Fast Fatty Acid Analysis by Core-Shell Reversed-Phase Liquid Chromatography Coupled to Evaporative Light-Scattering Detector

A high-performance liquid chromatography analysis method with an evaporative light-scattering detector has been developed for the separation and quantitative analysis of free fatty acids in biological matrices. Core-shell reversed-phase high-performance liquid chromatography separation of 10 free fatty acids is achieved within 10.5 min using a methanol/water (0.05% trifluoroacetic acid) eluent gradient. After optimization, the drift tube and nebulization temperature of the evaporative light-scattering detector was set at 35 ̊C, nitrogen flow-rate at 1.1 standard liter per minute and column temperature at 25 ̊C. All calibration curves showed good regression (r > 0.9975). A validation procedure following the International Conference on Harmonisation guidelines was implemented to certify the method. Relative standard deviations did not exceed 1.5% and 4.25% for repeatability and reproducibility respectively.


Introduction
The development of fast high-performance liquid chromatography (HPLC) methods for routine analysis is of interest for a rapid identification and quantification of free fatty acids (FFAs).
Gas chromatography (GC) is obviously the most commonly used method for FFA analysis. However, before analysis, FFAs must be methylated to obtain fatty Food and Nutrition Sciences acid methyl esters which is tedious and time consuming. Moreover, there is a risk of structural degradation during this methylation process since it requires high temperature under alkaline conditions. For this reason, accuracy of GC analysis can be greatly affected when unstable molecules like long chain polyunsaturated FFAs need to be quantified [1]. Compared to GC, HPLC analysis can be conducted at low temperature thus limiting the risk of heat degradation of the samples. Moreover, FFAs can be directly assayed without any laborious preparation step [2]. HPLC is also convenient because it is simple to implement and generally provides good reproducibility. The main drawback of using HPLC for FFA analysis when using standard UV detector is that only FFAs with double bonds can be detected. Moreover, the use of a non-absorbing solvent is required, thus limiting the choice of the mobile phase. To overcome this drawback, different alternative detection systems can be implemented such as an evaporative light-scattering detector (ELSD). ELSD is a universal detector only sensitive to the mass of the vaporized analyte and is not limited by light absorption characteristics of the individual components and/or eluent nature [3]. In ELSD, the solvent is evaporated and the only requirement is a low volatility of the compounds as compared to the mobile phase. It also allows stable baselines even when multi-solvent gradients are used. Nevertheless, for an optimal operation, several ELSD parameters need to be tuned even though ELSD optimization is seldom reported in literature.
In recent years, HPLC columns have also been considerably improved. For example, core-shell silica particles have been increasingly used for highly efficient separation with fast flow rate and relatively low back pressure [4].
To the best of our knowledge, HPLC methods specifically dedicated to fatty acids in their free form are rarely described. An overview of the recent published papers is presented in Table 1. A wide range of equipment including HPLC columns and detectors are used. Even though some of these methods show clear advantages, they are either quite expensive or not fast enough, or present a low resolution.
The purpose of this study is to propose an HPLC method which is at a time, fast, simple and accurate for determination and quantification of numerous FFAs that can be found in different lipids and food matrices. Thus, the developed HPLC method takes into account the advantages of both core-shell technology and ELSD. Firstly, this study focuses on optimization of ELSD operating parameters, namely drift tube and nebulization temperatures and nitrogen flow rate. Column temperature was also optimized in order to reach maximal peak resolution. Then, the method was validated following the International Conference on Harmonisation (ICH) guidelines [10].

Method Validation
Method validation was conducted according to ICH guidelines [10]. In this frame, the reliability of the HPLC method was established through its specificity, calibration, limit of detection (LOD), limit of quantification (LOQ), accuracy, precision, robustness and carry-over

Specificity
Sunflower lecithin, that does not contain FFAs, was chosen as the lipid matrix to investigate specificity.
Standard FFA solution (2 mM) was analysed in the absence and in the presence of sunflower lecithin (20 mM) to ensure that FFAs and lecithin would not interfere with each other. Results were analyzed considering the peak area and the retention time (RT) of each FFA.

Limit of Detection (LOD) and Limit of Quantification (LOQ)
For each FFA, the values of LOD and LOQ were determined using Agilent Chemstation Software. They correspond to the lowest concentrations giving rise to a particular signal-to-noise ratio. A signal-to-noise ratio between 3 or 2:1 is generally considered acceptable for estimating LOD and for the LOQ a typical signal-to-noise is 10:1.

Accuracy
The accuracy of an analytical method expresses the closeness between the ex-

Precision
Precision was evaluated in terms of repeatability and reproducibility. Intra-assay precision (repeatability) and inter-assay precision (reproducibility) were calculated respectively after six consecutive runs or six consecutive preparations for each studied FFA.

Robustness
Three selected analytical parameters (percentage of water in the mobile phase, flow rate and column temperature) were changed one by one and their effect was observed on each FFA. All assays were performed in triplicate and results were expressed as percentages of recovery.

Carry-Over
Carry-over was assessed by sequential injections of blank samples after a sample containing twice the amount of FFAs found in the highest points of the calibration curves.

Optimization of ELSD Parameters
Working with ELSD requires optimization of two parameters: i) drift tube and nebulization temperature, ii) nebulizing gas flow rate [11].

Optimization of Drift Tube and Nebulization Temperature
According to manufacturer recommendations, drift tube and nebulizer temperatures should be similar and should not exceed solvent boiling point (70˚C for methanol). Thus, the temperature range chosen for this study was from 25˚C to 55˚C. With increasing temperatures, an increase in peak area is observed, more due to a peak broadening than to a peak height increase. However, peak height is an important factor since peaks insufficiently high do not emerge from the baseline and therefore molecules cannot be detected. Peak symmetry is another criterion to take into account since a good symmetry indicates the absence of co-elution. In this frame, a peak symmetry in the range of 0.8 -1.2 is considered as suitable. Thus, peak height and symmetry are used to determine the best setting. For optimization of drift tube and nebulization temperature, nebulizing gas flow rate was kept constant at 1.1 Standard Liter per Minute (SLM). Presumably due to its short chain length, octanoic acid is not detected by ELSD in the concentration range studied whatever the temperature. Bravi et al. [12] showed that ELSD is only suitable for the analysis of C12 -C22 FFAs since they are not evaporated at this drift tube temperature. FFAs with shorter chain length are not detected since they have been evaporated totally or partially with the solvent. This is the main drawback when working with ELSD. As seen on Figure 1, peak height increases when the temperature rises up from 25˚C to 55˚C except for four fatty acids: lauric acid (C12:0), myristic acid (C14:0), arachidic acid (C20:0) and behenic acid (C22:0). The behaviors of arachidic and behenic acids are quite intriguing and the attempts of explanation  rather presumptive. Nevertheless this is not really crippling to consider an optimized setting for drift tube and nebulization temperature.
Above 50˚C, lauric acid is no longer detected. However, up to 35˚C, peak height is stable for lauric and myristic acid. As the peak height remains high for arachidic and behenic acids despite a decrease, it appeared that the temperature of 35˚C constituted the best compromise for an optimal detection of the 10 considered FFAs. It has also been checked that all FFAs have a peak symmetry value above 0.8 or very close to 0.8 at 35˚C. Drift tube and nebulization temperature was thus set at 35˚C.

Nitrogen Flow Rate Adjustment
Influence of nitrogen flow rate was then studied with drift tube and nebulization temperature set at 35˚C. According to manufacturer recommendations, nitrogen flow rate should be set between 1 to 2 SLM. Nebulization gas flow rate determines the size of droplets formed during nebulization. The highest signal is generally obtained with the lowest gas flow rate [13].
In terms of peak height, the best ELSD response is at 1.1 SLM for all FFAs except for C12:0, C20:0 and C22:0 ( Figure 2). As far as C12:0 is considered, increasing the flow rate above 1.1 SLM does not significantly improve peak height.
In case of C20:0 and C22:0, the response, even though not the highest at 1.1 SLM, is high. Hence, the flow rate was kept at 1.1 SLM.

Optimization of Column Temperature
According to manufacturer recommendations, column temperature and ELSD drift tube temperature have to be close. Nevertheless, the column temperature also affects efficiency of chromatographic separation and to optimize this para- As seen on Figure 3, efficiency decreases when the temperature rises up ex-  than at 25˚C. To choose between 20 and 25˚C, the peak resolution between the two worst separated component, C18:2 and C22:6, was considered. Since no difference is observed between the two temperatures, column temperature will be set at 25˚C to be closest to the drift tube temperature.

Fatty Acid Separation
The chromatographic separation of 10 FFAs using ELSD is shown on Figure 4.

Specificity
Specificity was assessed by analyzing a sample of sunflower lecithin supplemented with a standard FFA solution. Since a good separation is observed be- tween all components (data not shown), the method can be considered specific.

Calibration
Since ELSD response depends on the analyzed molecules, it is necessary to establish calibration curves for each FFA studied.
As seen on Figure 5, calibration curves are non-linear and the empirical model that best describes this response is a polynomial model [16]: Peak area a concentration = × Polynomial regression of calibration curves was performed with Excel program giving a and b values for each FFA. For each standard, calibration curve shows a r 2 value above 0.9975 (Table 2). It can also be seen that high molecular weight molecules saturate the ELSD with lower concentrations than those of smaller size.

LOD and LOQ
Limits of quantification and detection were determined according to a signal to noise approach. For fatty acids with a chain length of 14 or greater, satisfactory sensitivity is achieved with a LOQ value between 0.03 and 0.3 mM (Table 2).
With shorter chains such as C12:0, volatility is more important and LOD and LOQ values increase considerably. Quantification of C12:0 requires an amount of at least 0.7 mM.

Accuracy
The results presented in Table 2 show that the recovery values are between 95%

Precision
Repeatability and reproducibility were investigated to evaluate the precision.
Results are expressed as relative standard deviation (RSD%). RSD values do not exceed 1.5% for repeatability and 4.25% for reproducibility (Table 2).

Robustness
The results of robustness are presented in Table 3. Recovery rates vary between 87% and 114% when changing the percentage of water in the mobile phase. They vary between 93% and 105% when changing column temperature and between 96% and 106% for flow rate variation. The method was found to be robust when column temperature and flow rate were altered but the percentage of water in the mobile phase is a critical parameter that should be carefully controlled. It can also be seen that there is no significant change in the resolution of the two critical peaks of C22:6 and C18:2 which stays above the critical value of 1.5.

Carry-Over
No signal is observed in the blank solutions after the injection of twice the amount of FFAs used for the highest points of the calibration curves. This confirms the absence of memory effect in the chromatographic runs.

Conclusions
The HPLC method described in this paper allows the rapid separation of 10 FFAs from 12 to 22 carbon chain length including unsaturated FFAs within 10.5 min without baseline drift. Easy to implement, this method was also shown to be robust, accurate and sensitive with low LOD and LOQ values. Actually, greater sensitivity can be obtained with other methods, however they require either time consuming pre-analytical steps such as a fluorescence derivatization, or much more expensive equipment than ELSD (Table 1).
Finally, compared to existing protocols, the proposed method seems to combine numerous advantages making it particularly suitable for determination and quantification of FFAs.