5.7.1 Types of Ions Generated by MSI
As MSI is performed on complex biological samples, such as tissue sections, spectra can become more complex compared to those acquired by analysis of lipid extracts because of the formation of multiple adducts for each lipid. These can complicate the spectra and increase the probability of unresolved isobaric overlaps that primar- ily arise from endogenous salts, leading to simultaneous detection of [M+H]+, [M+Na]+, and [M+K]+ ions. This is particularly noticeable for some PLs, GSLs, and TAGs that are detected in the positive‐ion mode but may also complicate negative‐
ion mode spectra where lipids such as CL can be detected as [M‐2H+Na]−/ [M‐2H+K]− ions, in addition to the more common [M−H]−. Proper consideration and identification of different adducts formed are essential to correctly identify lipids detected during MSI experiments.
Different adducts from the same lipid species may give rise to different ion distri- butions that reflect Na+/K+ ratios throughout the tissue. For example, decreased levels of [M+K]+ ions and increases in [M+Na]+ ions of PC lipids have been observed using MALDI‐MSI of brain tissue following traumatic brain injury [149]
and ischemic stroke [150, 151] and attributed to loss of Na/K‐ATPase. It is critical that these effects are considered and changing adduct ratios not incorrectly assigned as alterations in lipid distribution. One approach to circumvent the influence of salt distributions is to wash the sample with aqueous buffer solution, typically ammo- nium formate or ammonium acetate, that removes salts from the tissue and favors the formation of [M+H]+/[M−H]− ions [22, 23].
The polarity in which the mass spectrometer is operated significantly influ- ences the lipid species detected. Many lipid species are preferentially ionized in one polarity. Neutral and basic lipids (e.g. PC, sphingomyelin (SM), TAGs, and some GSLs) are usually best detected in the positive‐ion mode. In contrast, acidic lipids such as PS, PI, GL, CL, FFA, and bile acids are better detected in the negative‐ion mode in the form of deprotonated species. Therefore, acquiring both positive‐ and negative‐ion mode MSI data from the sample to maximize lipid coverage can be advantageous. Examples of this include the sequential
acquisition of dual‐polarity data using MALDI [152, 153] and DESI [154] or fast polarity switching during MSI experiments using MALDI [155], IR‐MALDESI [109], and DESI [156].
5.7.2 In-source Fragmentation Considerations
Careful consideration of in‐source fragmentation is critical for accurate data interpretation and lipid assignments. While in‐source fragmentation is also an issue for ESI analysis [157, 158], it can become even more significant for MSI where the extent of in‐source fragmentation can be larger for methods such as MALDI and especially SIMS. SIMS analysis of tissue and cells often yields high abundance fragment signals corresponding to the phosphocholine headgroup at m/z 184 and FFAs. Other examples of in‐source fragmentation that may be encountered during MSI include headgroup losses from phospholipids that give rise to either phosphatidic acid (PA) or [DAG+H/Na/K‐H2O]+ ions and the forma- tion of dimethyl‐PE (DMPE) ions that arise from the loss of CH3 from PC lipids following anion adduction and are isomeric with PE [159]. Figure 5.4 summarizes the in‐source fragments observed during the MALDI analysis of 17 phospholipids and sphingolipid standards. Note that in the case of MALDI [160], the extent of in‐source fragmentation can be influenced by the laser energy and the choice of matrix.
While it is generally advisable to minimize in‐source fragmentation, it may also be exploited to aid lipid identification. For example, Angerer et al. used a 40 keV cluster ion beam consisting of 15% CO2 in argon for the SIMS‐MSI analysis of Phagocata gracilis flatworms, resulting in the generation of both intact lipids signals and their in‐source fragments [161]. The co‐localization of fragments with their pre- cursor lipids allowed fragment/precursor assignments to facilitate annotation of lipids species to the molecular lipid species level.
5.7.3 MSI Lipid Identification Using Accurate Mass
The procedure for identifying lipids detected in MSI follows the same approach used for shotgun lipidomics (see Chapter 3). Sum‐composition identification acquired using high mass resolution mass analyzers such as Orbitrap, FTICR, and orthogonal TOF analyzers is commonly performed in MSI studies. Sum‐
composition identification typically requires mass accuracies ≤3 ppm and can be aided by online annotation platforms such as Metaspace [162, 163]. However, it is important to manually curate automated identifications to minimize false identifications such as those inconsistent with lipid ionization or biosynthesis pathways.
An essential requirement of accurate sum‐composition identification is the reso- lution of peaks from isobaric interferences [164]. In many MSI studies, mass
5 Mass Spectrometry Imaging of Lipids 134
resolutions >100 000 are obtained, which are sufficient to resolve many isobaric peaks. For example, the recent coupling of SIMS with Orbitrap mass spectrometry enabled SIMS‐MSI of lipids at mass resolutions of 240 000 (@ m/z 200) with pixel sizes <2 μm using a 20 keV argon GCIB [76]. While such resolutions in the lipid m/z‐
region are sufficient in many cases, a higher resolution can be required [165]. For example, a mass resolution of >150 000 @ m/z 750 is needed to resolve common isotopologues such as a 13C2‐containing peak with the monoisotopic peak of the same lipid containing one less double bond [166]. Even higher mass resolving pow- ers (>300 000) are needed to resolve the difference between 12C2 and 23Na1H, such as that needed to resolve [PC(36:1)+Na]+ and [PC(38:4+H)]+ ions. Mass resolutions as high as 106 have been reported for MSI using FTICR coupled with DESI [167], MALDI [49], and LAESI [107]. For example, using 21 T FTICR with a mass resolu- tion >800 000 in the phospholipid mass range, it was revealed that peak splits of
<2 mDa are commonly observed during tissue imaging experiments (e.g.
[PE(O‐38:7)13C1‐H]− and [PG(34:1)‐H]−). Such small mass differences cannot typi- cally be resolved using conventional instrumentation and require additional meth- ods such MS/MS (see below) to resolve and identify.
Positive-ion mode Negative-ion mode
DG TG PC/PC ether PE/PE ether
PC/PC ether PE/PE ether PI
PS PG SM HexCer
PI PI-Frag
Cer
GM3
PS PG SM HexCer
LacCer
SFT SFT-Frag
-Sialic Acid-Hex Cer
-Sialic Acid -Lac
-Hex
CerP -GM3HG
CL CL-Frag
-RCOOH-RCOOH -OH
-RCOOH -RCOH -DG+H2O
2H2O
-DG -RCOOH-RCOH GM3
CL -P-Cho
-P-Cho -P-EtnA
-P-Inositol -Inositol
-Serine -Serine
-N(CH3)3 -CH2N(CH3)3
-C4H10O5
-N(CH3)3 -Cho
-CH2N(CH3)3 -N(CH3)3
-Cho -Cho
-N(CH3)3 Cer-H2O
DG-H2O PA/PA ether PA/PA ether
-RCOOH
-EtnA: ethanolamine -Cho: choline -RCOOH: fatty acid -P-Cho: phosphocholine -P-EtnA: phosphoethanolamine -GM3HG: GM3 head group
-Sialic Acid -DG+H2O
-DG
-Cho CerP
LacCer
CL-Frag -EtnA
Figure 5.4 Summary of common in-source fragmentation pathways that may occur during MALDI-MS analysis of lipids. Similar processes can also occur for other MSI techniques.
Source: Reprinted with permission from reference Garate et al. [160]. Copyright (2020), American Chemical Society.
5.7.4 Deploying MS/MS for Lipid Identification in MSI
Identification beyond the sum‐composition requires MS/MS to characterize lipids to the molecular species level (i.e. identifying a PI containing 18:0 and 20:4 fatty acyls). MS/MS spectra can be obtained manually from the tissue following MSI acquisition for the identification of ions of interest. MS‐MSI can also be per- formed whereby selected peaks are fragmented and the ion distributions of frag- ment ions used for MSI. Prentice et al. have demonstrated the acquisition of multiple MS/MS spectra from each laser shot using MALDI‐axial‐TOF instru- ment by rapidly switching the timed ion selector and enabling detection of frag- ments supporting the identification of potassiated PC(32:0), PC(36:1), and PC (40:8) during a single MALDI‐MSI acquisition. Fisher et al. described a new tan- dem‐TOF‐SIMS instrument allowing simultaneous acquisition of both MS1 and MS/MS spectra from the same ion packets [168]. This system has been used to localize vitamin E within neuronal cells and Mycobacterium marinum‐infected zebrafish [169]. MS/MS also has the advantage of allowing differential imaging of lipids that cannot be resolved at the MS1 level. This has been deployed for nano‐
DESI‐MSI, whereby a targeted list of 92 m/z values was sequentially fragmented within each pixel [170]. As one example, detection of characteristic fragments allowed the imaging and identification of [PE(P‐16:0/22:6)+Na]+ and [PC(32:1)+K]+ lipids within mouse uterine tissue that would otherwise require a mass resolving power >3 500 000.
Broadband MS/MS data can also be acquired during MSI acquisitions using data‐dependent acquisition (DDA) approaches [171, 172]. While these do not allow for MS/MS‐imaging (as there are no spatially repetitive MS/MS data), they allow large‐scale identification of detected ions. For example, using a hybrid ion trap/Orbitrap instrument, MALDI‐MSI and accurate mass data can be acquired using the Orbitrap while DDA‐MS/MS are acquired in parallel using the ion trap.
This method has been demonstrated for rat brain imaging at a spatial resolution of 40 μm whereby automated lipid identifications utilizing both accurate mass MS1 data and MS/MS are performed using ALEX123 software [172]. Across the positive‐ and negative‐ion mode, the spatial distribution of 100 molecular lipid species was revealed in addition to their high confidence identification using both accurate mass (MS1) and MS/MS. For example, the addition of MS/MS enabled identifying isomeric PS(18:0_20:1) and PS(18:1_20:0) located primarily within the white matter (WM).
5.7.5 Isomer-Resolved MSI
MSI combined with conventional collision‐induced dissociation (CID) MS/MS does not allow the resolution of many isomeric lipids such as those varying in the stereo‐
specific number (sn) of acyl chain on the glycerol backbone or lipids varying in the
5 Mass Spectrometry Imaging of Lipids 136
double‐bond (db) position of acyl chain double bonds. As different isomeric variants can reflect different biosynthetic processes and/or give rise to different functional properties (e.g. membrane fluidity), developing methods for isomer‐resolved MSI has become a significant area of research in recent years.
The coupling of ozone‐induced dissociation (OzID) with MALDI‐MSI has facili- tated imaging of both sn‐ and db‐isomers of PC throughout biological tissues [26].
OzID relies on the gas‐phase reaction of mass‐selected lipids with ozone. The prod- ucts of this reaction allow unambiguous assignment of db‐positional isomers. For sn‐isomers, CID is first performed to generate the [M+Na‐183]+ ion, which is then reacted with ozone in an MS3 experiment to yield sn‐specific fragments. The recent coupling of this OzID with MALDI‐MSI on a Waters Synapt G2Si mass spectrometer using the high‐pressure ion mobility region as the reaction chamber facilitates the rates of five pixels/second for both db‐ and sn‐isomers [173]. By fragmenting a popu- lation of selected precursor ions before the ion mobility region, isomer‐ specific frag- ments can be generated simultaneously for db‐ and sn‐isomers. Further, sequential ozonolysis on CID/OzID product ions (CID/OzID2) reveals db‐positions specific to the sn‐1 location. The technique was deployed to map PC db‐isomer distributions throughout human prostate cancer tissue [174, 175]. In one study by Young et al., the distribution of PC(34:1n‐9) was correlated with epithelial cells from potential pre‐
malignant regions, whereas PC(34:1n‐7) correlated with immune cell infiltration and inflammation [175]. The isomer‐specific distributions were rationalized as arising from the different ordering of desaturation (SCD‐1) and elongation (EVOLV5/
EVOLV6) reactions throughout the tissue. In addition to OzID, electron‐induced dis- sociation (EID) on mass‐selected PC ions is another gas‐phase method enabling imaging of db‐ and sn‐isomers from tissue [176]. EID involves the reaction of ionized lipids with energetic (2–70 eV) electrons, whose interaction leads to rich MS/MS spectra containing both even and odd electron fragment ions, some of which are diagnostic of db‐ and sn‐position. Gas‐phase charge inversion methods involving the conversion of cationized PC lipids into demethylated anions have also been used to study sn‐isomer populations in brain tissue using MALDI‐FTICR [177]. Subsequent CID of the [PC−CH3]− produces conventional acyl chain‐specific fragments, whose ratios can provide insights into the relative population of different sn‐isomers.
Derivatization of lipids before analysis is another approach for isomer‐resolved MSI. On‐tissue Paternò–Büchi (PB) reactions are an example of this and were first coupled with MSI using a MALDI‐MSI approach and benzaldehyde as the reactive reagent [178]. Following deposition of the PB reagent, ultraviolet (UV) activation, mass selection, and CID of PB‐derivatized lipids diagnostic fragments allowing dif- ferentiation of db‐isomers are formed. PB reactions have been applied to numerous PLs including PCs as well as hexosylceramides. For example, PS(18:1_18:1)n‐9 was elevated in the gray matter of mouse brain tissue relative to the n‐7 variant.
Benzophenone is another PB reagent used for MALDI‐MSI of db‐isomers in both brain tissue and the parasite Schistosoma mansoni [179]. Unlike benzaldehyde, the PB reaction is induced upon UV laser irradiation used to initiate MALDI. Benzophenone has also been used to promote PB reactions for DESI‐MSI analysis of FA(18:1) db‐
isomers brain tumor tissue from mice [180]. Furthermore, a variety of alternative
derivatization methods that allow generating of products yielding db‐specific frag- ments upon CID have recently been reported, including hydroperoxides for nano‐
DESI‐MSI [181] and surface epoxidation for DESI [182] and IR‐MALDESI [183].
Ultraviolet photodissociation (UVPD) can also be used for isomer‐resolved MSI and involves the irradiation of mass‐selected lipids with a 193 nm laser pulse. To date, only DESI‐MSI has been coupled with UVPD [184], and it has been applied to study both PC isomers in brain tissue, in addition to region‐specific db‐isomer popu- lations throughout human lymph node tissue containing thyroid cancer metasta- sis [184]. UVPD has also been used to study FFA db‐isomers using a charge inversion approach between FFA and 1,8‐ethyl DC [185]. Changing populations of n‐9 and n‐7 FA18:1 were correlated with metastatic, estrogen receptor, and progesterone sta- tus of ovarian cancer tissue.