Making normal more normal in quantitative lipidomics:  using the Sulfo-phospho-vanillin assay for quantitative LC-MS/MS lipidomics

Laura Bailey, Kari B. Basso

University of Florida

Sample normalization is vital for quantitation. Lipidomic LC-MS/MS has been normalized using mass/counts, DNA, or protein; however, a single method is not widely accepted nor have lipid-based normalizations been applied. While many targeted or semi-targeted lipid quantitation methods exist, entire classes may be neglected. Alternatively, the sulfo-phospho-vanillin assay (SPVA) yields a more complete measure of lipid concentration requiring a single lipid C=C [1].

We published the first SPVA application to lipidomic LC-MS/MS as a pre-quantitation method, similar to Bradford protein quantitation. Using biologically sourced standards, relative lipid quantitation showed smaller standard deviations and greater concentration differences [2]. We report additional work on true biological sample extracts to evaluate sample normalization via SPVA lipid quantitation to gravimetric measurement and protein concentrations.

Tissue and fluid extracts, including with pure lipid and/or protein spiking (0-100µL), were prepared using chloroform:methanol. Lipid content was measured using an optimized SPVA methodology in a 96-well glass-coated plate [2]. The absorbance was measured at 530nm, and the lipid concentration interpolated from Avanti lipid standards. Protein content was measured via Qubit.

LC-MS/MS measurements were conducted by reversed-phase chromatography (ThermoScientific Acclaim PepMap) on an Impact II Qq-TOF (Bruker) mass spectrometer. Separation was conducted using a 50-minute gradient of 60/40% acetonitrile/water and 90/8/2% isopropanol/acetonitrile/water, both containing ammonium formate and formic acid. Analyses were conducted in positive mode, employing data-dependent activated MS/MS. Non-quantified injections (3µL) were compared to 2-3µg lipid content based on SPVA and protein measurements. Metaboscape was used for bioinformatics.

Our recent publication used E.coli and brain Avanti standards as lipid sample normalization proof-of-concept. We optimized a 300µL microplate reaction, showing total lipid linearity in simple and complex lipid mixtures (R²=0.99). Concentration diverse (∆~100µg/mL) test samples showed good accuracy (14% E.coli and 9% brain) and precision (7% E.coli and 10% brain). Relative brain lipid concentrations showed smaller standard deviations of 30-90% (percent difference).

Currently, investigations are on plasma, E.coli, and mouse kidney, brain, and heart. Extracts were evaluated for total lipid using SPVA and total protein to estimate lipids using the accepted rule [lipids]≈4*[proteins]. Extracted lipids deviated from the 4-times lipid estimation, least to most deviation:  heart (4.5-times), E.coli (5.6-times), kidney (7.0-times), brain (11.6-times), and plasma (0.7-times). LC-MS/MS sample normalization via gravimetric (MQ), protein (PQ), and lipid (SPVA) showed more reproducible results (8%, 26%, 8% RSD, respectfully), greater lipid annotations (26% greater), and tighter statistical clustering (PCA) using SPVA. Additionally, PQ showed strange inverse lipid relative concentrations compared to SPVA and MQ.

To investigate these results, spiking experiments were devised to systematically increase the lipid and/or protein concentration from a pure, non-native source (E.coli into mouse). E.Coli lipid-spiked-kidney showed unchanged total protein (±0.02µg/µL) and linearly increasing total lipid (slope=0.0234µg/µL²) by PQ and SPVA, respectively. As lipid-spike concentration increased, volcano plots show MQ and PQ significant features increased non-linearly, revealing non-native-lipids over-estimation. Alternatively, SPVA accurately showed linearly increasing “up-regulated” non-native-lipids. MQ and PQ showed significant artificial “down-regulation” of native-lipids compared to SPVA, which showed no artificial down-regulation. Protein-spiked experiments are expected to show inverse total protein and lipid pre-concentration trends, with lipid concentrations unchanged and protein concentrations increasing with protein-spike. However, LC-MS/MS sample normalization should show primarily “down-regulated” lipids by PQ while SPVA significant lipids are unchanged with protein-spike concentration.