Analysis of linoleoyl and oleoyl macrogolglycerides by high performance liquid chromatography coupled to the atmospheric pressure photoionization mass spectrometry

– Linoleoyl macrogolglyceride (LM) and oleoyl macrogolglyceride (OM) are pharmaceutical ingredients, obtained from corn and apricot kernel oils respectively. This study aims to know the detailed chemical composition of LM and OM, in order to understand their roles in pharmaceutical formulations. These two products were analyzed by non-aqueous reversed phase high-performance liquid chromatography (NARP-HPLC), using Vintage Series KR C18 column (250 (cid:1) 4.6mm, 5 m m) and non-aqueous acetonitrile/ acetone mixture as mobile phase. The ionization source used was atmospheric pressure photoionization (APPI) and the analyzer was LTQ-Orbitrap ® (hybrid analyzer: double linear ion trap coupled to a Fourier transform orbital trap). LM and OM consist of complex mixtures, constituted of mono-(MG), di-(DG) and triglycerides (TG) and mono-(MPEGE) and di PEG-6 esters (DPEGE) of linoleic acid (18:2) for LM and of oleic acid (18:1) for OM. NARP-HPLC-APPI method allowed the separation and the identi ﬁ cation of the glyceride classes (MGs, DGs and TGs) and the PEG esters of different chain lengths (PEG-chain lengths of fatty acid moieties and number of units of ethylene oxide), at the same time and in one single run, for both products LM and OM. The comparative study between LM and OM showed that, a higher presence of linoleic esters for LM, and a higher presence of oleic esters for OM.


Introduction
Vegetable oils are products used for food and therapeutic purposes. They are used as raw material in the pharmaceutical, cosmetic, detergent industries... as active ingredient or excipient in the crude or transformed state. The transformation of vegetable oils allows to obtain compounds with improved physicochemical and/or organoleptic characteristics for examples: stability during storage, odor, color... (Soumanou et al., 2005).
These two ingredients could exhibit some differences of behavior during the manufacturing of complex pharmaceutical formulations such as nanocarriers. One hypothesis is that the physical stability of these formulations could be affected by prevailing fine chemical composition of these two ingredients.
The purpose of this paper was to develop a separative analytical method using an optimized gradient of elution that will allow the separation of each compound within glycerides subclasses and PEG-esters of LM and OM, to be able to establish a detailed identification of the composition of each one.
Non-aqueous reversed-phase high performance liquid chromatography (NARP-HPLC) is very used for mono-, di-, and triglycerides analysis (Mottram and Evershed, 1996;López-López et al., 2001;Holčapek et al., 2003Holčapek et al., , 2005Salghi et al., 2014). Especially due to the very low volatility and insolubility in aqueous solutions of triglycerides (TGs), NARP-HPLC is the most suitable technique for their analysis (Maloumbi et al., 2015). The C18 grafted columns are the most suitable for the analysis of TGs with fatty acids in the range C16 to 22. A stationary phase with chain lengths similar to those of TG acyl chains maximizes the interactions and offers the greatest efficiency (Rezanka et al., 2017). Regarding the mobile phase, acetonitrile/acetone or acetonitrile/isopropanol mixtures are commonly used.
The glycerides do not have chromophores, therefore, UV detection is rarely used. However, the glycerides absorb at low wavelengths (200-210 nm), but some solvents also absorb at these wavelengths. For this reason, the universal detectors, evaporative light-scattering detector (ELSD) and charged aerosol detector (CAD) are commonly used for the detection.
Tamba Sompila et al. (2014) used NARP-HPLC for the analysis of TGs, they demonstrated that the use of the retention laws allows to identify TGs of equatorial African oils (Tamba Sompila et al., 2014). Nevertheless, their approach requires a first step to determine the fatty acid composition of the oils, and to possess standard compounds. However, HPLC coupling with mass spectrometry (MS) is generally used. HPLC-MS is a powerful tool in lipid analysis; it provides structural information with great sensitivity. Atmospheric pressure chemical ionization (APCI) is the most frequently used ionization technique for glycerides analysis. This technique allows easy coupling to NARP-HPLC and high ionization efficiency for non-polar molecules (Jakab et al., 2002;Holčapek et al., 2003;Castilho et al., 2004;Fasciotti and Pereira Netto, 2010;Rezanka et al., 2010). Atmospheric pressure photoionization (APPI) and electrospray ionization (ESI) were also used for the analysis of vegetable oils (Gómez-Ariza et al., 2006). Gómez-Ariza et al. (2006) showed that APPI and ESI sources are complementary modes. APPI allows more sensitive detection of MG and DG fragment ions than ESI and ESI is more suitable for the detection of TGs. However, Imbert et al. (2012) demonstrated that APPI has a higher ionization capacity than APCI and ESI for lipids.
As for the characterization of PEG and PEG esters the methods used are:for PEG, matrix-assisted laser desorption ionization-Fourier transform mass spectrometry (MALDI-FTMS) (Pastor and Wilkins, 1998) and matrix-assisted laser desorption ionization-time of flight mass spectrometry (MALDI-TOF) (Jackson et al., 1997) -, for PEG esters, low energy collision-induced dissociation in ESI mass spectrometry (ESI-CID) with direct injection into the system (Chen et al., 2002). Chen et al. (2002) showed that low energy dissociation of PEG esters strongly dependent on the type of cation used for ionization. The structural information on the polymer chain and end groups is best obtained using Li þ , Ag þ , and other transition metal ions (Chen et al., 2002).
In the present work, we developed an analytical method by reversed phase high performance liquid chromatography for the simultaneous separation of all chemical entities within glycerides sub-classes and PEG esters of LM and OM. First, NARP-HPLC-CAD was performed using LM to optimize the experimental chromatographic conditions to reach the best selectivity. Afterwards, the developed method was applied to OM. Then, a mixture of standards molecules of MG, DG and TG was injected to check the reliability of the HPLC method. Second, NARP-HPLC-APPI-LTQ-Orbitrap was performed to allow the identification of each constituent. The same mixture of standards molecules was injected to determine the behavior fragmentation of MG, DG and TG. Then, the identification of main peaks of LM and OM was established. Finally, a comparative study between LM and OM was conducted.

Chromatographic method
HPLC instrumentation and the charged aerosol detector were from Thermo Fisher Scientific (Bremen, Germany). The nebulization temperature was set at 35°C and nitrogen pressure at 5 bars.
Vintage Series KR C18 column (250 Â 4.6 mm, 5 mm) was from Interchim (Montluçon, France). The column was thermostated at 25°C. The mobile phase was a mixture of acetonitrile and acetone. A mixture 97/3, acetonitrile/ acetone, v/v, was applied in the first ten minutes, an automatic gradient was applied from 10 min to 100 min until reaching 60/40 of acetonitrile/acetone, v/v, then maintained at 60/40 from 100 to 200 min. The flow rate was set at 1 mL/ min with automatic injection of 20 mL. All samples were diluted at 1/100 in a mixture of acetonitrile and dichloromethane 50/50, v/v.
In addition, to compare between LM and OM, two batches were analyzed, with three injections for each batch. In each sample the same chemical internal standard (ISTD) was added, N-Lauroyl-D-sphingosine at 1 mg/mL.

Mass spectrometry method
HPLC is coupled with LTQ-Orbitrap velos Pro mass spectrometer from Thermo Fisher Scientific (Bremen, Germany). The signal was acquired with Xcalibur software from Thermo Fisher Scientific. The spectrometer is a hybrid device incorporating two analyzers, a double linear ion trap (at high and low pressure) and a Fourier Transform orbital trap. The ionization source used was atmospheric pressure photoionization (APPI), using the positive-ion mode. Vaporizer temperature of the probe was set at 350°C. Sheath gas, auxiliary gas, and sweep gas flow rates were set at 40, 20, and 0 (arbitrary unit) respectively. Capillary temperature was set at 325°C and S-lens RF level at 60%. The data was acquired in the mass range m/z 150.00-1100.

Processing data
Raw data file was imported in MZmine software (MZmine 2, 2010). Displaying the total ion chromatogram (TIC) permitted to take note of the baseline level and the height of the smallest significant peak. The peak detection is a threestep process: mass detection, chromatogram building and peak deconvolution. For mass detection, we used centroid mode and noise level was set to10 5 and MS level to 1. For chromatogram building, selected minimum time span was 1 min and minimum height 10 6 with a m/z tolerance of 0.001 m/z or 5 ppm. Then, a deconvolution was conducted using the "Local minimum search" algorithm with a chromatographic threshold of 1%, a search minimum in retention time (RT) range of 1 min, a minimum relative height of 5%, a minimum absolute height of 10 6 , a minimum ratio of peak top/edge of 20 and a peak duration ranging from 0 to 5 min. After peak deconvolution, MZmine produces a resolved peak list where the most intense isotope was considered. Peak alignment is done with join aligner method with m/z tolerance set to 0.001, weight for m/z set to 20, and the maximum allowed relative distance between two retention time values was 5%, and score for perfectly matching RT values was 10. After, gap filling was applied using the option same RT and m/z range with m/z tolerance of 0.001 m/z or 5 ppm. As we chose to add a chemical standard to all samples, we normalized our data on peak area using the weighted contribution of standard compound with a m/z vs. RT balance of 3. Our chemical internal standard of exact mass 481.449 appears at m/z: 464.443-464.448

Data analysis: Descriptive statistics and multivariate analysis
From the matrix obtained after data processing by MZmine software, a new matrix was constructed, by the calculating the mean and standard deviation of m/z corresponding to each RT, for LM and OM.
In addition, principal component analysis (PCA) and orthogonal partial least squarediscriminant analysis (OPLS-DA) were performed using Simca Pþ (version 12.0.1.0, Umetrics). The analysis was carried out on 12 samples and 1327 variables (variables represent the couples RT and m/z).
Prior to analysis, our data set was centered and normalized. The variation of each variable was centered so that the mean of each column is equal to 0 and scaled to unit variance (1/SD). Principal component analysis (PCA) is a multivariate technique that analyses a data table in which observations (here different Labrafils) are described by several variables (RT and m/z). The goal is to extract the important information from the data matrix and to find a set of new orthogonal axes called principal components. However, PCA remains a descriptive chemometric method. Therefore, OPLS-DA is used to highlight the discriminant variables explaining the difference between the groups.

Optimization of chromatographic conditions
The optimization of chromatographic conditions was performed with LM. It was not performed with a mixture of standards as it was not possible to have all the standards because the richness of the chemical composition of LM and OM.
The gas chromatography was not considered because of the presence of the polymers (MPEGEs and DPEGEs). Nonaqueous phase mobile was used because LM and OM are not soluble in water.
Due to the low volatility, the presence of polymers and the no water solubility of LM and OM, liquid chromatography and non-aqueous phase mobile were selected for the analysis of the two ingredients. The initial operating conditions were based on the method of the TGs separation developed by Tamba Sompila et al. (2014). The first run was 60 min with an isocratic system 70/30, acetonitrile/acetone, v/v, using vintage Series KR C18 column (250 Â 4.6 mm, 5 mm) and charged aerosol detector In NARP-HPLC, acetonitrile is the weak solvent. It allows better resolutions and facilitates the separation of unsaturated glycerides through its interactions with electrons p of double bonds. Acetone (or isopropanol) is the strong solvent, added in a smaller amount than the weak solvent, it allows to modulate the analysis time and the polarity of the mobile phase (Maloumbi et al., 2015;Rezanka et al., 2017).
The isocratic system acetonitrile/acetone 70/30 v/v was not efficient for our purpose as PEG esters overlapped with MGs between 0 to 20 min.
A first experimental plan was established to optimize the separation of the peaks in isocratic mode, increasing the percentage of the weak solvent: a pitch of 5% of acetonitrile was considered (i.e. the mobile phase mixtures tested were: 75/ 25, 80/20, 85/15, 90/10, 95/5, acetonitrile/acetone, v/v). Indeed, acetonitrile, in larger quantities, allowed a good resolution of less apolar products but delayed the release of highly apolar products.
Then, a gradient mode was considered to improve the analysis time and elute all constituents. Several combinations of the mobile phase and slope of eluent strength were tested. Finally, the best gradient profile was the following, for the first 10 minutes, an isocratic step acetonitrile/acetone 97/3 v/v was set. Following by an automatic gradient applied from 10 min to 100 min, until reaching acetonitrile/acetone 60/40 v/v. Then another isocratic step was applied from 100 min to 200 min, with this last mobile phase composition acetonitrile/acetone 60/40 v/v.
As LM and OM as both mixtures of glycerides and PEGesters, this method had be applied for LM as well for OM. The major peaks for LM and OM were separated under these improved experimental conditions (Fig. 2).
3.2 Reliability of the developed method: relationships chemical structure/chromatographic behavior To highlight the relationships chemical structure/chromatographic behavior of glycerides, the same HPLC conditions retained for the analysis of LM and OM were used to analyze a mixture of standard molecules. First, MG (18:1), 1,2-DG (18:1/18:1), 1,3-DG (18:2/18:2), TG (18:2/ 18:2/18:2) standards were injected separately to make sure of the retention time (RT) of each one. Second, a mixture of all the standards was injected, the obtained chromatogram is shown in Figure 3. MG was eluted first followed by DGs then TGs. RT increased with increasing molecular weight. For molecules of near molecular weight, as DG (18:1/8:1) and DG (18:2/18:2), the molecule with more unsaturated bonds would have a shorter retention time.
Comprehensively, the retention increased with increasing of the total carbon number (CN). For the same CN, the retention decreased according to the number of double bonds (DB). It is known that NARP-HPLC mainly separates the different glycerides according to their equivalent carbon number (ECN), which is calculated as following, ECN = CN À (2 Â DB) (Holčapek et al., 1999;Lísa et al., 2007Lísa et al., , 2011Rezanka et al., 2017).
To highlight the relationships chemical structure/chromatographic behavior of PEG-esters, the NARP-HPLC-APPI-MS was applied to LM and OM as no standards of pure PEGesters were available. Using mass spectrometry (see Sect. 3.3.2), it has been possible to determine the elution order of the polymers. Mainly it depends first on the number of esters groups (mono-PEG-esters eluted before di-PEG-esters), second the number of double bonds (DB) on the fatty acid moieties (monounsaturated compounds eluted after diunsaturated compounds). Thirdly, MPEGEs with various number of ethyleneglycol units were eluted under the same chromatographic peak, while for the DPEGEs, the retention decreased according to the number of ethyleneglycol units.

Characterization of mono-, di-and triglycerides
APPI is recently introduced technique, where a highly energetic photon is used to effect generation of charged ion species. In comparison to APCI, APPI offered lower detection limits, highest sensitive and the same fragmentation pattern for mono, di and triacylglycerols (Holčapek et al., 1999;Gómez-Ariza et al., 2006;Rezanka et al., 2010).
The position of the substitution of fatty acids in TGs was determined from fragmentation of TG, considering the diglyceride fragments m/z. It has been demonstrated that the ratio of the abundance of these fragment ions could be used to identify the sn-2 and sn-1 (or sn-3) positions (Mottram and Evershed, 1996). The relative abundance of [M þ H -RCOOH] þ ion fragment corresponding to loss of fatty acid from sn-2 position is less than the relative abundance of [M þ H -RCOOH] þ ion fragment due to loss fatty acid from sn-1 (or sn-3) (Mottram and Evershed, 1996).

APPI-MS ionization and fragmentation profiles of PEG esters are highlighted as follows:
For each mono-PEG ester, the protonated molecular ion [M þ H] þ and the fragment ion [M -(nÀ1)(EO) þ H -H 2 O] þ were observed. The MPEGEs mass distribution was exhibited as adjacent protonated molecular ions [M þ H] þ separated by 44.026 Da, consistent to the mass of the repeat unit, ethylene oxide (EO), as observed with ESI mass spectrometry (Chen et al., 2002).

Identification of constituents of linoleoyl and oleoyl macrogolglycerides
MGs, DGs, TGs, MPEGEs and DPEGEs were identified for LM and OM as described above, exploiting the retention time, protonated molecular ion and typical fragment ions for each molecule.
The chromatograms of LM and OM can be divided into three segments, the first one being from 0 to 10 min, the second from 30 min to 70 min and the third from 115 min to 150 min. For both products, the first package represents MGs and MPEGEs, the second DGs and MPEGEs and the third TGs. Retention times were obtained after alignment and normalization of all peaks of LM and OM chromatograms (for more details see Supplementary data 1, 2 and 3).

Comparative study between linoleoyl and oleoyl macrogolglycerides
MG, DGs and TG injected at the same concentration (1 mg/ mL), had different response factors (R (i)): R (MG) < R (DG) < R (TG). The response factors of DG (18:1/18:1) and DG (18:2/18:2) had the same order. (Tab. 1) Therefore, in the samples of LM and OM, the relative amounts of the different lipid classes (MGs, DGs and TGs) cannot be compared. The comparison can only be done between the lipid sub-classes, within MGs or DGs or TGs.
The peak areas of LM and OM were normalized, to the peak area (ion fragment [M þ H -H2O] þ ) of ISTD (see the Sect. 2.4.1). From the matrix obtained after data processing, a new matrix was constructed, by calculating the mean and standard deviation of relative intensities of m/z corresponding to each RT, for LM and OM. The comparison of the two ingredients was performed on the majority ion (m/z) of each constituent (Fig. 7, Supplementary data 4) The Figure 7 shows that LM and OM have the same qualitative composition. The difference between the two ingredients is the relative amounts of each compound within each sub-classes of glycerides and PEG-esters.

PCA and OPLS-DA results
PCA was carried out for the two macrogolglycerides. As mentioned in the Section 2.2, for each macrogolglyceride, two batches were analyzed, with three injections for each batch. The first two principal components accounted for 66% of the variance. The 1st principal component PC1 explained 53% of variance and the 2nd principal component PC2 13% of variance.
The projection of samples (score plot) on the 1st and 2nd components is presented in Figure 8. 1st component represents the chemical composition, the 2nd component represents the batch. PC1 separated the two macrogolglycerides into two distinct groups and PC2 separated batch 1 and batch 2 for both LM and OM. These two pharmaceutical ingredients were obtained, from corn oil for LM and from apricot kernel oil for OM, by a manufacturing process standardized by the supplier.
As the raw materials are of vegetable-origin, then, there is a variability of the composition of corn and apricot kernel oils. Therefore, the batches of finished goods can have a relative variability of their compositions. This explains the separation, in the score plot, of the two batches analyzed for LM and OM.
For PEG esters, MPEGEs were eluted before DPEGEs. Then, PEGEs were eluted by decreasing the number of double bonds (DB) on the fatty acid moieties (monounsaturated Fig. 7. Comparative profile of LM and OM: relative intensities of majority ion m/z (mean and standard deviation) for each compound. A: corresponds to monoglycerides, diglycerides and triglycerides. B: corresponds to mono-and di-PEG esters. The relative intensities of m/z are the means of the normalized peak areas of the two batches of LM and OM injected in triplicate. compounds eluted after diunsaturated compounds). Regarding the mass distribution of PEG esters, it was detected under the same peak for MPEGEs (18:1), as well as for MPEGEs (18:2). This highlights that the number of ethylene glycol units did not impact the retention of MPEGEs. For Di-PEG esters, the presence of the two fatty acid moieties allowed a better separation compared to MPEGEs and the compounds of this sub-class were eluted in order of decreasing of number of units of ethylene oxide.
It can be concluded that NARP-HPLC combined with APPI-MS is a suitable tool to characterize the two macrogolglycerides. HPLC-APPI allowed us to obtain the detailed chemical composition of LM and OM. These two pharmaceutical ingredients have the same qualitative composition but differ by the relative amounts of each compound, depending on the analyzed macrogolglyceride.
As expected, the data analysis (descriptive statistics and multivariate analysis) of both the compositions, showed a  The European Pharmacopoeia monographs for the quality control of these ingredients (OM and LM) propose to access the overall fatty acid profile after hydrolysis of all the esters. The purpose of this global approach was to characterize the pharmaceutical raw material through its main fatty acid i.e. oleic acid (C18:1) for OM and linoleic acid (C18:2) for LM. Nevertheless, it remains very interesting to have a more detailed molecular approach of OM and LM for a best understanding of their behavior during the manufacturing process of some dosage forms (such as nanocarriers). Indeed, their differences in the molecular distribution of OM and LM could lead to various functionality-related characteristics of these ingredients.