SupplementaryInformation for
Absolute quantitative lipidomics reveals lipidome-wide alterations in aging brain
Jia Tu1,2, Yandong Yin1, Meimei Xu1,2, Ruohong Wang1,2, and Zheng-Jiang Zhu1,*
1 Interdisciplinary Research Center on Biology and Chemistry, and Shanghai Institute of Organic Chemistry, Chinese Academy of Sciences, Shanghai, 200032 P. R. China
2 University of Chinese Academy of Sciences, Beijing, 100049 P. R. China
Corresponding Author
Dr. Zheng-Jiang Zhu
*e-mail: , Phone: 86-21-68582296
Fig.S1 Red lines represent for the precisions of lipid identifications under various matched parametersin both positive and negative modes. The precision is calculated as true positives divided by the sum of true positives and false positives. The histograms represent for the true positives under various matched parameters, where “weight.mz” ranges from 0-3 and “weight.int” ranges from 0-1. The sampleswere the natural glycerophospholipid standards. All identification results were manually inspected.
Fig.S2Differences of MS/MS spectra obtained from different platforms and collision energy levels. (a) MS/MS spectra of lipid PC(16:0/16:0) obtained from the NIST database. (b) MS/MS spectra of lipid PS(17:0/17:0) acquired using different collision energy levels
Fig.S3 Some examples to demonstrate the poor similarity between the experimental MS/MS spectra obtain in our lab and the LipidBlast library. The NIST program (MS Search V2.0) is used for the match, and the match scores are given by the NIST program.
Fig.S4Detailed steps for the curation of an in-house spectral MS/MS library: step 1, acquisition of standard MS/MS spectra; step 2, generation of rules for fragmentation and relative intensity, and step 3, prediction of the MS/MS spectral library.
Fig.S5Evaluation of the false positive numbers of identified lipids using either our in-house MS/MS library or LipidBlast. Samples are the natural glycerophospholipid standards. All identification results were manually inspected.
Fig.S6An example to demonstrate the capability to distinguish sn1/sn2 positional isomers using our in-house MS/MS spectral library (a) instead of LipidBlast (b).
Fig.S7 The quantification dynamic ranges for different lipid classes.
Fig.S8 The reproducibility of the calculated RF values for different lipid species over months.
Fig.S9 The calculated recovery rates for each lipid class through spiking 34 lipid standards into a 13C-labeledE. colibacteria sample.
Fig.S10Numbers of the quantified lipids in aging mouse brain.
Fig.S11Pie diagrams for the mass percentages of 22 quantified lipid classes (992 lipid species) in mouse brain across different ages (12-week and 32-week).
Fig.S12The lipid species in the clusters A, C, E and D.
Fig.S13 Pie diagrams for the mass percentages of membrane-esterified fatty acids in glycerophospholipids across different ages (12-week and 32-week).
Fig.S14 The alterations of membrane-esterified fatty acids across five ages.
Supplementary Table 1. The calculated RF values for 34 lipid classes.
Lipid class / RF / Polarity / Lipid class / RF / PolaritypPC+H / 0.50 / POS / PC+HCOO / 1.51 / NEG
LPC+H / 0.33 / POS / PE-H / 1.00 / NEG
Sph+H / 0.84 / POS / PG-H / 0.90 / NEG
Cer+H / 3.34 / POS / PS-H / 3.20 / NEG
SM+H / 0.46 / POS / aLPC+HCOO / 0.93 / NEG
Cer-P+H / 2.55 / POS / LPE-H / 0.45 / NEG
PhytoCer+H / 3.02 / POS / pLPE-H / 1.00 / NEG
DG+NH4 / 0.93 / POS / LPG-H / 0.42 / NEG
TG+NH4 / 0.31 / POS / LPS-H / 0.67 / NEG
PhytoSph+H / 1.28 / POS / LPA-H / 0.72 / NEG
HexCer+H / 2.06 / POS / PIP2-H / 0.98 / NEG
ST+H / 7.26 / POS / CL-H / 4.61 / NEG
Car+H / 0.08 a / POS / LPI-H / 0.84 / NEG
MG+H / 13.32 a / POS / pPE-H / 0.70 / NEG
CE+NH4 / 263.78 a / POS / Hex2Cer+HCOO / 1.48 / NEG
SIP+H / 17.51 a / POS / PA-H / 1.35 / NEG
GM1+H / 53.45 a / POS / PI-H / 1.27 / NEG
Note: “a” refers to the RF value out of the expected range from 0.1 to 10, and is removed from the quantitative analysis.
Supplementary Table 2. The statistics of lipid number and spectra number in our in-house MS/MS spectral library.
Abbrev.name / Lipid
No. / MS/MS Spec. No. / [M+H]+ / [M+Na]+ / [M+NH4]+ / [M-H]- / [M+HCOO]-
PC / 5476 / 16428 / + / + / / / / / +
pPC / 222 / 666 / + / + / / / / / +
LPC / 74 / 222 / + / + / / / / / +
aLPC / 74 / 222 / + / + / / / / / +
PE / 5476 / 16428 / + / + / / / + / /
pPE / 222 / 666 / + / + / / / + / /
LPE / 74 / 222 / + / + / / / + / /
pLPE / 74 / 222 / + / + / / / + / /
PS / 5476 / 16428 / + / + / / / + / /
LPS / 74 / 148 / + / / / / / + / /
PG / 5476 / 16428 / / / + / + / + / /
LPG / 5476 / 10952 / / / + / / / + / /
PI / 5476 / 10952 / / / / / + / + / /
LPI / 74 / 222 / + / + / / / + / /
PIP2 / 5476 / 16428 / + / + / / / + / /
PA / 5476 / 16428 / / / + / + / + / /
LPA / 74 / 148 / / / / / + / + / /
CL / 25426 / 50852 / + / / / / / + / /
Sph / 74 / 74 / + / / / / / / / /
S1P / 64 / 64 / + / / / / / / / /
PhytoSph / 64 / 128 / + / + / / / / / /
Cer / 169 / 169 / + / / / / / / / /
CerP / 169 / 338 / + / / / / / / / +
PhytoCer / 168 / 336 / + / / / / / / / +
SM / 168 / 336 / + / / / / / / / +
Hex2Cer / 168 / 336 / + / / / / / / / +
HexCer / 168 / 168 / + / / / / / / / /
ST / 169 / 169 / + / / / / / / / /
GM1 / 168 / 504 / + / + / / / / / +
MG / 64 / 64 / / / / / + / / / /
DG / 1764 / 1764 / / / / / + / / / /
TG / 2640 / 2640 / / / / / + / / / /
CE / 74 / 74 / / / / / + / / / /
Car / 74 / 74 / + / / / / / / / /
Total / 76361 / 181300
S1
Supplementary Table 3. The comparison of the lipid coverage between our in-house MS/MS library and LipidBlast.
Our in-house library / LipidBlastNumber / LipidClass / Number of lipids / LipidClass / Number of lipids
1 / PC / 5476 / PC / 5476
2 / pPC / 222 / pPC / 222
3 / LPC / 74 / LPC / 80
4 / PE / 5476 / PE / 80
5 / pPE / 222 / pPE / 222
6 / LPE / 74 / LPE / 80
7 / PS / 5476 / PS / 5123
8 / PG / 5476 / PG / 5476
9 / PI / 5476 / PI / 5476
10 / PA / 5476 / PA / 5476
11 / CL / 25426 / CL / 25426
12 / CerP / 169 / CerP / 168
13 / SM / 168 / SM / 168
14 / ST / 169 / ST / 168
15 / GM1 / 168 / [glycan]-Cer / 880
16 / MG / 64 / MG / 74
17 / DG / 1764 / DG / 1764
18 / TG / 2640 / TG / 2640
19 / aLPC / 74 / MGDG / 5476
20 / pLPE / 74 / DGDG / 5476
21 / LPS / 74 / SQDG / 5476
22 / LPG / 74 / Ac2PIM1 / 144
23 / LPI / 74 / Ac2PIM2 / 144
24 / PIP2 / 5476 / Ac3PIM2 / 1728
25 / LPA / 74 / Ac4PIM2 / 20736
26 / Sph / 74 / LipidA-PP / 15625
27 / S1P / 64
28 / PhytoSph / 168
29 / Cer / 169
30 / PhytoCer / 168
31 / Hex2Cer / 168
32 / HexCer / 168
33 / CE / 74
34 / Car / 74
Total / 76361 / 113804
Supplementary Table 4. The information for7 SIL-IS used in the traditional approach for absolute quantitation.
Mixture Components / Conc. (µg/ml)15:0-18:1(d7) PC / 160
15:0-18:1(d7) PE / 5
15:0-18:1(d7) PG / 30
18:1(d7) LPC / 25
18:1(d7) LPE / 5
15:0-18:1(d7)-15:0 TG / 55
18:1(d9) SM / 30
S1
Supplementary Table 5.The molecular structure template used for the curation of our in-house spectral library.
Abbrev.name / Backbone/Sphingoid base / Head group / Side chain 1 / Side chain 2 / Side chain 3 / Side chain 4 / Sum formulaPC / C3H6O3PO3 / C5H12N / C(n)H(2n-2m-1)O / C(p)H(2p-2q-1)O / / / / / C(8+n+p)H(2n-2m+2p-2q+16)O8PN
pPC / C3H6O3PO3 / C5H12N / C(n)H(2n-2m-1) / C(p)H(2p-2q-1)O / / / / / C(8+n+p)H(2n-2m+2p-2q+16)O7PN
LPC / C3H6O3PO3 / C5H12N / C(n)H(2n-2m-1)O / H / / / / / C(8+n)H(2n-2m+18)O7PN
aLPC / C3H6O3PO3 / C5H12N / C(n)H(2n-2m+1) / H / / / / / C(8+n)H(2n-2m+18)O6PN
PE / C3H6O3PO3 / C2H6N / C(n)H(2n-2m-1)O / C(p)H(2p-2q-1)O / / / / / C(5+n+p)H(2n-2m+2p-2q+10)O8PN
pPE / C3H6O3PO3 / C2H6N / C(n)H(2n-2m-1) / C(p)H(2p-2q-1)O / / / / / C(5+n+p)H(2n-2m+2p-2q+10)O7PN
LPE / C3H6O3PO3 / C2H6N / C(n)H(2n-2m-1)O / H / / / / / C(5+n)H(2n-2m+12)O7PN
pLPE / C3H6O3PO3 / C2H6N / C(n)H(2n-2m-1) / H / / / / / C(5+n)H(2n-2m+12)O6PN
PS / C3H6O3PO3 / C3H6NO2 / C(n)H(2n-2m-1)O / C(p)H(2p-2q-1)O / / / / / C(6+n+p)H(2n-2m+2p-2q+10)O10PN
LPS / C3H6O3PO3 / C3H6NO2 / C(n)H(2n-2m-1)O / H / / / / / C(6+n)H(2n-2m+12)O9PN
PG / C3H6O3PO3 / C3H7O2 / C(n)H(2n-2m-1)O / C(p)H(2p-2q-1)O / / / / / C(6+n+p)H(2n-2m+2p-2q+11)O10PN
LPG / C3H6O3PO3 / C3H7O2 / C(n)H(2n-2m-1)O / H / / / / / C(6+n+p)H(2n-2m+2p-2q+13)O9PN
PI / C3H6O3PO3 / C6H11O5 / C(n)H(2n-2m-1)O / C(p)H(2p-2q-1)O / / / / / C(9+n+p)H(2n-2m+2p-2q+15)O13P
LPI / C3H6O3PO3 / C6H11O5 / C(n)H(2n-2m-1)O / H / / / / / C(9+n)H(2n-2m+17)O12P
PIP2 / C3H6O3PO3 / C6H13O11P2 / C(n)H(2n-2m-1)O / C(p)H(2p-2q-1)O / / / / / C(9+m+n)H(2n-2m+2p-2q+17)O19P3
PA / C3H6O3PO3 / H / C(n)H(2n-2m-1)O / C(p)H(2p-2q-1)O / / / / / C(3+n+p)H(2n-2m+2p-2q+5)O8PN
LPA / C3H6O3PO3 / H / C(n)H(2n-2m-1)O / H / / / / / C(3+n+p)H(2n-2m+2p-2q+7)O7PN
CL / C9H18O13P2 / H / C(n)H(2n-2m-1)O / C(p)H(2p-2q-1)O / C(x)H(2x-2y-1)O / C(j)H(2j-2k-1)O / C(9+n+p+x+j)H(2n-2m+2p-2q+2x-2y+2j-2k+14)O17P2
Supplementary Table 5. The molecular structure template used for the curation of our in-house spectral library (continued).
Abbrev. name / Backbone/Sphingoid base / Head group / Side chain 1 / Side chain 2 / Side chain 3 / Side chain 4 / Sum formulaSph / C(n)H(2n-2m+1)O2N / H / / / H / / / / / C(n)H(2n-2m+3)O2N
S1P / C(n)H(2n-2m+1)O2N / H2PO3 / / / H / / / / / C(n)H(2n-2m+4)O5NP
PhytoSph / C(n)H(2n-2m+1)O3N / H / / / H / / / / / C(n)H(2n-2m+3)O3N
Cer / C(n)H(2n-2m+1)O2N / H / / / C(p)H(2p-2q-1)O / / / / / C(n+p)H(2n-2m+2p-2q)O3N
CerP / C(n)H(2n-2m+1)O2N / H2PO3 / / / C(p)H(2p-2q-1)O / / / / / C(n+p)H(2n-2m+2p-2q+2)O6NP
PhytoCer / C(n)H(2n-2m+1)O3N / H / / / C(p)H(2p-2q-1)O / / / / / C(n+p)H(2n-2m+2p-2q+1)O4N
SM / C(n)H(2n-2m+1)O2N / C5H13O3PN / / / C(p)H(2p-2q-1)O / / / / / C(5+n+p)H(2n-2m+2p-2q+13)O6PN2
Hex2Cer / C(n)H(2n-2m+1)O2N / C6H11O5 / / / C(p)H(2p-2q-1)O / / / / / C(6+n+p)H(2n-2m+2p-2q+11)O8N
HexCer / C(n)H(2n-2m+1)O2N / C12H21O10 / / / C(p)H(2p-2q-1)O / / / / / C(12+n+p)H(2n-2m+2p-2q+21)O13N
ST / C(n)H(2n-2m+1)O2N / C6H11SO8 / / / C(p)H(2p-2q-1)O / / / / / C(6+n+p)H(2n-2m+2p-2q+11)O11N
GM1 / C(n)H(2n-2m+1)O2N / C37H61O28N2 / / / C(p)H(2p-2q-1)O / / / / / C(37+n+p)H(61+2n-2m+2p-2q)O31N3
MG / C3H5O3 / / / C(n)H(2n-2m-1)O / H / H / / / C(3+n)H(2n-2m+6)O4
DG / C3H5O3 / / / C(n)H(2n-2m-1)O / C(p)H(2p-2q-1)O / H / / / C(3+n+p)H(2n-2m+2p-2q+4)O5
TG / C3H5O3 / / / C(n)H(2n-2m-1)O / C(p)H(2p-2q-1)O / C(x)H(2x-2y-1)O / / / C(3+n+p+x)H(2n-2m+2p-2q+2x-2y+3)O6
CE / C27H45O / / / C(n)H(2n-2m-1)O / / / / / / / C(27+n)H(2n-2m+44)O2
Car / C7H14O3N / / / C(n)H(2n-2m-1)O / / / / / / / C(7+n)H(2n-2m+13)O4N
Note: “n/p/x/j” refers to the carbon chain length (2-26); “m/q/y/k” refers to the unsaturated degree (0-6).
S1
Supplemental Materials and Methods
2.1 Chemicals
Lipid standards were all purchased from Avanti Polar Lipids (Alabaster, USA). Abbreviations for all lipid classes were provided in Supplementary Table 1. Ammonium formate and 2-propanol (IPA) was purchased from Fisher Scientific (Morris Plains, NJ, USA). LC–MS grade water (H2O), acetonitrile (ACN), methanol (MeOH), GC grade dichloromethane (DCM), and LC grade tert-butyl methyl ether (MTBE) were purchased from Honeywell (Muskegon, MI, USA). The human plasma was purchased fromEquitech-Bio (Catalog No. HPH-0500, TX, USA). The Jurkat cells were donated by Professor Junying Yuan from Harvard Medical School.
2.2Sample preparation
The lipids from human plasma, Jurkat cell, mouse brain tissue were extracted using a modified MTBE extraction method (Matyash et al. 2008).Specifically, the human plasma sample (30 μL) was firstly diluted to 200μLusing water. Then, 480 μL of the extraction solvent (MTBE:MeOH =5:1, v/v) containing0.5μL of SIL-IS(d7-PE (15:0/18:1), 100 ppm, Catalog No.791638C, Avanti) was added. The sample was vortexed for 30 s, followed by the 10 min of sonication. The solution was centrifuged at 3,000 rpm for 15 min. The upper organic layer (i.e., MTBE layer) was collected into a new Eppendorf tube. Then, 200 μL of MTBE was added to the left aqueouslayer for further extraction. The solution wasvortexed, sonicated, and centrifugation again as previously described. Then, the upper organic layer was collected again. The re-extraction process was repeated twice to ensure the high extraction recovery. Finally, the pooledorganic layer was evaporated using a vacuum concentrator(Catalog No.7310038, LABCONCO). The dry extract was reconstituted using 100 μL of DCM:MeOH (1:1, v/v) prior to LC–MS/MS analysis.
The Jurkat cell pellet (~ 5106 cell/sample) was mixed with water (200μL). The sample was then vortexed for 30 s and incubated in liquid nitrogen for 1 min. Then it was thaw at room temperature and sonicated for 10 min. This freeze–thaw cycle was repeated three times in total.Protein concentration in the sample wasmeasured using the Pierce BCA Protein Assay Kit (Catalog No. 23225, Thermo Fisher).Then, 480 μL of the extraction solvent (MTBE:MeOH =5:1, v/v) containing0.5μL of SIL-IS(d7-PE (15:0/18:1), 100 ppm) was added. The sample was vortexed for 30 s, followed by the 10 min of sonication. And the rest procedures were the same as the preparation of human plasma samples.
The frozen mouse brain tissue was weighted and homogenized in H2O (~200 μLH2O for ~10mg tissue) using the homogenizer (Precellys 24, Bertin Technologies, France). Protein concentration in the homogenized solution wasmeasured using the Pierce BCA Protein Assay Kit (Catalog No.23225, Thermo Fisher). Then, 50 μL of homogenized solution was taken and diluted to 2000μLusingwater. Then, 4800 μL of the extraction solvent (MTBE:MeOH =5:1, v/v) containing 1 μL of SIL-IS(d7-PE (15:0/18:1), 100 ppm) was added. The solution was vortexed for 30 s, followed by 10 min of sonication, and centrifugation at 3000 rpm for 15 min. The upper organic layer was collected. And an additional 2000 μL of MTBE was added to the bottom layer for re-extraction. The re-extraction was repeated twice. The pooled organic layer was evaporated using a vacuum concentrator. The dry extract was reconstituted using 100 μL of DCM:MeOH (1:1, v/v) prior to LC –MS/MS analysis.
2.3LC–MS/MS data acquisition
The LC–MS/MS analyses were performed using an HPLC system (1290 series, Agilent Technologies) coupled to a quadrupole time-of-flight mass spectrometer (TripleTOF 6600, AB Sciex). Chromatographic separations were performed on a PhenomenexKinetex C18 column (particle size, 1.7 μm; 100 mm (length)2.1 mm (i.d.))with a column temperature of 55 °C. The mobile phases A = 10 mM ammonium formate in H2O:ACN (6:4, v/v), and B = 10 mM ammonium formate in IPA:ACN (9:1, v/v), were used for both ESI positive and negative modes. The linear gradient elutes from 40 to 100% B (0–12min), 100% B (12–14min), 100 to 40% B (14–14.2 min), then equilibrate at 40 % B until 18 min. And the flow rate was set as 0.3 mL/min.The mass spectrometry parameters were applied as follows: ion source gas 1 (GS1), 60 psi; ion source gas 2 (GS2), 60 psi; curtain gas (CUR), 30 psi; temperature, 600 °C; ionspray voltage floating (ISVF), 5000 V or -4500 V in positive or negative modes, respectively; declustering potential (DP), 100 V. Data-dependent acquisition (DDA) method was used for MS/MS acquisition.Each acquisitioncycle consists of one rapid TOF MS survey scan (200ms) followed by the consecutive acquisition of 11 product ion scans (50ms each). For TOF MS survey scan, the mass range is from 200 to 2000 Da, and collision energy (CE) is set as 10 V.For product ion scan, the mass ranges are from 100 to 2000 Da, and collision energy (CE) is set as 4525 V. Dynamic background subtraction was applied. Dynamic exclusion is set as 4 seconds after 2 occurrences to ensure thenon-repetitiveacquisitionof MS/MS spectra from the same ion.
To acquire the MS/MS spectra of lipid standards,lipid standards were first prepared at a concentration of 10ppm in either MeOH/DCM (1: 1, v/v) or MeOH/H2O(1: 1, v/v). Each lipid standard solution was infused at a flow rate of 150μL/min into the mass spectrometer without a LC column. Source parameters were set the same as above. In each acquisition,multiple collision energy (CE) levels were applied, including 9 fixed levels (10, 20, 30, 40, 50, 60, 70, 80 and 90 V) and 5 ramp ones (55±35, 50±30,45±25,40±20, and 35±15).
2.4Curationof an in-house MS/MS spectral library for lipid identification
An in-house lipid MS/MS spectral library was curated using the fragmentation rules and relative intensities of fragment ions derived from the experimentally acquired standard MS/MS spectra. Briefly, we firstly acquired the experimental MS/MS spectra of 71 lipid standards representing for 34 lipid classes at the optimized CE level. In our Q-TOF instrument, the CE level was optimized as 45 25 V, which provides abundant and characteristic fragment ions for most lipid species. For example, we have chosen four PS standards for the acquisition of experimental MS/MS spectra. Secondly, we generated the fragmentation rule for PS and recorded relative intensities of each fragment ion from the experimental MS/MS spectra. The relative intensities of fragment ions provide the positional information of fatty acyl chain, and are used to distinguish the positional isomeric lipids, such as PS (16:0/18:1) and PS (16:0/18:1). Finally, we predicted the MS/MS spectra of all 5,476 PS lipids. Lipid species within PS class have the similar molecular structures and the similar fragmentation patterns. Molecular structure templates for lipids were partly followed as LipidBLAST. In short, the structure of lipid specie consists of linear combination of backbone, head group and fatty acid chain (Supplementary Table 5). The structures of lipids within PS class vary in fatty acid chain, ranging from C2 to C26 in lengths of acyl carbon chains and from 0 to 6 in the number of double bonds in a single acyl chain. Thus the MS/MS spectra of all lipids in PS class have the conjectural fragment ions m/z, and the same relative intensities for each fragment ion, basing on the above fragmentation rules and relative intensities.The same method was applied to predict other lipid species. Several adductive forms, including [M+H]+, [M+NH4]+, [M+Na]+, [M-H]-, and [M+HCOO]- were considered.
2.5 Calculation of recovery rates
The recovery rate is defined as the fraction of the analyte recovered after sample extraction.It can be evaluated using the peak areas of the standard added before sample extraction and after extraction (Eq. 1).
= (1)
The calculation of recovery rateswas separated into two different protocols. In first experiment,5 μL of 34 lipid standard mixtures (5 ug/ml each, Supplementary Table 1) was added to the matrix (13C-labled bacteria sample)before the sample extraction (B). Then, another5 μL of 34 lipid standard mixtures (5 ug/ml each) was added to the matrix (13C-labled bacteria)after the sample extraction (A).To comprehensively evaluate recovery rates of each lipid, a flow injection analysis was performed to detect lipids besides of LC-MS experiment.
In lipidomics analysis of biological sample, to evaluate recovery rate in sample preparation, we usually prepared the pooled QC samples in duplicate. In one QC sample, the SIL-IS (d7-PE (15:0/18:1), Catalog No. 791638C, Avanti) was added before the sample extraction (B). In the other QC sample, the same SIL-IUS was added after the sample extraction (A). Then, the Eq. 1 is used to calculate the recovery rate of SIL-IS in each experiment. In our opinion, the recovery rate of SIL-IS between 60% an 140% is acceptable.
2.6 Statistical analyses
Lipid levels were expressed in terms of mass fractions normalized to the total lipids detected in each sample. Statistical comparisons of normalized mass fractions among mouse brains from five ages were performed using Kruskal-Wallis non-parametric analysis of variance (ANOVA). For all analyses, ***p value < 0.001;**p value < 0.01; *p value < 0.05.Hierarchical clustering analysis was performed to group lipids that displayed the similar alternation patterns across five ages using the pheatmap() function provided in R (R Studio 1.0.44, Boston, USA). Correlation matrix analysis was performed using the Pearson’s correlation using the cor() function in R.
References
Matyash, V., Liebisch, G., Kurzchalia, T. V., Shevchenko, A., & Schwudke, D. (2008). Lipid extraction by methyl-tert-butyl ether for high-throughput lipidomics. J Lipid Res, 49(5), 1137-1146, doi:10.1194/jlr.D700041-JLR200.
S1