Shotgun Lipidomics of Rodent Tissues

Although advances in MS have yielded a variety of methods to accurately monitor many lipid species, the previous lipid profiling of various mammalian tissues did not show consistent results and, consequently, the particular tissue-specific functions of lipids remain unclear. In comparison with protein functional analysis, for which more robust approaches to knock out the expression of particular compounds are available, the majority of lipids cannot be selectively turned off or overexpressed in tissues, making the functional evaluation of lipids difficult. Advanced profiling of tissue lipidome concentrations may provide an alternative approach to identify associations between circulating lipids and human diseases.

In evaluating comprehensive lipidomics methods capable of qualitatively and quantitatively covering the endogenous lipid profile of any mouse tissue, we gave preference to the shotgun lipidomics method. In general, two opposite types of sample analysis are possible: either a completely untargeted screening of lipids using liquid chromatography (LC)-based separation with further mass spectrometric detection to reveal the total lipid complexity of the sample, or targeted approaches, which mostly allow for very precise quantification of specific lipids of interest. In contrast, the strong feature of the presented shotgun lipidomics workflow is the rapid comprehensive coverage of hundreds of endogenous lipids from predefined lipid classes, which can still be performed in a robust semi-quantitative manner.

The ionization efficiencies of different lipid classes are structure-dependent and can vary drastically depending on the different experimental ionization conditions. In contrast to LC-based separation methods, shotgun analysis minimizes these differences due to the direct simultaneous infusion of the whole lipid extract under the same ionization conditions into the MS instrument. The use of isotopically labeled lipid analogs or structurally similar non-endogenous standards allows semi-quantification of all lipid classes. Shotgun lipidomics provides low inter- and intra-run variability during MS analysis; as a result, this method produces lower coefficients of variation²¹ than untargeted liquid chromatography-based methods, which require multiple standards for adequate quantification. Importantly, while no external calibration curve is used in the current method, the method is still considered fully quantitative³².

One level of labeled internal standard (or unlabeled standard that is not endogenously expressed) per lipid class is sufficient for the quantification of most lipids. Only a few publications have reported a partial method validation for a shotgun lipidomics method. For example, in Gryzbek et al.¹⁷ and Surma et al.²¹, reversed calibration curves were prepared using internal standards and a fixed amount of sample matrix. Linearity was assessed by the linear regression of log-transformed lipid amounts and their intensities and reported as R² and slope, respectively. The limit of detection (LOD) and limit of quantitation (LOQ) were determined by weighted linear regression based on a signal-to-noise ratio of 3 for LOD and 10 for LOQ. For most lipid classes, LOQ was defined between 2-9.8 pmol for adipose tissue and 0.05-5µM in plasma. In both cases, non-endogenous single internal standards per class were used to derive estimates for all lipids in the class. However, in this work, we do not provide LOD/LOQ due to several concerns: the endogenous matrix is not compound-free, and the surrogate matrix for tissues is unavailable-with this, the spike of small known amounts of standards is not possible. We do not perform a classical targeted quantification approach with the use of a calibration curve series of a particular compound normalized by its identical isotopically labeled standard due to the nonexistence of pure standards and the very limited availability of isotopically labeled lipids. Orbitrap detectors automatically perform the conversion of the transient signal by applying a Fourier transformation, and some signal is already substituted-as a result, the lower concentration range will be only linear down to some minimum signal, below which the molecule will not be detectable anymore. Xcalibur software signal-to-noise values depend on the m/z ratio of the molecule; as a result, the compounds of each lipid class, containing different combinations of fatty acids, will have different noise values. Moreover, LOQ/LOQ values are strictly linked to the type of matrix, and when quantification of the lipidome is done in different rodent tissues, it should be reflected by the assessment of LOQs for each tissue type individually.

In fact, the method offers a high linear dynamic quantification range of up to four orders of magnitude³³ and very good sensitivity to cover the most important endogenous structural lipids, which can be further increased by technical improvements in MS acquisition³². The coefficients of variation of the mean lipid concentrations were mostly below 15%, meaning that shotgun lipidomics complies with Food and Drug Administration (FDA) requirements as a method to be considered for good laboratory practice (GLP) and good clinical practice (GCLP) studies³⁴.

However, it must be noted that, due to their different polarity, some lipid classes become much more impacted by the contribution of specific conjugated FAs. This leads to the distortion in the intensity response in equimolar mixtures containing a broad range of conjugated FAs, resulting in quantification bias, as highlighted by Koivusalo et al.³⁵ for phospholipid classes. Of note, these authors presented data for a broad range of FAs, from 24-48 chain length, which likely does not reflect the situation in a real biological sample. The response for polyunsaturated species was 40% higher than for fully saturated species, but this effect was only observed for higher concentrations. When the mixture was progressively diluted, the effect of unsaturation gradually diminished and virtually disappeared at 0.1 pmol/µL per species. Additionally, all measurements were done on ion trap and triples quadrupoles instruments and not on Q-Exactive equipment.

Another advantage of the presented workflow is its technical flexibility, which allows adaptation to specific project requirements. For example, any lipid extraction protocol – such as the modified Bligh and Dyer³⁶, methyl tert-butyl ether³⁷, or butanol-methanol³⁸ methods – can be used for preparing a total lipid extract before MS analysis. The main limitation of chloroform-methanol extraction is that the lower phase contains the lipid fraction, which complicates routine work and especially automation; in addition, the toxicity of chloroform must be considered. Tert-butyl ether extraction is widely used for lipid analysis of plasma samples³⁷, and an automated version has been proposed²¹. In this case, we chose the BUME method because it provides even better recoveries for the spiked PG, PI, PA, and PS lipid classes³⁸, lower solvent consumption, and the possibility of automation³⁹, while the overall concentrations quantified for tissue samples extracted by all three methods were comparable. Additionally, while the sample extraction was performed manually in the current work, it is also possible to obtain reproducible and precise results from large sample sets by automated sample preparation and lipid extraction in a 96-well format^40,41. This allows one to implement lipidomics analysis in large-scale clinical and toxicological studies.

In the current work, we performed the MS acquisition of positive and negative modes separately without polarity switching as described by Schuhmann et al.⁴². The stability of the nanomate signal is better in negative mode for a slightly less concentrated solution than in positive mode. Additionally, we developed a fully traceable procedure with converter software from .raw files to mzML, which provides the values to be specified in the LipidXplorer software – with this, manual resolution slope calculations are not required. We also applied different noise setting substitutions because, in positive mode, the noise levels of the spectra are higher than in negative mode. All the steps were optimized for traceable, high-throughput routine analyses.

For identification, shotgun lipidomics analysis can exploit the unique behavior of different lipid classes, which form unique adducts in different polarity modes. In this method, PC and PE species with the same molecular mass that overlap in the positive ionization mode can be fully separated in the negative ionization mode, as PC forms acetate or formate adducts, and PE is ionized in a deprotonated form. Moreover, (partial) structural deconvolution is possible for the method utilizing not only the molecular formula but also the bulk fatty acid structure. For example, FA identification on the level of total carbon and total unsaturation count works for all phospholipids, DAGs, TAGs, and lyso-phospholipids. The bottom-up quantification of each isomeric form can be partially performed for some of the phospholipid classes⁴³ but is much more complex for DAGs and TAGs due to the unequal signal response of different FAs in the MS2 spectra.

It is also important to emphasize the need to implement appropriate quality control procedures, fully aligned with recent initiatives in this field⁴⁴. As we wish to ensure proper data traceability and reproducibility between and within the laboratory, a number of steps have been taken such as proper randomization of the samples for all steps of the analysis, working with supplier-certified standard mixtures, the inclusion of quality control samples, procedures to verify batch acceptance or rejection, and the creation of an internal database to track QC performance over the long term. Also consistent with these initiatives is the need for a standardized method to address sample stability. In general, the majority of structural lipids are not impacted by lipid oxidation, which is more relevant for oxilipins, oxidized lipids, and polyunsaturated fatty acids that are, therefore, much more sensitive to storage and handling conditions. However, the correct assessment of sample stability is still a technically challenging task.

However, this protocol has a limitation when it comes to determining submolecular levels of compounds. Considering that there is no separation of the total extract, all isoforms of lipids with the same molecular mass but different fatty acid compositions are merged in the MS analysis. For most classes, it is possible to achieve partial deconvolution of the structure by using the fragmentation ratios of residual fatty acid fragments in MS2. However, the independent quantification of each isoform remains a particularly challenging task owing to the big differences in the fragmentation behaviors of different isoforms and the fact that pure chemical standards are not available in large enough variety to define the compensation values. Another limitation is that the ESI process inevitably generates artifacts, resulting in the artificial generation of peaks for some lipids such as DAGs, Pas, and FAs, which can lead to erroneous quantification.

Next, we summarize the most critical parts of the protocol based on our experience. The first one is related to the fact that each mouse tissue type has a unique lipid profile in terms of both the lipid amount and class ratios. With this, the starting amounts of the tissue based on the total protein content before extraction must be carefully determined in order not to saturate the MS signal and not to leave the dynamic quantification range due to lipid aggregation at high concentrations³³ or – at the opposite extreme – to provide enough MS signal to cover the major lipid compounds for each lipid class.

The second critical aspect is to ensure a proper alignment of the position of the direct infusion nano-source chip outlet with the transfer capillary of the mass spectrometer. Considering that full calibration of the mass spectrometer in both modes is performed weekly, swapping between the calibration source and the nano-source chip setup can be the reason for dramatic variability in signal intensity due to misalignment during installation.

Another critical part of the protocol is the careful handling of the internal standard mix. As this mixture contains a significant amount of dichloromethane, once opened, it should be consumed quickly to avoid long storage and multiple uses leading to evaporation and artificial concentration change. Moreover, consistent handling of the standard mixture after removal from −20 °C storage is important, since temperature differences can lead to volume inconsistencies during the pipetting with air-cushion pipettes. An option would be to replace dichloromethane in the standard resuspension buffer with pure methanol, which could improve the handling convenience but might negatively impact the solubility of some lipid classes and, thus, decrease quantification accuracy for these lipid classes.

The final critical part is data processing. The data processing workflow is combining the Peak by Peak software conversion from .raw to .mzML format, applying MS2 scan averaging and MS1 and MS2 noise filtration, as well as peak picking and data compression. As an alternative, the Proteowizard software can be also used for data conversion, but, in this case, several settings in LipidXplorer have to be defined manually. All the complexity of shotgun lipidomics is particularly concentrated on the step of deconvolution of direct injection MS1 and MS2 spectra. The open-source LipidXplorer software imports the converted spectra from mzML file format based on mass accuracy, mass resolution, and the slope of its change with increasing m/z. The software merges multiple individual MS and MS/MS spectra acquired within an analysis run. Afterward, it aligns individual peaks within different sample runs, and, in each cluster of aligned peaks, it replaces their masses with their single intensity-weighted average mass, while their abundances in each data file remain untouched. Representative masses of aligned peak clusters and individual peak intensities are stored in a master scan database.The master scan database contains all MS1 and MS2 spectra generated for all samples in the batch and can be deconvoluted to lipid identifications by queries written in the molecular fragmentation query language (MFQL).

Overall, the method covers the identification of DAG, TAG, and SE lipids based on positive mode and PC, PE, PS, PI, PA, PG, SM, LPC, and LPE lipids based on negative mode acquisition. During lipid identification, isotopic correction is performed for MS1 and MS2, and adjusted intensities are reported in the .xlsx output file. Alternatively, several other software are available to process shotgun data, such as ALEX⁴⁵ and LipidHunter⁴⁶.

Fatty acids – major nuclear and cellular membrane components – are stored in the form of phospholipids for further conversion into bioactive molecules. They can be converted by lysophospholipid acyltransferase enzymes into lysophospholipids through the Lands pathway⁴⁷. The LPCAT3 enzyme, for example, is known to show high specificity to incorporate AA into lysophosphatidylcholine and lysophosphatidylserine intermediates. The relatively high expression of these enzymes was reported within inflammatory cells, such as alveolar macrophages and bronchial epithelial cells⁴⁷, leading to the release of larger relative proportions of AA-containing phospholipids in those cells. However, following their liberation from membrane phospholipids by phospholipase A2, the conversion of PUFAs into various types of lipid mediators is catalyzed by many enzymes⁴⁸. Further, these PUFAs can become the substrate for the production of pro-inflammatory and anti-inflammatory prostaglandins via the action of cyclooxygenase (COX)-1 and COX-2 enzymes⁴⁷. Phospholipids are one of the sources of lipid mediators released at particular locations where these mediators are required to demonstrate their biological effects.

The lung is a complex organ comprising multiple cell types, each playing overlapping and niche roles in facilitating normal lung development and function. Only a few studies have been done to isolate and profile different cell types in the mouse or human lung (e.g., alveolar type 2 cells)^49,50; other major lung cell types have not been characterized. Another interesting study⁵¹ was carried out to isolate and perform lipid analysis of endothelial, epithelial, mesenchymal, and mixed immune cells in the mouse lung. It was observed that the concentration of PUFAs incorporated in PCO and PG was enriched in the immune cells. Considering the fact that CS induces an increase in immune cells within the lung (4-5 fold), as assessed by bronchoalveolar lavage (BAL)⁵², the total lipid changes observed in this study could be explained by a cumulative increase in immune cell recruitment in the mouse lung.

In conclusion, the observed distortion of monounsaturated fatty acids and PUFAs incorporated in phospholipids can reflect the excess of certain FAs within the cell and/or constitute an intracellular resource of oxilipin precursors that are overproduced under oxidative stress and inflammatory conditions. However, additional data on free EPA, DHA, and other oxilipins are essential to clarify this point and are outside the scope of the current method application.