Carotenoids from Foods of Plant, Animal and Marine Origin: An Efficient HPLC-DAD Separation Method

Carotenoids are important antioxidant compounds, present in many foods of plant, animal and marine origin. The aim of the present study was to describe the carotenoid composition of tomato waste, prawn muscle and cephalothorax and avian (duck and goose) egg yolks through the use of a modified gradient elution HPLC method with a C30 reversed-phase column for the efficient separation and analysis of carotenoids and their cis-isomers. Elution time was reduced from 60 to 45 min without affecting the separation efficiency. All-trans lycopene predominated in tomato waste, followed by all-trans-β-carotene, 13-cis-lutein and all-trans lutein, while minor amounts of 9-cis-lutein, 13-cis-β-carotene and 9-cis-β-carotene were also detected. Considering the above findings, tomato waste is confirmed to be an excellent source of recovering carotenoids, especially all-trans lycopene, for commercial use. Xanthophylls were the major carotenoids of avian egg yolks, all-trans lutein and all-trans zeaxanthin in duck and goose egg yolk, respectively. In the Penaeus kerathurus prawn, several carotenoids (zeaxanthin, all-trans-lutein, canthaxanthin, cryptoxanthin, optical and geometrical astaxanthin isomers) were identified in considerable amounts by the same method. A major advantage of this HPLC method was the efficient separation of carotenoids and their cis-isomers, originating from a wide range of matrices.


Reagents and Standards
All the solvents used for sample preparation and extraction were of analytical grade and were obtained from Merck (Darmstadt, Germany). All solvents used for HPLC analysis (acetonitrile, 1-butanol and methylene chloride) were of HPLC grade and were obtained from Merck (Darmstadt, Germany). All-trans carotenoid standards (lycopene, β-carotene, lutein, zeaxanthin, canthaxanthin and astaxanthin) were purchased from Sigma Chemical Co. (Sigma-Aldrich Company, St. Louis, MO, USA).

Instrumentation
The HPLC (Hewlett Packard Series 1100, Waldbronn, Germany) system was composed of a HP 1100 Quaternary Pump, an Agilent 1100 Series Micro Vacuum Degasser, a Rheodyne model 7010 Sample Injector and a HP 1100 Series Diode Array Detector (DAD). The HPLC system was equipped with a YMC (Tokyo, Japan) C 30 column (250 × 4.6 mm I.D., 5 μm particle). The analysis of the chromatographic data was carried out on a ChemStation for LC 3D software (Agilent Technologies, Waldbronn, Germany).

Tomato Waste
Tomato processing waste (500 kg), composed of skin and seeds, was collected from NOMIKOS (Aliartos, Boeotia, Greece), a tomato-processing factory. Moisture content was determined at fresh tomato processing waste upon arrival at the laboratory and was found to be 80.48% ± 0.35%. The moisture content was determined according to the AOAC method 925.09 (1999). Tomato pomace was air dried at 25 °C, upon arrival, and the dry material (moisture content = 7.65% ± 0.21%) was subsequently vacuum packed in plastic bags and stored at −20 °C. Prior to each extraction, 60 g of dried tomato waste was homogenized in a domestic blender and finally powdered in a laboratory mill (Type ZM1, Retsch GmbH, Haan, Germany) equipped with different particle size sieves. Dry ground material was kept in glass jars wrapped with aluminum foil at −20 °C.
Carotenoids were extracted according to the method described by Strati and Oreopoulou [10]. Homogenized and ground dry tomato waste samples (10.00 g) were stirred with 100 mL acetone (tomato waste/solvent = 1:10) in a 500 mL extraction vessel equipped with a vertical water cooler and multiple-neck lid. The vessel was placed in a temperature-controlled (±1 C) water bath and continuously agitated with a propeller type stirrer. The extraction temperature and time were kept constant at 25 ± 1 °C and 30 min, respectively. The mixture was vacuum filtered (Filter funnel DURAN 25 852 34, SCHOTT DURAN, Mainz, Germany); the solid residue was collected and re-extracted two more times, with fresh extraction solvent under the same conditions. Extracts were combined and centrifuged at 3000 rpm (HERMLE centrifuge Z380, Gosheim, Germany) for 10 min to separate the supernatant. Then, the supernatant was evaporated to dryness in a rotary vacuum evaporator (Rotavapor RE 111, Flawil, Switzerland) at 32 °C, dissolved in 1 mL methylene chloride and transferred to a vial. To prevent oxidation, t-butyl-hydroquinone, at a concentration of 100 mg/L, was added to all samples. The new solution was filtered through a 0.45 μm membrane filter, and 20 μL were injected for HPLC analysis. All samples were analyzed within three days after extraction. The whole extraction procedure was performed under dimmed light.
2.3.2. Foods of Animal (Duck and Goose Egg Yolks) and Marine (Penaeus kerathurus Prawn) Origin Duck (Anas platyrhucus) and goose (Anser anser) conventional eggs, less than three days after laying, were obtained from a local producer in Lakonia, Greece (three sampling repetitions) [11]. Egg yolks were separated manually from their respective whites, then homogenized in a blender, vacuum-packed in plastic bags and stored at −20 °C. All samples were analyzed within three days after extraction. Before analysis, samples were allowed to achieve room temperature.
Fifty adult Caramote prawns (Penaeus kerathurus) were caught in the North Aegean Sea (near Platamona Bay) in October 2010. Prawns were brought to the laboratory alive and individually measured for weight and length (with an average of 28.5 g/prawn and 16.7 cm/prawn, respectively). Then samples of muscle and cephalothorax were collected, individually weighed and homogenized.
As carotenoids are lipid constituents, total lipids from the above samples (duck and goose egg yolks, as well as prawn muscle and cephalothorax) were extracted according to Bligh and Dyer [12]. After phase equilibration, the lower chloroform layer (total lipids) was removed and dried in a rotary vacuum evaporator (Rotavapor RE 111, Flawil, Switzerland) at 32 °C. The extracted lipids were redissolved in chloroform/methanol (9:1, by volume). To prevent oxidation, t-butyl-hydroquinone, at a concentration of 100 mg/L, was added to all samples. All samples were analyzed within three days after extraction. The whole extraction procedure was performed under dimmed light.

HPLC Analysis of Carotenoids
For the identification of the different carotenoids, tomato waste extract and total lipids from avian (duck and goose) egg yolks and Penaeus kerathurus muscle and cephalothorax were further analyzed by HPLC-photodiode array detection. Before being injected, lipid samples were dried under nitrogen gas and dissolved in acetone:hexane (2:3, by volume). Afterwards, samples were filtered through a 0.45 μm membrane filter to remove particulate residues. Twenty microliters of solution were injected for the HPLC analysis. The solvent systems selected were based on several previous studies [13,14] and preliminary experiments using carotenoid standards. The most appropriate solvent system was found to be composed of acetonitrile, 1-butanol and methylene chloride. A mobile phase of acetonitrile (A), 1-butanol (B) and methylene chloride (C) with the following gradient elution was used: 69.3% A, 29.7% B and 1.0% C, initially; increased to 67.2% A, 28.8% B and 4% C, in the first 10 min; 61.6% A, 26.4% B and 12% C, after 20 min; 49% A, 21% B and 30% C, after 40 min; and returned to 69.3% A, 29.7% B and 1% C, after 50 min. The UV-visible spectra were obtained between 250 and 600 nm. The flow rate was maintained at 2 mL/min and the column temperature at 25 °C. The separation efficiency was evaluated on the basis of capacity factor (k) and separation (selectivity) factor (α), as follows [15]: where t R is the retention time of the peak of interest, t 0 is the unretained peak's retention time and k A and k B are the capacity factors for peaks A and B, respectively [15].

Identification and Quantification of Carotenoids
The identification of trans and cis isomers of carotenoids was carried out by comparing the retention times and absorption spectra with reference standards and absorption spectra characteristics, as described in the literature [13,14].
Quantification was performed based on absolute calibration curves of all-trans lutein (447 nm), all-trans zeaxanthin (453 nm), all-trans canthaxanthin (452 nm), all-trans astaxanthin (478 nm), all-trans-β-carotene (455 nm) and all-trans-lycopene (476 nm), with a minimum of five concentration levels. For all-trans lutein, two standard curves were prepared, due to the great difference in lutein content found in the different samples. The concentration range for the carotenoid standard curves were: 1-20 μg/mL (low concentration) and 20-200 μg/mL (high concentration) for all-trans-lutein, 20-200 μg/mL for all-trans zeaxanthin, 50-500 μg/mL for all-trans canthaxanthin, 500-2000 μg/mL for all-trans astaxanthin, 2-40 μg/mL for all-trans-β-carotene and 10-100 μg/mL for all-trans-lycopene. The above concentration ranges were chosen as the most appropriate for determination of individual carotenoid content and were based on preliminary experiments.
The cis isomers of carotenoids were quantified using the standard curves of the corresponding all-trans carotenoids, because of the similarity in extinction coefficient [16]. Neolutein and cryptoxanthin were quantified using the curve of all-trans lutein and astaxanthin isomers and esters using the curve of all-trans astaxanthin. The carotenoids of all samples were quantified on a dry weight basis.

Determination of Limits of Detection (LOD) and Limits of Quantification (LOQ)
The limits of detection (LOD) and limits of quantification (LOQ) were determined using the calibration curves according to the method described by ICH [17].
The LODs and the LOQs were calculated according to the Equations (1) and (2): where S is the mean of the slopes of calibration curves and σ the standard deviation of the response [17].

Statistical Analysis
Three independent samples were analyzed and values were averaged and reported along with the standard deviation (SD). All statistical calculations were performed with the SPSS package (IBM SPSS Statistics, version 19.0, Chicago, IL, USA) statistical software for Windows.

Results and Discussion
Initially, the HPLC method developed by Lin and Chen was tested for the determination of carotenoids in processed tomato juice [7]. However, the analytical conditions did not show good separation efficiency (some overlapping peaks), and the analysis itself took very long (more than 60 min, data not shown). We, therefore, improved the analytical method in terms of solvent percentages of mobile phase in order to cope with the above problems. After several preliminary trials, a gradient mobile phase consisting of acetonitrile, 1-butanol and methylene chloride, as described in the method section, was applied for the analysis of carotenoids and their isomers from different carotenoid matrices.

HPLC Analysis of Carotenoids in Tomato Waste Extract
The HPLC chromatogram of carotenoids in tomato waste extract is presented in Figure 1. Seven carotenoids were separated and identified from the tomato waste acetone extract within 30 min. The chromatogram indicates that good separation efficiency and an adequate separation time were achieved for the analysis of carotenoids in tomato waste. Table 1 presents the chromatographic and the quantification data for the carotenoids in tomato waste. The k value (capacity factor) is used to assess the solvent strength of the mobile phase. The k values of all peaks ranged from 0.53 to 12.74, indicating that a proper solvent strength of the mobile phase was controlled. It has been reported that for optimum separation, the k values should range from 2 to 10, however, they can range between 0.5 and 20, when complicated compounds are to be separated [13]. The separation or selectivity factor (α) values for all the peaks were greater than 1.0, implying that a good selectivity of mobile phase to sample components was achieved. (4) all-trans-β-carotene; (5) 9-cis-β-carotene; (6) 13-cis-β-carotene; and (7) all-trans-lycopene. Three all-trans carotenoids, namely lutein, β-carotene and lycopene, were identified based on criteria described in the experimental section. The visible absorption spectrum of peak 1 was 423, 447 and 477 nm ( Table 1) and was identical to the all-trans lutein standard used in this study. Following the same approach, peaks 4 and 7 were identified as all-trans-β-carotene and all-trans lycopene, respectively. Peaks 2 and 3 were identified as 9-cis lutein and 13-cis lutein, respectively, because a low (at 350 nm) and a high (at 376 nm) intensity cis peak were observed. Additionally, a hypsochromic shift of 5 nm was observed for 9-cis lutein. It is reported that the mono-cis isomers of carotenoids result in a hypsochromic shift, when compared to the parent all-trans forms. Moreover, the presence of central isomers, such as 13-cis or 15-cis-carotenoids, would result in a significant absorption in the ultraviolet region (320-380 nm) [13]. Accordingly, and for the same reasons as above, peaks 5 and 6 were identified as 9-cis-β-carotene and 13-cis-β-carotene, respectively, because a hypsochromic shift of 5 and 3 nm occurred in the maximum absorption peak and two cis peaks (at 340 and 345 nm) appeared in the obtained spectra of the above mentioned peaks. The Q-ratio is defined as the ratio of height at cis-peak to the height at maximum absorption peak and may also be used to identify the cis-isomers [18]. As it can be observed in Table 1, all cis-isomers identified had Q-ratios very similar to those reported in the literature [13,14,16,18,19].
The carotenoid content of tomato waste (Table 1) was calculated based on the calibration curves of the respective standards, as described in the experimental section. All-trans lycopene predominated in tomato waste (64.84 ± 0.87 μg/100 g dry waste), followed by all-trans-β-carotene, 13-cis-lutein and all-trans lutein. Minor amounts of 9-cis-lutein, 13-cis-β-carotene and 9-cis-β-carotene were also detected ( Table 1). Considering the above findings, tomato waste is confirmed to be an excellent source of recovering carotenoids, especially all-trans lycopene, for commercial use.

HPLC Analysis of Carotenoids in Avian (Duck and Goose) Egg Yolks
Six carotenoids were separated and identified from duck and goose egg yolks within 10 min. Figure 2 presents the chromatograms of duck (a) and goose (b) egg yolk carotenoids. Both chromatograms show good separation efficiency and a very short resolution time. Table 2 presents the identification data of these major egg yolk carotenoids, based on the chromatographic information obtained from the spectra. The k values of all peaks ranged from 0.60 to 2.83, indicating the adequacy of solvent strength of the mobile phase and its separation capacity (α values for all the peaks were greater than 1.0).  RT: retention time; k: capacity factor; α: separation (selectivity) factor; three independent samples were analyzed; data are expressed as mean ± standard deviation (n = 3); * β-cryptoxanthin was quantified using the curve of all-trans lutein based on the similarity of its spectra characteristics with lutein; ** β-carotene amount was within the detection limits but lower than the quantification limits.
Peaks 1, 3, 4 and 6 were positively identified as all-trans zeaxanthin, all-trans lutein, all-trans canthaxanthin and all-trans-β-carotene, based on comparison of retention times and absorption spectra characteristics with those of the respective standard compounds used in this study. Additionally, the Q-ratios found were close to the ones reported in the literature [13,16,20,21]. The absorption maxima of peak 2 (identified as neolutein) showed a relatively small hypsochromic shift of 5 nm with respect to all-trans-lutein, which suggests that this is probably a cis isomer of lutein. The location of the cis double bond in this isomer is unknown; however, the presence of a strong cis peak in the UV region at 330 nm indicates that this cis double bond occupies a more central position, i.e., 13-, 13′-or 15-, 15′-cis isomer [13]. Due to the absence of a commercial standard, peak 5 was tentatively identified as all-trans-β-cryptoxanthin, based on spectral characteristics and Q-ratios reported in the literature [22], and it was quantified using the curve of all-trans lutein based on the similarity of its spectra characteristics with lutein.
The k (capacity factor) and the α (separation factor) values of all peaks ranged from 0.37 to 11.17 and 1.0 to 1.14, indicating a good separation capacity of the mobile phase. By comparing UV spectra of peaks 1-3 with those obtained from the corresponding standards and Q-ratios with reference values in the literature [20,21,23], peaks 1-3 were identified as zeaxanthin, all-trans-lutein and canthaxanthin. Peaks 4 and 5 were identified as β-and α-cryptoxanthin, respectively, based on the similarity of spectral characteristics and Q-ratios with those reported in the literature [24,25]. Peaks 9 and 11 had a maximum absorption wavelength of 478 and 479 nm, respectively. By comparing UV spectra of peaks 9 and 11 with those obtained from the corresponding standard and absorption wavelength with reference values in the literature [26], peaks 9 and 11 were identified as the optical isomers (3R,3′R)-trans-astaxanthin and (3S,3′S)-trans-astaxanthin. Astaxanthin possesses two identical asymmetric carbon atoms at C-3 and C′-3, producing three optical isomers with all-trans configuration, the two enantiomers (3S,3′S) and (3R,3′R) and the meso-astaxanthin (3S,3′R/3R,3′S) [26]. Turujman [27] reported that synthetic astaxanthin standard consists of 25% of each enantiomer and 50% of the meso form [27]; in accordance with the chromatographic profile of the astaxanthin standard used in the present study. The means of identifying the astaxanthin isomers was through the spectral shift or the shift of the maximum absorption wavelength. In comparison to the (3R,3′R) and (3S,3′S)-all-trans-astaxanthin (λ max of 478 and 479 nm, respectively), the four peaks (6-8 and 10 in Figure 3) showed a small hypsochromic effect of 3-4 nm and were possibly assigned as cis-astaxanthin isomers. Bjerkeng reported that astaxanthin consists of three chiral R/S isomers and 272 possible geometrical cis/trans isomers of which the quantitatively most important are 9-cis-, 13-cis-and 15-cis-isomer [28]. Finally, the fractions that elute later (peaks 12 and 13 in Figure 3) showed similar absorption spectra with trans-astaxanthin, indicating astaxanthin-derived compounds; possibly astaxanthin fatty acid esters.
The comparison among the muscle and cephalothorax carotenoids showed that all-trans canthaxanthin predominated in both tissues (30.98% and 33.18% of total carotenoids, respectively), followed by all-trans zeaxanthin, all-trans astaxanthin and all-trans lutein in muscle (23.63%, 19.09% and 14.54% of total carotenoids, respectively) and all-trans-astaxanthin, all-trans lutein and all-trans zeaxanthin in cephalothorax (20.30%, 19.92% and 17.22% of total carotenoids, respectively). Furthermore, lesser amounts of unidentified astaxanthin esters were also determined in both tissues. Cryptoxanthin was identified only in the cephalothorax. On a weight basis, it was observed that the cephalothorax presented a higher total carotenoid content than muscle (13.30 ± 0.10 and 2.20 ± 0.07 mg/100 g wet tissue, respectively) ( Table 3). Results showed that Penaeus kerathurus