Mass spectrometry images acylcarnitines, phosphatidylcholines, and sphingomyelin in MDA-MB-231 breast tumor models.

The lipid compositions of different breast tumor microenvironments are largely unknown due to limitations in lipid imaging techniques. Imaging lipid distributions would enhance our understanding of processes occurring inside growing tumors, such as cancer cell proliferation, invasion, and metastasis. Recent developments in MALDI mass spectrometry imaging (MSI) enable rapid and specific detection of lipids directly from thin tissue sections. In this study, we performed multimodal imaging of acylcarnitines, phosphatidylcholines (PC), a lysophosphatidylcholine (LPC), and a sphingomyelin (SM) from different microenvironments of breast tumor xenograft models, which carried tdTomato red fluorescent protein as a hypoxia-response element-driven reporter gene. The MSI molecular lipid images revealed spatially heterogeneous lipid distributions within tumor tissue. Four of the most-abundant lipid species, namely PC(16:0/16:0), PC(16:0/18:1), PC(18:1/18:1), and PC(18:0/18:1), were localized in viable tumor regions, whereas LPC(16:0/0:0) was detected in necrotic tumor regions. We identified a heterogeneous distribution of palmitoylcarnitine, stearoylcarnitine, PC(16:0/22:1), and SM(d18:1/16:0) sodium adduct, which colocalized primarily with hypoxic tumor regions. For the first time, we have applied a multimodal imaging approach that has combined optical imaging and MALDI-MSI with ion mobility separation to spatially localize and structurally identify acylcarnitines and a variety of lipid species present in breast tumor xenograft models

from breast tumor xenograft tissue sections. We present spatial distributions of sphingomyelin (SM) and multiple phosphatidylcholine (PC) species, which were colocalized with normoxic, hypoxic, or necrotic tumor regions. MSI also localized the distribution of two acylcarnitine molecular ions, which were identifi ed from tumor tissue by ion mobility separation (IMS) followed by tandem mass spectrometry. We have demonstrated that the incorporation of MSI into a multimodal imaging approach revealed the molecular complexity of tumor tissue at an unprecedented level.

Chemicals and materials
The matrix ␣ -cyano-4-hydroxycinnamic acid (CHCA) was purchased from Sigma-Aldrich (Germany). Water, acetonitrile (ACN), and trifl uoroacetic acid (TFA) were purchased from Biosolve (The Netherlands). Cresyl violet acetate, Ponceau S, and paraformaldehyde powder were purchased from Sigma-Aldrich. Gelatin Type A was purchased from Sigma. Mayer's hematoxylin was purchased from Sigma, and aqueous Eosin Y from EMD Chemicals, Inc. Cytoseal 60 Mounting Medium, Richard-Allan Scientifi c, was purchased from Thermo Scientifi c.

Breast tumor xenograft model
The MDA-MB-231 human breast cancer cell line was purchased from the American Type Culture Collection (ATCC) and used within 6 months of obtaining from ATCC. This cell line was tested and authenticated by ATCC using two independent methods: the ATCC cytochrome C oxidase I PCR assay, and short tandem-repeat profi ling using multiplex PCR. The MDA-MB-231 cell line was genetically modifi ed to express tdTomato red fl uorescent protein under the control of hypoxia response elements (MDA-MB-231-HRE-tdTomato) as previously described ( 42 ). Cells were injected into the upper thoracic mammary fat pad of athymic nude mice (2 × 10 6 cells/injection), and tumor growth was monitored using standard calipers. The research was conducted in conformity with the Public Health Service policy on Humane Care and Use of Laboratory Animals. All experimental animal protocols were approved by the Institutional Animal Care and Use Committee of the Johns Hopkins University School of Medicine.

Tumor dissection and optical imaging
When tumors reached a volume of approximately 500 mm 3 , mice were euthanized and tumors were removed. Three individual tumors were embedded into 10% gelatin blocks prepared using 15 mm × 15 mm × 5 mm cryomolds (Sakura Finetek). Cresyl violet acetate and Ponceau S were prepared in 10% warm (37°C) gelatin at a concentration of 10 mg/ml, mixed in a 1:1 ratio, and injected into 10% gelatin blocks next to the tumor tissue. Each block was sectioned into serial 2 mm-thick fresh tumor sections using an acrylic adjustable tissue slicer (12 mm depth up to 25 mm width, Braintree Scientifi c, Inc.; Braintree, MA), and tissue slicer blades (Braintree Scientifi c, Inc.) as previously described ( 43,44 ). These serial fresh tumor xenograft sections were each placed on individual microscope slides (Fisherbrand catalog number 12-550-34, Fisher Scientifi c; Pittsburgh, PA), and stored in an ice box containing ice on the bottom, with the slides located on a perforated plate at approximately 1-cm above the ice to minimize tissue degradation ( 43,44 ). These fresh sections were imaged by of such heterogeneity is tissue hypoxia, a state of low oxygen tension (pO 2 values р 2.5 mmHg) observed in many solid tumors ( 2 ). Hypoxia increases overall tumor aggressiveness, induces radio and drug resistance, triggers resistance to apoptosis, and enhances cellular migration and invasiveness, eventually leading to cancer metastasis ( 3 ). At the molecular level, hypoxia triggers many known and unknown metabolome, lipidome, and proteome changes that modulate tumor vascular expansion and provide an overall advantage for malignant growth, causing rapid disease progression ( 4 ). Severe, prolonged hypoxia can lead to cell death and tissue necrosis ( 4 ). Tumor expansion results in the formation of metabolically diverse, viable and necrotic, normoxic and hypoxic, vascularized and avascular tissue regions, which, to date, have not been imaged in a single experiment or by a single imaging technique. We hypothesize that these different tumor tissue regions have different distributions of metabolites and lipids. MSI has the capability of parallel detection and visualization of biomolecules present in these specifi c tissue microenvironments, which allow for a more-comprehensive analysis of tumor biology.
Unlike cancer cell cultures, human tumor xenograft models in mice provide a three-dimensional, native-like environment for studying different aspects of tumor growth ( 5 ). Additionally, tumor xenograft models allow for genetic modifi cations such as incorporation of genes encoding fl uorescent proteins for visualizing and monitoring selected processes such as the influence of tissue hypoxia on tumor extracellular matrix ( 6 ) or cancer metastasis ( 7 ).
Lipid-rich tumors have been associated with increased tumor aggressiveness and metastasis (36)(37)(38). To develop effective cancer treatment strategies, it is crucial not only to identify but also to localize the lipid molecules involved in cancer progression. To date, lipid distributions in breast tumors remain largely unknown due to limitations in lipid imaging methods, which often require incorporation of fl uorescent tags and provide limited specifi city (39)(40)(41). Here, we have adopted a novel multimodal imaging approach, which integrates microscopy and MSI, for a labelfree, ex vivo visualization of different phospholipids directly fragmented using 20-30 V collision energy applied in the trap cell (in TOF mode) or 30-40 V applied in the transfer cell (in IMS mode). Spectra were analyzed using MassLynx software (Waters, UK). Ion mobility-separated ions were visualized by Driftscope software (Waters, UK), and selected drift time windows were exported to MassLynx for spectral analysis. FA chains of lipids of interest were identifi ed from a combination of the mass of the lipid and the MS/MS fragmentation pattern obtained with a mass accuracy of 0.1 Da at m/z 1,000. The LIPID MAPS, Human Metabolome Database version 2.5 (www.hmdb.ca) and the METLIN Metabolite Database (metlin.scripps.edu) were used to search for possible lipid structures and MS/MS fragmentation spectra.

Overlap analysis
We analyzed three MDA-MB-231-HRE-tdTomato breast tumor xenografts that were imaged by the multimodal approach described above. Necrotic tumor regions were outlined in H and E-stained 10 µm-thick sections. Hypoxic tumor regions were detected from the fl uorescence images of tdTomato protein expression in 2 mm-thick sections. The remaining tumor regions were normoxic. Ion images of each individual m/z were overlaid with the corresponding optical images and coregistered by using the positions of fi ducial markers and tumor boundaries as previously described ( 43,44 ). The tumor region was masked by applying an active-shape model to the registered optical bright-fi eld images ( 48 ). The highest 5% of the area under the histogram of images was automatically thresholded as the ionic or the hypoxic region inside the tumor, respectively. The volume percentage of ion overlap with normoxic, hypoxic, and necrotic regions inside the tumor was calculated to quantify their colocalizations as previously described ( 43,44 ).Volume percentage of hypoxia was defi ned as the number of hypoxic voxels that overlap with a given ion normalized to total hypoxic voxels in the tumor. Similarly, volume percentage of ions with normoxic voxels and necrotic voxels was measured. The image registration and colocalization analysis program was written in Matlab 2012b (Mathworks, Inc.; Natick, MA). Mean and standard error of overlap volume percentage from tumor center sections (n = 3) were calculated for each m/z and tumor region using MS Excel 2003.

Multimodal imaging approach
We have employed a multimodal imaging approach for investigating the effects of tumor hypoxia and tumor necrosis on the regionally distinct, spatially heterogeneous lipid composition in a breast tumor model. In this approach, we have integrated optical microscopy with MSI. Bright-fi eld microscopy visualized the shape of the tumor, fl uorescence microscopy outlined the position of hypoxic tumor regions, and histochemical staining revealed the localization of necrotic tumor tissue. MSI employing MAL-DI-MSI visualized the distribution of intact phospholipid molecular ions. All modalities were combined through the presence of fi ducial markers visible in all imaging modalities ( 45 ). Figure 1 presents results obtained from a representative breast tumor xenograft using our multimodal imaging approach. Figure 1A shows a bright-fi eld image of a 2 mm-thick tumor section obtained from the tumor center, which was embedded in a gelatin block containing the bright-fi eld and fl uorescence microscopy with a 1× objective attached to a Nikon inverted microscope, equipped with a fi lter set for 528 to 553 nm excitation and 600 to 660 nm emission, and a Nikon Coolpix digital camera (Nikon Instruments, Inc.; Melville, NY). Bright-fi eld imaging captured the position of the fi ducial markers present inside the gelatin block, as well as the shape of the tumor tissue. The fl uorescence from tdTomato expression in hypoxic regions of these tumor sections was detected by fl uorescence microscopy. The GNU Image Manipulation Program (GIMP 2.6) was used for two-dimensional (2D) coregistration and overlay of bright-fi eld and fl uorescence images of 2 mm-thick tumor sections. All 2 mm-thick sections were snapfrozen immediately after microscopic imaging and stored at Ϫ 80°C until further analysis.

H and E staining method
Tissue sections were stained using a modifi ed H and E staining protocol as previously described (43)(44)(45). Briefl y, 10 µm sections attached to Superfrost Slides (VWR International, Cat. 48311-600) were washed with PBS, fi xed with freshly depolymerized 3% paraformaldehyde solution for 30 min, washed with distilled water (dH 2 O), and treated with Mayer's hematoxylin for 30 min at room temperature, followed by fi ve washes with dH 2 O. Sections were immediately immersed in aqueous Eosin Y for 30 min, followed by fi ve washes with dH 2 O, mounting with aqueous mounting medium, and attaching of a coverslip. Bright-fi eld images of H and E-stained sections were acquired using 1× or 20× objectives attached to a Nikon microscope equipped with a Nikon Coolpix digital camera (Nikon Instruments, Inc.; Melville, NY).

MALDI-IMS-MSI
Six tissue sections obtained from three individual tumors were subjected to MSI analysis. Before MSI analysis, tissue sections were placed inside a vacuum desiccator for 30 min. CHCA matrix was prepared at a concentration of 10 mg/ml in 1:1 ACN:H 2 O/0.1% TFA and was applied by an ImagePrep (Bruker; Germany) application system. Tissue sections were analyzed on a MALDI-Q-TOF (Synapt HDMS; Waters, UK) instrument in IMS mode detecting positive ions. Images were acquired with 150 m × 150 m spatial resolution and a 100 m laser beam spot size. For better detection of masses above m/z 700, the instrument was set to suppress all masses below m/z 700. For 2D MSI analysis and overlay of images, data were visualized using BioMap software (Novartis; Basel, Switzerland). Ion mobility data analysis and visualization were performed by using DriftScope software (version 2.1; Waters, UK) as well as high-defi nition imaging (HDI) software (version 31; Waters, UK). The LIPID Metabolites And Pathways Strategy database (LIPID MAPS, www.lipidmaps.org) was used to search for possible lipid structures ( 46,47 ).

Lipid identifi cation by MS/MS analysis
The identifi cation of selected lipid species was performed on tissue in positive-ion mode using a MALDI-Q-TOF instrument (Synapt; Waters, UK) after completing MSI experiments. Lipid ion precursors were selected within a 2 Da selection window and shows an adjacent 10 µm-thick section stained with hematoxylin and eosin (H and E) for visualization of the necrotic tumor regions. Figure 1H presents the shape of the necrotic regions inside the tumor, outlined by a dotted line. Figure 1I shows the coregistration of the necrotic tumor regions obtained from the H and E image ( Fig. 1H ) with the hypoxic tumor regions from the fl uorescence image ( Fig. 1B ). The fi ducial markers were not clearly visible in some H and E images, in which case, the coregistration was performed based on the tumor shape by using the Im-ageJ-TurboReg plug-in (http://rsb.info.nih.gov/ij). Magnifi ed images of viable and necrotic tumor regions, as indicated in Fig. 1G, are shown in Fig. 1J, K, respectively.

Mass spectrometric analysis of lipids
Mass spectrometric imaging was employed directly on thin tumor tissue sections for detection, visualization, and identifi cation of a number of lipid biomolecules. Figure 2 presents selected molecular structures of acylcarnitine and lipids detected by MSI. These biomolecules all have a quaternary amine group, which can attain a positive charge, resulting in good ionization properties.  Figure 1B presents the fl uorescence image of tdTomato red fl uorescent protein exclusively expressed in the hypoxic tissue regions of this genetically modifi ed tumor model ( 42 ). The coregistration of the bright-fi eld image ( Fig. 1A ) with the tdTomato fl uorescence image ( Fig. 1B ) was performed based on the position of the fi ducial markers, which were visible in both images. The red color of the tdTomato signal was changed to black for better visualization of the hypoxic tumor regions and subsequently overlaid on the bright-fi eld image of the tumor tissue as shown in Fig. 1C . Figure 1D presents the 10 µm-thick section obtained from this tumor for MSI. The position of the fi ducial markers as visualized by MSI (shown in Fig. 1E ) was used for coregistration of the MSI dataset with the fl uorescence image of tdTomato as shown in Fig. 1F . Figure 1G

Mass spectrometric imaging of lipids
MALDI imaging. Images obtained from optical imaging and MALDI-MSI were coregistered as previously described based on the position of the fi ducial markers species were detected in the mass range from m/z 450 to m/z 550. Heme was observed at m/z 616.1 [M] + as a singly charged ion. Intact lipid molecular ions were detected as protonated ions as well as sodium and potassium salt adducts in the mass range from m/z 700 to m/z 860. Figure 4 presents an expanded view of the spectrum shown in Fig. 3 showing multiple sphingomyelin and phosphatidylcholine species detected as proton, sodium, and potassium adducts. We detected and identifi ed protonated  Table 1 .
The structures of all described lipid species were confi rmed by MALDI-MS/MS profi ling experiments performed directly on tumor tissue sections. Figure 5 shows a typical MS/MS spectrum of protonated PC(16:0/18:1), which has a highly abundant fragment signal at m/z 184.0 arising from the protonated phosphocholine ion, a signature peak in the MS/MS of protonated lysophosphatidyl- To quantify the ion signals in different breast tumor regions, we performed an overlap analysis of each ion distribution with each of the three distinct regions, namely the normoxic, hypoxic, and necrotic tumor regions, which were visualized by our multimodal imaging approach from three MDA-MB-231-HRE-tdTomato tumor xenografts.  been separated from other ions of similar mass. Figure 9 shows an example of two ions at m/z 428.25 and m/z 428.38, which were analyzed in IMS mode. Separated precursor ions were fragmented, which resulted in the identifi cation of m/z 428.3 as stearoylcarnitine [M+H] + . Figure 9A presents the ion mobility graph from two fragmented ions. The left side, circled by the red box, shows the separated faster ion and its fragments, while the green box encircles the slower ion and its fragments. Figure 9B presents the corresponding drift time plot showing the intensity of the slow and fast ions and their drift times, which are 1.7 milliseconds and 2.5 milliseconds, respectively. Figure 9C presents a full spectrum of detected ions and their fragments, obtained without IMS. IMS performed prior to fragmentation resulted in two spectra of fast ( Fig. 9D ) and slow ( Fig. 9E ) ions. Subsequent collision-induced dissociation allowed for the identifi cation of the slower ion at m/z 428.3 as protonated stearoylcarnitine.
LPC(16:0/0:0) was highly enriched in necrotic tumor regions. The remaining PCs, SM, and acylcarnitines displayed the highest percentage of overlap with hypoxic tumor regions. Imaging of isobaric species in IMS mode . IMS enables the separation of ions on the basis of their collisional crosssection, which depends on their shape and charge. In our study, IMS separated the background ions (chemical noise, matrix clusters, etc.) from the biomolecular lipid species. Figure 8A presents a spectrum of peak-picked data acquired during an IMS imaging experiment, and Fig. 8B shows a drift plot of the separated ions. In Figs. 8A and 8B, background ions are highlighted in blue, whereas biomolecular lipid ions are highlighted in red. Ion image 1 in Fig. 8B shows the distribution of a background matrix ion, which was separated from two lipid ions, shown in ion images 2 and 3.
IMS not only improved the MSI data obtained from complex tumor tissue, but also enabled the identifi cation of several biomolecules, which otherwise could not have

Advantages of MSI-IMS over other techniques used in lipidomics studies
In the present study, we analyzed human breast tumor xenograft tissue using MSI-IMS, and localized multiple  performed on lipid extracts by HPLC or TLC prior to mass analysis ( 49,50 ). IMS, in which separation occurs in milliseconds, can be easily coupled with MSI and helps obtain ion images of individual lipid species without requiring labor-intensive sample preparation procedures or long acquisition times. The development of new HDI software (Waters, UK) incorporating drift plots and ion drift time made our data analysis fast and simple. The effects of acyl chain length, degree of unsaturation, lipid class, head group, and cation formation on an ion's collision crosssection in the gas phase and on its ionization effi ciency have been studied for direct tissue profi ling ( 51 ). Such information greatly improves our understanding of ion behaviors detected from a complex tissue sample during MSI-IMS.
achieved on the basis of their molecular mass and characteristic fragmentation patterns. Such an imaging approach requires very little sample modifi cation prior to imaging, and introduces minimal alterations in lipid localizations and structures. Phospholipid ion formation by MALDI-MSI, when performed directly on a tissue section, generates abundant alkali metal adduct ions, a phenomenon not observed when MALDI is performed on pure lipid extracts. Here, we have employed IMS, because tissue washing removes certain metabolites and lipids from the tissue or leads to their diffusion. Hence, we decided to follow a sample preparation procedure that introduces minimal changes to the composition of the tissue. IMS provided a separation step for our imaging experiments that is usually  tion and acts as a precursor of sphingolipid mediators ( 62,63 ). Sphingolipids such as SM(d18:1/16:0) are detected by MSI in the same mass range as PC species. However, the protonated ions of PC species have even m/z values, while those of SM species have odd m/z values; therefore, although they occupy similar mass ranges, the peaks do not overlap, as evident in Fig. 4 ( 51 ). Sphingolipids contribute to the pathogenesis of cancer as well as neurological and autoimmune disorders ( 64 ). Several sphingolipid-regulated functions are implicated in tumor initiation, promotion, progression, and response to chemotherapy due to their importance in regulating growth, survival, adhesion, and migration of cancer cells ( 65 ). Sphingolipids have been proposed as cancer markers ( 66 ), and tumors often display alterations in sphingolipid composition ( 66 ). Sphingolipid intermediates may trigger proliferation and impair the ability of damaged cells to undergo apoptosis ( 65 ). Signaling-induced hydrolysis of sphingolipids produces a series of messenger lipids, such as ceramides and ceramide-1-phosphate, which, as part of signaling cascades, act by recruiting cytosolic proteins. Sphingosine, sphingosine-1-phosphate, ceramide, and ceramide-1-phosphate have pivotal roles in apoptosis, infl ammation, and arachidonic acid signaling, all of which are pathways that are impaired in many cancers ( 67,68 ). Carnitine and acylcarnitines are essential compounds in FA metabolism ( 69 ). Carnitine assists with transport of FAs into mitochondria, where FAs are oxidized by ␤ -oxidation as a major source of energy. We have detected and imaged two acylcarnitines directly from MDA-MB-231-HRE-tdTomato breast tumor tissue sections by applying MSI together with on-tissue IMS-MS/MS analysis combined with a Human Metabolome Database search for their identifi cation. IMS was successfully employed for removing background ion fragments. We have identifi ed ions at m/z 400.3 and m/z 428.3 as protonated ions of palmitoylcarnitine and stearoylcarnitine, respectively, which both localized to hypoxic tumor regions. This accumulation of acylcarnitines in hypoxic tumor regions probably occurred because of a blockage of oxidative processes in these hypoxic tumor regions ( 70 ). The inhibition of ␤ -oxidation during hypoxia leads to an accumulation of acylcarnitines in the cytoplasm of hypoxic cells ( 69 ). The concentration of long-chain acylcarnitines increased rapidly within minutes after the onset of ischemia in vivo or hypoxia in vitro ( 71 ), as shown in plasma and hearts of hypoxic rats (72)(73)(74), a newborn rat model of hypoxia-ischemia ( 75 ), adult canine myocytes ( 76 ), rabbit fetal lungs ( 70 ), and blood samples from neonates ( 77 ). Acylcarnitines are highly polar, zwitterionic molecules and therefore tend to elute together with phospholipids such as phosphatidylcholine in many chromatographic systems. Here, we have demonstrated that due to the presence of their positively charged quaternary ammonium group, acylcarnitines can also be easily detected and localized by MALDI-MSI.

CONCLUSIONS
We detected, identifi ed, and visualized several important lipid species in a human breast tumor xenograft model To date, IMS has been applied to map the distribution of PC and cerebroside species from coronal rat brain sections ( 52 ) and for imaging of glucosylceramide in spleen sections from a mouse model of Gaucher disease ( 53 ).

Identifi cation of lipids and their biological role
Most of the lipids identifi ed in this study were SM and PC species detected in the mass range from m/z 700 to m/z 850. The presence of a positively charged quaternary ammonium group in both PC and SM species aids in the detection of these two classes of phospholipids in positiveion mode ( 51 ). Previous MALDI-MS studies of phospholipid mixtures have demonstrated that in positive-ion mode, PC and SM species can inhibit the detection of other classes of phospholipids, such as phosphatidylethanolamines, present in the tissue at similar concentrations ( 54,55 ). In this study, we focused on identifi cation and imaging of phosphocholine-containing lipids present in MDA-MB-231-HRE-tdTomato breast tumor xenograft models.
Analysis of MALDI mass spectra directly obtained from MDA-MB-231-HRE-tdTomato breast tumor sections revealed multiple PC species with FA chains ranging between 16 and 22 carbons and containing 0 to 2 unsaturated bonds ( Fig. 4 ). These PC species were detected primarily in viable regions of MDA-MB-231-HRE-tdTomato breast tumors, with PC(16:0/16:0) predominantly localized to normoxic tumor regions and all other PC species localized primarily to hypoxic tumor regions ( Fig. 7 ). The identifi cation of all lipid species involved collisional activation (see Fig. 5  , which could be due to breakdown of intact PC species. LPC species are formed during infl ammation, in cases of chemical exposure ( 56 ), and following heart ( 16 ) and brain ( 19,57 ) injury. Multiple different bioactive lysophospholipids exhibit pleomorphic effects on multiple cell lineages, including breast and ovarian cancer cells ( 58,59 ). LPC activates the ovarian cancer G protein-coupled receptor 1 family of G protein-coupled receptors , known to regulate cellular motility, growth, differentiation, and gene transcription, all of which are factors central to the biology of cancer ( 60 ). The effects of LPC on intracellular signaling in breast cancer cells is translated to functional changes such as increased production of multiple growth factors from breast cancer cells, including interleukin-6 and -8, which are potent regulators of neovascularization ( 61 ). The mechanisms regulating the production and degradation of lysophospholipids are just beginning to be elucidated.
MSI analysis of MDA-MB-231-HRE-tdTomato breast tumor tissue also revealed the presence of SM(d18:1/16:0), which was detected primarily in hypoxic tissue regions. SM(d18:1/16:0) is known to be involved in tissue infl amma-using MALDI-IMS-MSI. The detected lipids belonged to the lipid classes of PC and SM. For the fi rst time, different lipids were visualized in different microenvironmental tumor regions. Most PC species were found in viable normoxic and hypoxic breast tumor regions, whereas LPC was localized to necrotic regions, which could be attributed to PC breakdown triggered by tissue injury. Two acylcarnitines were localized in hypoxic tumor regions and identifi ed by using IMS. Their presence in the hypoxic regions suggests blockage of the ␤ -oxidation process of FAs inside mitochondria. In addition, SM(d18:1/16:0), known to be involved in cellular signaling processes, was enriched in hypoxic tumor regions. Our study demonstrated the fi rst visualization of phospholipids in a human breast tumor xenograft model and shed more light on processes occurring during tumor growth.