A Novel Experimental Workflow to Determine the Impact of Storage Parameters on the Mass Spectrometric Profiling and Assessment of Representative Phosphatidylethanolamine Lipids in Mouse Tissues

Abstract

Evaluation of signaling lipids is essential for measuring biological processes. There is a lack of experimental data regarding the proper storage of extracts for signaling lipid analysis, potentially impacting the procedures that can lead to accurate and reproducible evaluation. In this study, the importance of pre-analytical conditions for analyzing ion transitions for phosphatidylethanolamines (PEs), an abundant signaling phospholipid, was systematically assessed. A novel workflow was utilized involving an MRM-based experimental approach followed by statistical analysis. Specifically, lipids were extracted from the brain, heart, lungs, and serum of C57BL/6 mice. Extract subsets were resuspended in organic solvents prior to storage in various temperature conditions. Mass spectrometry analysis by multiple reaction monitoring (MRM) profiling was performed at four time-points (1 day, 2 weeks, 2 months, or 6 months) to measure relative amounts of PEs in distinct lipid extract aliquots. We introduce an innovative statistical workflow to measure the changes in relative amounts of PEs in the profiles over time to determine lipid extract storage conditions in which fewer profile changes occur. Results demonstrated that time is the most significant factor affecting the changes in lipid samples, with temperature and solvent having comparatively minor effects. We conclude that for lipid extracts obtained by Bligh & Dyer extraction, storage at −80.0°C without solvent for less than two weeks before analysis is ideal. By considering the data generated by this study, lipid extract storage practices may be optimized and standardized, enhancing the validity and reproducibility of lipid assessments.

Introduction

Lipids are known to be involved in a number of cellular processes, including cell signaling, energy storage, structural support, inflammation, apoptosis, and others [1-12]. Specifically, phosphatidylethanolamines (PE) serve as a precursor for phosphatidylcholines and a substrate for some post-translational modifications that influence membrane topology, cell and organelle membrane fusion, oxidative phosphorylation, mitochondrial biogenesis, and autophagy [13]. PEs have also been observed to have pro-phagocytic effects; lipids are translocated to the outer layer of the plasma membrane in the early stages of the apoptotic process, allowing for recognition and engulfment by phagocytes [3]. Given the importance of apoptosis in a wide variety of illnesses, including cancer, neurogenerative diseases (Alzheimer’s Disease, Parkinson’s Disease, Huntington’s Disease), and many others, it is vital to understand the factors which contribute to its induction and regulation, including the role of PE lipids. Although the importance of lipids in biological signaling and disease is beginning to be recognized, many experimental and analytical parameters have yet to be firmly established. To fully understand the role of lipids in biological processes, it is necessary for standard procedures to be identified that allow for accurate and reproducible sample assessment.

Currently, the most commonly used exploratory methods for the analysis of lipid extracts utilize lipid profiles obtained using mass spectrometry (MS). Subsequently, the resultant data is reduced and visualized using standard matrix algebra techniques such as principal component analysis (PCA) and partial least square discriminant analysis (PLS-DA). PCA and PLS-DA may be used separately or in combination with one another and are considered particularly suitable for analyzing large, complex datasets that are frequently seen in metabolomics studies [14, 15]. Our study utilized the MRM profiling approach for obtaining PE lipid fingerprints. A list of MRM scans was defined by the expected parent and product ions. Additionally, a set of MRMs which were compatible with peroxidized PEs was added. The intention of this MRM profiling method was not to provide secure lipid species identification but rather to assess, using the neutral loss diagnostic of PEs, whether the samples’ chemical profiles were changing over time.

While there is rich literature discussing the best practices in sample standardization (including storage), these publications rarely provide experimental data demonstrating the specific effects of different handling and storage parameters [16, 17]. The described study utilized a comprehensive statistical approach to examine the stability of the lipid MS profiles as a function of storage duration, temperature, wet/dry status, and tissue origin. The MRM-profiling method was selected as the most appropriate data collection strategy for two key reasons. The large number of surveilled samples (over 450 samples) prompted the choice of a method offering a high throughput (2 min per sample for data acquisition). This, combined with our interest in quantifying overall characteristics (sample profiles) rather than quantitative differences for specific lipid species, made MRM-profiling the most viable method of analysis. Consequently, the resultant data are represented by compositional vectors, which are appropriately pre-processed before utilizing PCA for dimensionality reduction, and generalized linear models for inference regarding the observed changes [18-21].

Lipid storage practices are frequently inconsistent between laboratories. Investigation of lipid assessment studies reported a wide variety in the duration and temperature of extract storage, with conditions ranging from 0.0°C to −80.0°C and three months to several years [22]. While molecular degradation over time has been determined to be unavoidable, techniques exist to slow to process [23]. Storage in organic solvents is a popular method to slow the lipid degradation rate during storage, though studies on its efficacy are limited [24]. Additionally, lipid degradation is known to be tissue-dependent, with even relatively small changes in composition impacting lipid hydrolysis and oxidation [25]. Despite the knowledge that such factors influence lipid extracts' integrity, there is a lack of research on their specific effects. This fact may contribute to interlaboratory variability and decreased data reliability. Thus, our current study utilizes a novel workflow involving an MRM-based experimental approach and statistical analysis to examine the impact of storage conditions (including temperature, presence of a solvent, and storage duration) on changes to the lipid profiles. PE lipids are used as a representative class in this study. Specifically, this examination seeks to systematically evaluate the impact of specific storage parameters on PE lipids' stability extracted from frequently utilized mouse tissue samples (lung, serum, brain, heart). Storage conditions varied in temperature (room temperature, −20.0°C or −80.0°C), sample solvent (methanol/chloroform 1:3 or dry lipid extract), and time (t₁=1 day, t₂=2 weeks, t₃=2 months, and t₄= 6 months). Identifying optimal storage conditions will allow for establishing standard operating protocols for lipidomic research that will enhance data quality and reproducibility.

Materials and Methods

Organ Collection

Lungs, brain, heart, and blood were collected from four healthy male C57BL/6J mice, obtained from Jackson Labs (Bar Harbor, ME). The organs were immediately frozen in liquid nitrogen before being transferred to a −80.0°C freezer. Blood was processed to serum via centrifugation at 3,500.0 rcf for 10.0 min at 4.0°C. Serum was stored with organs in −80.0°C conditions until lipid extraction.

Lipid Extraction

Similar to previous studies, organs were homogenized before the extraction of lipids via the Bligh-Dyer method [26-28]. Briefly, 250.0 μl chloroform and 450.0 μl methanol were added to samples prior to vortexing for approximately 10.0 sec. An additional 250.0 μl purified water and 250.0 μl chloroform were added, and samples were centrifuged at 16,000.0 rcf for 10 min. The bottom phase of each sample was isolated and divided equally among the 24 tested storage conditions: dry or resuspended in 200.0 μl CHCl₃:3 MeOH 1:1 (v/v) (“solvent” group), and stored at room temperature (25.0°C), −20.0°C or −80.0°C for 1 day, 2 weeks, 2 months, or 6 months. All extracts were concentrated to dryness. Extracts in the “solvent” condition were resuspended in 200.0 μl of CHCl₃:3 MeOH 1:1 (v/v) with 50.0 μg/mL butylated hydroxytoluene (BHT). Extracts were placed in their respective storage temperature conditions and analyzed at the corresponding time point. Each group consisted of 4 extracts per condition resulting in a total of 384 distinct extracts.

Lipid Analysis.

At selected time points (1 day, 2 weeks, 2 months, or 6 months), samples were prepared for lipid analysis via MRM-profiling. Specifically, lipid extracts in the “dry” condition were resuspended in 200.0 μl of 1 CHCl₃:3 MeOH prior to dilution. Dilutions were determined for each organ at the 1-day time point of analysis and were as follows: lungs 50.0×, brain 350.0×, heart 400.0×, serum 5.0×. These dilutions were utilized for all subsequent time points. Each extract was diluted in a 3:6.65:0.35 mixture of methanol, acetonitrile, and ammonium acetate immediately prior to loading into the mass spectrometer. Samples for each analysis condition were aliquoted in different vials so that no freeze-thawing cycle occurred before data acquisition. The instrumentation and conditions used were the same as reported in previous publications [29, 30]. For the four time-points evaluated, two minutes of data acquisition for MRMs related to PE lipids were used. The transitions were established from the Lipid Maps database and have been used in previous studies evaluating lipids [29-32]. In summary, the Lipid Maps database for PE lipids was downloaded, and the isomeric lipids were combined as the same parent ion mass. PE lipids (protonated), when fragmented using collision-induced dissociation, will lose their ethanolaminephosphate head groups as a neutral fragment of 141 mass units [33]. Therefore, combining the information of protonated PE lipids as parent ions and their respective neutral loss of 141 as the product ion yielded a list of MRMs to be used for a profiling experiment, which was both targeted and sensitive. Also, expected transitions for peroxidated (OOH) PE lipids were added to the assessment by adding a mass shift of 32 mass units [34] and maintaining expected neutral losses of 141 mass units. Due to the direct injection strategy, overlaps of MRMs related to native and peroxidated lipids can occur. The MRMs which were used are listed in Supplementary Information (ESM) Table S1.

Data representation and statistical analysis

Data analysis employed multiple linear models to evaluate, quantify, and link the changes in observed vectors of abundances to particular conditions. Initially, the values reflecting absolute quantities were filtered by removing any readout that was less than 50.0% higher than the negative (blank control). These readouts were deemed to be unreliable and were not used in further analysis. The data is represented by n=d×c×t×r extracts, where d is the number of storage conditions (e.g., dry vs. diluted), c denotes the tested temperatures, t is the number of time points at which the measurements were performed, and r is the number of examined organs (extract origin). The raw data matrix X_n,p is composed of one p×1 abundance vectors representing n extracts, where p is the number of identified lipids. The matrix X is normalized row-wise to 1; therefore, every row represents a compositional vector of fractional abundances. The data’s compositional nature means that each row of X lies in a p–1-dimensional nonnegative simplex $\mathbb{S}$, which could be approximated by the Dirichlet class of distributions [35]. As is, the compositional data should not be processed directly using dimensionality reduction or linear models because of the presence of spurious correlation and the closure problem [36-38]. Therefore, the X was mapped from $\mathbb{S}$-space to $\mathbb{R}$ using an appropriate α-transformation [39]. The resultant W matrix was used for downstream analysis. The values in W are distributed in an approximately Gaussian manner compatible with the downstream analytical approaches. In order to limit the data noise, we pre-selected the most informative lipid features (columns of W). The changes in lipids exposed to different storage conditions and times were expressed as dispersion index (variance to mean ratio). The 20 lipids showing the highest dispersion in their transformed compositional vectors (W matrix) were kept and used to compute compositional principal components (CPCs) [35, 40, 20]. The first CPC was used in a random-effects model, incorporating time, storage temperature, and extract origin. The post-hoc significance tests comparing the conditions pair-wise were performed using library emmeans for R [41].

Computing and visualizing the effect size of different conditions

The impacts of storage time and temperature were computed by creating a series of linear regression models for each transformed lipid abundance in each analyzed organ and noting the demonstrated effect size (expressed as η² and Cohen’s f) of time and temperature. If a storage condition (e.g., temperature) in a model explained a large percentage of variance (η²) in the lipid’s abundance, then the lipid could be considered susceptible to the change in temperature. The results were saved into two matrices (F₁ and F₂) containing the relevant tabulated effect sizes for each of the examined storage conditions. These collections of effect sizes (which, as mentioned above, de facto described the lipids’ susceptibility to a particular change in the conditions) were subsequently analyzed using multiway ANOVA and post-hoc test with Tukey adjustment to demonstrate differences in susceptibilities of lipids extracted from different organs.

Results

The analytical workflow used in this research is summarized in Figure 1. A large number of lipid extract aliquots were produced in order to avoid repeated freeze-thaw cycles of experimental samples. Lipid profiles were acquired by MRM profiling, a method of direct sample injection. Data analysis was based on a linear model linking the first principal component of α-transformed relative abundances with the descriptors of storage procedures (time, temperature, presence of solvent) and extract origin was used to analyze and visualize the overall samples’ variability profile. This dimensionality reduction strategy demonstrated that the degree of lipids’ changes (which, based on comparisons to previous lipid storage research, likely represents degradation) varies across different storage conditions. The values representing the lipids cluster according to the source organ and time-point of measurement. Time-point t₁ to t₃ measurements were very close in CPC space (particularly in extracts stored in −20.0°C and −80.0°C). However, the data representing time-point t₄ was visibly separated from the rest in all the storage conditions (Fig. 2). Differences may also be observed when the CPC-compressed data is separated by the organ of lipid origin vs. temperature. In the alternative view of the data (Fig. 3), it is clear again that the time point t₄ was represented by a distinct cluster in all the tested measurement conditions.

Fig. 1.

Schematics of the experimental workflow. (1) Brain, lungs, heart and blood samples were collected from four C57BL/6 mice immediately post-mortem and frozen at −80°C. (2) Samples were processed for Bligh & Dyer extraction and the organic phase was aliquoted in the appropriated number of vials to avoid freeze-thaw cycles for the phosphatidylethanolamine (PE) profiling of the four replicates of each organ. (3) Lipid extracts were stored in dried or diluted form at room temperature, at −20°C, or −80°C. The PE profile was acquired at the next day of the extraction, at 2 weeks, at 2 months, or 6 months of storage. (4). For data acquisition, diluted lipid extracts were directly injected to the ion source of a triple quadrupole mass spectrometer. (5) The profile of a set of MRMs related to PE lipids (ESM Table S1) was acquired over 2 minutes. (6) Values reflecting absolute quantities were filtered by removing any readout that was less than 50.0% higher than the negative (blank control) and processed for statistical analysis as described in the methods section.

Fig. 2.

Compositional principal component analysis showing the distribution of samples at 1 day, 2 weeks, 2 months, and 6 months after lipid extraction from four types of samples (mouse brain, heart, lung, and serum). Samples were stored in different conditions (room temperature, −20.0°C or −80.0°C) and diluted in an organic solvent or as dried lipid extracts.

Fig. 3.

Compositional principal component analysis shows a separation between measurements for ion transitions related to PE lipids collected at four different time points and originating from different types of samples (mouse brain, heart, lung, and serum) stored at three different temperatures (room temperature, −20.0°C, or −80.0°C) as diluted or dry lipid extracts..

The changes in time were also evident when visualizing the random-effects linear model linking the CPC-1 with the investigated conditions. The model demonstrated a moderate degree of PE profile changes between the 1 day and 2 week data, an increase at 2 months, and a large rise at 6 months (Fig. 4). PE lipids from lung tissue showed the biggest change captured by CPC-1 between t₁ and t₄, followed by brain, heart, and serum (ESM Table S2). The post-hoc tests demonstrated that the presence of solvent has a negligible overall effect, with dry conditions preserving lipids significantly better only at 25.0°C. Overall, there were no significant differences in outcome between 25.0°C, −20.0°C, and −80.0°C.

Fig. 4.

Visualization of the random-effect linear model linking the changes in time, with CPC-1 capturing the variance in lipid relative abundances for the four time points interrogated in this study. Each time point was evaluated for samples stored at room temperature, −20.0°C or −80.0°C and for lipid extracts stored in organic solvents or dried.

Specific ion transitions were observed to exemplify these patterns, with the predicted response from transition 774.5 → 633.5 associated with PEp (40:7), PE (36:3)_OOH, PE (38:1) serving as a compelling example (Fig. 4). Again, the lipids from lung showed the greatest changes, followed by brain, heart, and serum in ascending order of stability (ESM Table S3). There was a significant difference between changes observed at 25.0°C, while −20.0°C and −80.0°C were virtually identical. No significant differences between solvent conditions were observed. Some lipids followed more complex patterns. For instance, ion transition 648.5→507.5 associated with PEp (30:0) was uniquely stable in room temperate in dry storage conditions (Fig. 6). However, likely due to the instability and degradation of other lipids, the relative abundance of this analyte increased dramatically in other conditions. Also, the changes from t₁ to t₄ in the lipids originating from the heart were larger than the changes occurring for the same lipids in brain or lung tissue. However, serum remained the most stable extract of all tested specimens, as seen in ESM Table S4. Other PE lipids’ MRMs exhibited their own distinct patterns, unlike ion transition 648.5→507.5 or 774.5→633.5. For instance, ion transition 788.6→647.6 representing PEo (38:3)_OOH, PEo (40:1), PE (40:8) showed dramatically bigger changes in relative abundance from t₁ to t₄ for the lipids extracted from the heart (Fig. 7).

Fig. 6.

Changes over time in the α-transformed relative abundance of the ion transition 648.5→507.5 associated with PEp (30:0) according to the different temperatures of storage (room temperature, −20.0°C or −80.0°C) and the presence or absence of an organic solvent.

Fig. 7.

Changes in the transformed relative abundance for the four time points analyzed of the ion transition 788.6→647.6 compatible with PEo (38:3)_OOH, PEo (40:1), PE (40:8) under different temperatures and in the presence or absence of an organic solvent.

A series of linear models was built to determine alterations in the PE lipid extracts’ relative abundances exposed to separate sets of conditions, holding one condition of interest constant and varying the others. The effect sizes, computed as η² and Cohen’s f, were considered the measures of readout susceptibility to a change in the examined condition. For instance, an η² value from a linear model linking relative abundance and time, computed for extracts from heart, stored in dry condition and measured at −25.0°C, would indicate the susceptibility of these particular extractsto the changes in time. The distributions of measured susceptibilities were compared employing another ANOVA model and further post-hoc tests. In particular, the effects of time and temperature were the focus of the analysis. Box plots were used to visualize the mean and the variance of the effects. A shift towards high values indicates an elevated level of change in the extracts’ relative composition. In practical terms, it means that some PE lipids are easier to detect than others, forming a different extract “fingerprint.” Low values indicate higher stability of the extract in a particular set of conditions.

Differences in the means indicate that not all the extracts’ compositions were similarly impacted by the storage time (Fig. 8). For instance, heart extracts’ PE lipid profiles were significantly less susceptible to changes over time than lung-originating specimens (p<0.001). The brain extracts were less stable than the heart in a diluted setting (p=0.004), but not in dry conditions. Inversely, the serum extracts showed higher compositional stability than lung extracts in dry conditions (p<0.001), but not in diluted conditions. A summary of these results may be seen in ESM Table S5. Despite these differences, the solvent or dry lipid storage conditions had only a minor overall impact on compositional changes in the extracts, with only serum showing a significant difference (ESM Table S6).

Fig. 8.

Distributions of computed effect sizes demonstrating the overall susceptibility to PE profile changes of the four different sample extracts to changes in the overall composition when stored in the presence or absence of an organic solvent.

When stored in different temperatures, the differences between extracts in terms of their PE compositional characteristics were much more diverse. This observation was particularly evident for the lipid extracts stored in diluted conditions. It was observed that the effects of temperature had a lower variance than that of time, with a higher degree of clustering occurring for each organ tested (Fig. 9). Dry conditions resulted in a higher effect size than that observed in solvent conditions. In solvent conditions, the effect sizes of PE lipids from the heart and serum were found to the largest. Lungs had slightly smaller effect sizes, while the smallest effect size was observed for brain lipids. In dry conditions, the effect size was approximately the same for the PE lipids from brain, heart, and lungs, with serum having a slightly larger effect size. We observed a substantial (SMD>0.95) and significant difference in compositional stability between brain and heart, lungs, and serum extracts stored in solvent (p<0.001), as seen in ESM Table S7. The differences were less prominent in dry conditions. However, serum was still observed to be more susceptible to compositional changes than heart or brain (p<0.01).

Fig. 9.

Distributions of model effect sizes indicate the overall susceptibility of different lipid extracts to changes in the PE profile in the overall composition for the different types of samples stored in the presence or absence of an organic solvent.

The comparison between storage in solvent and dry conditions demonstrated that only heart extracts remained compositionally similar in both settings. The differences were biggest for brain, followed by lung and serum specimens (ESM Table S8). The presented data averaged the effect over temperatures (when considering changes in time) and averaged the impact of time when considering the effect of temperature. We also visualized the results in a fully multivariate fashion, separating time, temperature, and storage conditions.

A multivariate representation was used to illustrate the distribution of susceptibilities across different conditions and time points. The graph in Fig. 10 shows the changes for MRMs of PE lipids’ relative abundances in different temperatures. The dispersion was visibly tight at the 1 day time point, demonstrating that temperature did not dramatically affect the observed abundances. There was no difference between extracts’ origins, regardless of the presence of a solvent. At the two week time point (t₂), the disparities between extracts began to emerge (for instance, serum extracts’ changes were significantly greater than the changes in other organs). Also, there was a visible increase in the distribution width, suggesting that some lipids became dramatically more or less relatively abundant. At the two month time point (t₃) in the solvent conditions, the variability increased slightly for heart and lung PE lipids, with little change observed for brain or serum PE lipids. Dry conditions at two months were similar to those at two weeks. At six months (t₄), we continued to see the high variability as well as significant differences between extracts. The results are summarized in ESM Table S9.

Fig. 10.

Distributions of model effect sizes showing the overall susceptibility of lipid extracts from different organs and serum to PE profile changes in the overall composition over time (1 day, 1 week, months, and 6 months) for lipid extracts stored in the presence or absence of an organic solvent..

Another multivariate representation was used to illustrate the vulnerability of extracts to profile changes as a result of time (Fig. 11). At 25.0°C in the solvent condition, there were no significant differences between extracts. This observation does not apply to the dry conditions, where heart extracts seem to change their PE lipids’ relative abundances over time to a lesser degree than brain and lungs. At −20.0°C in the solvent condition, the differences were even more visible, with only brain and serum being indistinguishable (ESM Table S10). A similar pattern is observed in dry conditions. At −80.0°C, the highest level of changes also happens in lungs. There was no difference between the brain and heart.

Fig. 11.

Distributions of model effect sizes demonstrating the time-related susceptibility of different extracts to PE profile changes in the overall composition according to the tissue type for dry and diluted lipid extracts stored at −20.0°C, −80.0°C, or room temperature (25.0°C).

Discussion

While lipids have recently been recognized as essential biomarkers that could be employed to identify and diagnose disease, there remains much that is unknown regarding the analytical practices for their study. Studies regarding lipid structural stability and their profile changes over time and storage conditions are uncommon but necessary for the standardization in the handling and storage of lipid extracts for lipidomic analyses. The described study utilized several different storage conditions to examine the effects of temperature, time, and wet/dry storage state on the MRM profile alterations of a representative class of lipids in non-polar extracts obtained by the Bligh & Dyer method, as well as tissue dependence of alterations. A novel workflow and data analysis approach were utilized to assess the PE lipid profiles as a representative class, which provided enhanced rigor and more detail on the impacts of specific storage parameters as compared to previous studies.

Of the various storage conditions tested, time was determined to impact lipid profile alterations the most. It was determined that the composition of the PE lipids identified in extracts was relatively stable at the 1 day and 2 week time-points, with a slight decrease in stability at 2 months and a much larger decrease at 6 months. This result was expected, as additional storage time allows for more interaction between the lipids of interest and outside forces, increasing susceptibility to peroxidation and degradation of the extracts. While isobaric overlap may have occurred in the PE lipid MRM profiles due to direct injection and low mass resolution approach used in this study, changes in the profile were clearly observed. Also, although changes to lipid structure over time were not assessed, the inclusion of expected peroxidated lipids in the profiles and results align with the commonly accepted practice of analyzing extracts in a timely manner to avoid potential degradation of the lipid analyte(s). Other research has shown similar results, with longer storage times leading to an enhanced deterioration of lipids from the muscle, brain, kidney, and liver, as measured by increased lipid oxidation and hydrolysis [42, 43]. However, another study examined the impact of storage time on lipid extract integrity and found no significant degradation at seven months for lipids, which were stored at −80.0°C [44]. A number of factors could cause this apparent discrepancy. The study by Dill et al. only examined non-extracted lipids present in tissue sections of the brain and no other organs. Our research data has demonstrated that lipids extracted from the brain can be more resilient to changes in the lipid profile than those from other organs, particularly the lungs. While additional research is clearly needed regarding the precise impact of time on lipid profile changes and specific degradation pathways, it would appear that the amount of time spent in storage does significantly influence the integrity of lipids in an extract. In combination with published literature, our findings suggest the existence of time-dependent differences in lipid extract stability between tissue extracts. Therefore, the duration of appropriate extract storage may be tissue-specific and require modification based on the tissue of origin. Based on our study’s findings, we conclude that lipid extracts stored over two weeks prior to mass spectrometric analysis may present a considerable level of change in the profiles of PE lipids.

The temperature had less of an effect on the changes of PE profiles than anticipated, with its impact being smaller than that of time. Unsurprisingly, the lipid extracts stored at 25.0°C (room temperature) demonstrated a greater change in composition than those stored at cooler temperatures. There was some difference between the lipids stored at −20.0°C and −80.0°C, but contrary to expectations, a “deep freeze” at −80.0°C did not appear to confer any significant stability as compared to data from samples stored at −20.0°C. This result is supported by the results of previous studies, which have demonstrated that higher temperatures cause enhanced deterioration of lipids [45, 46]. Specifically, many of the lipids present in a lipid extract can be oxidized over time, forming peroxides. These peroxides can also propagate oxidation, contributing to the continued degradation of the lipids in the extract [47]. Given that temperature is well-known to increase peroxidation, it is expected that higher storage at 25.0°C causes enhanced lipid profile alteration when compared to extracts frozen at −20.0°C or −80.0°C. However, while our data indicate that cooler temperatures better preserve lipids, it should be emphasized that the effect of temperature was relatively small in comparison to the effect of time, with some lipids appearing to be entirely unaffected by temperature differences. Based on this, it may be concluded that lipid extract storage at −80.0°C better preserves the sample PE lipid composition, though storage at −20.0°C or even room temperature is acceptable for short periods. Extracts stored at −80.0°C are still subject to changes in PE lipids’ profile, indicating that degradation over time may play a significant role in the analysis of lipid extracts.

The presence of solvent had insignificant effects on lipid extract profile alterations. In most situations tested, dry and wet storage conditions preserved extracts equally well, with no significant differences between the two regarding PE profiles. Inversely, the organ of origin of extracted lipids had a significant impact on its susceptibility to profile alterations. Typically, PE lipids from the lung were most susceptible to profile changes, followed by the brain and heart. Serum PE lipids were typically the most resistant to changes. This outcome was unexpected, as the lipid structure, and thus susceptibility to changes in the PE profile, which can be related to degradation, should be equivalent regardless of the organ of origin. This organ-dependent vulnerability could be due to several factors. Each organ contained a different proportion of the specific PE lipids which were present in the extracted lipid mixture. It may be that the lung lipids mixture contained more lipids that are structurally more susceptible to degradation and thus were peroxidized sooner and propagated peroxidation of other lipids in the extract. Alternatively, an organ’s susceptibility to lipid degradation could be due to the relative number of peroxidized lipids present in the extract at the time of extraction. Lungs, given their exposure to outside air, are more likely to have higher levels of peroxidized lipids than organs enclosed within the body cavity, such as the heart and brain. Overall, organ type was determined to be a relatively minor factor in terms of PE profile changes compared to other parameters evaluated. However, special consideration may be needed for lipids extracted from the lungs, as they have been demonstrated to be more vulnerable to profile alteration than those from other organs.

The results of this assessment provide considerable insight into the optimal storage of lipid extracts; however, the study had some limitations. As is the case with all mass spectrometric equipment, the function of instruments can change over time, potentially impacting sensitivity. Therefore, lipid filtering was performed to include only the most abundant lipids and control for these effects, though even high-abundance lipids may be impacted to some degree. The most significant limitation in the described study stems from the format of extract preparation and data collection. The acquired data do not contain absolute, standardized lipid quantities. Instead, the collected and subsequently analyzed values report the relative abundances of different PE lipids in the extracts. Therefore, our observations captured the extracts’ overall status and the changes in their compositional characteristics, rather than the absolute abundance changes, and are restricted to one lipid class. Consequently, this experimental design does not allow us to pinpoint precisely the slowest or the fastest-degrading lipids. However, the compositional analysis does readily identify the unstable specimens and quantifies overall characteristics’ changes. Although we cannot trace and quantify the degradation of individual lipids, the presented approach can recognize the lipids in the mixtures that have relative abundances most likely to be affected by the chemical changes and breakdowns occurring within the extracts due to variations in storage parameters.

As mentioned previously, our current study evaluated PE lipids as a representative class, as they represent an important endpoint in experimental studies examining cellular functions related to membrane specialization. However, they are only a fraction of the many classes of lipids. Future studies should examine other classes of lipids that may be differentially affected by storage conditions. Sphingolipids in particular play an essential role in cell membrane formation and signaling and are especially important in the neural system [48, 47, 49-51]. Abnormalities in sphingolipid metabolism have been shown to cause several neurological disorders [52, 53]. Ceramides, a subclass of sphingolipid, have been shown to mediate many cellular functions, including proliferation, inflammation, autophagy, senescence, and differentiation [54, 55]. These and other lipid classes may be subject to variable degradation due to the presence of attached residues and groups, such as fatty acids and phosphocholines. Given the importance of these lipids in scientific studies as a measure of cellular functions and disease, they represent key lipids for future assessments of storage conditions.

Conclusion

By utilizing a novel workflow involving an MRM-based experimental approach and statistical analysis to identify the impact of critical storage factors such as time, temperature, and solvent presence, we have provided information that will allow other researchers may optimize their protocols to minimize extract degradation and enhance the quality of the produced results. Further, the presented analysis may be used as a compelling argument for standardization, which in turn would improve the reproducibility of the results between laboratories. Long term, the fundamental knowledge regarding extract stability and dissemination of the best practices in extract storage will allow the field of lipidomics to advance while enhancing research and clinical impacts.

Fig. 5.

Changes in the α-transformed relative abundance of the ion transition 774.5→633.5 transition associated with PEp (40:7), PE (36:3)_OOH, PE (38:1) over the four time points, considering the temperature of storage and the presence or absence of an organic solvent.

Data Availability.

All data is described within the manuscript. Raw data is available upon request by contacting Drs. Rajwa (brajwa@purdue.edu) or Shannahan (jshannah@purdue.edu).