Article pubs.acs.org/JAFC
Raman Spectroscopy of Fish Oil Capsules: Polyunsaturated Fatty Acid Quantitation Plus Detection of Ethyl Esters and Oxidation Daniel P. Killeen,*,† Susan N. Marshall,† Elaine J. Burgess,‡ Keith C. Gordon,§ and Nigel B. Perry‡,§ †
The New Zealand Institute for Plant & Food Research Limited, 300 Wakefield Quay, Nelson 7010, New, Zealand The New Zealand Institute for Plant & Food Research Limited, §Department of Chemistry, University of Otago, P.O. Box 56, Dunedin 9054, New Zealand
‡
S Supporting Information *
ABSTRACT: Fish oils are the primary dietary source of ω-3 polyunsaturated fatty acids (PUFA), but these compounds are prone to oxidation, and commercial fish oil supplements sometimes contain less PUFA than claimed. These supplements are predominantly sold in softgel capsules. In this work, we show that Fourier transform (FT)−Raman spectra of fish oils (n = 5) and ω-3 PUFA concentrates (n = 6) can be acquired directly through intact softgel (gelatin) capsules. These spectra could be used to rapidly distinguish supplements containing ethyl esters from those containing triacylglyceride oils. Raman spectroscopy calibrated with partial least-squares regression against traditional fatty acid methyl ester analyses by gas chromatography−mass spectrometry could be used to rapidly and nondestructively quantitate PUFA and other fatty acid classes directly though capsules. We also show that FT−Raman spectroscopy can noninvasively detect oxidation with high sensitivity. Oils with peroxide values of as low as 10 mequiv kg−1, which are on the cusp of falling outside of specification, could be readily distinguished from oils that were within specification (7 mequiv kg−1). KEYWORDS: fish oil, omega-3, Raman, polyunsaturated fatty acids, docosahexaenoic acid, eicosapentaenoic acid oxidation, softgel capsules
■
INTRODUCTION
produces short chain fatty acids, peroxides, volatile aldehydes, and conjugated dienes (Figure 1), some of which have been associated with negative health impacts.18−21 The vast majority of studies (2102 out of 2187 reviewed) that have investigated oxidation of commercially available PUFA supplements found oxidation levels within specification.22 That review stated that the studies that reported out of specification results for oxidation often do not consider volatile flavorings such as 4anisaldahyde (Figure 1), which give false high results for some measures of oxidation.22 Some studies have reported EPA and DHA concentrations lower than the stated label claims,11,23,24 but more often, the label claim values for these compounds are exceeded.12,22 Therefore, we wanted to establish a rapid, nondestructive method for analysis of the actual product taken by consumers, i.e., fish oils in softgel capsules. Raman spectroscopy is increasingly used for qualitative lipid analysis in conjunction with principal component analysis (PCA)25−30 and for quantitative analysis in conjunction with partial least-squares regression (PLS-R).31−33 In its most common form, Raman spectroscopy involves irradiation of samples with a single wavelength (laser) and collection of Stokes scattered radiation.34,35 The technique is rapid, noninvasive, nondestructive, requires no sample preparation for many applications, and has a high potential for portability.34,35 It has also been demonstrated that Raman spectroscopy with PLS-R can predict EPA, DHA, and total ω-3
The “Western diet” is deficient in long chain, ω-3 polyunsaturated fatty acids (PUFA) (Figure 1),1 which are linked to a wide range of potential health benefits, especially relating to cardiovascular diseases.2−4 This has led to a global demand for dietary supplements rich in PUFA, particularly docosahexaenoic acid (DHA) and eicosapentaenoic acid (EPA) (Figure 1). These compounds are produced by marine algae, pass through trophic levels in marine food webs, and accumulate in high concentrations in fish.5,6 Some commercial PUFA supplements are byproducts from fishing industries,7 but the primary source is the Peruvian sardine and anchovy fishery, which is mostly dedicated to the production of high PUFA fish oil (and fish meal).8 Oils from this fishery tend to contain about 18% EPA and 12% DHA as triacylglycerides (TAG),9,10 which is therefore a common composition of many commercial fish oil supplements.11,12 PUFA supplements are also commercially available as concentrates, which have approximately double the EPA and DHA content of standard fish oils.11,12 These concentrates are generally produced by generating ethyl esters (EE) from fish oil followed by physical (e.g., molecular distillation) or chemical (e.g., chromatography) fractionation and finally reconversion (usually enzymatically) of EE to TAG, as summarized in Figure 1.13 Some producers forego the final step and sell supplements directly as EE, which are more strongly regulated in some markets and may have lower bioavailability of DHA and EPA.14−16 The high degree of unsaturation of DHA and EPA make them vulnerable to autoxidation in air, especially at higher temperatures and when exposed to UV radiation.17 This © 2017 American Chemical Society
Received: Revised: Accepted: Published: 3551
January 8, 2017 April 16, 2017 April 18, 2017 April 19, 2017 DOI: 10.1021/acs.jafc.7b00099 J. Agric. Food Chem. 2017, 65, 3551−3558
Article
Journal of Agricultural and Food Chemistry
Figure 1. Chemical structures of the most common long chain polyunsaturated fatty acids, a summary of how they are concentrated industrially, and examples of oxidation products and flavor compounds.
Table 1. Label Claim EPA and DHA Concentrations in Eleven Commercial Encapsulated ω-3 Oil Supplements Plus Determinations by GC−MS Analysis of Fatty Acid Methyl Esters and Noninvasive FT−Raman Spectroscopy of Encapsulated Oils GC−MS determination (% of total area, n = 2)
label claim (%) sample fish oil 1 fish oil 2 fish oil 3 fish oil 4 fish oil 5 ω-3 concentrate ω-3 concentrate ω-3 concentrate ω-3 concentrate ω-3 concentrate ω-3 concentrate ω-3 concentrate a
1 2 3a 3b 4 5 6
EPA
DHA
EPA+DHA
EPA
DHA
EPA+DHA
14 18 18 18 18 16 33 36 36 36 36 39
9 12 12 12 12 31 23 24 24 24 24 19
23 30 30 30 30 47 55 60 60 60 60 58
14.2 17.9 18.3 18.8 18.3 20.9 34.7 31.0 34.3 39.7 41.7 45.4
8.9 9.5 10.0 10.2 10.7 32.7 22.0 20.1 22.5 25.6 27.8 24.2
23.1 27.4 28.3 29.0 29.1 53.7 56.6 51.1 56.9 65.3 69.4 69.5
FT−Raman/PLSR prediction (% of total area, average of n = 10 ± SD) EPA
DHA
EPA+DHA
16.1 16.0 18.1 18.1 18.4 23.6 32.8
± ± ± ± ± ± ±
1.2 1.0 0.9 0.6 0.9 1.0 1.4
7.5 10.1 10.5 11.0 10.5 31.8 24.0
± ± ± ± ± ± ±
1.1 1.0 0.8 1.0 1.1 0.8 0.8
24.8 27.7 30.5 27.1 26.6 54.1 59.5
± ± ± ± ± ± ±
1.3 0.4 1.3 1.1 0.9 0.5 0.6
34.6 41.7 40.0 44.8
± ± ± ±
0.9 1.4 1.9 0.8
22.5 24.7 26.6 25.2
± ± ± ±
0.7 0.6 1.0 1.1
57.7 65.5 65.3 69.7
± ± ± ±
0.6 0.4 1.1 0.2
High in EEs and residual EEs in GC−MS. bResults from longer trans-methylation treatment; no EEs were detected in GC−MS.
fatty acids in fish oil with standard errors of prediction of 2.73, 2.56, and 3.27%, respectively.37 Similarly, several recent reports have investigated mid-infrared (IR) and near-infrared (NIR) spectroscopy as tools for rapid determination of fatty acid compositions in fish oils,36−40 with one report showing the potential of a portable IR device.38 Specific advantages of Raman spectroscopy over IR and NIR spectroscopies include its insensitivity to water, simple optical set ups that allow straightforward sample presentation and, due to the selection rules associated with Raman scattering, its enhanced suitability for the analysis of compounds containing a high degree of unsaturation.34,35 Direct, noninvasive quantitative determination of fatty acid profiles in softgel capsules is highly desirable for quality assurance and quality control purposes. Rapidly discriminating EE from TAG would also be useful in markets that regulate against the former. Finally, a noninvasive measurement of oxidation in softgel capsules would also be highly useful for quality assurance, especially considering the current debate surrounding fish oil supplement quality. In this report, we investigate the suitability of 1064 nm Fourier transform (FT)−
Raman spectroscopy for simultaneously determining these three important fish oil parameters.
■
MATERIALS AND METHODS
Chemicals. Analytical grade hexane, methanol, chloroform, conc HCl, sodium thiosulfate, iodine, cyclohexane, conc formic acid, glacial acetic acid, iso-octane, starch, and potassium iodide were purchased from Sigma-Aldrich New Zealand. Commercial Encapsulated Oils Sample Set. Eleven commercial ω-3 PUFA oil supplements in softgel (gelatin) capsules were purchased from three different New Zealand vendors in November 2016, all within their labeled “use by” date. A summary of the label claim EPA and DHA concentrations of these samples is presented in Table 1. Oxidized Fish Oil Sample Set. A bulk sample (1 L) of freshly bleached and deodorized fish oil was obtained in July 2016 and stored at 8 °C for 3 months. In duplicate, 10 g aliquots of the oil were oxidized for 1, 2, 3, and 5 h using a Metrohm 892 Professional Rancimat. The temperature was fixed at 90 °C with an air flow rate of 20 L h−1. A control sample (0 h), which was not subjected to Rancimat oxidation, was subsampled in duplicate at the same time for a total of 10 samples. Commercial softgel oil capsules (30) were pierced with a needle and drained of their contents. In triplicate, capsules were first 3552
DOI: 10.1021/acs.jafc.7b00099 J. Agric. Food Chem. 2017, 65, 3551−3558
Article
Journal of Agricultural and Food Chemistry rinsed and then filled with the oxidized fish oils. Softgel capsules are primarily composed of gelatin.42 Fatty Acid Methyl Ester (FAME) Analyses. The FAME profiles of the 11 commercial oil samples were measured in duplicate. Replicates were sampled from two different capsules. A single subsample of each of the 10 oxidized samples was also analyzed. A single oil drop (10 ± 3 mg) was accurately weighed into a 2 mL GC vial and dissolved in 1.8 mL of CHCl3. An aliquot (300 μL) of each sample was added to 2.5 mL of 9:1 methanol:conc hydrochloric acid and heated at 70 °C for 10 min to facilitate hydrolysis and methylation of fatty acids. Samples were cooled; hexane (4.0 mL) was added, and each sample was vortexed for 1 min. An aliquot of the upper hexane layer was analyzed by GC−MS using a Shimadzu 2010 QP GC−MS equipped with a Restek Rxi-5Sil MS column (30 m × 0.25 mm id, 0.25 μm film thickness). Splitless sample injections (1 μL) were performed with an inlet temperature of 250 °C. The initial temperature of the column oven was 90 °C, rising at a rate of 20 °C min−1 to 190 °C, followed by a temperature gradient of 0.5 °C min−1 to 200 °C, then at a rate of 2 °C min−1 to 220 °C, and then at a rate of 5 °C min−1 to 250 °C. A final temperature ramp at 35 °C min−1 for 2 min brought the column temperature to 320 °C, which was held for 2 min to remove low-volatility sterols and hydrocarbons from the column. The interface between the oven and mass spectrometer was set at 230 °C; ions were generated by electron ionization (E = 70 eV) and detected in a m/z range of 50−550 atomic mass units. The results are reported as percent area of the total peak area in the chromatograms. Peaks that were smaller than 0.2% of the total area were not included in the analyses. Peaks were generated from the total ion current ions with m/ z of 50−550. Peak assignments were performed by comparison to a commercial FAME mix (Supelco) and with reference to the National Institute of Standards and Technology (NIST) mass spectral library. When peaks could not be definitively assigned using these methods, tentative assignment was performed by interpretation of mass spectral fragmentation patterns and by comparison to online spectral libraries of FAME.41 Peroxide Values. These were determined for the oxidized fish oil samples by titrating accurately weighed aliquots of oil (1 g) with standardized Na2S2O3 as per AOCS official method Cd 8b-90, on the same day as oil encapsulation and Raman analysis. Subsamples of these oxidized oils were stored at −20 °C prior to FAME analysis. FT−Raman Acquisition and Chemometric Analyses. FT− Raman spectra of gelatin-encapsulated fish oils were measured directly using a Bruker Optics MultiRAM spectrometer equipped with a D418T Germanium detector cooled with liquid nitrogen and controlled by Bruker Opus v7.5 software. A 1064 nm Nd:YAG laser was used with a power setting of 100 mW. Samples (softgel oil capsules) were held in place using a spring-loaded piston with a slit approximately 0.5 cm wide through which spectra were acquired using a 180° backscattering geometry. This setup allowed a (reasonably) consistent focus to be maintained for all samples without adjustment. The laser spot was approximately 0.3 mm diameter, and the spectral resolution was 4 cm−1. Raman Stokes shifted radiation was recorded from 4000−200 cm−1, and the digitized spectra consisted of 3800 data points. Spectra were the average of 32 scans, and the total acquisition time was about 40 s per sample. A single Raman spectrum was acquired from each of 10 oil capsules from the 11 commercial PUFA supplements described in Table 1. FT−Raman spectra were imported into “The Unscrambler” software and subjected to a Savitzky−Golay second derivative transformation (11 smoothing points, symmetric kernel, third order polynomial) to remove broad baseline features and then a standard normal variate (SNV) transformation to remove intensity differences arising from variable focus and path length effects. The spectral regions from 4000−3050, 2650−1800, and 700−200 cm−1 did not contain any vibrational bands and were removed from the spectral data set. PCA was performed on the remaining spectral regions using the noniterative partial least-squares (NIPALS) algorithm. PLS-R models were produced by relating Raman spectra to results from the FAME analyses (Table S1). For the oxidized fish oil samples, Raman spectra were acquired, processed, and subjected to PCA as described above. Spectra were related to FAME profiles using
PLS-R, and full, “leave-one-out” cross validation was performed on each model. The validation results were used to select an optimal number of PLS-R factors for each model.
■
RESULTS AND DISCUSSION FAME Profiles of Commercial Oil Capsules. The FAME profiles of two oil capsules from each of the 11 commercial oil samples were measured by GC−MS. Individual fatty acids are reported as a percent of the total integrated peak area in each chromatogram in Table S1. FAME results were classified in groups: saturated fatty acids, monounsaturated fatty acids, PUFA, ω-3 unsaturated fatty acids, and ω-6 unsaturated fatty acids (Table S1). These results provide a proportional measure of each fatty acid as a percent of total fatty acids detected in each oil. Measured FAME results were compared with the stated percent label claim (% LC) values for each product (Table 1). The 5 fish oil samples analyzed in this study contained 99− 106% LC for EPA values, 79−99% LC for DHA, and 91−103% LC for EPA+DHA content (Table 1). The 6 commercial PUFA concentrates analyzed in this study contained 95−134% LC for EPA, 94−126% LC for DHA, and 95−120% LC for EPA+DHA (Table 1). The largest deficit in measured EPA+DHA content versus label claim was in fish oil 2, which had only 91% LC EPA +DHA. This sample had a label claim of 30% EPA+DHA and a measured value of 27.4% of total area (Table 1). The most understated EPA+DHA content versus label claim was found in ω-3 concentrate 6, which had 120% LC EPA+DHA. This sample had a label claim value of 58% EPA+DHA and a measured value of 69.5% of total area. On average, the percent LC EPA+DHA was 104 across all 11 commercial samples. One sample of ω-3 concentrate 3 contained EE, not TAG, and needed to be reanalyzed following a more rigorous methylation treatment. This sample is discussed further below. A recent investigation of the quality of PUFA supplement available on the New Zealand market reported a large deficit in EPA+DHA content compared with the label claim with an average content of only 68% LC.11 The negative results of that study were much-publicized, prompting a follow up study investigating a similar sample set of 10 commercially available PUFA supplements for sale on the New Zealand (and Australian) market.12 That report assessed % LC of EPA +DHA and found that all samples met or exceeded their claimed concentrations of EPA+DHA.12 Results from the present study are in better agreement with the more recent study by Nichol et al.12 than those originally presented by Albert et al.11 However, some samples analyzed in this study did fall slightly below the label claim values for EPA+DHA (Table 1). The report by Albert et al. found that many oils were oxidized beyond specified limits, leading the researchers to assess if oxidation was the cause of the label claim deficiencies they measured for EPA+DHA.11 The authors reported that no correlation (r2 = 0.20/0.12) existed between oxidation and percent LC deficiencies of EPA/DHA.11 Nichols et al. also measured the extent of oxidation found in the oils from their study, finding that all oils met the official specifications for oxidation (peroxide value
