Mass Spectrometry Analysis Using Formalin-Fixed Paraffin-Embedded Pathological Samples

Abstract

Biomarker discovery using mass spectrometry (MS) plays a significant role in clinical medical research. However, proteomic analysis of formalin-fixed paraffin-embedded (FFPE) specimens using MS has been challenging because of the reduced solubility caused by fixation, leading to crosslinking reactions among amino acid side chains in proteins. This review presents the techniques employed for omics analysis of FFPE specimens to identify disease-specific biomarkers.

1. Introduction

1.1 Mass spectrometry (MS)

MS is a technique used to ionize molecules for measurement and determine their mass-to-charge ratios (m/z). Metabolites, such as nucleic acids, lipids, and small molecules, have been identified as biomarkers. However, there are challenges in treating them as biomarkers, including difficulties in interpreting the amount of change. By contrast, proteins directly reflect biomedical phenomena. For instance, increased levels of pro-inflammatory cytokines indicate local inflammation, whereas elevated levels of specific proteins, such as tumor markers, provide essential information about cancer progression. Changes in the protein levels of these pathologies are diagnostically significant and offer valuable insights into disease monitoring, follow-up, and relapse prevention.

Furthermore, the ongoing development of antibody drugs targeting specific proteins has significant implications for diagnosis and treatment selection. In recent years, advancements in analyzer accuracy and data analysis technology have enabled the comprehensive analysis of proteins (proteome analysis). Liquid chromatography-tandem mass spectrometry (LC-MS/MS) is widely used for protein identification based on spectra. In addition, mass spectrometry imaging (MSI) has recently been developed to identify in situ proteins in pathological lesions. In the field of clinical MS, it is necessary to understand the methods of protein extraction, ion generation, mass analysis algorithms, interpretation of mass spectra, pathological analysis, and clinical data evaluation (Figure 1).

Figure 1.

Multifaceted mass spectrometry (MS) and the interconnected steps contributing to clinical MS.

1.2 Formalin-fixed paraffin-embedded (FFPE) proteomics

Protein datasets contain large amounts of information. Therefore, it is necessary to efficiently use analytical tools to discriminate protein functions by ontology analysis and determine whether they are metabolic enzymes, structural proteins, chaperones or stress response, and apoptotic or proliferative proteins for data interpretation. Network analysis of proteins and cell biological functional data will provide insights into pathogenesis and disease development. In addition, a collection with pathological tissue findings is beginning. Pathological conditions, such as hemorrhagic necrosis, changes in cell distribution during inflammation and immune responses, cell invasion, tumor growth, and infection, lead to changes in the concentration and distribution of proteins in tissues. Protein distribution and location information in pathological samples are expected to be useful biomarkers by correlating them with histopathological findings associated with clinical findings. We have previously reported its use in colon cancer [1], acute myocardial infarction [2, 3, 4, 5], glioblastoma and metastatic lung cancer [6], systemic lupus erythematosus (SLE) and its associated diseases [7, 8, 9], and malignant mesothelioma [10]. FFPE-based proteomics is a promising method for identifying disease biomarkers.

2. Liquid chromatography-Tandem mass spectrometry (LC-MS/MS) of formalin-fixed paraffin-embedded (FFPE) specimens

2.1 Liquid chromatography-Tandem mass spectrometry (LC-MS/MS)

LC–MS/MS combines the physical separation capabilities of liquid chromatography (LC) with the two-stage mass analysis capabilities of MS. It is a powerful tool for analyzing proteins, allowing for the precise and sensitive detection of proteins and peptides in FFPE samples.

2.2 Sample preparation: Laser microdissection and proteomics

Laser microdissection (LMD) is used to precisely separate and extract specific cells from tissue sections, allowing for the accurate isolation of lesions. LMD is currently employed as a clinical test for cancer mutation analysis. However, this method has limitations in obtaining sufficient amounts and quality of proteins because proteins cannot be amplified unlike DNA. When working with specimens other than cancer cell tissues, extracting the minimum required amount while avoiding contamination from surrounding tissue components and the working environment becomes challenging. For instance, if a specific stain is used to identify lesions, it may interfere with subsequent analyses, requiring careful selection.

The quality of the samples and preprocessing significantly affected the reliability of MS omics data. Preprocessing involves extraction, purification, and digestion. In using the serum samples, identifying low-abundance proteins (especially potential disease biomarkers) is challenging in the presence of nonspecific, high-abundance proteins, such as albumin [2, 3, 4, 5]. Therefore, adequate purification using affinity columns is necessary in some cases.

2.3 Chemistry of fixation

Formaldehyde-fixed cells and tissues become resistant to degradation, maintaining their morphology during subsequent procedures, such as staining. Formaldehyde is a simple aldehyde that can potentially interact with biological molecules, especially proteins, through its highly reactive carbonyl group. This reactivity primarily results from the partial positive charge on the carbonyl carbon, which attracts electron-rich nucleophiles.

A detailed reaction mechanism between formaldehyde and amino acids is as follows:

Step 1: Nucleophilic addition. The primary amine group in lysine (−NH₂) acts as a nucleophile. It approaches the electrophilic carbon in the carbonyl group of formaldehyde. The nucleophilic attack forms a tetrahedral intermediate.

Step 2: Elimination of water. This intermediate then undergoes a rearrangement, leading to the elimination of a water molecule and forming an imine linkage as follows: Lys-NH-CH(OH)-H → Lys = NCH₂ + H₂O.

Step 3: A further reaction with another amino group. This imine is reactive and can undergo further reactions. A nucleophilic attack occurs when this imine comes into proximity with an amino group from another lysine. The nucleophilic NH₂ group of the second lysine attacks the imine carbon, leading to the formation of a methylene bridge (−CH₂−) between the two lysine residues as follows: Lys = NCH₂ + H₂N-Lys → Lys-N-CH₂-NH-Lys.

Recent MS studies have shown that although most of the FA crosslinks occur between lysine or arginine residues, a significant portion of the crosslinks also includes asparagine, histidine, aspartic acid, tyrosine, and glutamine residues [11, 12]. This methylene bridging reaction results in an increase in molecular weight of 12 Da. However, in MS analyses, there are instances wherein an increase of 24 Da is observed. This suggests that a complex state may be formed wherein dimerization occurs among imine groups; however, the details remain unclear [11].

This methylene bridging results in covalent crosslinks among protein molecules or within the same protein molecule, stabilizing the tertiary and quaternary structures of the protein. These crosslinks are the primary reason formaldehyde is an effective fixative, preserving the structural integrity of biological specimens. The methylene bridge reaction aggregates proteins within the sample into a large cohesive mass, creating a hydrophobic environment. Therefore, it inhibits proteolytic reactions that require water, thereby stabilizing the bulk protein mass. Crosslinking helps preserve cellular structures by “fixing” them, making it easier for pathologists to observe tissues under microscopes. However, this aggregation poses challenges in identifying individual protein components within the bulk, and fixed proteins can become protease-digestion resistant, such as trypsin-digestion resistant. The specific cleavage sites of trypsin, particularly those near lysine and arginine residues, may be rendered spatially inaccessible because of methylene crosslinking using formaldehyde. Therefore, preprocessing to enhance the digestion reaction is necessary.

2.4 Protein extraction and digestion

Trypsinization and fragmentation of this fixed protein mass necessitate a suitable hydrophilizing and swelling process. Repeated heating and cooling cycles help break the methylene or nonspecific bonds, diminishing the forces among structural proteins, such as collagen, and facilitating digestive reactions.

The most crucial process in FFPE protein extraction is the fragmentation of the bulk protein through sufficient physical crushing before digestion [1, 3, 11]. Figure 2 shows an example of the extraction protocol. Each microdissected sample was suspended and crashed using an ultrasonic homogenizer in 0.1 mol/L NH₄HCO₃ containing 30% (v/v) CH₃CN (Buffer A). The sample tubes were heated at 95°C for 90 min with shaking every 30 min. The samples were centrifuged at 10, 000 g for 1 min and cooled on ice. Trypsin and lysyl-endopeptidase were added for digestion, and samples were incubated at 37°C overnight for tissue swelling. Then, samples were heated at 95°C for a few minutes to deactivate trypsin. After drying, the samples were resuspended in trifluoroacetic acid containing 2% CH₃CN [1, 3]. Optionally, following trypsin digestion, peptides are often purified to remove nonpeptide components that could interfere with downstream analysis. This purification process involves various techniques, including desalting, solid-phase extraction, and spin column chromatography. However, the purification process may be omitted when the proteins extracted from the FFPE sample are present in trace amounts [13].

Figure 2.

Sample preparation scheme for liquid chromatography–mass spectrometry (LC–MS).

2.5 Liquid chromatography (LC) and ionization

The peptide mixture was loaded onto an LC column. The peptides interacted differently with the column material and were eluted at different times. The solvent (mobile phase) was pumped through a column containing a stationary phase. The components in the sample interact differently with the stationary phase, causing them to flow through the column at different rates to separate the mixture into individual components based on their chemical properties.

As the separated components exit the LC column, they are introduced into the mass spectrometer and ionized through electrospray ionization (ESI) to charge the molecules so that they can be manipulated by electric fields within the mass spectrometer.

2.6 First mass filter (MS1)

The ions were then passed through the MS1, which separated them based on their m/z values. The most common mass analyzers in LC-MS/MS are quadrupoles, ion traps, and time-of-flight (TOF) analyzers. In the quadrupole MS, four cylindrical metal rods were arranged in a vacuum chamber at equal intervals. By applying a DC voltage (V_d) and a high-frequency AC voltage (V_i cosωt; where ω is the high frequency), a rapidly changing electric field was generated in the quadrupole. The parameters U, V, and ω were adjusted so that ions within a specific range of m/z entered a stable oscillation state, passed through the quadrupole, and reached the detector. The ions of interest passed through the quadrupole field, whereas the others were expelled. An ion trap system based on the quadrupole principle was recently adopted.

The ions generated in the ionization section are accelerated by a high acceleration voltage (10–30 kV) and fly at a constant speed through a drift region without an electric field. As the ions reached a constant flight distance, ions with lower m/z values arrived at the detector earlier, whereas ions with higher m/z values arrived later. This time difference can be converted into a mass difference, allowing the generation of a mass spectrum.

2.7 Collision-induced dissociation (CID)

The ions selected by the first mass analyzer were fragmented into smaller ions through CID in the collision cell. This process involves accelerating the ions and colliding them with an inert gas. The resulting fragment ions varied depending on the cleavage of the peptide bonds among the amino acids constituting the protein. Fragment ions were named based on the position of the cleavage bond, with those containing the N-terminus referred to as a, b, and c ions. Those containing the C-terminus were referred to as x, y, and z ions (see the below chemical structure formula in Figure 3; R₁, R₂, and R₃ indicate alkyl groups). Because the spectrum contains a variety of fragment ions, it is crucial to accurately distinguish these ions and determine the amino acid sequence based on the mass differences among adjacent homoserine ions.

2.8 Second mass filter (MS2)

The resulting fragment ions were passed into a second mass analyzer to separate the ions based on their m/z values. The second stage of MS was used to analyze the fragments. The resulting spectra provided information on the amino acid sequence of the peptide. Each step of the mass filtering, coupled with an identification method involving two stages of mass filtering after LC, is referred to as LC-MS/MS.

2.9 Detection

Finally, the ions are detected, usually using an electron multiplier, and their abundance is measured. This information was used to create a mass spectrum for analysis. Peptide sequences were identified from the MS/MS spectral information and matched against an amino acid sequence database. Proteins were deduced based on the protein sequences to which the peptide sequences were assigned.

3. Data preprocessing

The false discovery rate (FDR) is often calculated and evaluated to assess the reliability of an analysis. In this method, reverse sequences (decoy sequences) with reversed amino acid sequences were intentionally included in the protein database for identification, allowing for the determination of the number of incorrectly identified decoy sequences. FDR is calculated as [(the number of identified decoy sequences × 2) ÷ (the total number of identified sequences)]. The desired FDR discovery rate was <1%.

In MS, “shared peptide” refers to peptides shared among different proteins. Different proteins may possess similar peptide sequences. Shared peptides are typically excluded from omics data to prevent false positive results. Furthermore, potential contaminants, such as tryptic autolysates, and worker-derived proteins, such as keratin, may be excluded.

Additionally, a correlation analysis between samples using protein profiles obtained from simultaneous extractions is necessary to confirm whether protein extraction was conducted correctly according to the protocol. This analysis is crucial because protein extraction involves multiple steps that can influence experimental conditions. If the correlation coefficient is 0.6 or lower, it indicates a questionable result, which might make proper comparisons among challenging samples.

Our previous studies [3, 11] suggest that the number of identifiable proteins in tissues is approximately 2.0 × 10³. Typically, seven to eight peptide fragments estimate the entire protein. However, to ensure statistical reliability, some proteins may only be identified from a single fragment and were excluded from the identified proteins. Moreover, because these proteins are initially present in low quantities when they emerge as biomarker candidates, they should not be ignored but rather reanalyzed or confirmed using other methods.

Many proteins identified using FFPE samples are metabolic enzymes, chaperones, heat shock proteins, ribosomal components, and immunoglobulins. Tissue-specific collagen and other proteins are also included. Because of the high abundance of these proteins, they should be excluded when considering biomarkers. Various measures are required to prevent contamination. The inclusion of known biomarkers is essential for validating proteomic analyses. In previous reports of myocardial infarction, biomarkers, such as troponin, were required. Confirmation of existing biomarkers is beneficial for validating proteomic analysis. The amount of the analyzed proteins that can be stably and simultaneously extracted include glyceraldehyde 3-phosphate dehydrogenase (GAPDH), commonly used as a control in western blotting. Alternatively, tissue-specific signature proteins could be used as baselines.

3.1 Data visualization

Statistical techniques are generally used to explore biomarkers when the fold change in abundance exceeds a reference value (e.g., 1.5 times) compared with the control group. In the case of clinical samples, wherein abundance does not follow a normal distribution, a volcano plot can be used with a p-value <0.05 from the nonparametric test on the vertical axis. If a plot corresponding to one protein is far from the others, it may be a candidate biomarker.

When separating the research subject group and the control group, it is possible to reduce dimensionality through principal component analysis (PCA). If the number of cases is small, this analysis can visually confirm how much the two groups are differentiated [11]. After using PCA for dimensionality reduction of the data, a common approach is to apply another method, such as logistic regression or support vector machines, for classification using the principal components. In doing so, the increasing or decreasing trend of each protein must be clear among groups. If a protein consistently shows a decrease in the subject group and an increase in the control group, and there is a significant difference between the two groups, then that protein could be considered a potential biomarker. This can be confirmed using methods, such as immunostaining, for the identified candidates using many cases, such as tissue microarray, which increases the probability that the observation is accurate.

Additionally, by performing network analysis on large-scale protein data, it is possible to conceive the formation of pathological conditions because of the interactions between these proteins. One method for analyzing protein networks is using a STRING program (https://string-db.org/), which has been recently utilized. When a robust network is formed, it can provide insights into disease pathology. For example, apoptotic inhibitor proteins may form interconnected networks in the context of the proteins involved in cancer growth. Similarly, networks can also arise among proteins involved in stress responses. Further analysis of these diverse networks will facilitate a deeper understanding of disease pathogenesis. Unlike in genomic analysis, the interactions among expressed proteins are straightforward, allowing for such interpretations.

4. Biomarker search

Biomarker proteins are expected to be released into the blood for testing, have sufficient sensitivity and specificity, and their dynamics must be understood. For example, creatine kinase and troponin are well-known markers of myocardial infarction; these biomarkers are degraded within 24–48 hours and rarely retained for long periods. Information on the localization of the candidate proteins is essential to identify a new biomarker. Examples of localization information include whether it is on the plasma membrane or exists in intracellular organelles, the cytoplasm, the nucleus, or the extracellular matrix. In addition, protein solubility should be determined by the amino acid composition. This includes whether the protein is hydrophilic or hydrophobic and the pH at which the overall charge of a solute or particle is zero (isoelectric point). Finally, the analysis and interpretation of the whole protein obtained, that is, proteomic data, are essential.

Additionally, the discovery of biomarkers related to the physical functions of proteins in pathological tissues, such as the heart, is expected to play a significant role in histopathological diagnosis. Combining proteomics data with immunostaining and phosphotungstic acid-haemotoxylin staining has made it possible to obtain diagnostic information with the expertise of pathologists in understanding pathology and capturing morphological changes. The use of FFPE is a powerful approach to the identification of protein biomarkers. However, the protocol for such an analysis has yet to be completely standardized, and although it is feasible, further research is needed. Nonetheless, techniques such as MSI have become possible and hold promise for future advancements in this field.

5. Application of mass spectrometry imaging (MSI) based on time-of-flight mass spectrometry (TOF-MS)

MSI involves sectioning, mounting on a slide glass, dehydrating, and direct matrix deposition on tissue surfaces to promote the ionizing of peptides in tissues, that is, matrix-assisted laser desorption/ionization (MALDI). When a laser irradiates the surface and measures the mass spectrum at each point of the tissue, ions are generated from the sample. Based on the obtained mass spectrum data, the spatial distribution of various molecules can be visualized as a two-dimensional (2D) image (Figure 3).

Figure 3.

Type of ions due to fragmentation by the collision.

5.1 Matrix chemistry

A matrix is a small molecule with aromatic rings (Figure 4). The reasons why structures with aromatic rings are favored as matrices include the following:

1. Crystallization: Matrices with aromatic rings tend to form homogeneous crystals with high reproducibility, enhancing the reproducibility and resolution in MS.

2. Strong ultraviolet (UV) absorption: Aromatic rings demonstrate strong UV absorption because of π-π* transitions. DHB (2,5-dihydroxybenzoic acid) primarily absorbs in the 337 nm UV range, corresponding to the emission of nitrogen lasers.

3. Rapid energy transfer: The matrix efficiently absorbs the laser energy and rapidly transfers this energy to the analyte molecules. During this process, the energy levels of the peptides increase, and the activation energy for ionization reactions decreases. The role of the matrix is to function as an energy bridge, aiding in the ionization of the peptides. Therefore, the matrix facilitates the protonation (in positive ion mode) or the deprotonation (in negative ion mode) of the peptide.

4. Sample protection: Matrices with aromatic rings protect the analyte molecules from direct laser irradiation. This helps prevent the thermal decomposition of biological molecules, such as proteins and peptides.

5. Ionization assistance: Some matrices are known to have their conjugated aromatic systems involved in the ionization mechanism. For instance, they can assist in ionization via proton transfer (Figures 5 and 6).

Figure 4.

Scheme of matrix-assisted laser desorption ionization time-of-flight (MALDI-TOF) mass spectrometry imaging (MSI). Following the deposition of the matrix (DHB: 2,5-dihydroxybenzoic acid), a laser was used to ionize the peptides, and an electric field was applied until they reached the detector. The ionized peptides were separated based on mass-to-charge ratio (m/z) values, and the spectrum revealed a two-dimensional (2D) distribution of the peptides within the tissues. For each laser shot, a full mass spectrum is recorded. Thus, there is a corresponding mass spectrum for each pixel position on the tissue. The laser in the instrument systematically raster scans the entire tissue section. At each raster position (pixel), the laser desorbs and ionizes the molecules, which are then detected by the time-of-flight (TOF) mass analyzer. The photos in Figure 4 depict the MSI of the neuron-related nestin protein in glioblastoma tissue [6]. The number above each image represents the m/z values corresponding to the protein. The outline demarcates the margin of the tissue, and the color heatmap indicates the signal intensity of the MSI. In these images, the upper portion of the tissue shows a higher intensity, corresponding to a greater abundance in areas with a higher tumor cell density.

Figure 5.

Representative matrices. The matrix effectively absorbs laser light of a specific wavelength. Most matrices have a chromophore, such as in aromatic rings, which allows the matrix to be electronically excited because of this chemical property.

Figure 6.

Matrix-assisted laser desorption/ionization (MALDI). This ionization method utilizes a matrix that absorbs laser energy to generate peptide ions. In MALDI, the matrix and proteins present in the tissue are cocrystallized. Subsequently, during the ionization process, when the matrix is ionized and sublimated, energy is transferred from the matrix to the coexisting peptides, resulting in their ionization.

5.2 Preprocessing of tissue sample

A method was developed for preprocessing tissue samples. This method involves heating the sample in acetone in an acetonitrile aqueous solution (Buffer A, containing acetonitrile and NH₄HCO₃; Figure 7) and immersing it at 37°C for an extended period to induce swelling and allow the molecules within the solid tissue to move to the surface. This process facilitates efficient interactions between the matrix and laser desorption/ionization, resulting in enhanced sensitivity for MSI. Acquiring such data requires knowledge and techniques of solid-phase surface chemistry rather than those of liquid-phase solutes. Evaporators of the matrix help generate homogeneous “mixed crystals” containing the matrix and peptides on tissue surfaces.

Figure 7.

A PAP pen was used to encircle the tissue on the slide glass. After applying resin around the tissue, the slide was immersed in buffer a. a cover glass was placed on top and secured with a bond. The slide glass was placed on an aluminum plate and heated briefly with enough intensity to produce bubbles, ensuring buffer a to not evaporate from the bottom. The steric structure of proteins becomes loose and fluctuates with heat fluctuations [1].

5.3 Matrix application

The matrix is applied to the tissue in order to assist in desorption/ionization. The “co-crystal” of peptides and matrix is a physical state wherein the crystal lattice of the matrix surrounds the proteins and peptides rather than a result of chemical interactions. However, the specifics of this interaction are still not fully understood. The shape and size of the crystal vary depending on factors such as the evaporation rate of the matrix solvent, the type and concentration of the matrix used, the concentration of the peptide, and the condition of the sample surface (which can be assessed by the wetting angle of the matrix solution, i.e., a large angle indicates that the solvent is repelled by the surface, suggesting low wettability or hydrophobicity of the surface). If the sample contains connective fiber components, it may repel the matrix solution, further influencing the outcome. This highlights the importance of the preprocessing step for higher hydrophilicity.

Furthermore, the co-crystals must be more uniform. Techniques such as using iMLayer (by Shimadzu Corporation, Kyoto, Japan) can achieve this. This is where the matrix is heated in a vacuum to sublimate and deposited using an automatic deposition device.

5.4 Sample ionization

The peptide in the sample is then irradiated by a laser, leading to desorption and ionization of molecules. An electric field accelerates the ionized molecules into a flight tube. Larger ions take longer to travel through the tube than smaller ions. The time difference between ions is measured and used to determine the m/z values of the ions. In a MALDI device, the sample stage can be designed to move relative to the terminal, ensuring as uniform laser ionization as possible on the sample wherein mixed crystals have formed on the surface. Electrons (e−), positive matrix ions (M+), hydrogen atoms (H), and peptide radical species (P) are generated during the initial stages of MALDI when using UV lasers (e.g., nitrogen lasers). The wavelength of the nitrogen laser (337 nm) corresponds to an energy of 3.7 eV per photon. The ionization energy of a single molecule of DHB is 8.0 eV, requiring the absorption of at least three photons.

However, when DHB forms crystals, stacking or clustering occurs, allowing the absorption of just two photons to cover the activation energy. Following the energy absorption, DHB is ionized on the crystal surface. Increasing the irradiation amount of the laser causes more vaporization of the DHB ions with cluster forms. Proton transfer reactions from the ionized DHB to non-ionized DHB occur within this cluster. It is crucial to note that the proton is provided not from the carboxyl group of the DHB side chain but from the hydroxyl group at the 5′-position (Figure 7) [14].

The sublimation of the DHB from the solid phase to the gas phase by photon energy and the ionization of the peptide by proton transfer from the DHB in the gas phase in DHB and peptide interactions are represented in Figure 8. There is a debate as to whether the peptide is already ionized in the solid phase mixed crystal. As aforementioned, it seems energetically favorable for ionization to occur in the solid phase when DHB clusters accompany the peptide. However, for the ionized peptide to begin flying to the detector, it must not react with surrounding anions and simultaneously be released from the sublimated DHB clusters, requiring a multistep process. Therefore, ionization is expected to occur at a very low probability; the ion yield in MALDI is estimated to be approximately 10⁻⁹ to 10⁻⁸ in terms of the ion/neutral ratio [15].

Figure 8.

Matrix ionization. The proton transfer is highlighted in red.

Another model does not assume pre-ionization of the peptide and accepts a proton from the ionized DHB in the gas phase to ionize. Both mechanisms might occur simultaneously. There is a debate on which mechanism is more dominant [16].

5.5 Image creation

The mass spectra from each discrete location are combined to create an image showing the distribution of specific molecules. Functionally or in pathological conditions, related proteins are expected to have spatial distributions that correlate.

5.6 Resolution

Current MSI techniques have inherent resolution limitations that restrict detailed analysis, particularly at the microtissue or cellular level. Resolutions on the order of 10 microns have been achieved thus far, which is approximately the size of large cells, enabling signal acquisition on a cell-by-cell basis. Ten microns corresponds to the diameter of large epithelial cells and monocytes, and the resolution is sufficient for histological evaluation at the tissue level, for example, cancer invasion range.

5.7 Region of interest (ROI) analysis in mass spectrometry imaging (MSI)

In pathology, immunohistochemistry is the only method to detect protein localization within FFPE pathological specimens. This method confirms known proteins but does not provide data on unknown proteins. In contrast, MSI possesses the potential to comprehensively analyze proteins within tissues, paving its way as an indispensable tool for future pathological research. Furthermore, advancements in artificial intelligence (AI) have enabled the analysis of segmentation impressions on histopathological specimens stained with hematoxylin and eosin, yielding findings consistent with lesions. In pathology, by defining an ROI on images, areas, such as the entire tumor, its background, regions with specific histological characteristics, or around blood vessels, can be predetermined [6]. Direct comparative correlation analysis of signal intensity between the ROI and other regions using MSI has become feasible. This approach has realized the identification of peptide proteins specific to lesions. Nowadays, acquiring images using virtual slides is commonplace, and various image analysis software tools, such as HALO-AI (https://indicalab.com/halo-ai/), come equipped with histological analysis functionalities for lesion segmentation and correlation analysis of 2D data within image plots.

6. Hierarchical cluster analysis in MSI

MSI provides detailed image data that reflect the abundance of each m/z at specific coordinates on tissue samples. Essentially, this data can be represented as a mathematical matrix, enabling the computation of signal distribution distances using matrix operations.

One powerful analytical technique applied to this data is hierarchical cluster analysis. This method groups observations within a dataset into a tree-like structure called a dendrogram (Figure 9). Observations or proteins that showcase similar 2D distribution patterns on this dendrogram are positioned close to each other. This closeness suggests potential functional and pathological relationships between these proteins.

Figure 9.

Peptide ionization by the transfer of a proton ion from ionized 2,5-dihydroxybenzoic acid (DHB).

Among the techniques for hierarchical cluster analysis, Ward’s method stands out. It initially views the signal intensity value of each pixel in an image for a single m/z as a distinct cluster. As the clustering progresses, it combines clusters to minimize the increase in variance stemming from this integration. The primary objective is to ensure strong cohesion within each cluster while maintaining clear distinctions between different clusters.

When hierarchical cluster analysis is performed on the 2D plot distribution pattern from MSI data, it is possible to identify proteins with similar spatial distributions. A noteworthy application of this was showcased by Hiratsuka et al., who found a correlation between the signal distribution patterns of glial fibrillary acidic protein (GFAP) and those of tubulin/histones in glioma [6]. GFAP, an intermediate filament protein, is predominantly found in astrocytes, a specific type of glial cell in the central nervous system. It is used in neuropathology as a diagnostic marker to identify astrocytic cells and differentiate astrocytic tumors from other brain tumor types. The identified correlation between GFAP and tubulin/histones points to the possible roles these proteins play in brain tumor pathology and development.

Furthermore, glioblastomas exhibit frequent hemorrhaging. Research suggests that the spatial distribution of hemoglobin subunits is closely associated with this characteristic. This observation underscores the value of MSI data, providing invaluable insights into the pathological conditions and characteristics of various diseases, such as glioblastomas (Figure 10) [6].

Figure 10.

Dendrogram illustrating hierarchical clustering based on matrix-assisted laser desorption ionization time-of-flight (MALDI-TOF) mass spectrometry imaging patterns, facilitating histopathological correlation analysis among proteins. The height at which two branches merge in the dendrogram represents the distance between two data points or clusters. The similarity in the distribution pattern between A and B is greater than that between C and D, indicating that A and B cooperate more closely in function than C and D.

7. Conclusions and future perspective

MS studies of FFPE pathological specimens are expected to be powerful tools for analyzing pathophysiology. Correlative analysis of proteins using general LC-MS and protein abundance using MSI and histological findings can provide valuable insights.

MSI of FFPE specimens can be combined with spatial biology methods, such as analyzing the spatial distribution of mRNA in gene expression, enabling more detailed pathological analysis. This approach integrates multiple data sources to build a multimodal database that facilitates statistical analysis. Therefore, MS is expected to become an increasingly powerful tool for identifying biomarkers.

Acknowledgments

We thank Takushi Yamamoto and Hideshi Fujiwake of the Shimadzu Corporation for their advice.

Conflict of interest

The authors declare no conflict of interest.

Acronyms and abbreviations

DHB	2,5-dihydroxybenzoic acid
ESI	Electrospray ionization
FFPE	Formalin-fixed paraffin-embedded
LC	Liquid chromatography
LC-MS/MS	Liquid chromatography-tandem mass spectrometry
MALDI-TOF	Matrix-assisted laser desorption ionization time-of-flight
MS	Mass spectrometry
MSI	Mass spectrometry imaging