Advances in methods to analyse cardiolipin and their clinical applications

Cardiolipin (CL) is a mitochondria-exclusive phospholipid, primarily localised within the inner mitochondrial membrane, that plays an essential role in mitochondrial architecture and function. Aberrant CL content, structure, and localisation have all been linked to impaired mitochondrial activity and are observed in the pathophysiology of cancer and neurological, cardiovascular, and metabolic disorders. The detection, quantification, and localisation of CL species is a valuable tool to investigate mitochondrial dysfunction and the pathophysiological mechanisms underpinning several human disorders. CL is measured using liquid chromatography, usually combined with mass spectrometry, mass spectrometry imaging, shotgun lipidomics, ion mobility spectrometry, fluorometry, and radiolabelling. This review summarises available methods to analyse CL, with a particular focus on modern mass spectrometry, and evaluates their advantages and limitations. We provide guidance aimed at selecting the most appropriate technique, or combination of techniques, when analysing CL in different model systems, and highlight the clinical contexts in which measuring CL is relevant.

CL exists with a wide variety of acyl-chain compositions that differ in length and saturation, and alterations in the molecular conformation of CL provide insights into its impaired synthesis/ remodelling and disease pathophysiology [13]. Studies exploring CL acyl chain composition in cultured cells and murine tissues confirm that CL composition is partially controlled by tetra-linoleic acid availability and, in most tissues, the tetra-linoleic form of CL (18:2) 4 is the most abundant species [14e16]. Notably, the brain exhibits a diverse CL acyl chain profile comprised of longer FA side chains (i.e., 20:4 and 22:6), possibly resulting from a reduced FA 18:2 import across the blood-brain barrier and subsequent incorporation of long-chained FA.
Understanding the diversity of tissue-specific CL acyl chain composition is crucial for the development of analytical assays and to interpret alterations in CL species under pathological conditions. The clinical significance of CL measurement is clearly established in the pathomechanism of Barth syndrome (BTHS), a monogenic, ultrarare disorder caused by mutations in the TAZ gene that encodes a mitochondrial transacylase, involved in the remodelling of MLCL into mature CL [17,18]. Patients with BTHS suffer from cyclic neutropenia, skeletal and cardiac myopathies, and growth retardation. Biochemically, BTHS patients have elevated MLCL levels and MLCL/ CL ratio, which is commonly used as a sensitive diagnostic marker [19e21]. This emphasises the utility of measuring different acylchain compositions and other CL-related phospholipids during the clinical evaluation of patients with suspected CL-related disorders. Despite the medical relevance of MLCL, the interplay between MLCL and CL is not fully understood. Recently, mutations in the CLS1 and TAMM41 genes have been identified to cause multisystem mitochondrial disease, further highlighting the clinical relevance of CL [22,23]. Increasing evidence links aberrant CL metabolism and content to human disease, supporting the importance of developing sensitive and specific high-throughput methods for CL analysis and quantification. Table 1 summarises human conditions associated with CL abnormalities, including neurological disorders [24], cancer [25], and cardiovascular and metabolic disorders [26].
CL is a pharmacological target for Elamipretide, a small molecule under investigation in several clinical trials reported to improve the mitochondrial respiratory function via CL binding and stabilization [27]. CL also has significant potential as a tissue-specific biomarker as exemplified by the presence of brain-specific CL species in the plasma of patients following cardiac arrest [28]. Thus, CL has potential to diagnose and monitor disease progression, in addition to measuring efficacy during clinical trials, for several disease states.
Detection and quantification of CL is a valuable tool for confirming the presence of mitochondrial dysfunction, characterising pathophysiological mechanisms of disease, and clinical diagnostics. CL assays require specificity, due to the high diversity of CL species, and sensitivity, given the low abundance of CL within the cellular lipidome. Consequently, in parallel to the quantitative analysis of CL, qualitative measurements are required to enable the distinction of individual sample components. Identification of the different CL species is an intricate task and different methods are available depending on the scope of the analysis. In this review, we summarise the techniques currently available for measuring CL and discuss the advantages and disadvantages for each method when attempting to characterise CL content, structure, and localisation. A primary focus is on recent advances in mass spectrometric analysis of CL. Finally, we highlight the clinical scenarios in which CL measurement has existing and potential importance. Abbreviations: Acyl-CoA, acyl-coenzyme A; CDP-DAG, cytidine diphosphate-diacylglycerol; CL, cardiolipin; CLS1, CL synthase; CMP, cytidinemonophosphate; CoA-SH, coenzyme A (unconjugated); CTP, cytidinetriphosphate; FFA, free fatty acid; G3P, glycerol-3-phosphate; H 2 O, water; IMS, intermembrane mitochondrial space; LPC, lyso-phosphatidylcholine; LPE, lyso-phosphatidylethanolamine; MLCL, monolysocardiolipin; MLCLAT1, monolysocardiolipin acyltransferase 1; OMM, outer mitochondrial membrane; PA, phosphatidic acid, PC, phosphatidylcholine; pCL, premature CL; PE, phosphatidylethanolamine; PG, phosphatidylglycerol; PGP, phosphatidylglycerol phosphate; PGS1, phosphatidylglycerol phosphate synthase; P i , inorganic phosphate; PLA 2 , phospholipase A2; PP i , inorganic pyrophosphate; PRELID1, PRELI Domain Containing 1; PTPMT1, protein-tyrosine phosphatase mitochondrial 1; TAMM41, TAM41 mitochondrial translocator assembly and maintenance homolog; TAZ, tafazzin; TRIAP1, TP53 regulated inhibitor of apoptosis 1.

Methods to detect and quantify cardiolipin
A wide range of methods are available to measure CL. However, the specificity and sensitivity vary according to the nature of biological sample, separation strategy, and detection technique. The different analytical strategies can be divided into five broad groups ( Figure 2): 1) liquid chromatography, usually combined with mass spectrometry (LC-MS); 2) mass spectrometry imaging (MSI); 3) shotgun lipidomics; 4) ion mobility spectrometry (IMS); and 5) fluorometry and radiolabelling.
The focus of this review will be on mass spectrometric techniques, which are becoming widely available and have many advantages, in terms of sensitivity and specificity, over older chromatographic and fluorometric techniques. In recent years, LC-MS-based multi-analyte/lipidomic assays have also been introduced in clinical practice (see Ref. [29] for review).

Lipid extraction prior to analysis
Prior to measurement, biological lipids must be extracted and  isolated. One exception is when fluorescent dyes are used, such as 10-N-Nonyl acridine orange (NAO), which can be applied directly to cells or tissue [30]. Extraction of lipids from biological samples is a crucial pre-analytical step to ensure optimal chromatographic separation of low abundant species, such as CL. Several lipid extraction methods have been described; however, the two most widely used are Folch (1957) and Bligh and Dyer (1959) protocols [31,32]. Both methods utilise chloroform and methanol in different ratios to partition sugar and proteins in the top aqueous phase, and lipids in the lower organic phase. The organic layer is subsequently removed, dried down, and resolubilised for analysis. Despite representing gold standard methods for lipid extraction, these two chloroform-based protocols present some disadvantages, including the use of a toxic and carcinogenic solvent (i.e., chloroform), and the presence of lipids in the bottom sample phase, which increases the risk of contaminating the lipid mixture during extraction. Various protocols using simpler lipid extraction techniques and fewer toxic solvents have therefore been developed. Among these, the methyl tert-butyl ether (MTBE) lipid extraction procedure has proved faster and safer than conventional methods [33]. In addition to lower toxicity compared with chloroform, the advantages of this method include good recovery for all lipid groups and the presence of an organic layer settled on top of the aqueous phase, thus enabling easier automation of the extraction procedure. One potential disadvantage is represented by water contamination of the organic phase, which results in longer drying down times and carryover of contaminant species that can cause ion suppression. Another extraction protocol is based on the BUME (Buthanol:Methanol) method, which uses butanol and methanol resulting in a single phase extraction [34]. The initial single phase extraction with Butanol:Methanol (3:1) is followed by addition of heptanol:ethyl acetate (3:1) and 1% acetic acid. The top organic layer contains the lipids and can be removed, dried down, and resuspended for analysis. Solid phase extraction (SPE) has also been used as an additional clean-up step before phospholipid/CL analysis. Helmer and colleagues used a simple hydrophilic interaction liquid chromatography (HILIC)-based SPE cartridge method to clean up lipid extracts and look at oxidised CL (CLox) and MLCLs [35]. This method is described as quick and simple and it avoids the use of strong organic solvents, such as hexane, which are typically used for normal phase SPE of lipids.

Analytical techniques for CL analysis
Following extraction, complex lipids (e.g., CL) can be separated by either LC or thin-layer chromatography (TLC) [36], and the individual FA side chains analysed by gas chromatography (GC). These techniques are not described in detail here as they have largely been superseded by modern hyphenated mass spectrometric methods. Details of these techniques are summarised in Table 2.

Liquid chromatography -mass spectrometry (LC-MS)
The sensitivity and accuracy of CL identification and quantification by chromatography has benefited from rapid advances in mass spectrometry. Most techniques involve separation of individual classes of lipid by LC followed by MS.
LC separation uses a non-polar or polar stationary phase and involves dissolution of the lipid mixture in a liquid mobile phase, enabling partitioning of single lipid components according to their polarity and molecular interactions with the stationary and mobile phase. The sample is injected and separated on a column with a densely packed stationary phase. A wide variety of different stationary phases, column parameters, and elution solvents are available. The separated compounds have a characteristic retention time and can be measured by the mass spectrometer as they are eluting from the column.
Lange et al. compared and described reversed-phase, normal phase, and HILIC for the separation of lipids [37]. Reversed phase is currently the most used technique for phospholipid separation; however, HILIC is becoming increasingly popular. HILIC separates the lipid mixture based on the polarity of the phospholipid headgroup and has the advantage to have greater compatibility with electrospray ionisation techniques when compared to normal phase. Figure 3 summarises the overall workflow to analyse CL by LC-MS, from sample collection to the statistical analysis. Several mass spectrometry ionisation and detection techniques are available. Following HPLC separation, the most commonly used ionisation method is electrospray ionisation (ESI). This is a 'soft' ionisation technique that enables ionisation of molecular species, without significant in-source fragmentation; singly or multiply charged ions are produced from the liquid eluate derived from the HPLC column. Many studies have used HPLC-ESI-MS to quantify CL in both singly and doubly charged states (see Refs. [21,38] for examples). The use of tandem mass spectrometry techniques has provided important structural information and improved the detection limits enabling analysis of lipids in the picomolar range.
'Ultra-high-performance' liquid chromatography (UHPLC) utilises smaller particles in the stationary phase and at higher pressure than HPLC. This enables faster and more sensitive measurements, thus facilitating higher sample throughput. The methodology for measuring CL with UHPLC has previously been described [39]. Importantly, UHPLC-MS is compatible with HILIC, thus supporting separation of polar metabolites [37]. Although this approach has been used for CL analysis [40], reversed-phase HPLC is usually used to separate CL, based on the hydrophobic nature of the FA side chains [14]. Normal phase chromatography has also been utilised and is recommended by some authors to prevent ESI matrix effects [41], although it can also suppress ionisation of analytes due to the nature of the organic solvents used. Table 2 summarises common LC-MS methods employed for the detection of CL in multiple biological samples. HPLC-MS has been used to explore CL species as a diagnostic biomarker for human disease, with different pathologies presenting specific changes. For instance, decreased CL levels have been identified in patients with frontotemporal dementia (FTD, serum) [42], traumatic brain injury (TBI, brain tissue) [43], heart failure (cardiac tissue) [44,45], and hepatocellular carcinoma (tumour tissue) [46]. The specificity of HPLC-MS is particularly relevant when total CL content is normal, and the pathology is linked with aberrant levels of specific CL species or the ratio of CL with other phospholipids. For example, CL enriched with palmitoleic acid (CL-16:1) was detected in five of six patients with prostate cancer, and might explain the higher proliferation rates within the tumour cells [47]. Similarly, despite normal total CL levels, an increase in specific subspecies is reported and confers potential as a biomarker [48]. Finally, the MLCL/CL 4 ratio is increased in BTHS [19,49e52] and used as a diagnostic assay, further emphasising the advantages of applying mass spectrometry to study CL species. Once separated, they may be further fractionated by adsorption, ion-exchange chromatography, or by combinations of both. Over the last decade, there has also been an increasing demand for methods to identify and quantify oxidised lipids (including CL), given their potential application as disease biomarkers. Reactive oxygen species (ROS) and reactive nitrogen species (RNS) readily modify the chemical structure of CL species and other phospholipids [53] due to the presence of unsaturated FAs. The reaction between ROS and CL leads to formation of CLox. The identification of CLox, and method standardisation, is a challenge because of the number and low abundance of such oxidised species. HPLC-MS has been crucial in elucidating mechanistical and clinical insights of CLox [54e57]. Importantly, CLox is implicated in the pathophysiological development of several conditions, including neurodegeneration, diabetes, myocardial infarction, and ageing [58]. Furthermore, HPLC-MS has facilitated the characterisation of cellular mechanisms that promote release of pro-apoptotic factors triggering the cell-death pathway; for example, the oxidation of polyunsaturated CL species by cytochrome c in fatally injured cells initiates the apoptotic process [59]. In addition, experimental TBI in  rats leads to accumulation of more than 150 newly oxidised molecular species of CL that precipitate neuronal death [60]. CL oxidation has been also observed in experimental cerebral ischemia-reperfusion in rodents using 2D-LC-MS [61], and delivery of mitochondrially-targeted small molecule inhibitors, as electron scavengers, has shown to prevent accumulation of CLox products and the pathological consequences of brain injury. Thus, CL oxidation represents a target for neuro-drug discovery [60]. The effects of RNS on CL species and cellular functions remain largely unexplored. However, a new HPLC-MS method can detect nitroso, nitrated, and nitroxidised CL products, offering new opportunities to advance understanding in the pathophysiological consequences of RNS [62]. Finally, ultra-high-performance supercritical fluid chromatography (UHPSFC) has been suggested as a potential lipidomic method for CL measurement. UHPSFC uses a supercritical fluid, such as carbon dioxide, as a mobile phase. This allows low back pressure, high flow rates, and good solubility of lipids. UHPSFC has used to measure CL in porcine brain extracts [63]. A recent review of UHPSFC describes the benefits of this approach for speed of analysis and improved separation of non-polar lipids, compared to more traditional HILIC and reversed-phase chromatography [64].

Mass spectrometry imaging (MSI)
CL abundance is relatively low in comparison with other cellular and tissue lipids. Consequently, there is limited understanding of the spatial distribution for CL using mass spectrometry in disease states; HPLC requires cellular homogenisation and precludes acquisition of spatial information.
In recent years, novel and label-free imaging techniques that are compatible with mass spectrometry have been developed to investigate mitochondrial content and acyl-chain composition, using a visualisation approach [65,66]. This has enabled mapping of lipid profiles and their contribution to cellular structure and pathology. Two major soft ionisation MSI techniques are used to visualise CL content: matrix-assisted laser desorption/ionisation mass spectrometric imaging (MALDI-MSI) and desorption electrospray ionisation mass spectrometry imaging (DESI-MSI).
Tissue imaging by MALDI-MSI is the most commercialised MSI technique. It offers label-free spatial resolution of diversified CL species and other phospholipids in tissues [65,67]. This tool allows the visualisation and mapping of diversified CL species across different brain areas and the structural MS/MS fragmentation and mapping of CL species, assuming use of a suitable mass detector. MALDI-MSI has shown a non-random distribution of individual oxidised and non-oxidised CL species across different brain tissues [67].
The selection and application of the matrix are critical for the absorption of the laser wavelength and the ionisation of lipids in the tissue sections. The optimisation of the most suitable MALDI matrix should be evaluated on the tissue sections and CL standards. Conventionally, 2,5-dihydroxy benzoic acid is the preferred matrix for lipid mapping and spectra using either spraying or sublimation protocols [65,68].
A variation of MALDI-MSI employs Fourier-transform ion cyclotron resonance (FTICR). MALDI-FTICR-MSI has been applied to interrogate different lipids in prostate cancer. Interestingly, the high intensity detection of specific CL species correlated with more severe phenotypes of tumour regions, suggesting CL as a promising biomarker for prostate cancer [69,70]. The technique has also been combined with time of flight mass spectrometry (MALDI-TOF-MSI) to measure CL [71]. This tool has been used to diagnose BTHS in patient-derived leukocytes, based on increased MLCL and pCL ratio over mature CL, and the changes in CL spectrum peaks in BTHS compared with healthy controls [72]. One advantage of this method is the simultaneous detection of CL and MLCL species using a single run of mass spectrometry analysis.
Another alternative technique is DESI-MSI, which uses an ionisation technique that directly electrically charges the sample surface [73]. DESI-MSI can be performed at ambient operating conditions and with minimum sample preparations. However, this technique has a lower spatial resolution (30e50 mm) than MALDI-MSI (<10 mm) [74]. MALDI-MSI has shown increased content and chemical diversity of CL species in oncocytic thyroid tumors [75].
Finally, tandem use of MALDI-MSI and DESI-MSI has shown high spatial and mass resolution of CL species, and other phospholipids and gangliosides, detecting multiple analyte classes from tissue samples [76].

Shotgun lipidomics
Shotgun lipidomics is the term used to described direct infusion mass spectrometry of lipid extracts without prior chromatographic separation [29,77]. Although it lacks the chromatographic resolution of LC-MS methods, when combined with a powerful mass spectrometer it can be used to undertake multiple mass spectrometry experiments in succession. It is generally used with high mass resolution mass spectrometers, including time of flight (TOF) and orbitrap mass spectrometers [29], and has been used for the analysis of CL species [29,77].

Ion mobility spectrometry (IMS)
Owing to the variety and complexity of lipids and their sidechains, it is not always possible to unambiguously identify components by their mass and fragmentation patterns alone. Many species are isobaric and have identical atomic composition. To enable identification of such isobaric mixtures, technologies such as IMS have been introduced. IMS, a gas-phase electrophoretic technique, allows separation of these isobaric ions in the gas phase according to their shape, charge, and size. It therefore adds an additional dimension to the separation of lipid components. There are a variety of IMS technologies depending on the type of mass spectrometer being used [78]. Further details with regards to the use of IMS in lipidomics are given in a review by Paglia et al. [79]. Addition of field asymmetry ion mobility spectrometry (FAIMS) has been shown to enrich and increase the sensitivity for low abundance doubly charged CL species [80].

Quantification of cardiolipin species by mass spectrometry
Quantification in mass spectrometry is usually performed using a stable isotope labelled internal standard with essentially identical chemical properties to the analyte of interest. This internal standard accounts for losses during sample processing and matrix suppression effects. The amount of analyte can be determined by the ratio of the analyte to internal standard and comparison to calibration curves. Quantification of CL and other lipids is complicated by the large number of species with different FA side chains present in each sample and the lack of appropriate stable isotope standards for each of these species. This is further complicated by the overlap of isotopic peaks (e.g. Mþ2) with those of CL species having saturated side chains (also an addition of two mass units for the loss of each double bond.). The problems and solutions to quantification of CL species are discussed in detail by Tatsuta [81]. The original CL detection method used a single non-naturally occurring internal standard (CL-(14:0) 4 ). Other CL species are now available to be used as internal and external standards (e.g., CL mix by Avanti lipids). These mixed standards enable better quantification as they account for differences in chromatographic separation, matrix effects and ionisation efficiencies between CLspecies with different side chains. There is still a need to increase the availability (i.e., number and type) of CL standards. This is particularly important for studies of CLox species where no commercial standards are currently available.
Analysis and quantification of CL mass spectrometry data can be complex. Various proprietary (e.g. LipidView (SCIEX) [81]) and open source software (e.g. MZmine 2/3 [35]) have been used to aid the analysis of these data.

Clinical diagnostics
Several techniques are available for CL detection and/or measurement. However, the choice of the analytical method depends on the experimental question, level of detail, and sensitivity required. Table 3 summarises specific advantages and disadvantages of each method. Additional factors that influence this decision include the tissue or cell type available for analysis, the clinical and/ or research question, and accessibility to the diagnostic facilities. Currently, few centres offer diagnostic CL analysis, and the focus is primarily aimed at diagnosing suspected BTHS using LC-MS. However, as the CL profile abnormalities linked with different human diseases have expanded, an urgent need to provide qualitative and quantitative diagnostic CL measurement has emerged. Examples of pathological states that would potentially benefit from detailed analysis of CL species include novel defects in CL synthesis and remodelling [22,82], mitochondrial disorders, FTD [42], TBI [43], cancer [46,47] and cardiovascular diseases [28]. Most of these disorders show changes in CL species in multiple biological tissues, which require modern mass spectrometric techniques for detection.
Several human biological samples can be utilised when investigating the role of CL in human health and disease, including cell lines (e.g., patient-derived fibroblasts), tissues (e.g., brain, liver, cardiac, and skeletal muscle), and biological fluids (e.g., blood, urine, and cerebrospinal fluid). In addition, induced pluripotent stem cells differentiated into disease-relevant cell types (e.g. cortical neurons, myotubes, or cardiomyocytes) enable researchers to characterise CL in a tissue-specific human disease model [83,84]. One important consideration is the choice of cell culture medium and growth conditions used, which may affect CL composition and mitochondrial function [16]. Importantly, high-end mass spectrometers are now increasingly available within diagnostic services. Consequently, any clinically relevant research findings are readily transferrable to diagnostic laboratories.

Conclusion and future perspectives
CL is emerging as a potential biomarker to diagnose and monitor disease progression, and as a potential pharmacological target, in several disease processes, including neurodegenerative disorders, cancer, cardiovascular and metabolic disorders. Despite advances in lipidomic techniques, challenges remain to ensure a complete understanding of CL metabolism is achieved. LC-MS and MSI techniques that assess individual phospholipid species have helped discern CL tissue-specificity and acyl chain composition. However, some of these methods are highly technical and require specialist equipment. Consequently, more cost-effective, scalable methods and/or probes that enable sensitive and reliable measures of CL would benefit diagnostic services, and selecting the most appropriate, disease-relevant assay is crucial. This may require the use of simplified, more targeted mass spectrometric assays, looking at subsets of CL species, in the clinical diagnostic setting. In future, characterising CL biosynthesis and remodelling pathways will provide additional insights into the pathophysiological implications of aberrant CL, while combining lipidomics and other stateof-the-art multiomics techniques will be required to fully appreciate the role of CL in human health and disease.

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Data availability
No data was used for the research described in the article.