This study used citrus seeds that are abundant in Indonesia, including citrus latifolia, C. limon (L) Burm f., citrus sinensis, C. paradise, citrus amblycarpa, C. maxima (Burm.) Merr., citrus reticulata, and citrus maxima. The eight citrus seeds were obtained from waste from the beverage industry, and some were purchased from farmers on March 20, 2023.

Among the eight samples, four of them could be identified as original Indonesian varieties, namely citrus maxima, C. maxima (Burm.) Merr., citrus paradise, and citrus amblycarpa. Meanwhile, citrus reticulata comes from China, citrus latifolia from Persia, and C. Limon (L.) Burm f. from India.

Oil extraction followed the method adapted from Aydeniz et al. (2014). The extraction process began with manual separation, washing, air drying, and freezing of citrus seeds at -20 °C. The seeds were then roasted in an oven at 30-35 °C for 30 minutes. Afterward, the seeds were cooled to room temperature, and the water content was measured using an Ohaus MB45 water content meter (OHAUS Instruments; Shanghai, Co., Ltd.). Cold pressing of the seeds was conducted using a laboratory-scale pressing machine. Fine particles and any remaining water in the pressed oil were separated through filtration. The centrifugation (Magal M16R, Shanghai.China) process operated at a speed of 6800 x g for 10 minutes. Oil samples were transferred into colored glass bottles and stored at −18 °C.

The extraction of polyphenolic compounds from citrus seed oil followed the procedure outlined by García-Villalba et al. (2010). Initially, all oil samples underwent sep pak C-18 cartridge filtration, assisted by a vacuum in the manifold. To initiate the process, 3 grams of seed oil were dissolved in 3 mL of hexane (ChemStationAsia, Malaysia), applied to the cartridge, and rinsed with 5 mL of hexane to eliminate non-polar components. The residual hexane in the cartridge was removed using a nitrogen stream. The remaining phenolic fraction was subsequently extracted with 10 mL of methanol, and the resulting solution was filtered through a 0.45 μm pore size PTFE (polytetrafluoroethylene) membrane from Fisher Scientific before inclusion in the HPLC chromatogram. All samples were stored at -18 °C and analyzed within 24 hours. The flow through each cartridge was consistently managed concurrently on the manifold.

The identification process for specific phenolic acids and flavonoids adhered to the methodology outlined by Yilmaz and Karaman (2017) with minor adjustments. A C18 reversed-phase column (Zorbax Eclipse Plus, 250 mm x 4.6 mm i.d. x 5 μm) facilitated separation, and the analysis was conducted at 25 °C at a flow rate of 0.55 mL/min, and 25-μL injection volume. The chromatographic analysis involved the concurrent monitoring of phenolic extracts at 280 nm. Mobile phase A consisted of 100% acetonitrile (Merck, Darmstadt, Germany), and mobile phase B was ultra-pure water with 0.2% (v/v) sulfuric acid (Merck, Darmstadt, Germany) 95-97%, analytical grade) following the guidelines of the National Institute of Standards and Technology (NIST, 2014). Separation was performed using the following gradient program: 80% A / 20% B for 0-20 min, 70% A / 30% B for 20-26 min, 60% A / 40% B for 26-32 min, 50% A / 50% B for 32-38 min, 40% A / 60% B for 38-42 min, 45% A /55% B for 42-47 min, 35% A / 65% B for 47-52 minutes, 20% A / 80% B for 52-54 minutes, 10% A / 90% B for 54-56 minutes, 100% B for 56-72 minutes, 45% A /55% B for 72-76 minutes, and 80% A/ 20% B for 76-80 minutes. The identification of individual phenolic compounds depended on their retention time, and quantification was performed by measuring the peak area at 280 nm, using a standard curve prepared from the corresponding standard, according to NIST guidelines. All analyses were performed in triplicate.

The assessment of carotenoid content, encompassing total carotenoids, β-carotene, and lutein followed the methodology proposed by Franke et al. (2010). For the extraction process, 0.5 g of each oil sample was combined with 2 mL of petroleum ether: acetone (1:1, v/v) until complete dissolution. Absorbance was measured at 445 nm using a spectrophotometer (Infitek, SP-IUV7; Shandong, China), with petroleum ether (MilliporeSigma, WGK Germany): acetone (Merck, Darmstadt, Germany) serving as a blank. The determination of carotenoid content employed specific equations, utilizing absorption coefficients determined in petroleum ether (Güneşer and Yilmaz, 2019)

TC= Total Carotenoid content (µg/g); A= Absorbance value at 445 nm; A 1 c m % = Specific absorption coefficients for carotenoids ( A 1 c m % = 2592, β-carotene extinction coefficient in petroleum ether); V = Volume of extraction solution (mL). P= sample weight (g).

In this study, TC was presented in units of mg/kg. To verify chlorophyll carotenoids as pheophytin-a, the AOCS Cc 13i-96 method (AOAC, 1997) was employed. This involved measuring absorbance at 630, 670, and 710 nm.

The determination of serotonin compounds in this study was made using HPLC and referring to Kruk et al. (2022). The mobile phase, composed of elution solvent A = 0.1% formic acid (BASF, China) in water and B (0.1% formic acid in acetonitrile), was employed for elution based on a scheme of 85-60% A and 15-40% B from 6-15 minutes, with a post-run phase of 5 minutes involving 60-20% A and 40-80% B. The detection of serotonin derivatives took place at 324 nm, using a C18 column (1.8 μm, 2.1 × 50 mm). Parameters such as injection volume (2 µL), flow rate (0.4 mL/min), and column temperature (25 °C) were standardized. The contents of CS and FS were determined by referencing standard curves developed from standard solutions.

Fatty Acid Methyl Esters (FAMEs) were synthesized following the Ce 2-66 protocol (AOAC, 1997) and subsequently quantified using a Gas Chromatograph (Shimadzu’s Nexis GC-2030; Maryland, USA) equipped with an HP 88 capillary column (100 m x 0.25 mm ID x 0.2 µm film thickness). The Gas Chromatograph was operated at various temperature settings: initially set at 120 °C for 1 minute, followed by an increase to 175 °C (10 °C/min) for 10 minutes, then to 210 °C (5 °C/min) for 5 minutes, and finally to 230 °C (5 °C/min) for an additional 5 minutes. For the analysis, a 1-µL injection volume was used with an injector split ratio of 1:50 and a flow rate of 2 mL/min, with hydrogen as the carrier gas. The injector and detector temperatures were maintained at 250 and 280 °C, respectively. Fatty acid identification was made through chromatography, employing a standard mixture of FAMEs for reference.

The volatile compound identification procedure was based on Guneser and Yilmaz (2017). Volatile compounds were gathered through headspace solid-phase microextraction (SPME) with specific fibers (2 cm to 50/30 µm DVB/Carboxen/PDMS; Supelco, Bellafonte). 2 grams of the oil sample, 1 gram of NaCl, and 20 μl of the internal standard (IS) (1 μl of 2-methyl-3-heptanone dissolved in 10 ml of methanol) were combined n a 40-ml SPME bottle, and stirred for 2 minutes. This mixture was then placed in a water bath at 45 °C for 15 minutes to stabilize the volatiles into the headspace. Subsequently, a needle was inserted into the bottle, and the needle fiber was introduced into the headspace at a depth of 2 cm for 10 minutes in a water bath. The volatiles collected on the needle fibers were then injected into a GC/MS equipped with an HP5 MS column (30-m x 0.25-mm i.d. x 0.25-µm). For the identification of volatiles, databases such as the National Institute of Standards and Technology (NIST, 2014) and The Wiley Registry of Mass Spectral Data (Wiley, 2006) were consulted, along with the Retention index (Kovats).

The statistical approach employed was analysis of variance (ANOVA), complemented by a post-hoc Tukey’s Honestly Significant Difference (HSD) test to discern variations in biochemical composition among different Citrus L seeds. The significance threshold was set at p < 0.05. The analysis was conducted in three replicates, and the results were presented as Mean ± Standard Deviation. This analytical framework facilitated the identification of noteworthy differences in the biochemical composition of citrus seeds. Additionally, we conducted a Principal Component Analysis (PCA) using Minitab 21 software. This facilitated a comprehensive exploration of both similarities and distinctions in the biochemical composition among various citrus varieties, providing a more detailed understanding of their profiles.

The flavonoid content, specifically catechin, in citrus latifolia seeds did not exhibit significant differences (p > 0.05) compared to C. maxima (Burm.) Merr. and citrus seeds. Similar findings were observed for citrus amblycarpa seeds against citrus reticulate and the C. paradise variety against citrus maxima. Notably, the variety C. limon (L.) Burm.f. displayed significantly different seeds (p < 0.05) across the seven samples.

Concerning eriocitrin compounds, no significant differences were observed in citrus sinensis, C. paradise, citrus reticulate, or citrus maxima seeds (p > 0.05). C. maxima (Burm.) Merr. seeds, however, exhibited the highest eriocitrin levels and were significantly different from the seven samples.

Citrus reticulate seeds demonstrated the highest rutin compound levels, significantly differing (p < 0.05) from the seven samples. Similar distinctions were observed for citrus maxima seeds in the naringin compound, C. limon (L.) Burm.f. (naringenin and neohesperidin), citrus latifolia (hesperidin), and citrus amblycarpa (kaempferol).

The phenolic acid compounds exhibited varying levels across the eight seeds, with citrus latifolia containing both the highest and lowest levels of gallic acid (tr-ferulic acid, rosmarinic acid, and tr-2-hydrocinnamic acid), showing significant differences from the other seven samples.

Additionally, citrus latifolia seeds demonstrated the highest levels of carotenoids, including total carotenoids, β-carotene, and lutein, and the lowest levels of total chlorophyll (pheophytin a). These differences were found to be significant (p < 0.05) across the seven samples. Detailed results regarding the levels of flavonoids, phenolic acids, and carotenoids are provided in Table 1.

The categorization of citrus seed varieties based on flavonoids, phenolic acids, and carotenoids is illustrated in Figure 2. Among the seed varieties, including C. paradise, C. limon (L.) Burm.f., citrus reticulate, C. maxima (Burm.) Merr., and citrus sinensis, there was a prominent influence on component 1, indicating their similarity in terms of the tested compounds. Conversely, the citrus maxima, citrus amblycarpa, and citrus latifolia varieties displayed dissimilarities, positioned distinctly apart in different quadrants (Figure 2A).

The grouping of flavonoids, phenolic acids, and carotenoids in citrus varieties is further depicted in Figure 2B. A narrow angle, indicative of similarity, was observed for compounds such as naringin, naringenin, total chlorophyll (pheophytin a), rosmarinic acid, rutin, tr-2-hydrocinnamic acid, neohesperidin, tr-ferulic acid, and eriocitrin. These compounds demonstrated a strong correlation in component 1.

Total carotenoid compounds, lutein, β-carotene content, and hesperidin exhibited similarities in component 2, albeit with a weak correlation. In contrast, syringic acid and catechin were positioned closely, indicating a strong correlation.

All seed varieties exhibited N-Acylserotonin compounds that did not display significant differences (p > 0.05) for < C21, ai-C21, Me-C20, n-C21, n-C25, and ai-C27. The prevalence of N-Acylserotonin compounds was found in the C22 to C24 homologs for all varieties, and the distribution of values was consistent across all samples. Notably, there was a substantial difference for Me-C22, with the citrus maxima variety recording the highest level (12 ± 0.3 mg/kg oil), which was found to be significantly different (p < 0.05) when compared among the seven samples. Comprehensive results regarding the composition of N-Acylserotonin for each variety are presented in Table 2.

The grouping of citrus seed varieties based on the N-Acylserotonins compound did not reveal a strong correlation between varieties. Each variety displayed a wide angle, but C. paradise and citrus sinensis varieties exhibited similarity compared to other varieties. As illustrated in Figure 2C, the even distribution of N-Acylserotonins compounds in seed varieties did not indicate any noticeable similarity among the samples.

Strong correlations with N-Acylserotonin compounds were observed, particularly in component 1, for ai-C25, Me-C24, and iso-C28. Additionally, in component 2, there were strong correlations for compounds < C21, Me-C20, n-C24, and Me-C21. The compounds Me-C23 and ai-C23 also exhibited similarities, albeit in the negative (-) area.

Furthermore, the n-C21 compound demonstrated similarity to ai-C27, n-C25, n-C23, and Me-C22. As illustrated in Figure 2D, the levels of N-Acylserotonins were evenly distributed into four quadrants, indicating similarity despite having distinct strong correlations for each N-Acylserotonin compound.

None of the samples exhibited significant differences (p > 0.05) for the compounds 3-Methoxy-1-butanol, 3-Carene, α-Ocimene, and Phenylethyl alcohol. The D-Limonene compound predominated, with the highest level (5902.07 ± 62 ppm) observed in the citrus latifolia variety, significantly differing (p < 0.05) from other varieties (except citrus sinensis).

The second dominant compound was b-Myrcene, reaching its highest levels at 124.89 ± 0.4 ppm in the citrus latifolia variety, and it was significantly different (p < 0.05) from other varieties. Comprehensive results for the composition of volatile compounds are provided in Table 3

The grouping of citrus seed varieties based on aromatic volatile compounds did not reveal a strong correlation between varieties. All samples exhibited very wide angles between each other, and the distribution of aromatic volatile compounds was uniform across all varieties, indicating a lack of similarity among them (see Figure 3A).

A strong correlation, indicated by a narrow angle, was observed for the compounds 4-carnomenthol, butyl iso butyrate, 2,5 dimethylpyrazine, 1-octenol, and phenylethyl alcohol. Similar properties were also noted in the compound group 3-Methylbutanol, benzaldehyde, and (E)-limonene oxide in component 1 (see Figure 3B)

In component 2, a narrow angle was observed for the compounds butyrolactone, isopropyl pentanoate, (z)-linalooxide, and α-thujene. A negative correlation was seen for the furfural, hexanal, and 2-furan menthol compounds.

None of the varieties exhibited significant differences (p > 0.05) for the compounds lauric acid, arachidic, and behenic. However, palmitic compounds showed significant differences (p < 0.05) among varieties, except for citrus latifolia and citrus maxima.

Polyunsaturated fatty acids (PUFA) dominated, reaching 53.27 ± 0.8 ppm in the citrus seed variety, and were significantly different from other varieties. The second dominant group was saturated fatty acids (SAFA), which exhibited significant differences in Citrus maxima varieties. The highest levels of monounsaturated fatty acids (MUFA) were observed in the C. limon (L.) Burm.f. variety (27.8±0.4 mg/kg oil), and it was significantly different from other varieties. The complete composition of fatty acid compounds is detailed in Table 4.

Based on the composition of fatty acid compounds, the citrus latifolia and C. limon (L.) Burm.f. varieties exhibited similarity in the component 1 area, characterized by very narrow angles. In contrast, other varieties demonstrated different conditions, showing no strong correlation with each other (see Figure 3C).

In the composition of fatty acid compounds, palmitic, linoleic, and stearic exhibited a narrow angle or strong correlation in component 1. Another group displaying a strong correlation consisted of behenic, linolenic, lauric acid, and margaric. In component 2, the myristic acid compound was very strongly correlated (narrow angle) with PUFA, while this condition did not apply to oleic or arachidic. Complete results can be observed in the biplot of the fatty acid composition test (Figure 3D).