Proton transfer reactions for the gas-phase separation, concentration, and identification of cardiolipins

Abstract

Cardiolipin (CL) analysis demands high specificity, due to the extensive diversity of CL structures, and high sensitivity, due to their low relative abundance within the lipidome. While electrospray ionization mass spectrometry (ESI-MS) is the most widely used technology in lipidomics, the potential for multiple charging presents unique challenges for CL identification and quantification. Depending on the conditions, ESI-MS of lipid extracts in negative ion mode can give rise to cardiolipins ionized as both singly and doubly deprotonated anions. This signal degeneracy diminishes the signal-to-noise ratio while in addition (for direct infusion) the dianion population falls within a m/z range already heavily congested with monoanions from more abundant glycerophospholipid subclasses. Herein, we describe a direct infusion strategy for CL profiling from total lipid extracts utilizing gas-phase proton transfer ion/ion reactions. In this approach, lipid extracts are ionized by negative ion ESI generating both singly deprotonated phospholipids and doubly deprotonated CL anions. Charge-reduction of the negative ion population by ion/ion reactions leads to an enhancement in singly deprotonated [CL – H]¯ species via proton transfer to the corresponding [CL – 2H]²¯ dianions. To concentrate the [CL – H]¯ anion signal, multiple iterations of ion accumulation and proton transfer ion/ion reaction can be performed prior to subsequent interrogation. Mass-selection and collisional activation of the enriched population of [CL – H]¯ anions facilitates the assignment of individual fatty acyl substituents and phosphatidic acid moieties. Demonstrated advantages of this new approach derive from the improved performance in complex mixture analysis affording detailed characterization of low abundant CLs directly from a total biological extract.

Graphical Abstract

Introduction

Cardiolipins (CLs) are unique dimeric phospholipids found in bacterial and mitochondrial membranes that exhibit distinct chemical and physical properties.¹ In eukaryotes, following biosynthesis from cytidinediphosphate-diacylglycerol and phosphatidylglycerol, CLs almost exclusively localize in the inner mitochondrial membrane.^{1, 2} Possessing two phosphatidyl moieties linked to a central glycerol backbone and four fatty acyl substituents, CLs are among the most structurally complex classes of phospholipids. While the roles of CL have not been fully elucidated, CLs are known to serve central functional roles in membrane dynamics, cellular bioenergetics, and apoptosis.^2–5 For example, CLs participate in mitochondrial energetic pathways by providing stabilization to mitochondrial enzymes and are required for optimal enzymatic function, as alternate phospholipid classes cannot substitute for CLs.⁶ Thus, CLs are essential to mitochondrial function, and in turn, mitochondrial dysfunction can arise from alterations in CL content and composition. Moreover, recent studies link CL deficiencies and acyl chain remodeling with numerous pathologies⁵ such as Barth syndrome⁷, heart failure⁸, diabetes⁹, and several types of cancer.^{10, 11}

Unfortunately, assessing modifications to CL structure and production presents considerable challenges, notably due to the high number of CL structural analogues. For instance, thousands of discrete molecular CL structures can be derived from permutations of the four acyl chains, including variations in esterification site along the glycerol backbones (i.e., sn-positional isomers) and fatty acyl composition (i.e., total number of carbons and degree of unsaturation).¹² Over recent decades, mass spectrometry (MS) has emerged as the most widely adopted analytical platform for lipid analysis, and in particular, two distinct MS-based platforms have been deployed in the detection, identification, and quantification of CLs.⁴ In the first approach, a lipid extract is admitted to a (tandem) mass spectrometer via liquid chromatography (LC). While CL class isolation from crude lipid extract can be readily achieved using LC-MS, separation of individual CL molecular structures can be chromatographically taxing, as isomeric CL structures frequently exhibit similar chromatographic behaviors.^{13, 14} In a second approach, a lipid extract is directly infused into the mass spectrometer using electrospray ionization (ESI). This method, referred to as direct infusion ESI-MS or shotgun lipidomics, has gained considerable attention.^15–17 Due to the two acidic phosphatidyl moieties, CLs generate abundant deprotonated anions upon direct infusion negative ESI with the potential to form either singly deprotonated [CL – H]¯ or doubly deprotonated [CL – 2H]²¯ dianions, depending on instrument geometry and experimental conditions. To characterize CL molecular species, low-energy collision-induced dissociation (CID) of the CL anion (either singly or doubly deprotonated) has been largely successful. In particular, the tandem (MS²) mass spectra of deprotonated CL anions (i.e., [CL – H]¯ or [CL – 2H]²¯) facilitate the assignment of individual phosphatidic acid (PA) moieties linked at the 1’ or 3’ position of the central glycerol backbone and fatty acyl substituents.¹⁸ Re-isolation and collisional activation of the [PA – H]¯ product ions derived from CL anions yield MS³ product ion spectra indicative of the fatty acyl composition and regiochemistry of individual PA moieties. Notably, regiochemical assignments based on MS³ product ion relative abundances do not preclude contributions from minor sn-positional isomers, and therefore such assignments should be made with caution and ideally confirmed with authentic CL standards.¹⁹ In an alternate approach, Hsu and Turk described fragmentation patterns of the [CL – 2H + Na]¯ ion, reporting analogous product ion spectra to that of the [CL – H]¯ ion.²⁰ Yet, the principle advantage of interrogating [CL – 2H + Na]¯ anions over deprotonated CL ions is the lack of extensive sample preparation prior to analysis, as the generation of [CL – H]¯ anions upon direct ESI is often reliant on multiple sample desalting steps. As the methods described above are reliant upon low-energy CID, subtle structural features, such as carbon-carbon double bond locations and geometry, cannot be discerned. In a recent study, Brodbelt and colleagues employed a combination of CID and 193 nm ultraviolet photodissociation (UVPD) on an orbitrap mass spectrometer to pinpoint the site(s) of unsaturation in CLs derived from a biological extract.²¹ To date, CID/UVPD has achieved the highest level structural identification for CL. However, this platform requires high resolving power, yields congested product ion spectra, and suffers from low product ion abundances, primarily due to the limited fragmentation efficiency of UVPD, complicating analysis of low abundance CLs in biological extracts.

Despite successes, significant challenges associated with CL detection and characterization using shotgun lipidomics remain. First, as CLs represent minor components of the cellular lipidome, CL ions generated via direct infusion negative ESI can be at or below detection limits for many mass spectrometric methods.²² Furthermore, the identification of CL molecular species can be exceedingly difficult when examining total lipid extracts. In particular, doubly charged [CL – 2H]²¯ anions appear in the same mass-to-charge (m/z) region that is typically dominated by singly charged anions derived from alternate phospholipid classes (i.e., phosphatidylethanolamine (PE), phosphatidylglycerol (PG), phosphatidylserine (PS), etc.). While high resolving power can aid in the mass assignment of [CL – 2H]²¯ anions, CL characterization requires tandem-MS (MS/MS). As most MS² instruments are unit mass selective at best, [CL – 2H]²¯ dianions are frequently co-isolated and collisionally activated with isobaric singly charged acidic phospholipid anions leading to composite product ion spectra and ambiguous assignments. Therefore, fractionation of the CL class via condensed-phase separations prior to ESI-MS² analysis is often required to achieve unambiguous CL identification. Furthermore, accurate identification of minor and/or overlapping CL molecular species remains an exigent task, especially for analytical platforms employing low-resolution mass spectrometers.⁴ Notably, Han and co-workers established a two-dimensional MS approach, exploiting the marked linoleate enrichment in CL and the doubly charged nature of CL molecular species to both identify and quantify low-abundance CLs directly from biological samples.²³

Recently, gas-phase ion/ion reactions have been demonstrated in strategies for lipid characterization in complex biological mixtures, including near-complete structural elucidation of phospholipids.^24–28 Moreover, gas-phase proton transfer ion/ion reactions have been implemented for complex mixture analysis, effectively resolving ions detected at similar mass-to-charge ratios but differing in both mass and charge.^{29, 30} In this study, we apply proton transfer ion/ion reactions on a modified hybrid triple quadrupole/linear ion trap mass spectrometer to separate, concentrate, and identify CLs from a total lipid extract, relying exclusively on gas-phase chemistries.

Experimental

HPLC-grade methanol and water were purchased from Fisher Scientific (Pittsburgh, PA). Acetic acid and 8-bis(dimethylamino)naphthalene (i.e., Proton-Sponge®) were purchased from MilliporeSigma (St. Louis, MO). All CL standards and E. coli. lipid extract (polar) were purchased from Avanti Polar Lipids, Inc. (Alabaster, AL). Solutions of CL standards were prepared in methanol to a final concentration of 1 μM. A solution of Proton-Sponge® solution was prepared in water/methanol/acetic acid (79.5/19.5/1.0) to a final concentration of 0.1 mg/mL.³¹ To prepare the lipid extract solution, E. coli. extract was diluted 50-fold in methanol.

Mass Spectrometry

Experiments were conducted on a Sciex QTRAP 4000 hybrid triple quadrupole/linear ion trap mass spectrometer (SCIEX, Concord, ON, Canada) that has been modified to perform ion/ion reactions.³² Lipid anions and Proton-Sponge® reagent cations were sequentially injected via alternately pulsed nESI.³³ First, lipid anions generated via negative nESI were mass-selected during transit through Q1 prior to storage in the high-pressure collision cell, q2. Next, positive nESI produced singly protonated Proton Sponge® reagent cations, which were subsequently isolated in transit through Q1, were accumulated in the reaction cell q2. Together in q2, the reagent cations and lipid anions were mutually stored for 300 ms. Next, the q2 radio frequency (RF) amplitude was raised to remove low-mass anions, and the remaining charge-reduced lipid anions were transferred to the low-pressure linear ion trap (LIT), Q3, for storage. Multiple iterations of the ion fill and ion/ion reaction steps were conducted, and charge-reduced product ions were accumulated in Q3 until the charge capacity was reached, noting that a single fill cycle requires approximately 800 ms. To perform MSⁿ experiments in Q3, charge-reduced CL anions were isolated with unit resolution and collisionally activated via single frequency resonance excitation (q = 0.2). Product ions were analyzed by mass-selective axial ejection (MSAE).³⁴

Nomenclature

When possible, we adopt lipid nomenclature described by Liebisch et. al.³⁵ For example, the cardiolipin subclass is abbreviated as CL. The shorthand notation for CL sum composition indicates the combined number of carbons in the four fatty acyl chains, followed by a colon and the number of carbon-carbon double bonds (e.g., CL 70:3 implies a total of 70 carbons and 3 carbon-carbon double bonds amongst the four fatty acyl substituents). To describe CL composition at the fatty acyl level, each fatty acyl substituent is listed sequentially utilizing literature recommendations for fatty acyl chain shorthand notations. Furthermore, fatty acyl composition is represented by the total number of carbons and the degree of unsaturation (e.g., 18:1 indicates a fatty acyl with 18 carbons and 1 carbon-carbon double bond). When known, the regiochemical assignments of the fatty acyl substituents are separated by a forward slash, and when unknown, fatty acyl constituents are separated by an underscore. We also employ analogous modifications to the shorthand CL notation as described by Macias et al.²¹ Briefly, these modifications are used to indicate when fatty acyl composition at each PA substituent is known. For instance, CL (16:0_18:0)_(18:1_20:4) denotes that while individual fatty acyl and PA moiety regiochemical assignments cannot be made, it is known that one PA constituent contains the 16:0 and 18:0 fatty acyl groups, while the other PA is defined as PA 18:1_20:4.

Results and Discussion

Proton Transfer Reactions to Facilitate Characterization of Synthetic CL

As described above, although [CL – 2H]²¯ anions can be exploited for CL identification, the assignment of doubly-charged CL precursor ions can be compromised when examining total lipid extracts due to (1) limited resolving power of quadrupole mass selection, (2) isomeric/isobaric overlap with more abundant phospholipid ions, (3) the inherently low abundances of CL molecular species and (4) the generation of degenerate signals across both possible charge states.^{4, 22} Furthermore, while interrogation of the singly deprotonated [CL – H]¯ anions enables CL structure elucidation, the abundances of these CL precursor ions are normally low in complex mixtures and thus challenging instrument dynamic range.¹⁸ In order to separate CLs from other phospholipids on the m/z scale, we employ gas-phase proton transfer reactions.

We first demonstrate this process with synthetic CL 16:0/18:1/16:0/18:1. Negative nESI of CL 16:0/18:1/16:0/18:1 predominantly generated the doubly deprotonated [CL 16:0/18:1/16:0/18:1 – 2H]²¯ anion (m/z 702), though a low abundance singly deprotonated [CL 16:0/18:1/16:0/18:1 – H]¯ anion (m/z 1404) was also detected, as shown in Figure 1a. Next, mass-selected [CL 16:0/18:1/16:0/18:1 – 2H]²¯ anions were reacted with the singly protonated Proton-Sponge® cations in the high-pressure collision cell, q2, for 300 ms yielding the product ion spectrum shown in Figure 1b. The proton transfer reaction resulted in charge reduction of the CL dianion, generating a dominant charge-reduced [CL 16:0/18:1/16:0/18:1 – H]¯ product ion (m/z 1404) as illustrated in Scheme S1 (Supporting Information). In Q3, subsequent mass-selection and ion-trap CID of the charge-reduced [CL 16:0/18:1/16:0/18:1 – H]¯ (m/z 1404) anion facilitated CL molecular structural identification. Illustrated with Figure 1c, ion-trap CID of the [CL 16:0/18:1/16:0/18:1 – H]¯ species gave rise to a prominent tricyclic glycerophosphate ester product ion (m/z 809) and a dehydrated phosphatidylglycerol ion (m/z 729). Also depicted in Figure 1c are product ions corresponding to the neutral losses of the fatty acyl substituents as acids (i.e., [PA 16:0/18:1 – H – R₂COOH]¯ (m/z 391) and [PA 16:0/18:1 – H – R₁COOH]¯ (m/z 417)) and ketenes (i.e., [PA 16:0/18:1 – H – R’₂CH=C=O]¯ (m/z 409) and [PA 16:0/18:1 – H – R’₁CH=C=O]¯ (m/z 435)). Consistent with previous observations¹⁸, losses of the sn-2 (or sn-2’) fatty acyl group were favored over those of the sn-1 (or sn-1’), as indicated by higher relative abundances of the [PA 16:0/18:1 – H – R₂COOH]¯ (m/z 391) and [PA 16:0/18:1 – H – R’₂CH=C=O]¯ (m/z 409) product ions compared to the [PA 16:0/18:1 – H – R₁COOH]¯ (m/z 417) and [PA 16:0/18:1 – H – R’₁CH=C=O]¯ (m/z 435) anions. We note that the CID spectra of the charge-reduced [CL 16:0/18:1/16:0/18:1 – H]¯ ion and the unreacted, doubly deprotonated CL ion are similar, yet the [PA 16:0/18:1 – H]¯ (m/z 673) product ion is notably absent from the [CL 16:0/18:1/16:0/18:1 – 2H]²¯ product ion spectrum (c.f. Figure 1c and Figure S1). Furthermore, dissociation of the charge-reduced CL 16:0/18:1/16:0/18:1 anion enabled assignment of CL fatty acyl composition, as the [16:0 – H]¯ and [18:1 – H]¯ carboxylate anions were detected at m/z 255 and m/z 281, respectively. Lastly, the dominant product ion observed in the CID spectrum of [CL 16:0/18:1/16:0/18:1 – H]¯ reflects the [PA 16:0/18:1 – H]¯ (m/z 673) ion, arising from ester bond cleavage at the 1’- or 3’-position of the central glycerol backbone. Re-isolation and collisional activation of the [PA 16:0/18:1 – H]¯ product ion (m/z 673) confirmed the diacylphosphatidyl moiety assignment, as the 16:0 and 18:1 carboxylate anions were observed in the resulting CID spectrum (see Supporting Information Figure S2). In agreement with previous studies¹⁸, the sn-1 (or sn-1’) [16:0 – H]¯ fatty acyl carboxylate anion (m/z 255) was more abundant than sn-2 (or sn-2’) [18:1 – H]¯ fatty acid anion (m/z 281) in the CID spectrum of [PA 16:0/18:1 – H]¯ (m/z 673) shown in Figure S2, suggesting preferential formation of the acyl anions from the sn-1 position. Collectively, proton transfer ion/ion reactions readily and rapidly transform doubly deprotonated CL anions in the gas-phase, producing charge-reduced [CL – H]¯ anions that fragment to yield structurally informative product ion spectra that permit CL molecular structure identification.

Figure 1.

Demonstration of proton transfer reactions for the identification of synthetic CL 16:0/18:1/16:0/18:1. (a) Negative nESI of CL 16:0/18:1/16:0/18:1. (b) Mutual storage product ion spectrum resulting from the proton transfer reaction between Proton-Sponge® cations and CL dianions. (c) Ion-trap CID spectrum of charge-reduced CL 16:0/18:1/16:0/18:1.

Proton Transfer Reactions for CL Class Separation

The proton transfer charge reduction approach was used to examine CLs from an E. coli. lipid extract. The negative nESI mass spectrum of E. coli extract is shown in Figure 2a and contains a variety of ions – mostly singly charged – in the phospholipid region (m/z 650 – 900). An enlargement of this mass spectral region illustrates the challenge of distinguishing CL dianions from singly charged isobaric phospholipid ions (see Figure S3). Also, the inset provided in Figure 2a highlights the low abundances of singly deprotonated [CL – H]¯ anions formed upon direct negative ion nESI. Note that interrogation of individually mass-selected [CL – H]¯ anions is plausible following direct negative ion nESI, at the expense of extremely long integration/averaging times, particularly for the lower abundance CL components.

Figure 2.

Demonstration of gas-phase proton transfer ion/ion reactions for the identification of CL from E. coli. extract. (a) Direct negative nESI mass spectrum of E. coli. extract. (b) Product ion spectrum resulting from the gas-phase proton transfer ion/ion reaction between anions shown in (a) and Proton-Sponge® reagent cations. Note that both spectra represent an average of 20 scans.

To separate [CL – 2H]²¯ anions from more abundant phospholipid monoanions, all ions derived from direct negative ion nESI of E. coli extract were allowed to react in the gas-phase with Proton- Sponge® reagent cations. The resulting product ion spectrum is illustrated with Figure 2b. The ion/ion reaction resulted in the charge reduction of doubly charged [CL – 2H]²¯ anions to yield [CL – H]¯ anions, as highlighted with the inset shown in Figure 2b. Importantly, the ion/ion reaction resulted in nearly a ten-fold increase in the [CL – H]¯ anion signal and dramatically reduced spectral complexity in the low-mass region by removing much of the dianion population (c.f. Figure 2a and 2b). Accounting for isotopic contributions³⁶, a normalized abundance CL profile for E. coli extract at the sum compositional level was readily obtained even at nominal mass resolution, as summarized in Figure 3. Relevant m/z values for identified CLs are reported in Table S1. Examples of abundant charge-reduced CLs species identified in the E. coli. extract include [CL 70:3 – H]¯ (m/z 1430), [CL 68:3 – H]¯ (m/z 1402), and [CL 66:2 – H]¯ (m/z 1376). Highlighting the dynamic range of the developed technique, minor components such as [CL 73:4 – H]¯ (m/z 1470) representing just 1% of the CL profile, could also be reliably identified, noting that we restrict CL identification to CL molecular species yielding charge-reduced [CL – H]¯ ions post-ion/ion reaction with signal-to-noise (S/N) greater than 10 (Figure 3). The CL profile of E. coli shown here is in good agreement with previous reports.^{18, 21}

Figure 3.

Normalized abundance cardiolipin profile for E. coli extract at the sum compositional level generated via proton transfer ion/ion reactions.

As demonstrated above, proton transfer ion/ion chemistry separates CL dianions from co-existing phospholipid monoanions while concentrating the CL signal into a single charge state. Importantly, this enables the detection of low-level CLs that under conventional workflows are often obscured by much more abundant phospholipid monoanions of the same nominal m/z ratio. For example, direct negative ion nESI of E. coli extract generates an ion at m/z 693.5 that can be assigned at nominal mass to [CL 67:3 – H]²¯ (theoretical m/z 693.4788). Ion-trap CID of the mass-selected ion population at m/z 693.5 generates the product ion spectrum shown in Figure 4a. The CID spectrum of m/z 693.5 exhibits evidence for at least two isobaric lipid structures, including PG 30:0 and CL 67:3 (Figure 4a). Explicitly, product ions likely resulting from fragmentation of [PG 30:0 – H]¯ and [CL 67:3 – H]¯ are indicated with the green and orange circles , respectively. In turn, the presence of multiple isobaric lipid anions complicates CL identification, though tentative structural assignments of PG 14:0_16:0 and CL 16:1_18:1_17:1_16:0 can be proposed from Figure 4a. As demonstrated with this example, proton-transfer ion/ion reactions have the major advantage of alleviating ambiguities arising from composite tandem mass spectra. To illustrate, the mass-selected precursor ion at m/z 693.5 was reacted with Proton-Sponge® reagent cations, giving rise to the product ion spectrum presented in Figure 4b. The proton-transfer reaction resulted in a minor product ion population observed at m/z 1388 composed entirely of charge-reduced [CL 67:3 – H]¯ ions (Figure 4b). While a minor charge-reduced CL 67:3 anion was observed following the ion/ion reaction, the major lipid contributor to the precursor ion at m/z 693.5 can be assigned as [PG 30:0 – H]¯, as a dominant ion at m/z 693.5 is observed after the ion/ion reaction. Following the proton-transfer ion/ion reaction, interrogation of the residual ions at m/z 693.5 yields the CID spectrum shown in Figure 4c. Notably, this product ion spectrum confirms PG 14:0_16:0 as the predominant contributor to the original m/z 693.5 precursor ion population, and importantly, shows no signals characteristic of the CL 67:3 structure. Note that interrogation of charge-reduced CL product ions will be discussed in more detail below. In general, this example confirms the utility of proton-transfer ion/ion chemistry to effectively resolve the CL class from other acidic phospholipid classes, permitting the gas-phase chemical separation of CL ions on the m/z scale.

Figure 4. (a).

(a) Ion-trap CID spectrum of m/z 693.5 from E. coli lipid extract. (b) Mutual storage product ion spectrum resulting from the ion/ion reaction between mass-selected m/z 693.5 from E. coli lipid extract and Proton Sponge® reagent cations. (c) Ion-trap CID spectrum of m/z 693.5 from E. coli lipid extract after the proton-transfer ion/ion reaction.

Proton Transfer Reactions and Refill Experiments to Concentrate Charge-Reduced CL Anions

We note that ion/ion reaction rates have previously been shown to be proportional to the square of the charge on the analyte ion provided the reagent ions are in great excess, and therefore doubly charged CL anions formed from the E. coli extract will react four times faster than singly charged phospholipid anions.³⁷ Thus, the proton transfer ion/ion reaction will preferentially charge reduce the dianions over neutralization of singly charged phospholipid anions in the reaction time available. As a result, these monoanions will continue to dominate the mutual storage product ion spectrum and occupy the largest fraction of the ion storage capacity of Q3 (Figure 2b). While the ion/ion reaction successfully separates the CL class from other acidic lipid classes, the resulting [CL – H]¯ anions represent a minor fraction of the total ions observed in Figure 2b, in part due to their naturally low-levels in lipid extract. Therefore, even after the ion/ion reaction, relatively long integration/averaging times would be required to obtain CID spectra for mass-selected charge-reduced [CL – H]¯ anions, particularly for minor CL species. However, the extent of averaging was minimized here by conducting multiple ion fill and ion/ion reaction steps with intervening removal of residual low m/z singly charged ions and transfer of the CL monanions to an adjacent ion trap. This process allows for the accumulation trap to be filled to capacity with the CL anions of interest, thereby eliminating deleterious space charge effects arising from the chemical noise. We note that a similar multiple fill and ion/ion reaction approach has been used hybrid Orbitrap platforms in whole protein characterization applications.^{38, 39} The net effect in the present context, however, is the gas-phase separation and concentration of the CL anions.

To concentrate charge-reduced CL product ion signal, we developed an approach employing multiple ion fill and ion/ion reaction steps. Charge-reduced [CL – H]¯ product ions are first generated using the ion/ion reaction between E. coli lipid anions and proton transfer reagent cations as outlined above. Next, low-mass anions are removed from q2 by increasing the q2 RF value after the ion/ion reaction while charge-reduced CL product ions are retained and transferred to the LIT for storage (Figure S4). To accumulate charge-reduced CL anion signal, these steps can be repeated iteratively prior to MSⁿ experiments and mass analysis in Q3. Figure 5 displays the results of this gas-phase concentration strategy. In comparison to averaging, in which S/N increases as the square root of the number of averages, refill experiments offer the primary benefit of a linear improvement in S/N with the number of refills. For instance, roughly the same S/N can be obtained from a 10-refill experiment or 100 averages of a single fill. A single mutual storage ion/ion reaction step between E. coli lipid anions and Proton Sponge® reagent cations (i.e., a single refill experiment) yields the product ion spectrum shown in Figure 5a. In comparison to a single fill experiment, the peak area of m/z 1430 increased by a factor of 1.9 with two fill cycles (Figure 5b), while three fills yielded a 2.6-fold increase in m/z 1430 peak area (Figure 5c). In general, these results are in good agreement with theory, as S/N should improve linearly with refill experiments. After total of three fill cycles, the peaks corresponding to the low-mass [CL – H]¯ product ions are exhibiting significant broadening which is a hallmark of space-charge effects, indicating that the charge capacity of Q3 has been reached (Figure 5c). While data presented in these experiments utilized only a total of three fills, the theoretical limit of this repetitive process is dependent on the ion storage capacity of the LIT. In other words, once the charge capacity of Q3 is reached, there is no additional benefit to continue refilling the trap. However, depending on initial ion abundances, variations in the total number of Q3 refills may be required to attain the LIT charge capacity, and, in principle, refill experiments can be conducted as many times as software limitations permit.

Figure 5.

Demonstration of refill experiments to concentrate charge-reduced [CL – H]¯ product ion signals. All mutual storage product ion spectra result from ion/ion reaction between lipid anions derived from E. coli. extract and Proton Sponge® reagent cations, followed by an increase in q2 RF value. Note that all spectra represent an average of 20 scans. Product ion spectra obtained via (a) 1, (b) 2, and (3) 3 fill cycles.

Identification of CLs in E. coli Extract Using Proton-Transfer Ion/Ion Reactions and Refill Experiments

To assign the fatty acyl composition of CLs present in E. coli extract, charge-reduced [CL – H]¯ ions were interrogated. Following the proton transfer ion/ion reaction, [CL – H]¯ product ions were accumulated in Q3 using a total of three fill cycles (Figure 5c). After the separation and concentration process, individual charge-reduced CL anions were mass-selected with unit resolution and subjected to ion-trap CID. In this work, we examined the charged-reduced [CL 70:3 – H]¯ (m/z 1430) and [CL 67:3 – H]¯ (m/z 1388) anions. As shown in Figure 6, CID of the mass-selected [CL 70:3 – H]¯ ion reveals the presence of at least three isomers. Product ions related to the major CL isomer are denoted with orange circles, while those attributed to the minor isomeric contributors are indicated with the green and purple circles (see Figure 6). The lower m/z region of the CID spectrum contains the 16:0, 17:1, 18:1, and 19:1 carboxylate fatty acid anions, noting that the carboxylate anions assigned to 17:1 and 19:1 could also represent cyclopropyl fatty acids.²¹ However, as cyclopropyl FAs cannot be differentiated from their isomeric straight-chain unsaturated FA counterparts based on observed m/z ratios alone, we resort to reporting these fatty acyl substituents as the sum composition equivalents. While there is evidence for an additional CL 70:3 isomer, as indicated by the [16:1 – H]¯ and [18:0 – H]¯ ions, these carboxylate anions and the corresponding product ions generated by the neutral losses of the 16:1 and 18:0 fatty acyl groups were not well-defined. Therefore, we only report the presence of three structural isomers of CL 70:3. From Figure 6, the product ion observed at m/z 699 could correspond to either [PA 18:1_18:1 – H]¯ or [PA 19:1_17:1 – H]¯, while the ion at m/z 673 likely reflects the [PA 16:0_18:1 – H]¯ ion. Further dissociation of these PA ions facilitated confident assignment of the fatty acyl constituents (Figure S5 and Figure S6).

Figure 6. (a).

(a) Ion trap CID spectrum of the charge-reduced [CL 70:3 – H]¯ ion (m/z 1430) that has been concentrated in Q3 using a total of three fill cycles. (b) enlargement of m/z 550 – 950 mass spectral region shown in (a). The lightning bolt indicates the ion subjected to ion-trap CID.

The CID spectrum of m/z 673 displays prominent [16:0 – H]¯ (m/z 255) and [18:1 – H]¯ (m/z 281) product ions, confirming the PA 16:0_18:1 structure (Figure S5). Similarly, re-isolation and ion-trap CID of the product ion at m/z 699 corroborated the PA 18:1_18:1 assignment, as a dominant 18:1 carboxylate anion was observed (Figure S6). Therefore, the major CL 70:3 isomer was identified as CL (18:1_18:1)_(18:1_16:0). As described by Hsu and Turk¹⁸, relative abundances of PA and fatty acyl constituent product ions can be exploited to assign regiochemistry. While these assignments should be interpreted with caution in the absence of authentic standards,¹⁹ these data suggest CL 18:1/18:1/16:0/18:1 is the major isomeric contributor to the CL 70:3 sum composition in E. coli extract. Additional minor contributors to CL 70:3 in E. coli. extract were identified as CL (18:1_17:1)/(19:1_16:0) and CL (18:1_19:1)/(17:1_16:0). Note that sn-positions were not assigned in the case of minor isomeric components, as individual [PA – H]¯ ions arising from CID of the charge-reduced CL anions cannot be re-isolated and collisionally activated due to exceedingly low product ion abundances. While these data were obtained through repeated cycles of ion/ion reaction (i.e., Q3 refill), for comparison, analogous information could was acquired through signal averaging of repetitive single-fill experiments (Figure S7). To achieve a comparable S/N however, signal averaging of single fill spectra required acquisition times 4–5 times longer highlighting the critical advantages of the multiple-fill sequence.

In an additional example, we examined a minor CL component present in E. coli extract. From Figure 3, CL 67:3 represents approximately 2.5% of the CL fraction in E. coli extract. As highlighted above with Figure 4, the precursor ion at m/z 693.5 contains a mixture of isobaric species present within a nominal m/z unit. Specifically, the major contributor to the precursor ion population at m/z 693.5 is [PG 14:0_16:0]¯, while the minor component was identified as [CL 67:3 – 2H]²¯. In turn, unambiguous CL identification is not easily achieved when relying on MS/MS of the [CL 67:3 – 2H]²¯ precursor anion alone (Figure 4a). Therefore, we utilized the separation/concentration steps described herein to both separate CL 67:3 from other lipids and enrich the resulting [CL 67:3 – H]¯ product ions in the gas-phase. Once more, Q3 was filled with proton-transfer product ions a total of three times, and the charge-reduced [CL 67:3 – H]¯ (m/z 1388) anions were mass selected with unit resolution and interrogated via ion-trap CID. The resulting [CL 67:3 – H]¯ (m/z 1388) CID spectrum is shown in Figure 7. In general, CID of the charge-reduced [CL 67:3 – H]¯ anion provides acyl chain composition and sum composition of the constituent PA moieties. The low-mass region of the charged reduced CL 67:3 CID spectrum displays abundant [16:1 – H]¯ (m/z 253), [16:0 – H]¯ (m/z 255), [17:1 – H]¯ (m/z 267), and [18:1 – H]¯ (m/z 281) product ions, while minor [14:0 – H]¯ (m/z 227) and [19:1 – H]¯ (m/z 295) anions were also observed (Figure 7a). The PA composition of the dominant isomer was given by product ions observed at m/z 671 and 659, indicating the PA 34:1 and PA 33:1 constituents, respectively. In turn, dissociation of the charge-reduced CL 67:3 product anion indicates that the dominant isomer is CL (18:1_16:1)_(17:1_16:0). Ultimately, this assignment is in good agreement with previous reports.^{20, 21} The data also show evidence for multiple minor isomeric structures, as the [PA 31:1 – H]¯ (m/z 631), [PA 32:2 – H]¯ (m/z 643), [PA 32:1 – H]¯ (m/z 645), [PA 34:1 – H]¯ (m/z 673), [PA 35:2 – H]¯ (m/z 685), [PA 35:1 – H]¯ (m/z 687), and [PA 36:2 – H]¯ (m/z 699) product ions were detected (Figure 7b). Explicitly, CL (16:0_18:1)_(16:1_17:1) was identified as a minor isomeric contributor to CL 67:3. Further structural identifications could not be made, as low abundances of constituent PA product anions arising from minor isomeric contributors hindered subsequent re-isolation and interrogation. Importantly, we note that these minor isomeric contributors to CL 67:3 have only been previously identified when employing condensed-phase fractionation of the CL class prior to analysis, highlighting not only the dynamic range of the developed platform but also the ability to improve S/N entirely within the mass spectrometer. In total, proton-transfer ion/ion reactions permit rapid identification of CL from total lipid extract without the need for prior chromatographic separation of the CL class. The approach outlined herein provides an alternative approach to CL profiling conducted entirely in the gas-phase. More so, CID of the charge-reduced CL anions enabled detailed structural characterization, including the assignment of fatty acyl groups, PA constituents, and, in some cases, regiochemistry. Furthermore, application of the methods described above enabled CL isomeric discrimination, revealing the presence of multiple isomeric contributors to a CL species in E. coli extract.

Figure 7. (a).

(a) Ion trap CID spectrum of the charge-reduced [CL 67:3 – H]¯ ion (m/z 1388) that has been concentrated in Q3 using a total of three refill experiments. (b) enlargement of m/z 625 – 705 mass spectral region shown in (a).

Conclusions

Localized almost exclusively in the inner mitochondrial membrane of eukaryotes, CLs are complex diphosphatidylglycerol phospholipids that carry four fatty acyl chains and serve vital biological roles in mitochondrial energetics and membrane dynamics.^1–5 While recent studies link CL dysregulation with numerous pathologies, including several types of cancer^{3, 5, 7, 9–11}, CL analysis is challenging, in part due to their vast structural complexity. Owing to their dimeric nature, direct negative ion ESI favors formation of doubly deprotonated [CL – 2H]²¯ anions, though minor singly deprotonated [CL – H]¯ anions can be observed. Without prior chromatographic isolation of the CL class, identification and subsequent characterization of [CL – 2H]²¯ anions generated via direct negative ESI of total lipid extract can be problematic. Explicitly, the resulting mass spectra are typically dominated by the more abundant singly charged phospholipid anions, and therefore doubly charged CL anions are often overshadowed by these coexisting lipid monoanions, hindering confident CL identification.

Herein, we described a shotgun lipidomics approach for the separation, concentration, and identification of CLs from total lipid extract utilizing multiple iterations of gas-phase proton transfer ion/ion reactions. Specifically, [CL – 2H]²¯ anions were transformed in the gas phase upon reaction with proton transfer reagent monocations, giving rise to charge-reduced [CL – H]¯ anions. Exploiting this entirely gas-phase approach, CL dianions are effectively separated on the m/z scale from coexisting singly charged phospholipid anions, permitting rapid identification of CLs without recourse to solution-based separations prior to MS analysis. However, as CLs represent minor fractions of the cellular lipidome, charge-reduced CL anions are still observed in low-abundance. To concentrate [CL – H]¯ product ions, we selectively refill the LIT with proton-transfer product ions. Importantly, this iterative process allows for the accumulation of the desired CL product ions and dramatically reduces the averaging time needed to achieve good S/N product ion ratios. Application of this strategy to the analysis of E. coli. lipid extract permitted relative quantification of CLs at the sum compositional level with high sensitivity. Furthermore, interrogation of charge-reduced CLs derived from E. coli extract enabled the identification of numerous isomeric contributors to a single CL molecular species. Collectively, the primary advantage to CL profiling utilizing proton transfer ion/ion chemistry is the elimination of condensed-phase CL class isolation prior to MS-analysis, reducing both sample volume requirements and analysis time. Consequently, the presented approach provides an alternative strategy to CL identification based entirely on gas-phase chemistries and offers improved mixture analysis performance, explicitly as CL anions are effectively resolved from more abundant, isobaric phospholipid monoanions, and subsequently concentrated in the gas-phase. Lastly, as demonstrated herein, gas-phase proton transfer ion/ion reactions in combination with MSⁿ can be readily applied to elucidate the structural complexities of CL molecular structures, as interrogation of charge-reduced CL anions revealed the presence of multiple CL isomers within a single mass-selected CL precursor anion population.