Matrix-Assisted Laser Desorption/Ionization (MALDI) Imaging Mass Spectrometry (IMS): A Challenge for Reliable Quantitative Analyses

Abstract

Matrix-assisted laser desorption/ionization (MALDI) imaging mass spectrometry (IMS) is capable of determining the distribution of hundreds of molecules at once directly from tissue sections. Since tissues are analyzed intact without homogenization, spatial relationships of molecules are preserved. The technology is, therefore, undoubtedly powerful to investigate the molecular complexity of biological processes. However, several technical refinements are essential for full exploitation of MALDI-IMS to dictate dynamics alteration of biomolecules in situ; these include ways to collect tissues, target-specific tissue pretreatment, matrix choice for efficient ionization, and matrix deposition method to improve imaging resolution. Furthermore, for MALDI-IMS to reach its full potential, quantitative property in the IMS should be strengthened. We review the challenges and new approaches for optimal imaging of proteins, lipids and metabolites, highlighting a novel quantitative IMS of energy metabolites in the recent literature.

Introduction

To understand the mechanism of any biological processes, it is vital to spatially resolve the distributions and abundance of candidate molecules responsible for the event in situ. Matrix-assisted laser desorption/ionization (MALDI) imaging mass spectrometry (IMS) technology is powerful in that it can determine the localization and abundance of hundreds of molecular species on the tissue slices at once.¹⁾ It provides comprehensive profiles of molecular distributions of proteins,^1–3) peptides,^4–6) lipids,^7,8) drugs,⁹⁾ and metabolites^10,11) with high spatial resolution. Readers are referred to more detailed review for more comprehensive account on MALDI-IMS technology.^12–15) Instead of giving a broad review, we describe the technical difficulties to reveal specific types of biomolecules especially focusing of small sized molecules (e.g. lipids, and metabolites) and update their solutions, including sample preparation, matrix choice and/or its deposition method found in the recent literature. We also highlight recent approaches allowing investigators not only to identify kinds of compounds on the samples but also to quantify exactly how much of each compound is there on the spot; i.e. quantitative MALDI-IMS.

Choice and Deposition of Matrix

The selection of matrix is an important step in the sample preparation for MALDI-IMS. Table 1 summarizes an overview for recommended matrices for the types of analytes. MALDI-IMS was initially applied to protein,¹⁾ then to lipids.

Table 1. Matrices used in MALDI-IMS studies.

Analyte	Application	Matrix	Reference
Proteins and peptides	Mass<5 kDa	α-Cyano-4-hydroxycinnamic acid (CHCA)	16)
		Ionic matrix (CHCA)/Aniline	17)
		Ionic matrix (CHCA)/N, N-dimethylaniline (DANI)	17)
	Mass>5 kDa	2,5-Dihydroxybenzoic acid (DHB)	4)
		Sinapinic acid (SA)	1)
Lipids	Positive mode	2,5-Dihydroxybenzoic acid (DHB)	4)
		2,6-Dihydroxy acetophenone (DHA)	7)
		Ionic matrix (CHCA)/butylamine (B)	18)
		Ionic matrix (DHB)/butylamine (B)	18)
		1,5-Diaminonaphthalene (DAN)	19)
		Nanoparticles	20)
	Negative mode	2,5-Dihydroxybenzoic acid (DHB)	21)
		9-aminoacridine (9-AA)	22)
		1,5-Diaminonaphthalene (DAN)	19)
Small molecules	Positive mode	2,5-Dihydroxybenzoic acid (DHB)	9)
	Negative mode	9-Aminoacridine (9-AA)	10)
Oligosaccharides		Ionic matrix 3-aminoquinoline (3-AQ)/(CHCA)	23)

It has been restricted to relatively high-molecular-weight compounds for a while, but not to low-molecular-weight “metabolites.” This is because conventional matrices such as 2,5-dihydroxybenzoic acid (DHB) and α-cyano-4-hydroxycinnamic acid (CHCA)²⁴⁾ produce a large number of interfering ions at low m/z range (150<m/z<600), which overlap with the range of m/z for small metabolites.²⁴⁾ Utilization of 9-aminoacridine (9-AA), which was found not to produce a large number of the cluster ions, opened ways to analyze metabolites.²⁴⁾ MALDI employing 9-AA matrix successfully identified multiple metabolites from the extracts of biological samples.²⁵⁾ Sun et al. identified 285 negatively charged ions from mouse heart extract, and 90 of the identified peaks were confirmed by tandem mass spectrometry.²⁶⁾ Subsequently 13 intrinsic metabolites were detected and identified from rat brain sections by MALDI-IMS at a 50 µm resolution.²⁷⁾ The ways to apply the matrix determine several factors. First, the lateral resolution depends on sizes of crystal. Second, extraction efficiency of analytes from the tissue depends on the accessibility of matrix solutions to the tissue interior and the process of recrystallization of matrix together with the extracted analytes during the solvent evaporation. Third, reproducibility of IMS data depends on the degree of homogeneous distribution within one tissue section and that of consistency among different tissue sections. Table 2 summarizes the different deposition methods of matrix.

Table 2. Matrix deposition methods used in MALDI-IMS.

Deposition method	Spatial resolution on tissue (µm)	Reference
Manual application	1,000<
Piezoelectric-based inkjet printer	150<	4,28,29)
Automated acoustic deposition	130<	30)
Electrospray deposition	100<	31)
Robotic sprayer	∼50	27)
Nebulized spray coating	∼10	32)
Sublimation	∼5	33)
Sublimation/recrystallization	∼10	34, 35)

Protein (<30 kDa) Imaging by MALDI-MS

The detection of intact proteins is still challenging. Although proteins constitute a larger fraction of cell mass (∼18% of total wet weight of the cell) than phospholipids (3%), their large molecular weight and species abundance (∼3,000 per cell) make the concentrations of one specific kind of proteins much lower than those for phospholipids.³⁶⁾ Indeed difficulties with identifying and imaging proteins by MALDI-IMS can stem from such properties which cause the inefficient extraction from the tissue; thereby, the poor ionization. In this section, we review recent advances to overcome the problem.

Since phospholipids are more easily ionized than proteins, removal of lipids from a specimen is a reasonable strategy for more efficient detection of the proteins.^34,37) To this end, several “tissue-rinsing” protocols have been developed. In these protocols, lipids are washed out with organic solvent prior to matrix deposition so as not to hamper the proteins to be ionized. Seeley et al.³⁷⁾ used isopropanol for rinsing and achieved a 6.5-fold increase in total ion currents of proteins (integrated m/z from 2,000 to 25,000) as compared to the unrinsed control. Use of alcohol increased the number of protein targets.³⁷⁾ Yang et al. refined such a protocol further using multi-step sequential rinsing with graded alcohols and Calnoy’s fluid to remove most of lipids and salts.³⁴⁾

Novel matrix deposition method developed by Yang et al.³⁴⁾ contributes to high-spatial resolution imaging of proteins. Sublimation of the matrix that is effective for lipid imaging achieving a spatial resolution of 5 µm,³³⁾ has not been adopted to the protein imaging because dry deposition of matrix limits the efficient extraction of the proteins. To incorporate the proteins into the matrix crystals effectively, the matrix is most often deposited on a tissue section by spraying or microspotting.⁴⁾ By wetting the tissue, these processes gain the sufficient extraction, but at the same time, cause potential delocalization/diffusion of the analytes. Yang et al. came up to a novel deposition method utilizing a recrystallization step after sublimation.³⁵⁾ These authors first coated the tissue surface by sublimation with sinapinic acid and exposed the tissue to solvent vapor in a small chamber with 1 mL of 5% acetic acid in water kept at 85°C for 3.5 minutes for recrystallization. With this, they achieved protein imaging (up to m/z 30,000) of the rodent brain tissue at a spatial resolution of ∼10 µm.

Here, we introduce recent biological implications unraveled by spatial profiling of proteins by MALDI-IMS of tissues collected from patients under various disease conditions.^1–3,38,39) Bauer et al. analyzed biopsy samples of breast cancer taken from patients before chemotherapy by MALDI-IMS. Three subtypes of α-defensin were found to be >30-fold overexpressed in tumor tissues from patients who responded to paclitaxel/radiation therapy compared to those from nonresponder patients.³⁹⁾ The study showed a clear relationship between biomarker expression and therapeutic efficacy. The anatomically oriented identification of protein biomarker of this kind should become a useful way to forecast the outcomes of the chemotherapy.

Lipid Imaging by MALDI-MS

Phospholipids are other target molecules to be visualized by MALDI-IMS.^{7,8,13,18,20,33,40,41)} Phospholipids consist of polar head group and two fatty acids (Fig. 1), and are classified into several groups depending on their chemical structure of headgroups (e.g. phosphatidylcholine (PC), phosphatidylethanolamine (PE), sphingomyelin, phosphatidylserine, phosphatidylinositol, and phosphatidylglycerol). A difficulty in imaging phospholipids is to separate the signal from isobaric (same molecular mass) molecules. Burnum et al. showed different distributions of isobaric PE (18 : 0/18 : 2) and PE (18 : 1/18 : 1) in murine day 8 embryos by MS/MS imaging^40,42) (Fig. 1).

Fig. 1. Different distribution of isobaric species separated by MS/MS imaging.

MS imaging of m/z 742.54 shows a ubiquitous pattern from top to bottom of the day 8 murine embryos. However MS/MS imaging of this ion shows that it is made up of two distinct isobaric structures that have very different localization patterns. PE (36 : 2; 18 : 0/18 : 2) is localized to the antimesometrial pole (lower), whereas PE (36 : 2; 18 : 1/18 : 1) is localized to the mesometrial pole (upper). Adapted by permission from the American Society for Biochemistry and Molecular Biology: Seeley et al. J. Biol. Chem. 286: 25459–25466, 2011.⁴²⁾

Since tissue samples contain considerable amounts of Na⁺ and K⁺, PCs are typically found in the forms of [M+K]⁺ and/or [M+Na]⁺ (Fig. 2 upper panel).^8,18) These alkali-metal adducts derived from a single molecular species make it difficult to characterize the abundance of the molecule on the tissue. This is because the salt spot-by-spot distribution per se is heterogeneous in the tissue, a resulted image might reflect the abundance of these salts rather than that of the target species. Furthermore, a single peak often contains multiple types of the molecular ion adducts; e.g., sodiated PC (16 : 0/18 : 1) and protonated PC (16 : 0/20 : 4) ions that appears at the same m/z 782 in the positive-ion mode. To solve this problem, Sugiura et al.⁸⁾ developed a novel method in which they intentionally added 20 mM potassium acetate to the matrix solution in order to force generate [M+K]⁺ ions (Fig. 2 lower panel). By so doing, these authors effectively minimized the number of the molecular ion adducts and showed unique spatial distributions of PC (16 : 0/18 : 1) and PC (16 : 0/20 : 4) in rat brain sections.

Fig. 2. Spectral change of mouse brain homogenate by potassium acetate addition.

Generation of multiple molecular ions from a single PC molecular species was reduced by adding an alkali-metal salt to the matrix solution. The use of the salt-added matrix solution allowed multiple molecular ion-forms of PCs (upper panel) to be reduced to a single potassiated molecular ion form (lower panel). Adapted by permission from the American Society for Biochemistry and Molecular Biology: Sugiura et al. J. Lipid Res. 50: 1776–1788, 2009.⁸⁾

Besides the structural lipids discussed above, a short lived lipid mediators such as prostaglandins (PGs) and leukotrienes, have been desired to be analyzed their spatial-temporal abundance on the tissue. Concentrations of PGs typically fall into the order of nM when measured for the concentrated extract from the inflamed tissue by liquid chromatograph mass spectrometer (LC-MS) or enzyme-linked immunosorbent assay (ELISA). To improve the detection efficiency of such low-concentration analytes, an attempt was made to improve the ionization efficiency by converting a ketone group of PGs to a quaternary amine group that has a permanent positive charge. Although the imaging of PGs have not been reported yet, such a derivatization approach⁴³⁾ will open door to MALDI-IMS analyses of PGs in the near future.

Metabolite Imaging by MALDI-MS

Although MALDI-IMS technology intends to reveal complexity of biological events by resolving spatio-temporal dynamics of various molecules on the tissue, the experimental cautions should be taken as to the ways to collect the tissues. It has been recognized as a problem for many years that tissue constitutes (e.g. proteins, lipids, and nucleic acids) rapidly change during sample collection due to autolysis and/or enzymatic reactions. It is, therefore, imperative to optimize fixation conditions for metabolites that reflect what is going on in situ not what has been happened after tissue removal. Corder et al. reported that decapitation before fixation caused a rapid decrease in ATP content, reaching approximately 25% of control at first 1 minute. Likewise, phosphocreatine, norepinephrine, epinephrine, and dopamine decreased rapidly after decapitation, reaching a plateau of 9.1%, 8.7%, 3.0%, and 6.1% of control by 2 minutes, respectively.⁴⁴⁾ To avoid such postmortem degradation of metabolites, activities of enzymes should completely be inactivated by quick fixation methods including the focused microwave irradiation or in situ freezing.^10,45) See the reviews for more comprehensive account on this issue.^46,47) Providing the spatial information of adenylates contents of the mouse brain, Hattori et al. showed how rapidly the degradation of adenylates takes place when it was not properly prepared.¹⁰⁾ Apparent contents of ATP, ADP, and AMP ([ATP]_app, [ADP]_app, and [AMP]_app, respectively) in the in situ freezing method (ISF) varied considerably among different cerebral architecture (Fig. 3A, left). [ATP]_app in the cortex was greater than those in amygdala and hypothalamus, while the opposite was the case for [AMP]_app. Such structure dependence in [ATP]_app is consistent with regional heterogeneity of local cerebral metabolic rate for glucose validating this method.^48,49) On the other hand, post-mortem freezing (PMF; decapitation before freezing) caused unacceptable autolytic reduction in ATP and increases in ADP and AMP. In this case, the heterogeneity disappeared (Fig. 3A, right).

Fig. 3. Degradation of adenylates caused by delayed fixation of tissue constituents.

(a) Brain is extremely susceptible to postmortem changes in contents of cerebral labile metabolites such as adenylates. Thus the best available methods must be employed to trap the metabolites, as they exist in vivo and to minimize autolytic changes. To achieve this, the in-situ freezing (ISF), which enables suspension of metabolic processes by rapidly lowering the tissue temperature while maintaining blood flow and oxygenation during the freezing process, was employed. Note that values of [ATP]_app are much higher with in situ freezing (ISF) versus postmortem freezing (PMF; freezing after decapitation). Content maps for ATP, ADP, and AMP are constructed on the same tissue. (Top) Hematoxylin and eosin (H&E) staining after imaging mass spectrometry (IMS). Am; amygdala, Hp; hypothalamus. Scale bar=1.0 mm. (b) Contents of adenylates determined by capillary electrophoresis electrospray ionization mass spectrometry (CE/ESI/MS). EC_AVE, averaged energy charge. (Arrowheads) Experiments depicted in (a). Gray bars: mean values. * p<0.05. Adapted by permission from Mary Ann Liebert, Inc. publishers: Hattori et al. Antioxid. Redox Signal. 13: 1157–1167, 2010.¹⁰⁾

It is worth noting that development of a novel strategy is prerequisite for achieving such an intergroup comparison of adenylates contents between ISF and PMF groups since the MALDI/IMS per se still needs further efforts to be supported for quantification. In the following section, we introduce a new tactics to grant the IMS technology a more quantitative property.

Current Attempts for More Quantitative IMS

MALDI-IMS has strengths in analyzing the distribution of many metabolites in discrete areas with a single laser ablation.^1,11,41) However, as it stands, it is not sufficient to provide quantitative information for the analytes. By contrast, CE-MS excels in quantification of metabolites.^50–52) However, it removes spatial distribution of molecules due to tissue homogenization to extract metabolites. Hattori et al.¹⁰⁾ combined MALDI-IMS and capillary-electrophoresis (CE)-MS. Combining MALDI-IMS with CE-MS complements each other’s weakness and enables to transform acquired mass signals of a metabolite in absolute terms such as tissue content in µmol/g. Thus, it is possible to construct maps of small-molecule metabolites whereby abundance of metabolites was assigned in the tissue. Such assignment of contents makes it possible to directly compare patterns of biochemical derangements in the tissue at different time points; which may help determine the multimodal-reaction points of gaseous mediators in the tissue. Such a novel approach enabled to reveal the physiologic role of endogenous carbon monoxide in the central nervous system. The gas mildly suppresses ATP production during normoxia, which gives way to the rise in dynamic strength of compensatory ATP maintenance upon hypoxia.⁵³⁾

Kubo et al. subsequently modified the above approach to characterize metabolic systems on a human-derived solid tumor in vivo¹¹⁾(Fig. 4). Analyses of metabolism in human solid tumor have been difficult because the methods to fix the metabolites in situ (e.g. in situ freezing, microwave irradiation) applicable for animal experiments are not allowed for live human subjects. To solve this problem, these authors utilized super-immunodeficient NOD/SCID/IL-2Rγ^null (NOG) mice in which xenograft transplantation of human-derived colon cancer HCT 116 cell line was conducted to induce hepatic micrometastasis of the solid tumor.⁵⁴⁾ Thus the metabolic dynamics of human colon cancer can be dictated in this mouse model.

Fig. 4. Glutathione and UDP-HexNAc as marker metabolites enriched in colon cancer metastasis.

(a, g) Light-microscopic photographs of intrasplenically injected HCT116 colon cancer cell xenografts in the liver of NOG mice. Scale bar: 500 µm. (b, h) Green fluorescence images of the same specimen shown in (a, g). (c–f) Representative imaging mass spectrometry showing spatial distribution of apparent UDP-HexNAc concentration ([UDP-HexNAc]_app), the reduced from of glutathione ([GSH]_app), oxidized glutathione ([GSSG]_app), and ([GSH]_app)/([GSSG]_app) ratio in the same microscopic field plotted as a heat map respectively. (i) A heat map of apparent energy charge calculated from apparent contents of adenylates of each spot. Adapted and modified by permission from Springer: Kubo et al. Anal. Bioanal. Chem. 400: 1895–1904, 2011.¹¹⁾

The cryosections with 5-µm thickness of snap-frozen tumor-bearing livers thaw-mounted on indium–tin oxide glass were imaged using 9-AA by MALDI-MS with 10-µm spatial resolution in negative ion mode; such a high lateral resolution of 10 µm is necessary to detect small-size tumors during an early phase of metastasis. The strategy permits comparisons of metabolic changes occurring between tumor and non-tumor regions at different time courses during the growth of tumor. The study provided evidence that UDP-N-acetyl hexosamine, reduced forms of glutathione (GSH), and the value of energy charge, i.e. ([ATP]+1/2 [ADP])/([ATP]+[ADP]+[AMP]), were significantly elevated in the tumor, suggesting significance for these enriched metabolites for cancer survival⁵⁵⁾/growth in vivo (Fig. 4).

Unlike the aforementioned approach which utilizes the combination of two different types of mass spectrometry, Goodwin et al.⁵⁶⁾ made an attempt to obtain quantitative data directly from the MALDI TOF spectra. These authors set up to compare the signal intensities from the tissue section collected after an intravenous injection of the target compound versus those obtained from quantitation-control spots of the compound which was applied to vehicle-control tissue sections. By analyzing both spots during the same experiment, these authors showed that it is feasible to assign the contents of exogenously applied compounds such as raclopride and SCH 223390, two widely used reference ligands for positron emission tomography.

Outlook

MALDI-IMS is a certainly a technical breakthrough to examine changes in multiple molecular behavior to control organ function at a most comprehensive manner in situ. It is, however, obvious many more challenges are awaiting this cutting-edge technology to be exploited at its full potential. These include improvement in sensitivity and resolution, which counteract each other, protocols for more effective sample preparation, and instrumental development for robust quantitative MALDI-IMS. To make it useful for intraoperative diagnosis/prognosis at the clinical setting, speed of analysis must be improved. Furthermore, as the technology provides a large amount of data, which is often not intuitive for us to comprehend, development for purpose-specific computational analyses will help us to reveal mechanisms of complex biological processes.