Achieving Comprehensive Lipid Profiling with a CCS, Retention Time, and MS/MS Library

In this application note, a standardized protocol is used to measure RT and CCS information from a sample set consisting of 100 certified standards from multiple lipid classes to validate a predictive TWCCSN₂ Model by Broeckling et al.

Benefits

• An easy to deploy method and library, providing greater confidence for the identifications returned and enabling increased productivity
• A curated lipidomic database containing over 3200 lipid species which have been previously reported for human plasma and tissue samples
• Lipid standards allowed for the validation of a CCS predictive model, which allowed for the incorporation of in-silico values to increase coverage
• Assigned measured RTs to reduce false-positive identifications
• Comprehensive MS/MS information for multiple adducts built into the library for greater confidence
• Fast data processing and visualization using either UNIFI or Progenesis QI Informatics for maximum flexibility

Lipids are well recognized as important biological molecules, playing a diverse role in various molecular functions, including signal transduction, membrane anchoring, as well as energy storage. To gain insight into the lipidome, analytical technologies such as liquid chromatography coupled with mass spectrometry (LC-MS) have become a routine means of analysis, providing comprehensive, high throughput, and robust solutions. Discovery-based workflows utilizing high resolution strategies, allow complex samples to be profiled, providing a large number of potential lipid identifications. A significant hurdle encountered for discovery lipidomics is the ability to identify lipids of interest accurately and confidently from database searches. Additional physiochemical properties such as retention time (RT) and collisional cross section (CCS), as well as diagnostic MS/MS information can be incorporated into the search space to increase confidence in the identifications returned.   

CCS is an important physicochemical property of lipid species that is closely related to its chemical structure and three-dimensional conformation, yet is independent of the type or complexity of sample matrix.¹ Therefore, CCS values can be used in addition to other molecular identifiers, such as RT and m/z as an orthogonal attribute to increase the specificity of lipid identifications.¹ Several in-silico and experimentally derived CCS databases have been published to-date as reported by Paglia et al.¹ but suffer a high false-positive rate and often identifications lack biological relevance. Ideally, experimental measured CCS values should be used to populate CCS databases, but due to a limited number of commercially available lipid standards, in-silico measurements are needed to enable adequate coverage.

Here we used a standardized protocol to measure RT and CCS information from a sample set consisting of 100 certified standards from multiple lipid classes to validate a predictive TWCCSN₂ Model by Broeckling et al.² The high correlation established between the experimentally measured and predicted values then allowed for the remainder of the library to be populated with in-silico CCS values. Fragment ion information was also appended to the library, accounting for multiple adduct types. Extracts resulting from human plasma and tissues (i.e. heart and liver) were used as additional test samples for identifying endogenous species.

Sample Information

To assess the accuracy of the CCS prediction tool, a variety of lipid standards were acquired. These consisted of 69 individual lipids (Avanti Lipids, Alabama, USA) from across 14 classes, which were constituted into sample groups of no more than 7 lipids/mix with distinct reversed-phase (RP) chromatographic behaviors (i.e. to prevent co-elution), in IPA. Samples were injected at three different concentration levels in triplicate as a custom mix of standards. Additionally, Odd Chain LIPIDOMIX, Light Splash LIPIDOMIX, Differential Ion Mobility System Suitability LIPIDOMIX, Ceramide LIPIDOMIX, Splash LIPIDOMIX, and Deutrated Ceramide LIPIDOMIX were diluted 1:10 and 1:100 in IPA and injected as replicates (n=3). 

Library Building

Additional class-based lipid extracts (PS, PE, PG, Cer and SM mixes) were prepared at 0.1 mg/mL and 0.01 mg/mL and acquired as technical triplicates. These were used to verify and add confidence in the biological relevance of the lipids selected. To evaluate the utility of the library, extracts resulting from human plasma, liver, and heart tissues were screened.

Informatics

MassLynx, UNIFI, and Progenesis QI

Validating predicted CCS values

Over 100 individual lipids from custom mixes and commercially available sources were used to validate the in-silico CCS predictions.² Lipid standards were chosen to cover all the major lipid classes and to ensure a wide cross section of lipid structures were accounted for. The mean observed CCS values of measured lipid standards is compared to the predicted values by calculating the percentage difference:

Comparing the measured CCS with in-silico predicted values, provided high correlation for all classes. Figure 1 represents three example lipid classes, highlighting excellent linearity with r² >0.96 being generated. 

Peaks which were either saturated, had an error ≥5% between measured and predicted CCS values, or a retention time deviation ≥0.1 min were excluded. After correction for retention time, to reduce false-positive hits, 96.8% (positive ion) and 95.5% (negative ion) of the experimental data were found to be within ±5% CCS tolerance when compared with their in-silico values, confirming the findings of previously reported data.²  

A total of 32 and 16 lipid classes are represented with CCS values for positive and negative ion respectively. The CCS distribution ranges from 143–349 Å (positive ion) and 124–334 Å (negative ion). Exemplar CCS distributions for a subset of lipid classes corresponding to positive ion are provided in Figure 2. 

Database Overview

The availability of comprehensive, well-curated databases is important for ensuring compounds are accurately and confidently identified. Therefore, to address these needs, the comprehensive lipid database presented here, consists of 3200 lipid entries and include the most common adducts for both positive and negative ESI. Figure 3 shows the distribution of lipid entries between the ionization modes, whilst Figure 4 represents the adducts accounted for in both ionization modes. Of the 3200 lipids presented, 43% of the entries are represented by both ionization modes. This has the advantage of providing higher identification confidence. For example, when considering phospholipids, the ability to characterize both the head group and fatty acyl chain(s) is possible. 

The chromatographic method employed for generating the library consisted of a 12-minute cycle time, which has previously been demonstrated as being specific, robust, and reproducible for large scale lipidomic studies.³ Additionally, this methodology also utilized the benefits of the hybrid surface technology from the ACQUITY Premier System, whereby the sensitivity and recovery of phosphorylated and carboxylated lipid species are significantly improved (Application Note 720007092). Retention time values relating to the 100 lipid standards which were used for CCS validation, were also used for verification and appended to the database. Example chromatography for a plasma extract is provided in Figure 5, along with the RT distribution for the experimentally measured lipid standards.

Utility with Progenesis QI and UNIFI for Maximum Flexibility

To illustrate the utility of the library, database searches were conducted for a variety of endogenous matrices. The library has been constructed to provide maximum flexibility, enabling searches to be performed using either Progenesis QI or UNIFI Processing Software. Figure 6 shows an endogenous TG(44:2) identification resulting from a Progenesis QI Search. Comparing searches between the Waters Lipid Profiling Library and LipidMaps shows a 42% increase in fragment ion scores, whilst the overall identification score (inclusive of CCS and MS/MS) resulted in a 17% increase when using the Waters Lipid Library. The increases observed in identification score(s) is a result of the enhanced specificity provided with the inclusion of CCS and MS/MS information. Figure 7 highlights how implementing CCS can help to filter the number of identifications returned following a database search. 

Similarly, data corresponding to human plasma spiked with Odd Chain LIPIDOMIX was processed through UNIFI (Figure 8). The MS/MS relating to SM(d18:1/12:0) is shown for both positive and negative ion data with the library matched fragment ions clearly marked.

• A predictive algorithm which generates in-silico CCS values has been successfully validated using over 100 lipid standards across 14 lipid classes. 
• Incorporation of CCS, MS/MS, and retention time information is shown to increase specificity, resulting in increased identification scores and thereby providing greater confidence.
• Extensive database curation based on biological relevance and measured values from extracts.
• The versatility of the library allows it to be used with UNIFI and/or Progenesis QI Informatics.

1. Paglia, G., Smith, A. J., Astarita, G. (2021) Ion Mobility Mass Spectrometry in the Omics Era : Challenges and Opportunities for Metabolomics and Lipidomics. doi:10.1002/mas.21686.
2. Broeckling, C. D. et al. Application of Predicted Collisional Cross Section to Metabolome Databases to Probabilistically Describe the Current and Future Ion Mobility Mass Spectrometry. J. Am. Soc. Mass Spectrom. (2021) doi:10.1021/jasms.0c00375.
3. Giorgis Isaac, Nyasha Munjoma, Lee A. Gethings, Lauren Mullin, Robert S. Plumb. A Robust and Reproducible Reversed-Phase Lipid Profiling Method for Large Sample Sets, Waters Application Note, 720006959EN.
4. Giorgis Isaac, Robert S. Plumb. ACQUITY Premier LC Technology Significantly Improves Sensitivity, Peak Shape, and Recovery for Phosphorylated and Carboxylate Lipids, Waters Application Note, 720007092EN.