen) shows correlation with δH 3.79 (H-25b) [16], which further confirms our structural reassignment of garxanthochin A. Since garxanthochin B has been assigned the same carbon skeleton as garxanthochin A and their NMR data are similar except for the C-1 acyl substituent, we also revise the structure of garxanthochin B (Scheme 4).

Structural revision of garxanthochins A and B

Since we revised the relative configuration of garcicowins C and D in 2022 [13], we have further noticed that their derivative, 13,14-didehydroxygarcicowin C [36], also exhibits the same configurational error. As shown in Table 7, the 1H and 13C NMR data of the core carbon skeleton of 13,14-didehydroxygarcicowin C shows excellent agreement (particularly around the C-30 and C-34 stereocenters) with those of xanthochymusone H, whose structure was confirmed by single-crystal X-ray diffraction data [13]. So we firstly revise the relative configuration of C-23 and C-27 of 13,14-didehydroxygarcicowin C. Furthermore, the opposite optical rotations of 13,14-didehydroxygarcicowin C and xanthochymusone H ([α]D = –68.6 and + 60, respectively) and almost opposite experimental CD data indicate that their absolute configurations are opposite [λmax (Δε) 223 (+ 5.2), 267 (− 8.5), 311 (+ 2.1) nm for 13,14-didehydroxygarcicowin C; λmax (Δε) 224 (− 8.8), 258 (+ 7.2), 302 (− 0.25) nm for xanthochymusone H] [13, 36]. Ultimately, we reassign the absolute configuration of 13,14-didehydroxygarcicowin C as shown (Scheme 5).

Structural revision of 13,14-didehydroxygarcicowin C

The antiproliferative activities of all the isolates against two human hepatocellular carcinoma cell lines Huh-7 and Hep G2 through CCK-8 assay were preliminary evaluated [37]. Sorafenib was chosen as the positive control (IC50 9.7 and 16.6 µM, respectively). Compound 1 exhibited moderate cytotoxic activities against Hep G2 cells with IC50 value 7.3 µM, while other compounds did not show obvious activity (IC50 > 50 µM).

In conclusion, we have isolated six new polyprenylated acylphloroglucinols (1–6) and two known analogues from the fruits of G. xanthochymus and G. subelliptica. Compound 1 exhibits moderate antiproliferative activity against HepG2 cell lines. Furthermore, the structures of seven polyprenylated acylphloroglucinols (hyperselancins A and B, garcinielliptones F and G, garxanthochins A and B, and 13,14-didehydroxygarcicowin C) are revised by analysis of NMR and other spectroscopic data. Our structural confirmation work has cleared the obstacles for subsequent research on these natural products.