Enhanced lipid isomer separation in human plasma using reversed-phase UPLC with ion-mobility/high-resolution MS detection

An ultraperformance LC (UPLC) method for the separation of different lipid molecular species and lipid isomers using a stationary phase incorporating charged surface hybrid (CSH) technology is described. The resulting enhanced separation possibilities of the method are demonstrated using standards and human plasma extracts. Lipids were extracted from human plasma samples with lipid isomer separation plasma reversed-phase UPLC with ion-mobility/high-resolution MS re- versed-phase; extracted chromatogram. involve hundreds of samples with multiple replicate injections, and therefore, a shorter chromatographic run of 10 min was explored as an alternative to reduce the analysis time for isomeric separation. For this short method, all conditions were the same except the gradient. Initial gradient conditions started with 40% B, and immediately a linear gradient (curve 6) was started from 40% to 43% B in 1 min. In 0.1 min, mobile phase B was increased to 50% B. Over the next 4.9 min, the gradient was further ramped to 54% B, and the amount of mobile phase B was increased to 70% in 0.1 min. In the ﬁ nal part of the gradient, the % B was increased to 99% in 2.9 min. The eluent composition returned to the initial conditions in 0.1 min, and the column was equilibrated at the initial conditions for 0.9 min before the next injection.


Chemicals and materials
HPLC-grade methanol originated from Actu-All Chemicals BV (Oss, The Netherlands). HPLC-grade chloroform, ultra liquid chromatography (ULC)-MS-grade water, 2-propanol (IPA), acetonitrile (ACN), and 99% pure ULC-MS formic acid were purchased from Biosolve BV (Valkenswaard, The Netherlands). Ammonium formate, dichloromethane (DCM), leucine enkephalin (Leu-Enk), and poly-DL -alanine (product number P9003) were from Sigma Aldrich (St. Louis, MO). All lipid standards (as shown in supplementary Table I), and extracts of bovine heart, liver, and brain were purchased from Avanti Lipids (Alabaster, AL) and from Nu-Chek Prep (Elysian, MN). Human plasma with heparin as anticoagulant was obtained from healthy volunteers who had given their informed consent for use of their plasma for method development.

Preparation of lipid standards
A mixture of 66 lipids was prepared as shown in Table 1 . Stock solutions of 1 mg/ml were prepared in chloroform-methanol (2:1, v/v) and stored at Ϫ 20°C. Stock solutions were diluted prior to analysis in IPA-ACN-water (2:1:1, v/v/v).

Lipid extraction from plasma
Lipids were extracted from human plasma with a slightly adapted protocol from Bligh and Dyer ( 36 ). In brief, to 25 µl of plasma 100 µl of cold ( Ϫ 20°C) chloroform-methanol (2:1, v/v) was added. The sample was vortex mixed for 30 s at ambient temperatures and allowed to stand for 5 min at ambient temperatures. After vortex mixing for 30 s, the sample was centrifuged at 12,000 g for 5 min. The lower organic phase was transferred to a clean tube and evaporated to dryness under a gentle stream of nitrogen. Immediately prior to analysis, the sample was reconstituted in 25 µl chloroform-methanol (1:1, v/v) and diluted to 20 and 10 times the original volume of plasma in IPA-ACN-water (2:1:1, v/v/v) for positive and negative ion mode, respectively. Five microliters was injected into the UPLC system for positive ion mode and 10 µl for negative ion mode.

UPLC analysis
Separation of lipids was carried out on an Acquity™ UPLC system (Waters Corporation, Milford, MA). Mobile phase A consisted of 10 mM ammonium formate with 0.1% formic acid in water-ACN (40:60, v/v), and mobile phase B was 10 mM ammonium formate with 0.1% formic acid in ACN-IPA (10:90, v/v). As weak wash, ACN-water-IPA (30:30:40, v/v/v) was used, and the strong wash was a mixture of IPA-water-formic acid-DCM (92:5:2:1, v/v/v/v). Gradient elution was applied at a fl ow rate of 0.4 ml/min through a CSH C 18 column [100 × 2.1 mm inner diameter, particle size 1.7 µm (Waters Corporation, Milford, MA)] and thermostatted at 55°C. Initial conditions started with 40% B, and immediately a linear gradient (curve 6) was started from 40% to 43% B in 2 min. In the following 0.1 min, the percentage of mobile phase B was increased to 50%. Over the next 9.9 min, the gradient was further ramped up to 54% B, and the amount of mobile phase B was increased to 70% in 0.1 min. In the fi nal part of lipid isomers, separation needs to be executed. LC is a widely applied technique for separation of lipids, and additionally, it can be readily coupled to electrospray ionization MS. Focusing on only the triacylglycerols, nonaqueous reversed-phase (NARP) LC or silver-ion LC has been successfully applied (12)(13)(14)(15). However, this technique cannot be used for analysis of all lipid classes. Different chromatographic systems have been described for analysis of multiple lipid classes on the same platform. Normal-phase LC or hydrophilic interaction LC (HILIC) have been described to separate different lipid classes (16)(17)(18)(19). However, run times are often more than 30 min, which is not favorable for high-throughput analysis, and additionally, separation of one lipid class into different molecular species is not possible. A long run time is also a disadvantage for two-dimensional LC [HILIC combined with reversed-phase (RP) LC] ( 20 ). RP chromatography (RP-LC) has been widely applied for lipidomic analyses using different stationary phases [e.g., diphenyl ( 21 ), C 8 (22)(23)(24), and C 18 ( 23,(25)(26)(27)(28)(29)(30)(31)(32)(33)(34)]. For several years, our group has used high strenght silica (HSS) T3 columns (C 18 material) for our lipidomic analyses ( 33 ). Although we have extensive experience with this type of chromatography, and in separating both inter-and intralipid classes, optimal separation of different isomers has not been achieved.
Recently, Bird et al. ( 27 ) described the separation of cis / trans phospholipids, exploring different RP C 18 columns. An Ascentis Express C 18 (2.7 µm particles) column gave identical separations and peak shape for cis / trans isomers as a charged surface hybrid (CSH) C 18 column (1.7 µm particles). However, due to the lack of an ultrahigh-pressure LC system [ultraperformance LC (UPLC)], the CSH C 18 column was run at suboptimum conditions; hence, it was suggested to increase the fl ow rate to optimum conditions and thus improve peak shape and separation and shorten analysis time. The CSH material contains a low-level positive surface charge, in acidic mobile phases, in order to enhance the separation, in addition to increasing the loading capacity.
Herein, we demonstrate the use of CSH C 18 material for the separation of different lipid molecular species and different isomers within these classes in complex biological samples, in combination with ion-mobility spectrometry coupled with high-resolution MS.
Ion mobility is, besides chromatographic separation, adding an additional separation dimension located within the mass spectrometer and therefore enhancing structural elucidation of lipids. Using ion mobility, collisional crosssections (CCSs) of ions can be calculated, and besides accurate mass, fragmentation information, and retention time (Rt), these CCS values can be added to a searchable library for a routine work fl ow to increase the identifi cation confi dence. It was recently demonstrated that the CCS values are highly reproducible even using different machines ( 35 ).
The chromatographic separation in combination with the ion-mobility separation is of pivotal importance for structural elucidation of these lipid isomers. acquisition software to collect a reference scan every 10 s lasting 0.3 s. The LockSpray internal reference used for these experiments was Leu-Enk, and both were used as a lock mass as a lock CCS at a concentration of 0.1 mg/l in water-ACN (50:50, v/v) containing 0.1% formic acid. The reference calibrant was introduced into the lockmass sprayer at a constant fl ow rate of 10 µl/min using the fl uidics system of the mass spectrometer. A single point lock-mass calibration at m/z 556.2771 in positive ion mode and m/z 554.2615 in negative ion mode was used for the complete analysis.
Acquisition of the data were performed using MassLynx software version 4.1 SCN916 (Waters Corporation, Wilmslow, United Kingdom). The alignment of the low-and elevated-energy spectra and calculation of the CCS values using the mobility data were performed using UNIFI Research Edition, MS E dataviewer version 1.2 (both Waters Corporation, Wilmslow, United Kingdom) and Progenesis QI version 1.0 (Nonlinear Dynamics, Newcastle, United Kingdom). Lipid identifi cation of these time-aligned data was performed using LIPID MAPS online searching tools (http://www.lipidmaps.org) and Progenis QI version 1.0 (Nonlinear Dynamics).

Separation of different lipid subclasses
Our current method for routine lipidomic profi ling using RP-UPLC on an HSS T3 column shows separation of different lipid molecular species ( 33 ). However, multiple lipid classes are coeluting, and in biological samples, different lipid classes have completely different concentrations. If different lipid classes are coeluting, this can lead to ion suppression effects that obscure the detection of the low abundant lipids. Therefore, enhanced separation of lipid classes is very much advantageous.
A mixture of 66 lipids as shown in Table 1 was injected on the CSH C 18 UPLC column with the 20 min method. For the purpose of demonstrating the separation, all lipids, except the FFAs were analyzed in positive ion mode; however, it should be mentioned that multiple lipid classes [e.g., phosphatidylinositol (PI), phosphoethanolamine (PE), phosphatidic acid (PA), etc.] ionize much more effi ciently in negative ion mode. Fig. 1 shows the separation of this mixture, demonstrating separation of different lipid classes. Within each class, an additional carbon atom increases the Rt, while an additional double bond reduces the Rt. This is demonstrated in more detail subsequently .
To demonstrate the applicability in biological material, extracts of human plasma were injected in both positive and negative ion mode using HDMS E acquisition as shown in Fig. 2A , B , respectively, for the 20 min gradient and in Fig. 2C, D both positive and negative ion mode for the 10 min gradient. Both low-and elevated-energy spectra are available from a single injection, and therefore, it is possible to focus on specifi c lipid classes by searching for the class-specifi c fragments in the elevated-energy chromatogram. In the elevated-energy trace for positive ion mode ( Fig. 2A, C ), for example, the most abundant fragment ion is in many cases m/z 184.08, which corresponds to the phosphatidylcholine (PC) head group. All of the gradient, the % B was increased to 99% in 5.9 min. The eluent composition returned to the initial conditions in 0.1 min, and the column was equilibrated at the initial conditions for 1.9 min before the next injection, leading to a total run time of 20 min. Sample injections of 5 µl of both the lipid standard mixture and human plasma samples were carried out, and the autosampler temperature was set at 10°C. For both positive and negative ion mode, the same chromatographic conditions were used.
Lipidomics studies often involve hundreds of samples with multiple replicate injections, and therefore, a shorter chromatographic run of 10 min was explored as an alternative to reduce the analysis time for isomeric separation. For this short method, all conditions were the same except the gradient. Initial gradient conditions started with 40% B, and immediately a linear gradient (curve 6) was started from 40% to 43% B in 1 min. In 0.1 min, mobile phase B was increased to 50% B. Over the next 4.9 min, the gradient was further ramped to 54% B, and the amount of mobile phase B was increased to 70% in 0.1 min. In the fi nal part of the gradient, the % B was increased to 99% in 2.9 min. The eluent composition returned to the initial conditions in 0.1 min, and the column was equilibrated at the initial conditions for 0.9 min before the next injection.

MS
The UPLC system was coupled to a traveling wave ion-mobilityenabled hybrid quadrupole orthogonal acceleration time-offl ight mass spectrometer (SYNAPT G2-S HDMS, Waters Corporation, Wilmslow, United Kingdom). Detailed descriptions of this mass spectrometer can be found elsewhere ( 37,38 ). Electrospray positive and negative ionization modes were used. A capillary voltage and sampling cone voltage of (±) 0.6 kV and 30 V were used respectively for both polarities of electrospray ionization. The desolvation source conditions used nitrogen gas at 700 l/h with a constant desolvation temperature of 450°C. The source temperature was set at 120°C. Data were acquired over the m/z range of 50-1,200 Da. The mass spectrometer was operated in ion-mobility (HDMS E ) mode for acquisition in both polarities. During this acquisition method, the fi rst quadrupole Q1 was operated in a wide band radio frequency (RF) mode only, allowing all ions to enter the T-wave collision cell. The "trap" T-wave was operated at 4 V causing no fragmentation of the lipids. The intact lipid ions entered the helium cell region of the ion-mobility spectrometry (IMS) cell that was operated at 180 ml/min; the main function of the helium cell was to reduce the internal energy of ions and minimize fragmentation. The lipid ions then entered the IMS cell, held under 80 ml/min fl ow of nitrogen, to separate species according to their charge, mass, and CCS area. As the separated ions exited the IMS cell, they entered the "transfer" T-wave where two discrete and alternating acquisition functions were used. The fi rst function, typically set at 2 eV, collects lowenergy or unfragmented data, while the second function collects elevated-energy or fragment ion data, typically operated using a collision energy ramp from 30 to 55 eV. In both instances, argon gas is used for collision induced dissociation (CID).
The trap T-wave, IMS T-wave, and the transfer T-wave all carried different wave velocities; these were 314, 600, and 190 m/s, respectively. The Stepwave was operated at default settings with a wave velocity of 300 m/s and a wave height of 15.0 V.
Calibration of the ion-mobility cell for CCS calculations was performed using poly-DL -alanine at a concentration of 10 mg/L in water-ACN (50:50, v/v). The calibration procedure is performed automatically by the acquisition software using the fl uidics system of the mass spectrometer and based on the principle as described in detail in the recent paper from Paglia et al. ( 35 ).
The system was equipped with an integral LockSpray unit with its own reference sprayer that was controlled automatically by the   Table 1 ) on CSH C 18 acquired in positive ion mode in the 20 min gradient.
peaks with this specifi c fragment belong to the same class. In negative ion mode, lipids fragment differently, not giving class-specifi c fragments but giving direct acyl chain fragments (e.g., m/z 279.23 corresponds to an 18:2 chain, while m/z 281.25 corresponds to an 18:1 chain). The elucidation of the lipid acyl chains using the chromatographic method in combination with the ion-mobility function to separate coeluting species will be demonstrated subsequently.

Separation of isomers using CSH chromatography
In order to elucidate the structures of lipid isomers in an untargeted platform, it is of importance to separate the individual isomers as much as possible. Humans only synthesize cis (Z) FAs endogenously; however, due to dietary intake, synthetic trans (E) FAs are present in the human body as well, and it is well known that they play an important role in various metabolic diseases (39)(40)(41). Therefore, it is of great interest to separate cis / trans isomers in the chromatographic method. To test the ability of the UPLC and CSH technology to separate cis / trans isobaric lipids and to separate structural isomers, lipid standards were injected either alone or as a mixture. Using the elevated-energy trace, it appears that the fi rst and third peaks are indeed PC 18:1/18:1. Besides the same Rt as in the standards (7.74 and 8.49 min, respectively), the species also have a quite similar CCSs (302.8 and 305.9 A 2 , respectively) as obtained with the standards. As our method is not suited for double-bond localization, we cannot claim with full certainty that the peaks are PC 18:1 (9Z)/18:1 (9Z) and PC 18:1 (6Z)/18:1 (6Z), respectively. However, besides an Rt, the additional value of the mobility cell is that CCS values are obtained that give an additional confi rmation value. Besides the two PC 18:1/18:1 peaks, in the middle a third peak is visible (Rt 8.03 min). When using the elevated-energy trace, it appears that this peak is either PC 18:0/18:2 or PC 18:2/18:0, with CCS 301.8 A 2 .
As shown with the examples in Fig. 3 , the cis / trans isobaric isomers are baseline separated in chromatography, while the PG structural isomer PG 18:0/18:2 (9Z, 12Z) is baseline separated from the trans isomer and almost baseline separated from the cis isomer. Additionally, with the PC example, it is demonstrated that also structural isomers with a different double-bond location are fully baseline separated. This information would typically not be available using an infusion or traditional HPLC method. In comparison, this separation was not achieved using the HSS T3 chromatographic method ( 33 ) as will be demonstrated with examples in human plasma later in this manuscript.    Fig. 4 . Additionally, for TGs the separation of cis / trans isomers, doublebond positional isomers, and linear/branched isomers has also been described ( 13,20 ) using NARP, and as demonstrated in Fig. 3 , this is also possible for RP-LC for other lipid classes than TGs.
To demonstrate the applicability of this new chromatographic method in real biological samples and to compare it with our previous method using an HSS T3 column ( 33 ), we analyzed the same human plasma extracts on both platforms as shown in Fig. 5 . The XIC for PC 36:3 ([M+H] + m/z 784.5851) clearly shows that on the T3 column ( Fig. 5A ) there are at least two peaks visible that are not separated at all. Most likely, these peaks consist of multiple components as the peaks shape tends to indicate this. The same XIC on the CSH column in the 20 min method results in two near baseline separated peaks. The second peak has a shoulder that is also PC 36:3, and additionally, a fourth peak is visible. For the CSH column in the 10 min method, the separation and peak shapes are comparable with the 20 min method. This example demonstrates the superior separation of different isomers of the CSH method compared with the HSS T3 method.
For PE 36:2 ([M-H] Ϫ m/z 742.5392) using the HSS T3 column ( Fig. 5B ), two separate peaks are visible; however, the peak shape of the second peak already indicates that it consists of two components. On the CSH column with the 20 min method, it is clearly visible that three different isomers are present as peaks 2 and 3 are completely baseline separated, demonstrating again the superior separation of isomers that can be achieved with the CSH method compared with the T3 method. With the 10 min method on the CSH column, the separation of peaks 2 and 3 is slightly better than on the HSS T3 column but is defi nitely not comparable with the 20 min method. In the 20 min method, the amount of mobile   phase is increased from 50% to 54% from 2.1 to 12 min. On the 10 min method, this same increase in % B is from 1.1 to 6 min. Peaks that elute in the 20 min method relatively late in this part of the gradient appear not to elute in this part in the 10 min method. They elute in the next part (70-99% B from 6.1 to 9 min), and because of this steep gradient, a separation loss is observed. However, the 10 min method was developed as an alternative method for the analysis of hundreds of biological samples, and depending on the research question, it might be a useful method. We recommend using the 20 min method when an optimal separation of isomers is required. If, however, separation of isomers is not the major point of interest, but the separation of different lipid classes is of importance, then the 10 min method might be a good alternative for higher-throughput studies.

Robustness of the method
To test the robustness of the Rts on the CSH chromatography, human plasma extracts were injected in both positive and negative ion mode for both gradients. Samples were injected in each run and each condition for fi ve times, and this was repeated in six separate runs. Between the runs, the UPLC-MS system was used for different application. From each lipid class, a single lipid was chosen as an example of the entire class. If the lipids of one class elute over a wide Rt window, then a lipid eluting approximately halfway through this elution window was chosen.
Detailed results of this robustness experiment can be found in supplementary Tables I, II for the 20 and 10 min gradients, respectively.
To obtain these stable Rts from run to run, preparation of the mobile phase is of pivotal importance. We experienced that switching from 99% pure formic acid to 90% pure formic acid caused a shift in Rt of ‫ف‬ 3 min later (in the 20 min gradient). Additionally, fl ushing out the column with, for example, water-ACN (50:50, v/v) after a run and reconditioning of the column for the next run with mobile phase causes a slightly earlier elution time (maximum 0.15 min) for some lipids. Most likely, these effects are due to charge changes of the column material. However, this can be circumvented by storing the column in the mobile phase and not fl ushing with water-ACN (50:50, v/v). We have practical experience of leaving the column stored in the mobile phase for more than 5 months. After this period, the column can be used without any problem, back pressure and peak shape are identical to those obtained before, and additionally the Rts are also exactly the same as before. Besides using a high quality of formic acid and always leaving the column in the mobile phase, care should be taken with the preparation of especially mobile phase B. It is of critical importance that the ammonium formate is completely dissolved in the mixture of ACN-IPA (10:90, v/v), and this can be achieved by continuously stirring and heating the solvent during ‫ف‬ 1 h.
To test the applicability of the method in different matrices besides plasma, extracts of bovine liver, brain, and heart tissue were injected. Fig. 6 shows an extracted ion chromatogram of PC 34:2 using the 20 min chromatographic method. It can be clearly seen that the Rt of this lipid is the same in all three matrices demonstrating the applicability of the chromatographic method in various biological matrices.

APPLICATION OF THE METHOD: ELUCIDATION OF LIPID ISOMERS
The goal of developing the described powerful method for separation of different lipid classes and different isomers within these classes is to elucidate the structures of the individual lipids. With the Synapt G2-S system, when operated in HDMS E mode, a spectrum obtained with low collision energy and a spectrum with elevated collision energy are obtained from a single injection of the sample. The fragmentation of lipids using MS has been extensively described ( 33,(42)(43)(44)(45). From the fragment ion spectrum, the acyl chain information can be obtained. In positive ion mode, the most abundant fragment is often the head group. However, acyl chain-specifi c fragments are also visible and can be observed as M-SN1 and/or M-SN1-water, M-SN2 and/or M-SN2-water. In negative ion mode, the elucidation of the acyl chains is even more straightforward as the most abundant fragments are the  Only the fi rst peak at 15.41 min corresponds to the accurate mass of TG 54:6 (CCS 330.2 A 2 ). In Fig. 7B, C , the associated low and elevated spectra are shown, respectively, when the mobility function is not used. It is clearly visible that at this Rt not only TG 54:6 elutes but also other highly abundant lipids. As all ions enter the "transfer" region of the mass spectrometer, all the lipids with the same Rt are fragmented. The obtained spectrum is thus a mixture of all fragments derived from the different precursor ions, thus making structural elucidation a challenge, especially with this class of compounds where fragments of various precursor ions have a high degree of similarity. However, due to the use of the ion-mobility cell and exploiting its additional orthogonal separation technique, intact lipid ions are separated on basis of the CCS, mass, and charge. The fragment ions derived in the transfer cell can therefore be time aligned with the precursor lipid ions, not just by Rt but also using the ionmobility CCS, signifi cantly "cleaning up" the spectrum, showing only the fragment ions of that specifi c drift time. Fig. 7D + . Most likely, these adducts require higher energies to induce fragmentation than applied in this experiment. The use of the ion-mobility function allows this lipid structure to be elucidated as TG 18:2/18:2/18:2.
To demonstrate an example from another lipid class and additionally in negative ion mode, PI 36:1 as shown in Fig. 4 is elucidated. This peak has an Rt of 7.66 min, and when a spectrum is made of the low-energy trace without using mobility (supplementary Fig. IIIA), it is visible that other lipids elute at the same Rt. The elevated spectrum of all these lipids leads to a mixture of many different fragments (supplementary Fig. IIIB) making structural elucidation. However, if at this Rt only the peak with CCS 303.0 A 2 is selected by using the mobility information in the low-energy trace, only PI 36:1 is visible (supplementary Fig. IIIC), and as the elevated-energy spectrum (supplementary Fig. IIID) is time aligned, all obtained fragments are derived only from this parent making the elucidation of PI 18:0/18:1 or PI 18:1/18:0 possible. For PI lipids, the elevated-energy ramp of 30-55 V appears to be on the low side as quite some parent ion is left in the elevated-energy trace. However, all lipid-specifi c fragments are obtained, and thus acyl chain elucidation is possible.
With this method, it cannot be concluded which tail is actually SN1 and which one is SN2, and therefore, both PI 18:0/18:1 and PI 18:1/18:0 are possible.
Next to these two examples for both positive and negative ion mode, this feature is as well exploited for all other lipid subclasses and allows us to assign an identity to a large number of the lipid structures.

CONCLUSION
In this paper, the development and application of an UPLC method for the separation of lipid molecular species and lipid isomers using a CSH C 18 column has been described. We have coupled the chromatographic system to a traveling wave ion-mobility-enabled hybrid quadrupole orthogonal acceleration time-of-fl ight mass spectrometer. As lipid isomers are chromatographically separated, structural elucidation of the lipids can be obtained. The additional separation in the ion-mobility part of the mass spectrometer is of pivotal importance for this process as was demonstrated by the structural elucidation of the TG and PI species. Additionally, a with the associated lowenergy (B) and elevated-energy (C) spectra without mobility data used for the peak at 15.41 min (± 0.08 min) and the same low-energy (D) and elevated-energy (E) spectra using the mobility data making structural elucidation possible.
CCS value is available that gives an additional confi rmation for the compound. The improved peak capacity and separation effi ciency afforded by the chromatography and ion-mobility are complementary, and without one or the other, structural elucidation is a complex process. The method described here opens up future opportunities for the structural elucidation of many different lipids in biological samples. Future work will focus on the creation of a lipid database including Rts, acyl chain information, and CCS to be implemented for routine lipidomic screening.