On Mass Ambiguities in High-Resolution Shotgun Lipidomics

Article pubs.acs.org/ac

On Mass Ambiguities in High-Resolution Shotgun Lipidomics Chris Bielow,*,† Guido Mastrobuoni,‡ Marica Orioli,‡ and Stefan Kempa*,†,‡ †

Berlin Institute of Health Technology Platform Metabolomics, Max-Delbrück-Centrum for Molecular Medicine, Robert-Rössle-Straße 10, 13125 Berlin-Buch, Germany ‡ Berlin Institute for Medical Systems Biology, Max-Delbrück-Centrum for Molecular Medicine, Robert-Rössle-Straße 10, 13125 Berlin-Buch, Germany S Supporting Information *

ABSTRACT: Mass-spectrometry-based lipidomics aims to identify as many lipid species as possible from complex biological samples. Due to the large combinatorial search space, unambiguous identification of lipid species is far from trivial. Mass ambiguities are common in direct-injection shotgun experiments, where an orthogonal separation (e.g., liquid chromatography) is missing. Using the rich information within available lipid databases, we generated a comprehensive rule set describing mass ambiguities, while taking into consideration the resolving power (and its decay) of different mass analyzers. Importantly, common adduct species and isotopic peaks are accounted for and are shown to play a major role, both for perfect mass overlaps due to identical sum formulas and resolvable mass overlaps. We identified known and hitherto unknown mass ambiguities in high- and ultrahigh resolution data, while also ranking lipid classes by their propensity to cause ambiguities. On the basis of this new set of ambiguity rules, guidelines and recommendations for experimentalists and software developers of what constitutes a solid lipid identification in both MS and MS/MS were suggested. For researchers new to the field, our results are a compact source of ambiguities which should be accounted for. These new findings also have implications for the selection of internal standards, peaks used for internal mass calibration, optimal choice of instrument resolution, and sample preparation, for example, in regard to adduct ion formation.

M

lipidome,9 have been developed. Furthermore, multidimensional mass spectrometry (MDMS) methods,10 such as intrasource separation by polarity switching,11 are widely used. Derivatization introduces additional sample handling steps and thus increased time and cost and potentially sources of error, along with the risk of incomplete labeling. With the advent of the latest MS instrument technology, it is common to solve ambiguities via MSn evidence, as either NLS/PIS or full MS/MS.12 For data analysis and interpretation, it is thus paramount that mass ambiguities are understood comprehensively. Ideally, such ambiguities are avoidable, e.g., ambiguities due to salt adducts through a salt removal step during sample preparation.13 However, certain mass overlaps might be intrinsically present within the sample of interest and cannot be avoided. The criteria for confident identification of lipids at varying degrees of structural specificity using accurate precursor mass and MS/MS are not widely agreed on. Historically, precursor ion scans (PIS) or neutral loss scanning (NLS) evidence is widely used to detect defined lipid fragments.14 Almeida et al.12 suggested at least three specific MS/MS fragments; Kind et al.15 used spectral scoring and recommend a NIST dot score >600.

ass spectrometry-based analysis of (complex) lipid matrices has evolved into a powerful technology, enabling identification and quantification of over 500 lipid species per sample.1 Alterations in the lipidome have been linked to various human diseases, including Alzheimer’s,2 cancer, and metabolic syndrome.3 Shotgun lipidomics aims at identifying and quantifying as many lipid species as possible from direct injection of complex lipid mixtures via mass spectrometry. Advantages compared to LC-based methods include experimental simplicity and efficiency, absence of sample carry-over (if combined with a chip-based electrospray ionization), speed of analysis, and reproducibility. On the downside, shotgun methods suffer from limited ability to discern isomers, reduced sensitivity for low abundant species due a limited dynamic range, ion suppression, and mass ambiguities. The number of theoretically possible lipid species in lipid matrices is overwhelming, due to the huge number of combinations of head groups, acyl side chain length and position, double bond number and position, hydroxy groups, etc. This complexity alone leaves little hope that perfect mass separation of intact molecules is possible, making MS/MS a prerequisite for confident lipid identification. To alleviate the overlap problem, several approaches, including derivatization, aiming to analyze only particular subclasses,4−7 structural elucidation,8 or the full © 2017 American Chemical Society

Received: November 12, 2016 Accepted: February 7, 2017 Published: February 7, 2017 2986

DOI: 10.1021/acs.analchem.6b04456 Anal. Chem. 2017, 89, 2986−2994

Article

Analytical Chemistry Table 1. Adduct Composition of Ten Lipid Classes Confirmed Using High Purity Standardsa

a

internal standard

LipidMaps [lipid subclass]

positive mode

negative mode

PC (17:0/17:0) PE (17:0/17:0) PS (17:0/17:0) PA (17:0/17:0) SM 17:0 (d18:1/17:0) Cer C17 (d18:1/17:0) DG (8:0/8:0) TG (17:1/17:1/17:1) PI (8:0/8:0) LPC (17:0/0:0) ST-CholE (19:0)

[GP0101] [GP0201] [GP0301] [GP1001] [SP0301] [SP0201] [GL0201] [GL0301] [GP0601] [GP0105] [ST0102]

+H+, [+Na+] +H+, (+Na+) +H+, +Na+ +NH4+, +Na+, (+2Na+−H+) +H+, (+Na+) +H+, +H+−H2O, (+Na+) +NH4+, +H+−H2O, +H+ +NH4+ +H+, +NH4+, +Na+ +H+, (+Na+) +NH4+, (+Na+)

+HCOO−, [+Cl−] −H+, (+Cl−), (+NaCl−H−) −H+, (+Na+−2H+), (+HCOO− + Na+−H+) −H+, +HCOO− + Na+−H+, +2Na+−2H+ + HCOO− +HCOO−, [+Cl−] +HCOO−, [+Cl−], [−H+] Ø Ø −H+, [+HCOO− + Na+−H+] +HCOO− Ø

Legend: ( ), low ( 115000 since R1000 > 2 × 1000/0.02744 ∼72900. For smaller mass differences such as “H2 [−2]”, the situation is more complicated. In Figure 4, the resolution of a TOF and three common Orbitrap-based mass spectrometers and their ability to resolve small mass differences are shown. TOFs at 50k resolution are practicaly incapable of resolving DBA. For Orbitrap-based instruments with Rmax = 100k, the DBA can always be resolved below m/z 400, possibly up to m/z 650 (red band). Note that equal abundance in the case of DBA means that the actual lipid species have a ratio of about 1:6.5, such that the M + 2 isotope of the left species (with +1 DB) is equally high as the M + 0 isotope of the right species (see Figure S-1). At R = 450k, the DBA is 2990


Article

Analytical Chemistry

Figure 5. Percentage of monoisotopic peaks which have a potential overlapping peak from another lipid species within 1× or 2× fwhm for different base resolutions. For example, a value of 10% for DG means that 10% of the monoisotopic peaks from all DG species in the LipidMaps database have a nearby peak from an arbitrary class (within 1× or 2× fwhm).

capable of achieving resolutions R > 450k are FTICRs. However, FTICR resolution declines linearly, i.e., much faster, with increasing m/z compared to Orbitrap instruments (square root decline; see Methods); thus, higher base resolutions are required. Not surprisingly, higher base resolution results in less ambiguities, with noticeable gains when moving from 70k to 100k and 280k. Interestingly, most monoisotopic peaks are interfered by another peak, even at infinite resolutions (black line). When 100k vs 280k resolution is compared, all lipid classes benefit, especially TG at optimal abundance ratios (1× fwhm). Also, assuming arbitrary abundance (2× fwhm), all lipid classes show a significant improvement, especially ceramides, DG, and sterols. However, even at 450k resolution, the TG class potentially remains problematic. Most of these cases arise due to the well-known DBA (H2 [−2]) issue (see Figure 5). The sterol class seems to perform exceptionally well compared to other classes but has a lot of species with equal sum formulas (which were condensed into a single representative since they belong to the same subclass as described in the Methods), even though they are structurally different (e.g., within the cholesterol and derivatives [ST0101] subclass). DG, SM, PS, PA, and PI species show a remarkable resistance to resolvability even at exceedingly high resolutions (>450k). This is not due to perfect overlaps with other classes but very small mass differences. PC, ST, Cer, and PE to some extent are similarly resistant even at infinite resolution, indicating perfect sum formula equivalence with other classes (see “Identical Sum Formula” section below). Therefore, these classes (as well as TG) might be used for internal calibration, since the masses of their monoisotopic peaks can be assumed to be reliable. The Role of Sodium. Sparked by the observation that sodium (Na) is present in a number of MDDs, thus causing mass ambiguities, we also tested how many peaks can be resolved if sodium peaks could be avoided. Han and Gross recommend a modified Bligh and Dyer technique using LiOH to reduce the salt burden.13 The resolvability of monoisotopic peaks (as investigated above) improves ∼10% (median) if sodium-adducted lipids are removed (averaged across all resolutions ≤450k and fwhm).

always discernible irrespective of peak abundances well beyond m/z 900 (green band). In summary, the severity of the DBA depends largely on the instrument used and the abundance of the two lipid species involved. If the rightmost species (R) is at least 1/6.5 as abundant as the leftmost (L), then R’s monoisotopic peak will be correct (and only the M + 2 of L will be lost); hence, R can be successfully identified. The “C2H−1Na−1 [0]” overlap can almost never be resolved, no matter the instrument, and since this is a full overlap (i.e., the full isotope patterns overlap), no other witness peaks can be interrogated to resolve the problem in MS1, thus requiring MS/MS. General Resolvability by Lipid Classes. To elucidate the value of high-resolution instruments and to obtain a more global picture across all MDDs, we investigated which lipid classes are most commonly affected by mass ambiguities, depending on instrument resolution and abundance. Lipid classes with few overlaps are prime candidates for internal mass recalibration, reducing the need to rely on a few spike-in standards. We assume the abundance ratio of any two peaks is 1:1 (1× fwhm case) or an arbitrary abundance ratio (2× fwhm). The latter requires a larger distance (hence 2× fwhm) in order to be resolvable. The results are based on the assumption that all lipid species from the database are actually present, which is clearly an exaggeration. Hence, the results here must be regarded as an upper boundary (worst case scenario) for the resolvability across all species. Which cases occur in practice depends on the actual sample and the biological condition as well as contamination and (to a minor extent) spike-ins. The results for Orbitrap-based instruments are shown in Figure 5. It shows the percentage of monoisotopic peaks for each lipid class which are overlapping with another interfering peak. We chose the monoisotopic peak instead of all peaks, because M + 0 is usually used during database searches. Higher (theoretical) resolutions (600k, 800k, 1mio) are included here to predict performance of future instrument generations. The “inf” point references the baseline of inseparable lipids, i.e., with identical sum formulas across lipid classes (MDD “−[0]”). Today, the only other commercially available instrument type 2991


Article


Figure 6. Fraction of isotopic peaks (M + 0, M + 1, M + 2, M + 3) affected by an overlap within either 1× fwhm or 2× fwhm at different resolutions based on all species in the LipidMaps database.

The highest gain is achieved at R = 450k with 40% improvement (across all lipid classes) at 1× fwhm. For details, see Figure S-6. Thus, sodium-free samples may gain a moderate (R ≤ 280k) or significant (R = 450k) advantage in terms of resolvability, subject to practical feasibility of removing sodium. Ambiguity by Isotopic Peak Position. Most lipidomics software includes a deisotoping step to reduce the number of peaks to their monoisotopic representation. We already established that monoisotopic peaks can be highly contested (and thus shifted in mass and intensity) and that DBA can significantly contribute to an overlapping signal of an M + 2 and M + 0 peak. To obtain a more holistic picture, we investigated the ambiguity of isotopic peaks analogously to the monoisotopic analysis from above. Our results (Figure 6), indicate that M + 0 is highly contested, whereas M + 1 and M + 3 are significantly less overlapped. The performance of M + 2 highly depends on resolution, mostly due to the DBA as described before. Applying a database search to real data from rat tissues directly to M + 1 peaks did not yield significantly more identifications (data not shown) and has the disadvantage of targeting a lowerintense peak (M + 1 vs M + 0), whose mass accuracy becomes unreliable when being below a certain intensity threshold (∼70k counts on our setup, data not shown). However, the M + 1 search is able to disambiguate several cases of DBA occurring in an M + 0 search (mostly TG species) in positive mode. Identical Sum Formula. In the previous section, we established that a very common MDD is the perfect overlap of isotopes, due to identical sum formula (MDD “−[0]”). Since we collapsed intraclass species with identical sum formulas prior to our analysis, the remaining overlaps are interclass species. For example, PC(41:3) and PE(44:3) both have a sum formula of C49H92NO8P and thus an identical isotopic distribution and monoisotopic mass of 853.6561 amu. In MS1 positive ion mode, we cannot even hope to differentiate ion species based on mutually exclusive adducts, since both are protonated (H+). In negative ion mode, the deprotonation is common for PE, whereas PC acquires a HCOO− group, which could help to unambiguously identify it. To obtain a holistic picture, we list the equivalence rules for all identical sum formulas as described above (Methods section) and generate a mass equivalence graph (Figure 7). Two larger components can be observed, both representing phospholipids (PA/PC/PE/PS/CerP). Smaller components with two nodes cover DG/Cer, SM/PE-Cer, and PI/PI-Cer. Some transitions require odd carbon side chains, which are commonly regarded as low abundant (but highly interesting) in biological matrices of mammalian origin.28

In summary, PC species are the main hubs, allowing for numerous transitions to other lipid classes. This supports the findings from Figure 5, where PC species do not benefit from increasing resolution (since PC species can have identical sum formulas with lipid species from other classes). During this analysis, we identified five wrongly annotated PC lipid species in LipidMaps (wrong common and systematic name; see Table S-1 for details).

■

DISCUSSION Lipidomics is an emerging field and gains increasing interest in clinical applications. In regard to the importance of lipid data in diagnosing metabolic disorders, quality of identification and quantification of lipids is of high importance.29 Thus, we have taken the effort to analyze the possibilities of lipidomics in regard to available databases and also instrumentation. Specifically, the direct infusion lipidomics methods may have a high potential to be transferred to clinical practice in a reasonable time frame. Therefore, we inspected current possibilities and limitations of mass ambiguities in shotgun lipidomics. We have derived a mass difference description (MDD), consisting of a chemical difference formula and isotope offset for any two pairs of lipid species as well as mass equivalence rules (and graphs) for perfectly overlapping lipid species. While knowledge of these rules does not guard against cases where lipid species assignment can only be made when correcting for external factors, such as uncontrolled in-source conversion, e.g., PS → PA or LysoPC → LysoPA,30 it represents a powerful knowledge base which can be used to optimize lipid identification and quantification. Knowledge of exact mass overlaps, and odd chain carbon changes in particular, allows one to formulate requirements for internal standards, which are often chosen as odd side chain carbon species. In practice, any odd chain PC species overlaps with even chain PE species (see Figure 7); thus, the presence of these species must be checked and (if required) quantitatively corrected for in MS and MS/MS. For inexact mass overlaps, the most prominent MDD, the double bond ambiguity (DBA), has been described before.23,25 Here, we find that when considering real-world adducts a plethora of smaller mass differences are possible, warranting a mass resolution as high as possible. However, when considering all MDDs, the benefits from increasing the mass resolution beyond R = 450k become smaller. Other high-frequency ambiguities, which were previously described (“C1H2O−1 [2]”, “C1H1N−1 [1]”), are well resolvable with modern instruments. 2992


Article


Figure 7. Ambiguity graph for perfect mass overlaps. Nodes are lipid classes; edges represent mass-neutral transitions. Lipid classes with multiple occurrences carry a unique color. Directed edges (arrows) are weighted (thickness) by the number of occurring pairs. Each edge is annotated with the delta set of descriptors: “C” denotes the change in the number of methylene groups in the side chain(s), “DB” the difference in the number of double bonds, “o” the number of alkyl chains, and “h” the number of hydroxy groups. Edges without adduct annotation assume identical adducts. If an adduct is specified explicitly, this implicitly determines the ion mode. An example lipid pair is given (gray color) for each edge. Dashed edges enforce lipid species with an odd number of side chain carbons (odd “C”). All pairs without “lyso” (e.g., PC → PE) represent the corresponding lyso-pair as well (e.g., LysoPC → LysoPE).

target lists (ALEX,33 LipidXPlorer)17 is especially prone to identifying an impartial lipid if used incorrectly. The authors of LipidBlast reported that identification using dot-product scoring against their database was unsuccessful on as much as 24% of clean MS/MS spectra.15 The situation is expected to worsen considerably for shotgun lipidomics where most spectra are affected by coisolation. Recently, “Greazy”, a lipid specific search engine for LC-MS data, was made public.19 It offers a valuable concept of FDR-controlled scoring for glycerophospholipids. We expect that some modifications are required before it is applicable to shotgun MS. Scoring schemes for lipid species like glycerolipids (e.g., TG, DG) and sterols (e.g., cholesterols), which are highly abundant in common matrices, are yet to be devised. We expect our work to help establish requirements for scoring functions targeting more complex spectra. Ceramides might pose a challenge, since they lack a class-specific headgroup, but share identical precursors with DG or PA (no headgroup) and PCs (with headgroup). Thus, specific losses of the parent species have to be checked in MS/MS, to disentangle potential mixtures. Direct-injection shotgun lipidomics needs rigid quality control in order to exclude overlapping species; e.g., if a PA species is

Fortunately, our analysis did not yield an hitherto overlooked MDD of high frequency but several dozen potential MDDs albeit with low prevalence. Problematic (unresolvable) ambiguities mostly involve sodium, which is omni-present, arguing for careful sample preparation and control of ammonium or lithium modifiers, possibly affecting fragmention patterns.31,32 Mass ambiguities across lipid species in MS1 do not necessarily imply that two species are not discernible. Several methods to resolve this ambiguity are available without changing the experimental setup (e.g., by using LC, multidimensional MS, or MS/MS methods confirming class-specific fragments). In shotgun lipidomics, the ionization mode can be used to gain additional information, even in MS1 alone; e.g., in positive ion mode, PC vs PS species both ionize as H+ (and are therefore ambiguous; see Figure 7), but in negative ion mode, PC + HCOO− vs PS − H+ is unambiguous. Due to the high overlap potential of lipid species, especially for phospholipids, particularly PC species (see Figure 7), MS/MS scoring functions need to be chosen carefully. Existence of common fragments in overlapping lipid species can easily lead to partial or even misguided database hits. Software working with 2993


Article

Analytical Chemistry positively confirmed via MS1 precursor and MS2 headgroup, this does not exclude the presence of PC. We suggest the attachment of a “% signal explained” metric to each MS/MS and the number of co-occurring mass traces in MS1 within the isolation window, e.g., “PA(36:3); 58% MS/MS signal explained; 2 mass traces”. If only positive filters (i.e., presence, not absence of peaks) are used for lipid identification, careful postprocessing is required to distribute abundances. This should be possible using the rules for perfect mass overlap as derived above. Deisotoping before identification is not strictly required, since isotope peaks are not prone to overlap (depending on resolution used; see Figure 6). True M + 1 masses will usually not yield a database hit and are thus harmless. On the contrary, given that M + 2 can easily be an overlap target, a potential identification might be thwarted by collapsing the signal. Also, approaches to alleviate the DBA (e.g., ref 25) rely on the presence of isotopic peaks for consistency checks.

■

(2) Mapstone, M.; Cheema, A. K.; Fiandaca, M. S.; Zhong, X.; Mhyre, T. R.; MacArthur, L. H.; Hall, W. J.; Fisher, S. G.; Peterson, D. R.; Haley, J. M.; et al. Nat. Med. 2014, 20, 415−418. (3) Quehenberger, O.; Dennis, E. A. N. Engl. J. Med. 2011, 365, 1812− 1823. (4) Han, X. J. Lipid Res. 2005, 46, 1548−1560. (5) Leiker, T. J.; Barkley, R. M.; Murphy, R. C. Int. J. Mass Spectrom. 2011, 305, 103−108. (6) Wang, M.; Fang, H.; Han, X. Anal. Chem. 2012, 84, 4580−4586. (7) Wang, M.; Hayakawa, J.; Yang, K.; Han, X. Anal. Chem. 2014, 86, 2146−2155. (8) Wang, M.; Han, R. H.; Han, X. Anal. Chem. 2013, 85, 9312−9320. (9) Ryan, E.; Reid, G. E. Acc. Chem. Res. 2016, 49, 1596−1604. (10) Han, X.; Gross, R. W. Expert Rev. Proteomics 2005, 2, 253−264. (11) Schuhmann, K.; Almeida, R.; Baumert, M.; Herzog, R.; Bornstein, S. R.; Shevchenko, A. J. Mass Spectrom. 2012, 47, 96−104. (12) Almeida, R.; Pauling, J. K.; Sokol, E.; Hannibal-Bach, H. K.; Ejsing, C. S. J. Am. Soc. Mass Spectrom. 2015, 26, 133−148. (13) Han, X.; Gross, R. W. Anal. Biochem. 2001, 295, 88−100. (14) Hartler, J.; Tharakan, R.; Kofeler, H. C.; Graham, D. R.; Thallinger, G. G. Briefings Bioinf. 2013, 14, 375−390. (15) Kind, T.; Liu, K.-H.; Lee, D. Y.; DeFelice, B.; Meissen, J. K.; Fiehn, O. Nat. Methods 2013, 10, 755−758. (16) Schwudke, D.; Oegema, J.; Burton, L.; Entchev, E.; Hannich, J. T.; Ejsing, C. S.; Kurzchalia, T.; Shevchenko, A. Anal. Chem. 2006, 78, 585− 595. (17) Herzog, R.; Schwudke, D.; Schuhmann, K.; Sampaio, J. L.; Bornstein, S. R.; Schroeder, M.; Shevchenko, A. Genome Biol. 2011, 12, R8. (18) Herzog, R.; Schuhmann, K.; Schwudke, D.; Sampaio, J. L.; Bornstein, S. R.; Schroeder, M.; Shevchenko, A. PLoS One 2012, 7, e29851. (19) Kochen, M. A.; Chambers, M. C.; Holman, J. D.; Nesvizhskii, A. I.; Weintraub, S. T.; Belisle, J. T.; Islam, M. N.; Griss, J.; Tabb, D. L. Anal. Chem. 2016, 88, 5733−5741. (20) Fahy, E.; Subramaniam, S.; Murphy, R. C.; Nishijima, M.; Raetz, C. R. H.; Shimizu, T.; Spener, F.; van Meer, G.; Wakelam, M. J. O.; Dennis, E. A. J. Lipid Res. 2008, 50, S9−S14. (21) Han, X.; Gross, R. W. Mass Spectrom. Rev. 2005, 24, 367−412. (22) Hyötyläinen, T.; Orešic, M. Anal. Bioanal. Chem. 2015, 407, 4973−4993. (23) Schwudke, D.; Schuhmann, K.; Herzog, R.; Bornstein, S. R.; Shevchenko, A. Cold Spring Harbor Perspect. Biol. 2011, 3, a004614− a004614. (24) Röst, H. L.; Sachsenberg, T.; Aiche, S.; Bielow, C.; Weisser, H.; Aicheler, F.; Andreotti, S.; Ehrlich, H.; Gutenbrunner, P.; Kenar, E.; et al. Nat. Methods 2016, 13, 741−748. (25) Wang, M.; Huang, Y.; Han, X. Rapid Commun. Mass Spectrom. 2014, 28, 2201−2210. (26) Makarov, A.; Denisov, E.; Kholomeev, A.; Balschun, W.; Lange, O.; Strupat, K.; Horning, S. Anal. Chem. 2006, 78, 2113−2120. (27) Fauland, A.; Kofeler, H.; Trotzmuller, M.; Knopf, A.; Hartler, J.; Eberl, A.; Chitraju, C.; Lankmayr, E.; Spener, F. J. Lipid Res. 2011, 52, 2314−2322. (28) Jenkins, B.; West, J.; Koulman, A. Molecules 2015, 20, 2425−2444. (29) Surma, M. A.; Herzog, R.; Vasilj, A.; Klose, C.; Christinat, N.; Morin-Rivron, D.; Simons, K.; Masoodi, M.; Sampaio, J. L. Eur. J. Lipid Sci. Technol. 2015, 117, 1540−1549. (30) Zhao, Z.; Xu, Y. J. Chromatogr. B: Anal. Technol. Biomed. Life Sci. 2009, 877, 3739−3742. (31) Wang, W.; Liu, Z.; Ma, L.; Hao, C.; Liu, S.; Voinov, V. G.; Kalinovskaya, N. I. Rapid Commun. Mass Spectrom. 1999, 13, 1189− 1196. (32) Wang, C.; Wang, M.; Han, X. Mol. BioSyst. 2015, 11, 698−713. (33) Husen, P.; Tarasov, K.; Katafiasz, M.; Sokol, E.; Vogt, J.; Baumgart, J.; Nitsch, R.; Ekroos, K.; Ejsing, C. S. PLoS One 2013, 8, e79736.

ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.analchem.6b04456. Isotope intensity, peak coisolation and MS peak count, isotope delta mass, Gaussian mixture model, Cauchy/ Lorentzian mixture model, sodium-free isotopic peak overlap, corrected LipidMaps species, removal of redundant molecular species, and confusion tables for the most frequent MDDs (PDF) CSV databases (ZIP)

■

AUTHOR INFORMATION

Corresponding Authors

*E-mail: [email protected]. Tel: +49 30 9406 3114. Fax: +49 30 9406 49164. *E-mail: [email protected]. Tel: +49 30 9406 3114. Fax: +49 30 9406 49164. ORCID

Chris Bielow: 0000-0001-5756-3988 Present Address

M.O.: Department of Pharmaceutical Sciences, Università degli Studi di Milano, Via L. Mangiagalli 25, 20133 Milan, Italy. Author Contributions

The manuscript was written by C.B., with contributions of all authors. All authors have given approval to the final version of the manuscript. Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS We thank Tobias Opialla for proofreading the draft. Henning Langer provided the lipid phase of rat tissues and blood. G.M. and S.K. gratefully acknowledge funding by BMBF and the Senate of Berlin via the Berlin Institute for Medical Systems Biology.

■

REFERENCES

(1) Quehenberger, O.; Armando, A. M.; Brown, A. H.; Milne, S. B.; Myers, D. S.; Merrill, A. H.; Bandyopadhyay, S.; Jones, K. N.; Kelly, S.; Shaner, R. L.; et al. J. Lipid Res. 2010, 51, 3299−3305. 2994


On Mass Ambiguities in High-Resolution Shotgun Lipidomics

Recommend Documents