Dried red pistils containing stigmas (called “saffron”) from saffron plants (Crocus sativus L.) that had been cultivated in Spain were purchased at a spice-special store in Tokyo, Japan. As for freesia (Freesia x hybrida), we used cultivars “Ishikawa f2 go” (“f2”), “Kayak,” and “Aradin.” Cultivars “f2” and “Kayak” were grown in a greenhouse, and their flowers were harvested in middle and late March, respectively. Cultivar “Aladin” was purchased at a floral shop in Ibaragi, Japan.

Microscopy pictures of the “f2” petal were recorded by BH2 (Olympus) after cutting it in 3% agar with MicroSlicer ZERO-1 (Dosaka EM, Kyoto, Japan).

Dichloromethane (CH₂Cl₂), methanol (MeOH), acetonitrile (CH₃CN), ethyl acetate (EtOAc), and chloroform of analytical grade were purchased from Wako Pure Chemicals Industries, Ltd. (Osaka, Japan) or Kanto Chemical Co. Inc. (Tokyo, Japan). The other chemicals were purchased from Sigma-Aldrich (St. Louis, MO, United States).

In order to detect volatile components of saffron and freesia “f2” flowers, we carried out a gas chromatography-mass spectrometry (GC-MS) analysis. Each 0.1 g of their dried materials was ground with an addition of 0.3 ml of water (H₂O) and 0.3 ml of MeOH, subsequently extracted with 0.6 ml of chloroform (by mixing and centrifugation), and the chloroform extraction was repeated by adding 0.5 ml of H₂O, and finally, each chloroform phase was concentrated to threefold by decompression. The prepared samples were subjected to GC-MS using Shimadzu GCMS-QP5050 (Shimadzu, Kyoto, Japan), as described (15).

Nuclear magnetic resonance (NMR) spectra (¹H and ¹³C NMR, ¹H-¹H DQF COSY, HMQC, HMBC, and NOESY) were measured using an AVANCE 400 MHz NMR spectrometer (Bruker, Rheinstettern, Germany) utilizing standard programs in TopSpin1.3. Chemical shifts were referenced to solvent signals. HR-ESI-MS (+) was measured using a JMS-T100LP mass spectrometer (JEOL, Tokyo, Japan) and accumulated mass calibration was performed using reserpine [C₃₃H₄₁N₂O₉, m/z 609.2812044 (M+H)⁺].

As a boiled cooking model, saffron (dried; 0.08 g) or freesia petals [1.8 g (fresh weight)] were added to boiled water (50 ml) and boiled for 20 min. As a grilled cooking model, saffron (dried; 0.08 g) or freesia petals [1.8 g (fresh weight)] were added to olive oil (7.5 ml) and grilled at 180°C for 5 min.

A paella-type cooking was adopted as an example of the boiled dishes with saffron, as follows: Saffron (0.08 g)-including water (300 ml), polished rice (500 g), and olive oil (30 ml) were put together on a frying pan, and the materials were boiled under medium heat for 20 min. A pilaf-type cooking was adopted as an example of the grilled dishes with saffron, as follows: Olive oil (30 ml), polished rice (500 g), and saffron (0.08 g) were grilled on a frying pan under medium heat for 10 min. Then, water (300 ml) was added to the frying pan and the materials were boiled for 15 min.

Rice was cooked in a rice cooker, using polished rice (Japonica; 75 g), water (90 ml), and “f2” flowers (5 g fresh weight) (Figure 2E). After cooking, the flower debris was eliminated (Figure 2F right). Bread was baked as follows: Wheat powder (100 g), water (60 ml), NaCl (1.5 g), budding yeast (powder; 1 g), and dried powder from “f2” flowers (10 g fresh weight) by microwave were mixed and left for 2 h 20 min. The material was incubated for 1 h at 37°C, and baked for 21 min at 220°C (Figure 2D right).

Each sample obtained by cooking had an addition of 100 μL of 9,10-dimethylanthracene solution (10 mg/1 ml CH₂Cl₂) as an internal standard to confirm the recovery rate of apocarotenoids. Each of the cooked materials as the cooking models was extracted by CH₂Cl₂-MeOH (1:1) (V:V) (20 ml), MeOH (20 ml), and 50% (V/V) MeOH (20 ml) in a stepwise manner by stirring for 30 min at 20°C. All the extracts (total 60 ml) were combined and concentrated to dryness in vacuo.

Each of the actually cooked materials [the paella (boiled) and pilaf (grilled) types] was freeze-dried and powdered by a mill. The powder was extracted by CH₂Cl₂-MeOH (1:1) (1 L), MeOH (1 L), and 50% (V/V) MeOH (1 L) in a stepwise manner by stirring for 30 min at 20°C, and all the extracts (total 3 L) were combined and concentrated to dryness in vacuo.

Each extract was dissolved into MeOH (10 ml), and 2.5 μL (saffron) or 20μL (freesia) of the supernatant was analyzed by high-performance liquid chromatography-diode array detector (HPLC-DAD) [column, PEGASIL ODS SP100 (4.6 × 150 mm, Senshu Scientific Co. Ltd., Tokyo, Japan); solvents: A 5% (V/V) CH₃CN containing 20 mM phospholic acid, B 95% (V/V) CH₃CN containing 20 mM phospholic acid. 0 → 5 min 100% A, 5 → 18 min 100% A → 100% B linear gradient, 18–30 min 100% B, flow rate, 3.0 ml/min; detection, DAD; monitored at 250–600 nm].

Saffron (0.22 g) was extracted for 30 min at 20°C by stirring with CH₂Cl₂-MeOH (1:1) (V:V) (100 ml), MeOH (100 ml), and 50% (V/V) MeOH (100 ml) in a stepwise manner. CH₂Cl₂-MeOH (1:1) extract was concentrated to dryness to give a yellow oil, and the yellow oil was applied on a diol column chromatography (10 × 150 mm, CHROMATOREX DIOL MB 100-40/75, Fuji Silysia Chemical Ltd., Aichi, Japan) with n-hexane-CH₂Cl₂ (5:1) (V:V) and developed with the same solvent. The elution fractions containing yellow pigment were combined and concentrated to dryness (4.7 mg), and further purified by ODS preparative HPLC [column, Develosil C30-UG-5 (10 × 250 mm, Nomura Chemical Co. Ltd., Aichi, Japan); solvent, CH₂Cl₂-CH₃CN (95:5) (V:V); flow rate, 3.0 ml/min; detection, DAD; monitored at 250–600 nm]. The peak eluted at 9.5 min was collected to afford pure 4 (0.4 mg).

MeOH and 50% MeOH extracts were combined and concentrated to dryness (38.2 mg), and 1–3 in the extract were purified by ODS preparative HPLC [column, ADME ODS (10 × 250 mm, Osaka Soda Co. Ltd., Osaka, Japan); solvent, 50% (V/V) CH₃CN containing 0.1% (V/V) trifluoro acetic acid (TFA); flow rate, 3.0 ml/min; detection, DAD; monitored at 250–600 nm]. In this HPLC, the mixture of 1 and 2 was eluted at 3.0 min, and 3 was eluted at 5.0 min. The peak of 3 eluted at 5.0 min was collected and concentrated in a small volume to remove CH₃CN and finally freeze-dried to afford pure 3 (7.1 mg). The peak eluted at 3.0 min was also collected and freeze-dried to give yellow oil containing 1 and 2. This yellow oil containing 1 and 2 was further purified by ODS preparative HPLC [column, PEGASIL ODS SP100 (20 × 250 mm); solvent, 30% (V/V) CH₃CN; flow rate, 8.0 ml/min; detection, DAD; monitored at 250–600 nm]. In this HPLC, 1 and 2 were eluted at 8.0 min (12.8 mg) and 13.0 min (11.4 mg), respectively, as pure compounds.

Saffron (0.22 g) in olive oil (10 ml) was grilled for 1 min in a frying pan with medium heat and extracted with CH₂Cl₂-MeOH (1:1) (V:V) (100 ml) twice. The extract (200 ml) was concentrated to dryness, added n-hexane (30 ml), and mixed well. The precipitate containing apocarotenoids was purified by ODS preparative ODS HPLC [column, Develosil C30-UG-5 (20 × 250 mm); solvent, 85% (V/V) CH₃CN containing 0.1% (V/V) TFA; flow rate, 8.0 ml/min; detection, DAD; monitored at 250-600 nm]. In this HPLC, 5 and 6 were eluted at 12.0 min (3.2 mg) and 17.0 min (2.1 mg), respectively, as pure compounds.

Freesia (Freesia x hybrida) “f2” flowers (179.3 g) were freeze-dried and powdered by mill (18.6 g). This powder was extracted for 30 min by stirring with CH₂Cl₂-MeOH (1:1) (V:V) (1 L), MeOH (1 L), 80% (V/V) MeOH (1 L), and 50% (V/V) MeOH (1 L) in a stepwise manner.

The CH₂Cl₂-MeOH (1:1) and MeOH extracts were combined and concentrated to dryness in vacuo (14.50 g). This extract was dissolved in H₂O (300 ml) and partitioned with EtOAc (300 ml) twice, and a water layer containing yellow pigment was collected. The water layer was concentrated to 200 ml to remove EtOAc, 1 M HCl (10 ml) was added to adjust at pH3, and adsorbed to HP-20 column chromatography (50 × 150 mm, Mitsubishi Chemical Corporation, Tokyo, Japan). The HP-20 column was washed with H₂O (750 ml) and 50% (V/V) MeOH (750 ml), and the yellow compound was eluted with MeOH (750 ml). The MeOH eluate was concentrated to dryness to obtain yellow oil (1.11 g). The yellow oil was mixed with 15 ml of CH₂Cl₂, and the insoluble yellow sediment was collected (repeated twice). The obtained red sediment was dried (0.91 g), dissolved in 3 ml of 50% MeOH, and further purified using preparative ODS HPLC (column, Develosil C30-UG-5 (10 × 250 mm); solvent, 30% (V/V) CH₃CN + 0.1% (V/V) TFA; flow rate, 3.0 ml/min; detection, DAD; monitored at 200–600 nm). The yellow compound was eluted at 17.8 min as pure compound (7) (27.1 mg).

The 80% MeOH (including compound 7 also) and 50% MeOH extracts were combined and concentrated in a small volume (300 ml) to remove MeOH and partitioned with EtOAc (300 ml) twice, and the water layer containing yellow pigment was collected. The water layer was concentrated to 200 ml to remove EtOAc, and adsorbed to HP-20 column chromatography (50 × 150 mm). The HP-20 column was washed with H₂O (750 ml) and the yellow compound was eluted with 50% (V/V) MeOH (750 ml). The 50% MeOH eluate was concentrated to dryness to obtain yellow oil (0.75 g). The yellow oil was then purified using preparative ODS HPLC [column, Develosil C30-UG-5 (10 × 250 mm); solvent, 20% (V/V) CH₃CN + 0.1% (V/V) TFA; flow rate, 3.0 ml/min; detection, DAD; monitored at 200–600 nm]. The yellow compound was eluted at 15.3 min as pure compound (8) (14.0 mg).

HR-ESI-MS (+) m/z 837.31834 (C₃₈H₅₄O₁₉Na calcd. for 837.31570, Δ 3.16 ppm). UV-Vis λmax (ε) in MeOH 254 (9200), 315 (6800), 407 (43000), 428 (62000), 452 (56000). ¹H NMR (CD₃OD) δ: 1.96 (3H, s, H-19), 2.00 (9H, s, H-20, H-19', and H-20'), 3.15 (1H, m, H-5”'), 3.17 (1H, m, H-2”'), 3.21 (1H, m, H-2””), 3.24 (2H, H-4”” and H-5””), 3.36 (2H, H-3”' and H-3””), 3.37 (1H, m, H-4”'), 3.54 (1H, m, H-4”), 3.57 (1H, m, H-5”), 3.65 (2H, H-6”'b and H-6””b), 3.73 (1H, m, H-2”), 3.78 (1H, m, H-3”), 3.80 (1H, m, H-6”b), 3.83 (2H, H-6”'a and H-6””a), 4.15 (1H, m, H-6”a), 4.32 (1H, d, J = 7.6 Hz, H-1””), 4.59 (1H, d, J = 7.8 Hz, H-1”'), 5.65 (1H, d, J = 7.8 Hz, H-1”), 6.47 (3H, H-14, H-9', and H-14'), 6.62 (1H, m, H-11'), 6.64 (1H, m, H-11), 6.70 (1H, m, H-8'), 6.78 (1H, m, H-12), 6.80 (2H, H-15 and H-15'), 7.29 (1H, d, J = 9.1 Hz, H-10'), 7.42 (1H, d, J = 11.9 Hz, H-10). ¹³C NMR (CD₃OD) δ: 12.7 (C-19), 12.7 (C-19'), 12.9 (C-20), 12.9 (C-20'), 62.2 (C-6”'), 62.6 (C-6””), 69.4 (C-6”), 70.6 (C-4”), 71.0 (C-4”'), 71.5 (C-4””), 75.0 (C-2””), 76.0 (C-2”'), 77.6 (C-3”), 77.6 (C-3””), 77.7 (C-5”), 77.9 (C-5”'), 77.9 (C-5””), 78.0 (C-3”'), 82.4 (C-2”), 94.6 (C-1”), 104.5 (C-1””), 105.6 (C-1”'), 124.6 (C-11), 125.0 (C-11'), 126.3 (C-9'), 127.5 (C-14'), 127.7 (C-9), 132.6 (C-12'), 133.1 (C-12), 136.7 (C-14), 137.9 (C-13), 138.2 (C-13'), 140.4 (C-10'), 142.0 (C-10), 145.0 (C-15'), 146.4 (C-15), 168.1 (C-8). 171.9 (C-8').

Methylene Blue (80 μL, 25 μM in ethanol) and linoleic acid (100 μL, 0.24 M in ethanol) were added to 5 ml glass test tubes with and without 40 μl of an apocarotenoid (final concentration, 1–100 μM in ethanol). The tubes were mixed well and then illuminated at 7,000 lux and 22°C for 3 h in a Styrofoam box, after which 50 μl of the reaction mixture was removed and diluted to 1.5 ml with ethanol. OD₂₃₅ was then measured to estimate the levels of conjugated dienes formed in the reaction (16). OD₂₃₅ in the absence of the carotenoid was also measured as a negative control (no ¹O₂-quenching activity), and the ¹O₂ quenching activity of each apocarotenoid was then calculated relative to this reference value. The activity was determined as the IC₅₀ value, representing the concentration at which 50% inhibition was observed.

Figures 3A,B show the results of GC analysis of the MeOH-chloroform extracts of saffron (A) and freesia “f2” flowers (B). Peak 1, one of the main peaks in saffron, was found to exist in freesia as peak 3. MS spectra of peak 1 and peak 3 had the highest similarity to safranal in the database of the GC-MS system (Figure 3C). Peak 4 appearing in freesia, which corresponded to peak 2 in saffron, was considered 3-hydroxy-β-cyclocitral (C₁₀H₁₆O₂; MW 168.23) by their MS spectra (C), which should be biosynthesized from zeaxanthin with carotenoid cleavage dioxygenase 2 (CCD2) (3, 17), as shown in Figure 1. It is thus likely that these two volatile compounds are present not only in saffron but also in freesia yellow flowers.

We extracted apocarotenoids from saffron (dried; 0.08 g) as well as its boiled and grilled samples by the cooking models, and analyzed them by HPLC-DAD (Figures 4A–C). Six apocarotenoids corresponding to peaks 1–4 and peaks 5, 6 were isolated from saffron and the grilled saffron, respectively, by the methods described in Material and methods, and subjected to HR-ESI-MS, ¹H and ¹³C NMR analyses. The results, 1–6 were identified as crocin (2), tricrocin (18), crocin-3 (18), crocetindial (19), crocetin (20), and 13-cis crocetin (21), respectively (their ¹H and ¹³C NMR spectra of 1–6 are shown in Supplementary Figures S1–S12). The chemical structures of 1–6 are exhibited in Figure 5.

The predominant apocarotenoid 1 and the other main apocarotenoids 2 and 3 showed the same UV-Vis spectra that possess λmax at 440 nm depending on the same aglycone, and minor apocarotenoids 4–6 also exhibited λmax around 440 nm (Supplementary Figure S13). Thus, we calculated the peak areas of individual apocarotenoids (1–6) and total apocarotenoids (1 + 2 + 3 + 4 + 5 + 6) at 440 nm in saffron, the boiled saffron, and the grilled saffron in HPLC-DAD analyses, as listed in Table 1. In the boiled cooking model of saffron, the amount of crocin (1), tricrocin (2), and crocin-3 (3) almost did not change, while crocetindial (4) disappeared, presumably due to its decomposition or a chemical reaction to other compounds. In the grilled cooking model of saffron, the amount of crocin (1) and tricrocin (2) decreased, the amount of crocin-3 (3) slightly increased, and crocetin (5) and 13-cis crocetin (6) appeared (Figure 4C). In addition, the total amount of apocarotenoids decreased to about 1/3 (Table 1). Supplementary Figure S18 also shows that the purified crocin (1) and crocetin (2) were partially changed into crocetin (5) and 13-cis crocetin (6), and into 6, respectively, at 180°C for 5 min. These observations indicated that hydrolysis of the ester linkage between glucose—the crocetin aglycone occurred, presumably by conversion from crocin and tricrocin to crocin-3 and further to crocetin, along with the subsequent trans to 13-cis isomerization of a part of crocetin. It is also likely that the degradation of the apocarotenoids occurred through strong heat cooking like grilling.

In order to confirm whether the same chemical changes occur in actual cooking, saffron was mixed with rice and a typical boiling and grilling cooking process to prepare dishes was performed. The apocarotenoids contained in the dishes were extracted and subjected to HPLC-DAD analyses (Figures 4D,E; Table 1). These results apparently showed that the same chemical changes occurred in the actual cooking. However, the total peak area of detected apocarotenoids in both dishes has decreased to about 1/5-−1/10. It is currently under investigation what happens to such disappearing apocarotenoids.

The apocarotenoids in freesia “f2” flowers (7 and 8) were isolated through chromatographic methods as described in Material and method.

The structures of 7 were analyzed by HR-ESI-MS and 1D (¹H and ¹³C) and 2D (¹H-¹H DQF COSY, HMQC, and HMBC) NMR spectra. HR-ESI-MS analysis of 7 showed (M + Na)⁺ at m/z 837.31834, and the molecular formula of 7 was determined as C₃₈H₅₄O₁₉ [(M + Na)⁺ C₃₈H₅₄O₁₉Na calcd. for 837.31570 (Δ3.16 ppm)]. The molecular formula of 7 was identical to that of tricrocin, and the ¹H and ¹³C NMR spectra of 7 in CD₃OD (Supplementary Figures S13, S14) were closely related to tricrocin but not identical. The carotenoid aglycone of 7 proved to be identical to that of tricrocin (=crocetin) by comparison the ¹H and ¹³C NMR with those of crocetin. Since three anomeric signals [δ_H 4.32 (J = 7.6 Hz), δ_H 4.59 (J = 7.8 Hz), δ_H 5.65 (J = 7.8 Hz)] with β configuration were observed in the ¹H NMR spectrum of 7 and since the molecular formula of 7 (C₃₈H₅₄O₁₉) was [crocetin (C₂₀H₂₄O₄) + 3 hexose (C₁₈H₃₆O₁₈)−3H₂O], 7 was proposed to possess a structure in which 3 β-hexoses were bound to crocetin. To determine the type of hexose, 7 (5.0 mg) was hydrolyzed in 2 M HCl (5 ml) under reflux for 2 h, and the hydrolysate was partitioned with EtOAc (5 ml) twice. The H₂O layer containing hexose was concentrated to dryness (2.0 mg), and analyzed by ¹H and ¹³C NMR in D₂O, and [α]_D [+68.3° (c 0.1, H₂O)]. These data clearly proved that hexoses in 7 were all D-glucose. The linkages of 3 β-D-glucose to crocetin in 7 were analyzed by ¹H-¹H DQF COSY, HSQC, and HMBC, and the linkages among glucoses were proved to be glucoseB(1 → 2)glucoseA and glucoseC(1 → 6)glucoseA and the ester linkage of glucoseA to crocetin was proved by the key ¹H-¹³C long-range couplings observed in HMBC spectrum (from H-1”' (δ_H 4.59) to C-2” (δ_C 82.4), from H-1”” (δ_H 4.32) to C-6” (δ _C 69.4), and from H-1” (δ _H 5.65) to C-8 (δ _C 168.1) as shown in Figure 6. From these observations, the structure of 7 was determined to be crocetin neapolitanosyl ester (Figure 6).

The production of 7 by artificial manufacturing processing was reported in two previous studies (22, 23), while this is the first report finding 7 in natural flowers.

HR-ESI-MS analysis of 8 showed (M-H)⁻ at m/z 1299.47367, and the molecular formula of 8 was determined as C₅₆H₈₄O₃₄ [(M-H)⁻ C₅₆H₈₄O₃₃ calcd. for 1299.47657 (Δ 2.23 ppm), and this formula was identical with that of crocetin dineapolitanosyl ester which was reported as a minor yellow pigment produced by Crocus sativus. Thus, we compared ¹H and ¹³C NMR spectrum of 8 (Supplementary Figures S15, S16) with those of reported crocetin dineapolitanosyl ester (2), and identified 8 as crocetin dineapolitanosyl ester (Figure 6).

We extracted apocarotenoids from the yellow petals of freesia cultivar “Kayak” [1.8 g (fresh weight)] as well as their boiled and grilled samples by cooking models, and analyzed them by HPLC-DAD (Figure 7). Figure 7A (Table 2) revealed that the petals of freesia “Kayak,” like the case of “f2,” possessed the dineapolitanosyl ester and (mono)neapolitanosyl ester of crocetin dominantly, which were 48 and 52% of whole apocarotenoids when calculated by peak areas at 440 nm, respectively. In the boiled cooking model in freesia, the amount of crocetin dineapolitanosyl ester was reduced to 59% and a small amount of crocetin appeared, while the level of crocetin (mono)neapolitanosyl ester was not changed (Table 2; Figure 7B). In the grilled cooking model in freesia, the amounts of crocetin dineapolitanosyl ester and crocetin neapolitanosyl ester were reduced to 11 and 23%, respectively, and a small amount of crocetin also appeared (Table 2; Figure 7C). These results indicated that hydrolysis of the ester linkage between glucose—the crocetin aglycone occurred in both cooking models, presumably by conversion from crocetin dineapolitanosyl ester to crocetin neapolitanosyl ester and further to crocetin. The degradation of the apocarotenoids is likely to occur specifically through strong heat cooking like grilling.

We also extracted and analyzed apocarotenoids from the flowers of freesia cultivar “Aladin” in the same way, and found that freesia “Aladin” possessed the dineapolitanosyl ester and (mono)neapolitanosyl ester of crocetin dominantly, like “f2” and “Kayak.”

This study revealed that the main pigments in the yellow flowers of freesia (Freesia x hybrida) were unique apocarotenoids, crocetin (mono)neapolitanosyl ester, and crocetin dineapolitanosyl ester. Plants that include the (mono)neapolitanosyl ester of crocetin have not been reported so far. The stigmas of Crocus sativus (saffron) and the fruits of Gardenia jasminoides, retain crocin predominantly (1, 2). Recently, fully open mature flowers of G. jasminoides were shown to accumulate crocin-3 and crocetin as main apocarotenoids (24). These findings suggest that, among plants producing crocetin glycosides, only freesia retains UDP-glucose transferase(s) [UGT(s)] that has activity sufficient for constructing the neapolitanosyl group composed of three molecules of glucose, as shown in Figure 1B. The microscopy image of the petal of cultivar “f2” showed that such yellow apocarotenoids accumulate not in plastids (chromoplasts) but vacuoles (Figure 2G).

In addition, we demonstrated that the freesia apocarotenoids can keep yellow color after bread and rice cooking, as shown in Figures 2D,F. This study opens the feasibility of freesia yellow flowers as new edible flowers with beneficial functions for human health.

Many carotenoids were reported to possess ¹O₂-quenching activity depending on their olefin structures (14). Although such activity of crocin (1) was reported (25), ¹O₂-quenching activity of 2–8 has not been reported so far. Thus, we examined these ¹O₂-quenching activities and proved that 1–8 possessed almost equivalent moderate ¹O₂-quenching activity (IC₅₀ 24–64 μM), presumably due to their same conjugation number (=9) (Table 3). On the other hand, apocarotenoids 1–8 did not show DPPH radical scavenging activities and lipid peroxidation-inhibiting activities even at 100 μM, according to our experiment.