Structural identification of triacylglycerol isomers using electron impact excitation of ions from organics (EIEIO)

Abstract

Electron-induced dissociation or electron impact excitation of ions from organics (EIEIO) was applied to triacylglycerols (TAGs) for in-depth molecular structure analysis using MS. In EIEIO, energetic electrons (∼10 eV) fragmented TAG ions to allow for regioisomeric assignment of identified acyl groups at the sn-2 or sn-1/3 positions of the glycerol backbone. In addition, carbon-carbon double bond locations within the acyl chains could also be assigned by EIEIO. Beyond the analysis of lipid standards, this technique was applied to edible oils and natural lipid extracts to demonstrate the power of this method to provide in-depth structural elucidation of TAG molecular species.

Triacylglycerols (TAGs) are among the most abundant lipids in the human body, located primarily within adipose. TAGs are also found in plants and are concentrated in seeds as a source of energy during embryonic development. It is from these various seeds that a majority of edible oils are derived. The physiological role of TAGs is to serve as a reserve of fatty acids for energy generation via β-oxidation. However, these same stored fatty acids may also serve as precursors, either directly or indirectly, to bioactive oxidized metabolites such as eicosanoids that play a key role in mediating the cell-signaling pathways that control inflammation (1). High levels of TAG in plasma have been implicated in a host of pathophysiologies, including arteriosclerosis, nonalcoholic fatty liver disease, and metabolic syndrome (2).

Understanding the exact roles of individual TAG molecules in these diseases is not clear, due mainly to the difficulty in characterizing these TAGs at the molecular level. For example, while current MS technology can account for the total number of carbon atoms in a TAG, as well as the approximate acyl chain composition (3), other critical information such as the exact regioisomerism of fatty acids along the glycerol backbone, the position(s) of the acyl chain double bonds, the stereochemistry of the double bonds, and so forth, cannot currently be obtained using conventional bioanalytical techniques. For example, a common method of TAG analysis is the chromatographic detection of FFAs or their derived methyl or ethyl esters, which are obtained by hydrolysis from TAGs (4). Although convenient, this method does not allow for consideration of the structural aspects of the intact TAG molecular species (i.e., regioisomeric structural information). This information is critical to understand the dynamic functions of TAGs, their digestion, and their metabolism. A conventional TAG characterization method in MS is based on low-energy (5, 6) or high-energy collision-induced dissociation (CID) (7, 8). These methods provide TAG structural information, including acyl chain lengths and double bond number; however, the double bond positions and the acyl chain regioisomeric arrangement are not easily defined due to the limited information derived from the resultant fragments. HPLC has also been used to resolve regioisomers of TAGs (9), but it requires long analysis times, authentic standards as references, and extensive method development, potentially making this method unattractive to a high-throughput lab. A recent study of TAG regioisomers was reported using differential mobility spectrometry (DMS) without HPLC (10). Despite advances in the goal to characterize lipid molecular species, including using ozone-induced dissociation (OzID) (11, 12) or a Paternò-Büchi reaction (13, 14) to determine double bond positions, an efficient method to characterize lipid structural details has been lacking. Recently, a study using OzID found that TAG molecular species containing C18:1, n-7 versus C18:1, n-9 correlated with clinical variables of dyslipidemia and are proinflammatory (15). These results highlight the need for a method to fully characterize TAG molecular species in order to relate TAG structural details with human disease.

In order to acquire more in-depth structural information for lipids, we recently developed a near-complete structural analysis using electron-induced dissociation (16) or electron impact excitation of ions from organics (EIEIO) (17) for phosphatidylcholines (PCs) (18) and SMs (19). Singly protonated precursor lipid ions generated by an ESI were introduced into a branched ion trap (20) wherein they were reacted with a 10 eV electron beam. This produced information-rich fragment ions that revealed lipid class, respective acyl chain lengths, the number and locations of double bonds in the fatty acid chains, and regioisomer specificity. Here, we applied EIEIO to reveal key structural details, at the molecular species level of TAGs. We identified and developed an optimized analytical strategy in conjunction with diagnostic rules to analyze TAGs by EIEIO. This resulted in the complete characterization of the TAG lipidome in undefined TAG mixtures without the need for authentic standards.

METHODS

Materials

TAGs were analyzed from solutions prepared using a mixture of HPLC-grade dichloromethane (DCM)-methanol (50:50, v/v) with 0.2 mM sodium acetate or 10 mM ammonium acetate. These solvents were purchased from Caledon Laboratory Chemicals (Georgetown, Ontario, Canada), while the salts were purchased from Sigma-Aldrich Canada Co. (Oakville, Ontario, Canada).

Synthesized TAG standards, including TAG 16:1/18:1(n-9Z)/16:1 and TAG 16:1/16:1/18:1(n-9Z), were obtained from Larodan Fine Chemicals AB (Malmo, Sweden). Another group of TAG isomers differing only in their carbon-carbon double bond positions, TAG 18:1(n-12Z)/18:1(n-9Z)/18:1(n-12Z), was obtained from Avanti Polar Lipids Inc. (Alabaster, AL). Finally, TAG standards containing polyunsaturated acyl chains, including tridocosahexaenoyl glycerol [TAG(22:6(n-3Z,n-6Z,n-9Z,n-12Z,n-15Z,n-18Z), or DHA-TAG], tri-α-linolenoyl glycerol [TAG(18:3(n-3Z,-n6Z,n-9Z), or ALA-TAG], and tri-γ-linolenoyl glycerol [TAG(18:3(n-6Z,n-9Z,n-12Z), or GLA-TAG] were purchased from Nu-Chek Prep Inc. (Elysian, MN). All of the TAG standards were used without further purification. The working solutions of these standard samples were prepared at 1 µg/ml in the standard working solvent.

Tri-linolenoyl glycerol (CLA-TAG) was synthesized in-house followed by a method described by Medina et al. (21). Conjugated linolenic acids, CLA(9E,11E), CLA(9Z,11E), and CLA(10E,12Z), suspended in ethanol were purchased from Cayman Chemical (Ann Arbor, MI). Immobilized lipase from Candida antarctica and molecular sieves, 4 Å, were purchased from Sigma Aldrich. Glycerol and HPLC-grade hexane were obtained from Fisher Scientific Co. (Ottawa, Ontario, Canada) and Caledon Laboratory Chemicals, respectively. Ethanol was evaporated in nitrogen gas flow from each CLA to obtain FFA of 40 mg. Each CLA were diluted in hexane (900 µl) with glycerol (4 µl), beads of the lipase (10 mg), and the molecular sieves (100 mg) in a 5 ml glass vial with a Teflon cap. The mixture was incubated for 20 h using a shaking bath (300 rpm) at 50°C. The beads were removed from the solution after centrifugation, then the liquid phase was evaporated by nitrogen gas flow. The residue was suspended in chloroform as stock (nominal concentration ∼10 mg/ml), from which a stock solution was prepared using standard working solvent to a concentration of ∼10 µg/ml.

Analytical standards of edible oils, including olive oil, coconut oil, and linseed oil, were purchased from Sigma-Aldrich. Certified ultrapremium extra virgin olive oil, Cobrançosa (Portugal) harvested in 2014, and extra virgin olive oil, Solon (Greece), were purchased from a specialty olive oil store and a grocery store in Toronto, respectively. These oils were diluted in the working solvent to a concentration of ∼100 µg/ml total lipid without purification.

Avocado oil and fish oil were extracted in-house from a fresh avocado fruit and wild sock-eye salmon fillet, respectively. A ripe avocado and a filet of the salmon were purchased from grocery stores in Toronto. Deionized water (975 µl), methanol (2,000 µl), and DCM (900 µl) were added to individual samples of avocado and salmon (both ∼0.5 g). Each mixture was shaken well for 1 min in a glass vial with a Teflon cap. Liquid phase was moved to another vial to remove solid. Additional deionized water (1,000 µl) and DCM (900 µl) were added to the liquid and gently mixed. After the mixture was centrifuged for 10 min (1,200 rpm), the DCM phase (bottom layer) was collected by a pipette and evaporated under nitrogen. This dual phase purification process was repeated to each residue. The amount of the extracted oil was determined from the increase of the mass of each vial with the evaporated residue compared with the empty vials. The washed residue was suspended in chloroform as stock. These stocks were diluted in the standard working solvent with concentration of 100 µg/ml.

Instrumentation

We used a TOF mass spectrometer system equipped an electron-ion reaction device (ExD cell) (20) for this work, which was the same instrument as in our previous report on SM analysis (19).

The detail of ionization method on lipid analysis using a Turbo V^TM ion source (Sciex) was described previously (18). Infused flow rate of working solution in this work was 0.3 ml/h, and ESI voltage was +5,500 V. The detail of DMS (SelexION^® Technology, Sciex) for lipid application was described before (19, 22). Separation voltage (SV) of 3,900 V peak to peak, modifier gas of 2-propanol, and DMS temperature of 200°C were continuously applied. Counter nitrogen gas flow for resolution enhancement (DR) (0–40 psi) and compensation voltage (CoV) were optimized for each measurement.

The previously reported ExD device (20) was used in EIEIO condition (i.e., electron kinetic energy at 10 eV, typically). Quasi-flow-through mode or simultaneous electron ion injection mode (20) for 150 ms reaction time was used in this work as the optimized product yield condition (19). We used helium as cooling gas in the ExD and the Q2 collision cell. The TOF mass analyzers were operated in high-sensitivity mode (23). Accumulation time was given between 1 and 5 min per single precursor m/z in this work. Roughly separated groups of ions by DMS were further separated by m/z by the Q1 mass filter placed between the ion source and the ExD device. Mass window of the m/z separation was ∼1 m/z unit.

Nomenclature

For the identification of the TAG species that we analyzed in this work, we have adopted a modified nomenclature first suggested by Marshall et al. (24) and based on commonly accepted lipidomics terminology (25). Briefly, we identify TAGs with known sn positions of given acyl chains using slashes (e.g., TAG 16:0/18:0/18:1 implies known regioisomerism at all sn positions); however, where there is a question as to the sn-1 and sn-3 substitution, we use the underscore terminology (e.g., TAG 18:1_/18:0/_16:0 refers to a TAG with a known sn-2 substitution, but interchangeable sn-1 and sn-3 substitution). As Marshall et al. (24) state, we identify synthesized TAG standards using TAG A/B/C nomenclature given the known composition of such species.

RESULTS AND DISCUSSION

EIEIO provides characteristic fragments for TAGs

In previous studies (18, 19), we have demonstrated the utility of EIEIO for providing detailed structural information for different types of polar lipids (i.e., bearing polar head groups). However, TAGs are somewhat different that these lipids, as they are nonpolar, bearing no easily ionizable moiety themselves. For example, in order to form TAG ions for MS analysis, all lipidomics work flows must rely on adduct formation with some added salt, such as sodium or ammonium cations. Despite this difference, TAG species fragmented easily under EIEIO conditions (Fig. 1A–C), just like PCs and SMs measured previously. Sodiated TAGs were more promising EIEIO targets than ammoniated species because of the simplicity of dual chain loss ions of the sodiated species. In contrast, the ammoniated TAGs tended to fragment primarily at the NH₄⁺ group, resulting in uninformative chain loss (supplemental Fig. S1). Hence, sodiated TAG ions were used throughout the remainder of this EIEIO study.

Fig. 1.

EIEIO spectra of sodiated standard mixed acyl TAGs. A: TAG 16:0/16:0/18:1(n-9Z). B: TAG 16:0/18:1(n-9Z)/16:0. C: TAG 18:1(n-12Z)/18:1(n-9Z)/18:1(n-12Z). Dual acyl chain loss fragments, which work as TAG diagnostic peaks, are annotated by color. Cleavage site of the diagnostic peaks are shown in sn-1 doublet (blue) (D), sn-2 singlet (red) (E), and sn-3 doublet (green) (F) as bonds that intersect the arrowed lines.

When subjected to EIEIO, sodiated TAGs undergo dual acyl chain loss (fragments appear between m/z 300 and 350) and single acyl chain loss (fragments appear between m/z 500 and 650) as major product ion channels. Also appearing in the EIEIO mass spectrum, in the m/z region between the intact TAG precursor ion and the single acyl chain loss product ion, one can find a complete series of acyl chain fragments. Similar to our previous studies (18, 19), each carbon-carbon single bond cleavage in the acyl chains was represented by two ions: an even-electron (nonradical) product ion and its matched radical ion pair (bearing one fewer hydrogen atom than its partner). The relative intensities of these diagnostic peaks were typically ∼1% for the dual and single chain loss fragments and ∼0.1% for the acyl chain fragments compared with the residual precursor TAG ions (base peak, 100%; Fig. 1A–C). Overall, the EIEIO dissociation efficiency for TAGs was comparable to our previous PC and SM experiments when using the simultaneous ion-electron injection mode and a reaction time of 150 ms (19).

Characterizing TAG structures using EIEIO: from brutto level to regioisomeric specificity

The EIEIO spectra of TAGs contain various levels of information that can be used to characterize their structures. Beginning with the intact m/z of the sodiated TAG, we can obtain a brutto level assessment of the lipid [e.g., TAG (50:1) is the TAG investigated in Fig. 1A by virtue of its observed intact m/z]. Beyond this base level of characterization, we could also perform CID on these TAGs to reveal rudimentary and approximate knowledge about the acyl chain regioisomerism and unsaturation; such assessments are available to many other conventional MS-based analyses. However, EIEIO provides much more structural information for TAGs than do these methods, beginning with regioisomeric assignment of acyl chains.

The structural characterization of TAGs by EIEIO begins by surveying the spectrum in the dual acyl chain loss region (m/z 250–400); it is here that the most information is contained on acyl group chain length, the number of double bonds, and acyl chain regioisomerism. Conversely, the single acyl chain loss region contains fewer characteristic peaks and is generally of lesser utility. Using the spectra in Fig. 1C as examples, we observe that within the dual acyl group loss region there are two distinct peak groups: singlets (e.g., m/z 331.25) and paired doublets (e.g., m/z 345.26 and 347.24). High-energy CID experiments also yield similar fragments (7, 8), confirming our assignments here. Each singlet peak (e.g., m/z 305.24 in Fig. 1A; m/z 331.26 in Fig. 1B, C) maps to the acyl chain located at the TAG’s sn-2 site. These singlet peaks are formed by two possible bond cleavage patterns as indicated in Fig. 1E. The other set of diagnostic fragments are the doublet peaks, with the lower m/z member of the pair being more intense than the higher m/z peak. The difference in m/z between the doublet peaks (1.98 m/z units) reveals that the two fragment ions differ only slightly, with the lighter fragment containing a methylene (-CH₂-) group, while the heavier fragment contains an oxygen atom instead of that -CH₂- group (Fig. 1D, F).

Just as the singlet peak identifies the acyl chain at the sn-2 position of TAGs, so the doublet peaks identify the acyl chains at the sn-1 and sn-3 positions. For example, Fig. 1B displays a set of doublet peaks at m/z 319.26/321.24, which was related to a 16:0 acyl chain at the sn-1 and sn-3 sites; no other doublets are visible in the EIEIO spectrum, revealing identical substitution at both sn positions. A similar outcome was observed in Fig. 1C, where one set of doublet peaks at m/z 345.26/347.24 show that the sn-1 and sn-3 positions are both substituted with 18:1 acyl chains for this TAG molecule. However, Fig. 1A presents a more complex scenario than the other two EIEIO spectra, as it contains two sets of doublet peaks. One set appears at m/z 319.26/321.24 (a 16:0 acyl chain) and at m/z 345.26/347.24 (related to an 18:1 acyl chain). Because the sn-1 and sn-3 positions in TAGs are not distinguished in chirality using our present EIEIO method, we use the interchangeable “_” notation, such as TAG 16:0_/18:1/_18:1.

More importantly, these singlet and doublet peaks are consistent for given acyl chain lengths and are independent of the TAGs from which they originate. This allows us to predict the m/z values for any acyl chain’s singlet and doublet peaks for use as diagnostic fragment ions (Table 1; supplemental Table S1).

TABLE 1.

Diagnostic peak m/z for TAG regioisomerism identification

	Double Bond No.
Length	0	1	2	3	4	5	6
sn-1, sn-3 Diagnostic peak m/z (sodiated)
16	319.261	317.246	315.230	313.214
	321.240	319.225	317.209	315.194
18	347.293	345.277	343.261	341.246	339.230
	349.272	347.256	345.240	343.225	341.209
20	375.324	373.308	371.293	369.277	367.261	365.246
	377.303	375.287	373.272	371.256	369.240	367.225
22	403.355	401.339	399.324	397.308	395.293	393.277	391.261
	405.334	403.319	401.303	399.287	397.272	395.256	393.240
sn-2 Diagnostic peak m/z (sodiated)
16	305.246	303.23	301.214	299.199
18	333.277	331.261	329.246	327.23	325.214
20	361.308	359.293	357.277	355.261	353.246	351.23
22	389.339	387.324	385.308	383.293	381.277	379.261	377.246

Complete list of diagnostic peaks is shown in supplemental Table S1.

Interestingly, the peak intensity ratio between the singlet peak and the doublet peaks also represents the number of specific acyl groups in a TAG molecule. For example, Fig. 1A displays equivalent intensities for the singlet peak (m/z 305.24) and two doublet peaks (m/z 319.26/321.24 and m/z 345.27/347.25), indicating unique acyl chain substituents at each sn position, with 16:0 at the sn-2 position (the singlet peak) and different acyl chains at sn-1 and sn-3(16:0 and 18:1). For Fig. 1B, C, the same acyl groups are bonded at sn-1 and sn-3 such that the intensities of the doublet peaks were twice that of the singlet peak. These intensity ratios are ultimately used in the deconvolution of regioisomer constituent calculations (vide infra).

Characterizing TAG structures using EIEIO: identifying double bond locations

As stated previously, obtaining information on the number of carbon-carbon double bonds present in constituent TAG acyl chains is somewhat facile using MS-based methods. Generally, after a TAG ion is fragmented using conventional CID, acyl chain fragment losses can be detected, and the degree of total unsaturation within those acyl chains can be estimated (6). However, the identification of the specific locations of these carbon-carbon double bonds is much more difficult, requiring the use of ion/molecule reactions (11–14). With EIEIO, we are able to perform this identification in TAG molecules much the same as we previously reported for PCs (18) (Figs. 2, 3, and 4; supplemental Figs. S2, S3, and S4). The rule of thumb is that single carbon-carbon bonds are fragmented, but carbon-carbon double bonds are not, resulting in a characteristic fragmentation pattern in the EIEIO mass spectra. Each cleavage site produces an odd-electron radical fragment and an even-electron nonradical fragment. For saturated acyl chains (e.g., 16:0), radical fragments appear as a series of ions evenly spaced by 14.01 m/z units (i.e., a methylene or -CH₂- unit). For acyl chains containing double bonds, the m/z of the radical fragments will be shifted by +2H for every double bond present in the fragmented part of the acyl chain. Because double bonds are not favored cleavage sites in EIEIO, the intensities of the aforementioned fragmentation pattern drop at the double bond position, forming a “V” shape in the spectra (Fig. 2A).

Fig. 2.

EIEIO spectrum of tri-α-linolenin(ALA)-TAG (A) and its radical peak intensity profile of acyl chain fragments produced by EIEIO (B). Full range spectrum is shown in supplemental Fig. S2A. The radical acyl fragment intensities are classified by two hydrogen m/z shifts across double bonds. Color transitions indicated by arrows show the double bond locations.

Fig. 3.

Radical peak intensity of acyl chain fragments produced by EIEIO on standard polyunsaturated acyl TAGs, tri-γ-linolenin(GLA)-TAG (A) and tri-DHA-TAG (B). Each original spectrum is shown in supplemental Fig. S2B, C. Red arrows represent exceptional n-6 enhancement in polyunsaturated omega-3 acyls.

Fig. 4.

Radical peak intensities of acyl chain fragments produced by EIEIO on standard mixed acyl TAGs (A, B) and CLA-TAG (C). Each original spectrum is shown in supplemental Fig. S3. Color transitions indicated by arrows show the double bond locations.

To indicate the carbon-carbon double bond positions clearly, we classified the intensities of the diagnostic radical fragments by the 2H shifts. This classification was determined using the following calculations, where N = the site of cleavage as counted from the terminal methyl carbon of the acyl chain and m( ) = the molecular mass of the lost neutral fragment:

$$\begin{array}{l}0\text{H group}(black):\text{radical fragment at}\hspace{0.17em}\mathit{m}/\mathit{z}=(\text{precursor}\hspace{0.17em}\mathit{m}/\mathit{z})-\text{m}\left(\text{H}\right)-\text{m}\left({\text{CH}}_2\right)\times \text{N}\\ 2\text{H group}\left(\text{red}\right):\text{radical fragment at}\hspace{0.17em}\mathit{m}/\mathit{z}=(\text{precursor}\hspace{0.17em}\mathit{m}/\mathit{z})-\text{m}\left(3\text{H}\right)-\text{m}\left({\text{CH}}_2\right)\times \text{N}\\ 4\text{H group}\left(\text{green}\right):\text{radical fragment at}\hspace{0.17em}\mathit{m}/\mathit{z}=(\text{precursor}\hspace{0.17em}\mathit{m}/\mathit{z})-\text{m}\left(5\text{H}\right)-\text{m}\left({\text{CH}}_2\right)\times \text{N}\\ 6\text{H group}\left(\text{violet}\right):\text{radical fragment at}\hspace{0.17em}\mathit{m}/\mathit{z}=(\text{precursor}\hspace{0.17em}\mathit{m}/\mathit{z})-\text{m}\left(7\text{H}\right)-\text{m}\left({\text{CH}}_2\right)\times \text{N}\\ 8\text{H group}\left(\text{blue}\right):\text{radical fragment at}\hspace{0.17em}\mathit{m}/\mathit{z}=(\text{precursor}\hspace{0.17em}\mathit{m}/\mathit{z})-\text{m}\left(9\text{H}\right)-\text{m}\left({\text{CH}}_2\right)\times \text{N}\\ 10\text{H group}\left(\text{orange}\right):\text{radical fragment at}\hspace{0.17em}\mathit{m}/\mathit{z}=(\text{precursor}\hspace{0.17em}\mathit{m}/\mathit{z})-\text{m}\left(11\text{H}\right)-\text{m}\left({\text{CH}}_2\right)\times \text{N}\\ 12\text{H group}\left(\text{cyan}\right):\text{radical fragment at}\hspace{0.17em}\mathit{m}/\mathit{z}=(\text{precursor}\hspace{0.17em}\mathit{m}/\mathit{z})-\text{m}\left(13\text{H}\right)-\text{m}\left({\text{CH}}_2\right)\times \text{N}\end{array}$$

The intensities of the diagnostic radical fragments are displayed in Fig. 2B, which was calculated from EIEIO spectrum of ALA-TAG (Fig. 2A), GLA-TAG (Fig. 3A), and DHA-TAG (Fig. 3B). For example, in Fig. 2B and Fig. 3A, three double bonds starting from n-3 and n-6 are consistent with ALA (omega-3) and GLA (omega-6), respectively. In Fig. 3B, six double bonds starting from n-3 is consistent with DHA. While these TAGs were composed of polyunsaturated acyl chains, their EIEIO spectra and their acyl chain diagnostic peaks (i.e., dual acyl chain loss fragments) were rather simple; each of the three acyl chains in each species were the same (18:3, 18:3, and 22:6, respectively).

Another interesting observation from the insets of Fig. 2B and Fig. 3B are the nonzero intensities for the green (4H) plot at the n-6 position. We also observed this pattern during the EIEIO analysis of salmon fish oil extract that contained omega-3 polyunsaturated TAGs (vide infra). Interestingly, such exceptional behavior was not observed if the acyl groups were not omega-3 (Fig. 3A). One possible mechanism to explain this exception is preference to construct a stable ring molecule with six carbons and two double bonds (i.e., 1,4-cyclohexadiene from a radical precursor that lost a hydrogen at the omega carbon). Therefore, we automatically set the 4H intensity to zero at the n-6 position to avoid complications from this unusual occurrence.

For analyses of TAGs containing a mixture of different acyl chains, the EIEIO spectra and subsequent data analysis becomes more complex. This is because all fragmenting acyl chains of a TAG, while physically distinct, present their resulting fragmentation patterns within the same m/z space of the precursor TAG’s EIEIO spectrum. For example, when the fragmentation pattern of one acyl chain overlaps with the pattern of another, the contributions both single bond and double bond fragmentation overlap for a specific location along the identified acyl chain locations (e.g., n-6). In such cases, the V-shaped pattern (that reveals the presence of a double bond) does not drop to zero due to the contribution from another acyl chain’s single bond contribution (Fig. 4A, B). However, a distorted V-shaped profile still appears at the double bond locations, making double bond identification possible. For example, in the EIEIO analysis of TAG 16:0/16:0/18:1(n-9Z) (Fig. 4A), a transition appears in the black (0) to red (2H) profiles at the n-9 position. In Fig. 4B, two black (0) to red (2H) transitions occur at the n-9 and n-12 positions, with the V-shaped pattern intensity greater at the n-12 position than at n-9 (see the red line). This observation is consistent with TAG 18:1(n-12Z)/18:1(n-9Z)/18:1(n-12Z), which contains an acyl chain with a double bond at n-9 and two acyl chains with a double bond at n-12.

For TAGs composed of acyl chains bearing conjugated double bonds (CLA-TAG) (Fig. 4C; supplemental Fig. S4), the EIEIO fragmentation provided unique insights upon comparing the 0H (black) and 4H (green) plots. Here, we observe three acyl chain positions (n-7 through n-9) as having zero intensity values for both 0H and 4H plots, and also for the 2H (red) plot. The zero intensity for the 2H plot is consistent with the fact that the π electrons of conjugated double bonds interact and therefore behave differently than pure single or isolated (nonconjugated) double bonds.

Natural sample analysis strategy by DMS and EIEIO

While the EIEIO analysis of standard TAG samples was initially promising, we understood that more complex samples (i.e., extracts from natural sources) would present a much greater challenge. For instance, we realized that some form of separation would be required to minimize the overlap of targeted TAGs from interfering isobaric species, other molecules having the same nominal molecular mass (m/z in MS) but different chemical structure. In addition, other TAG species could be present as potential interferences based on their isotope patterns. For example, a hypothetical TAG of m/z 1,000 (monoisotopic, all ¹²C, ¹H, and ¹⁶O) could experience interference from the isotopomer of a TAG that only differs by one double bond (monoisotopic m/z of interfering TAG is m/z 998, with a 2 × ¹³C contribution at m/z 1,000). While this contribution is only ∼20% of the monoisotopic base peak of the interfering TAG, if it is present in a very large excess compared with the targeted TAG, convolution of two TAG EIEIO spectra will occur. Hence, we used DMS (22, 26, 27) to separate analyte TAGs from species such as interfering isotopomers, oxidized TAGs, and other unknown isobaric interferences. Such efforts reduce complexity of the EIEIO spectra and increase the confidence of the deconvolution process for separating contributions from multiple species with similar molecular structures.

Figure 5 shows a flow chart outlining the process for analysis of natural lipid extracts using DMS and EIEIO. As the first step, sample is infused into the ESI source, producing a multitude of ionized TAGs and other lipid species. All of these ions pass though the DMS cell while the CoV is scanned (step 1) from −5 V to +10 V in 0.1 V steps (SV set at 3,900 V). During each CoV step, TOF-MS spectra are acquired (step 2) (no precursor selection or EIEIO at this stage) to correlate potential TAG precursors to optimal CoVs for transmission and separation from interfering species; further refinement of this separation will occur in a later step (vide infra). Ultimately, these settings are compiled into a list of potential TAG targets (steps 3 and 4) for EIEIO analysis, using DMS at given CoV to provide cleaner and simpler precursor ion populations.

Fig. 5.

Procedure flow and examples of DMS-EIEIO MS analysis for natural lipid samples. The example is olive oil.

With the preliminary DMS conditions set for the targeted TAG precursors, further refinement of the DMS conditions (step 5) are found for each targeted TAG precursor to obtain the most intense precursor ions with the lowest levels of interference from species such as ¹³C isotopomers and oxidized species. The first choice should be lower DR value because it provides better ion transmission with shorter EIEIO accumulation time required. If interferences are detected, the resolving gas is increased to obtain better DMS separation, with retuning of the CoV value. For example, Fig. 5, step 5 shows a scenario where we aimed to separate TAG 54:2 (m/z 909.781) from two interfering species: the third isotope (i.e., 2 × ¹³C) of TAG 54:3 (m/z 909.714) and TAG (52:4) +O₂ (also at m/z 909.71). Given the isolation window of the Q1 quadrupole (∼1 m/z unit), all three species would be selected for EIEIO, producing a convoluted data set. To remove the interferences, we increased the resolving gas to 30 psi and retuned the CoV to +5.9 V to reduce the intensity of the interferences dramatically.

Deconvolution of mixed populations of TAG regioisomers using EIEIO

In the demonstration of EIEIO using standard synthesized samples (vide supra; Figs. 1–4), the TAG molecule’s structures were successfully analyzed and provided information on brutto level, the acyl chains present, the number and locations of carbon-carbon double bonds, and the acyl chains’ regioisomerism. However, natural lipid samples will have a high probability of containing regioisomers or double bond isomers for a given m/z value of an ionized TAG molecule. These isomers are incredibly difficult to separate using DMS (or any other analytical separation technique), and so, we have implemented some strategies to deconvolute the EIEIO spectral data to simplify the identification.

First, we identified all of the diagnostic singlets and doublets (vide supra) appearing in an EIEIO spectrum and used them to reconstruct the TAG precursor mass to obtain a brutto level identification. We also used the diagnostic peak’s intensities to assist in the narrowing of the identification of the unknown TAG candidates using the following relationship:

$$\left(\text{diagnostic peak intensities}\right)=\left(\text{assignment matrix of diagnostic peak}\right)\times \left(\text{intensity of TAGs}\right)$$

One simple example of this preliminary process is shown in Fig. 6, which displays the EIEIO spectrum of potential TAG precursors of m/z 905.758. Here, we observed two singlets and two doublets, which required the high-resolution measurement capability of the TOF mass spectrometer to separate the nearly overlapped peaks at m/z 345.2. Based on these data, we identified (at least) two distinct TAGs [both brutto level TAG (54:4)], one bearing an 18:1 acyl chain at the sn-2 position, the other 18:2 (based on the singlets rule, vide supra). The sn-1/sn-3 positions were assigned as either 18:1 or 18:2, meaning that (at least) two isomeric TAGs were present: TAG 18:2_/18:1/_18:1 (designated TAG1) and TAG 18:1/18:2/18:1 (designated TAG2). The contributions to the experimental diagnostic intensities from two TAGs were determined as follows:

$$\text{i}\left(\text{18:2s}\right)=0.754\times \text{I}\left(\text{TAG2}\right)$$

$$\text{i}\left(\text{18:1s}\right)=0.754\times \text{I}\left(\text{TAG1}\right)$$

$$\text{i}\left(\text{18:2c}\right)=1.000\times \text{I}{\left(\text{TAG1}\right)}^{}$$

$$\text{i}\left(\text{18:2o}\right)=0.529\times \text{I}\left(\text{TAG1}\right)$$

$$\text{i}\left(\text{18:1c}\right)=1.000\times \text{I}\left(\text{TAG1}\right)+1.000\times 2\times \text{I}\left(\text{TAG2}\right)$$

$$\text{i}\left(\text{18:1o}\right)=0.529\times \text{I}\left(\text{TAG1}\right)+0.529\times 2\times \text{I}\left(\text{TAG2}\right)$$

where i(acyl chain) represents a theoretical diagnostic intensity of singlet(s)/doublet (CH₂ type)/doublet (O type) of the designated acyl chain, and I represents constituent of designated TAG. The coefficients: 0.754, 1.000 and 0.529 represent nominal production ratios for the singlet, doublet (CH₂ type) and doublet (O type), respectively. These equations can be solved using the least mean square method that is obtained by minimizing S in the following expression:

$$\text{S}=\sum {[\left(\text{experimental diagnostic peak intensities}\right)-(\text{theoretical diagnostic intensities}:\text{i})]}^2$$

Fig. 6.

EIEIO spectrum of m/z 905 precursor ions in olive oil. This is an example of deconvolution of acyl chain regioisomers.

The TAG species identified using this process and their normalized intensities [%] are listed in Table 2 and supplemental Fig. S8.

TABLE 2.

Identified TAGs in Portuguese olive oil

Precursor m/z (Sodiated)	Composition [%]	Brutto Level	Found TAG	Double Bond Location
827.709	0.06	TAG 48:1	TAG 16:0/16:1/16:0[100%]	(n-7):78%
				(n-12):22%
855.738	4.59	TAG 50:1	TAG 16:0/18:1/16:0 [95%]	(n-9):83%
				(n-12):17%
853.724	1.48	TAG 50:2	TAG 16:0/18:2/16:0 [37%]	(n-9):43%	(n-6,-9):45%
			TAG16:1_/18:1/_16:0[32%]	(n-12):12%
			TAG 16:0_/16:1/_18:1 [20%]
851.712	0.17	TAG 50:3	TAG 16:0/18:3/16:0[32%]	(n-9):10%	(n-6,-9):37%	(n-3,-6,-9):30%
			TAG 16:1_/18:2/_16:0[31%]	(n-7):9%		(n-6,-9,-12):7%
			TAG 16:1_/16:1/_18:1[15%]	(n-12):8%
			TAG 16:0_/16:1/_18:2 [14%]
869.757	0.05	TAG 51:1	TAG 16:0_/18:1/_17:0[100%]	(n-9):55%
				(n-12):45%
867.739	0.18	TAG 51:2	TAG 16:0_/17:1/_18:1[37%]	(n-9):100%
			TAG 16:0_/18:1/_17:1[34%]
			TAG 17:1_/16:0/_18:1 [13%]
865.727	0.02	TAG 51:3	TAG 15:1_/18:1/_18:1 [55%]	(n-9):100%
			TAG 17:1_/16:1/_18:1 [45%]
881.759	24.05	TAG 52:2	TAG 16:0_/18:1/_18:1 [92%]	(n-9):89%
				(n-12):11%
879.743	5.12	TAG 52:3	TAG 16:0_/18:2/_18:1[36%]		(n-6,-9):47%
			TAG 16:1_/18:1/_18:1[26%]	(n-9):44%
			TAG 16:0_/18:1/_18:2 [21%]	(n-12):9%
			TAG 18:1/16:1/18:1 [11%]
877.730	0.89	TAG 52:4	TAG 16:0_/18:3/_18:1 [30%]		(n-6,-9):46%	(n-3,-6,-9):27%
			TAG 16:0_/18:2/_18:2 [22%]	(n-9):23%		(n-6,-9,-12):7%
			TAG 16:0_/18:1/_18:3[19%]	(n-12):3%
			TAG 16:1_/18:2/_18:1 [14%]
875.714	0.09	TAG 52:5	TAG 16:1_/18:2/_18:2 [38%]	(n-9):12%	(n-6,-9):35%	(n-3,-6,-9):48%
			TAG 16:1_/18:3/_18:1 [32%]	(n-12):5%		(n-6,-9,-12):3%
			TAG 16:1_/18:1/_18:3 [30%]
895.773	0.25	TAG 53:2	TAG 17:0_/18:1/_18:1 [52%]	(n-9):100%
			TAG 17:1_/18:0/_18:1 [15%]
			TAG 17:1_/18:1/_18:0 [12%]
			TAG 18:1/17:0/18:1 [11%]
893.758	0.28	TAG 53:3	TAG 17:1_/18:1/_18:1 [55%]	(n-9):92%	?
			TAG 18:1/17:1/18:1 [32%]	(n-12):8%
			TAG 17:1_/18:0/_18:2 [10%]
891.741	0.04	TAG 53:4
913.816	0.06	TAG 54:0
911.806	0.83	TAG 54:1	TAG 16:0_/18:1/_20:0[52%]	(n-9):100%
			TAG 18:0/18:1/18:0 [48%]
909.778	13.94	TAG 54:2	TAG 18:1_/18:1/_18:0 [67%]	(n-9):93%
			TAG 18:1/18:0/18:1 [33%]	(n-12):7%
907.773	37.46	TAG 54:3	TAG 18:1/18:1/18:1 [100%]	(n-9):94%
				(n-12):6%
905.757	7.49	TAG 54:4	TAG 18:2_/18:1/_18:1 [58%]	(n-9):47%	(n-6,-9):50%
			TAG 18:1/18:2/18:1 [42%]	(n-12):2%
903.744	1.92	TAG 54:5	TAG 18:2_/18:2/_18:1 [33%]		(n-6,-9):34%	(n-3,-6,-9):22%
			TAG 18:3_/18:1/_18:1 [32%]	(n-9):32%		(n-6,-9,-12):8%
			TAG 18:1/18:3/18:1 [25%]	(n-12):3%
			TAG 18:2/18:1/18:2 [10%]
939.832	0.17	TAG 56:1
937.820	0.57	TAG 56:2	TAG 18:1_/18:1/_20:0 [100%]	(n-9):100%
933.790	0.09	TAG 56:4	TAG 18:2_/18:1/_20:1 [31%]	(n-9):34%	(n-6,-9):43%	(n-6,-9,-12):14%
			TAG 18:1_/18:2/_20:1 [30%]			(n-3,-6,-9):8%
			TAG 18:1_/18:3/_20:0 [15%]
			TAG 18:2_/18:2/_20:0 [13%]
			TAG 18:3_/18:1/_20:0 [11%]
965.854	0.11	TAG 58:2	TAG 18:1_/18:1/_22:0 [100%]	(n-9):100%
979.864	0.03	TAG 59:2	TAG 18:1_/18:1/_23:0[100%]	(n-9):100%
993.877	0.06	TAG 60:2	TAG 18:1_/18:1/_24:0 [100%]	(n-9):100%

Deconvolution of multiple double bond locations on multiple acyl chains

In addition to the complexity that stems from analyzing mixtures of TAG regioisomers, we are also presented with TAGs that are isomeric only in the positioning of the double bond(s) along their constituent acyl chains. Here, we also used a deconvolution strategy based on the presence of the characteristic “V”-shape pattern among the EIEIO fragment ions, using a similar intensity-based, least mean square fitting strategy as the regioisomer deconvolution analysis.

While the locations of double bonds can be at almost any location along a TAG’s constituent acyl chains, we made the assumption that the double bond locations were consistent to common fatty acid structures in biological samples. For example, we set rules for possible acyl chains in our TAG analyses as follows: (1) saturated acyl chains could be of any length; (2) omega-3 acyl chains contained three or more double bonds; (3) omega-6 acyl chains contained two or more double bonds; and (4) omega-5, -7, -9, -12, and -15 acyl chains contained only a single double bond. We assumed a two C-C single bond spacing between two double bonds. We also included both even and odd number of carbons in the possible TAGs acyl chains.

Ultimately, while we can deconvolute regioisomeric mixtures of TAGs, our method is presently to assign double bond locations to each constituent acyl chain for a mixture of isomeric TAGs. Hence, we have listed double bond location patterns and their amount in percentages in the result tables (Table 2; supplemental Fig. S8). One future strategy may involve the fragmentation of ionized TAGs via CID and subsequent EIEIO of the fatty acid fragments to determine double bond position.

Identifying TAGs in natural lipid samples

With an optimized DMS-EIEIO work flow in place, combined with data deconvolution strategies, and an understanding of our method’s limitations, we characterized several natural lipid extracts and edible oils. For brevity, we focused on the analyses of olive oils (8) in the main manuscript, with analyses on other extracts presented in supplemental Figs. S6–S10.

By comparison, the Portuguese and Greek extra virgin oils we analyzed were only moderately oxidized as shown in Fig. 7 (for the Portuguese oil) though the standard olive oil was oxidized (supplemental Fig. S5). The DMS-EIEIO analyses of these olive oil samples yielded many identified TAG species, as demonstrated by the TAGs identified in the Portuguese oil alone (Table 2). From the identified TAG profiles, we could calculate the FFA composition (Fig. 7B), which showed that the main FFAs in the standard (supplemental Fig. S5B) and Portuguese oils (Fig. 7B) were consistent each other, as well as the certificate data sheet provided by the vendor of the standard.

Fig. 7.

A: MS spectra of Portuguese olive oil. B: Acyl constituents calculated from identified TAGs by EIEIO.

Using DMS-EIEIO, we were also able to compare the TAG speciation in both olive and avocado oils (Fig. 8; and supplemental Fig. S8). While the major FFA constituents of the two different plant species were similar in both samples, 16:0 and 18:1 (Fig. 7A, B). Figure 8 conveys the regioisomerism of the olive oil (Fig. 8C) and avocado oil (Fig. 8D) in a contour plot, with the acyl chain in the sn-1 and sn-3 positions on the y-axis, and the acyl chain in the sn-2 position on the x-axis (all determined using the DMS-EIEIO work flow; Table 2 and supplemental Fig. S8C). In both oils, there is a strong preference for an 18:1 acyl chain to be present in the sn-2 position, while the sn-1 and sn-3 positions bore the 16:0 and 18:1 acyl chains. The olive and avocado oils did not show any significant difference in these profiles. Figure 7E, F show 3D distribution, where sn-2 was fixed at C18:1 and distribution of sn-1 and sn-3 were displayed. In this stage, significant difference between olive and avocado appeared, (i.e., TAG 16:0_/18:1/_18:1 was less abundant in olive but was abundant in avocado oil). Supplemental Fig. S6 shows 2D and 3D profiles of the three sources of the olive oil. In the same plant species, these profiles were identical or quite similar. In addition to the conventional FFA analysis, this 3D acyl distribution in TAGs provides more confident information to distinguish the origin of oils.

Fig. 8.

Comparison between olive oil and avocado extract. A, B: FFA constituents calculated from identified TAGs by EIEIO. C, D: The 2D display of sn-1 and sn-3 versus sn-2. E, F: The 3D display of sn-1 and sn-3 profile when sn-2 is 18:1.

In the supplemental information, results of EIEIO analysis of other oil samples were presented. Coconut oil (supplemental Fig. S7) is an example of TAGs with highly saturated and medium-length acyl groups. Strong 12:0 preference at sn-2 and strong paired preference of 8:0 and 12:0 at sn-1 and sn-3 were observed (astonishingly less abundant simultaneous 8:0 at sn-1 and 8:0 at sn-3). Linseed oil is an example of long chain with rich linolenic acids (supplemental Fig. S9) (28). This standard sample was also oxidized strongly. As acyl components, ALA was dominant, but GLA was less dominant. Arrangement of acyls was less preferential at any sites.

The example of a crude natural sample in animals is the salmon extract (supplemental Fig. S10). In its MS spectra, 80 TAG precursors were listed. By applying DMS-EIEIO, 282 TAGs were identified (mixed double bond positions). In the TAGs, 46 species of FFA were included (supplemental Fig. S10B) (length and double bond number are identified, but mixed double bond positions). Polyunsaturated omega-3 acyls (DHA, EPA) were detected elsewhere in the list. In the salmon extract, 18:1, 20:1, and 20:5 preferences at the sn-2 site were observed.

CONCLUSIONS

In this study, we have demonstrated the utility of combining DMS and EIEIO to provide in-depth characterization of TAGs. With EIEIO, energetic electrons (∼10 eV) fragmented TAG ions to allow for regioisomeric assignment of identified acyl groups at the sn-2 or sn-1/3 positions of the glycerol backbone. In addition, carbon-carbon double bond locations within the acyl chains were also identified by EIEIO. Beyond the analysis of lipid standards, this technique was applied to edible oils and natural lipid extracts to demonstrate the power of this method to provide in-depth structural elucidation of TAG molecular species. This study proved that DMS-EIEIO is a powerful tool for lipidomics structure analysis of TAGs, as well for PCs, SMs, and other lipid classes. Several challenges remain for this technique, including identification of the chirality at the TAG’s C2 position (i.e., sn-1 vs. sn-3), as well as distinguishing cis/trans double bond isomerism.