Mass spectrometry imaging (MSI) is not only the subject of extensive studies in academics, but also of wide interest in the pharmaceutical industry [1, 2]. MSI can directly acquire the molecular profile and visualize the spatial distribution of each ionized compound across a sample, particularly biological tissues. As one of the most promising and developing techniques in exploring spatial multi-omics, MSI is able to locate the drug and metabolite distribution but also provides insight into the phenotype changes underlying disease progression and drug intervention. In this article basic MSI workflow and technical advances will be introduced. Thereafter, the application of MSI in pharmaceutical research and future directions will be further elaborated and prospectively discussed.
2. GENERAL WORKFLOW OF MSI
The general MSI workflow for a preclinical pharmaceutical study is shown in Figure 1 . Using a whole-body section from a dosed rat as an example, the test rat will be euthanized at the setup time point after drug administration. The body will be immediately harvested, frozen, and fixed into embedding gel for the follow-up cryo-sectioning process. After sufficient dehydration, the cryo-sectioned whole-body sample will be mounted on a 2D moving stage and hit by an in situ ionization probe either in spot-by-spot scanning with a pulsed laser/ion beam or in line-by-line scanning with continuous electrospray/plasma. The exogenous drug along with endogenous components located in each impact region can be instantly desorbed, ionized, and transported into an MS system for MSI data acquisition. Thereafter, the MSI software will reconstruct any molecular image according to the ion intensity from each physical location, given a defined mass bin centered at the exact m/z of the target ion with a specified mass tolerance.
3. TECHNICAL ADVANTAGES OF MSI
The use of mass spectrometry for signal readout outperformed whole-body autoradiography (WBA) in drug distribution studies because MSI easily differentiates the parent drug and its metabolic product due to the m/z difference [3, 4]. In addition, MSI has thousands of available m/z channels for molecular profile recording compared to fluorescence or infrared-based spectroscopic imaging. Moreover, fluorescence or infrared-based spectroscopic imaging usually have very limited wavelength choices for multiplex display because of the bandwidth overlap. Liquid chromatography-mass spectrometry (LC-MS) continues to be the gold standard method to investigate the pharmacokinetic properties of a drug, including absorption, distribution, metabolism, and excretion (ADME); however, LC-MS is time- and labor-intensive with respect to sample pre-processing, including extraction, purification, and enrichment. This bulk measurement only reflects average levels of the drug and its metabolites in each organ or tissue region. More detailed spatial information associated with the efficacy or toxicity will be inevitably sacrificed because of the vigorous tissue disruption process. In contrast, MSI achieves direct desorption and ionization of analytes from the biological tissue sample surface. More precise drug localization information can be preserved at the sub-organ, tissue microcompartment, or cellular level depending on the physical size of the impact region by an ionization probe.
4. AVAILABLE CHOICES FOR IN SITU IONIZATION
The development of in situ ionization, as the core of MSI, is always the central task to drive the progress of MSI advances both in sensitivity and spatial resolution. Nevertheless, in most situations, sensitivity and spatial resolution are a pair of trade-off metrics when choosing a suitable in situ ionization method. Higher spatial resolution means less impact area and desorbed sample amount for feeding the follow-up detection. In turn, an increase in the impact area may help to raise the detection sensitivity, but at the cost of poor spatial resolution. Extensive efforts are still urgently needed to pursue an in-depth coverage of chemical species and precise observation of a drug’s process within intricate physiologic structures or organelles at a few microns and even nanometer scale.
To date, various in situ ionization methods have been extensively developed; representative in situ ionization methods are shown in Table 1 . According to the types of desorption and ionization source, major impact probes include laser (e.g., matrix-assisted laser desorption/ionization [MALDI]), ion beam (e.g., secondary ion mass spectrometer [SIMS]), plasma (e.g., dielectric barrier discharge ionization [DBDI]), and charged microdroplet spray (e.g., desorption electrospray ionization [DESI]), The former two typically work in a closed, vacuum environment. The desorbed and ionized species have less of a chance to collide with the air and introduce the external interfering components during the travel from the sample surface into the MS inlet. Intricate instrumentation design and relatively higher cost are needed for laser/ion beam focusing, transmission, and vacuum maintenance. Development of the DESI was an evolutionary breakthrough and opened a new era of ambient ionization . Ambient ionization-based MSI liberates the sample analysis from the vacuum condition, avoids the use of a matrix, and greatly lowers the technical barrier and analytical cost for an MSI experiment . Versatile ambient ionization methods are under explosive development in the past decade .
|Ionization||Matrix||Vacuum||Typical resolution||Chemical species||Ref.|
|MALDI (Matrix-assisted laser desorption ionization)||Y||Y||25-50 μm||Proteins, organic drugs|||
extreme: 200 nm
|Proteins, small organic drugs|||
|AP-MALDI (Atmospheric pressure-MALDI)||Y||N||10-30 μm|
extreme: 1.4 μm
|Proteins, small organic drugs|||
|LA-ICP (Laser ablation-inductively coupled plasma)||N||Y||5-200 μm||Metal-labels on antibody or organometallic drug; <200 Da|||
|LAESI (Laser ablation electrospray ionization)||N||N||150-400 μm|
extreme: 5 μm
|Small organic drugs||[137, 138]|
|IR-MALDESI (Infrared matrix-assisted laser desorption electrospray ionization)||Y||Y||45-200 μm||Proteins, small organic drugs||[139, 140]|
|DESI (desorption electrospray ionization)||N||N||100-200 μm||Small organic drugs||[5, 6, 141]|
|AFA-DESI (air flow-assisted DESI)||N||N||100-200 μm|
extreme: 50 μm
|Small organic drugs|||
|Nano-DESI||N||N||10-25 μm||Proteins, small organic drugs|||
|LESA (Liquid extraction surface analysis)||N||N||>200 μm||Proteins, small organic drugs||[143, 144]|
|SIMS (secondary ionization mass spectrometry)||N||Y||50 nm||Metal-based organic drugs|
|nanoSIMS||N||Y||<50 nm||Metal-based drugs|||
Each ionization method has unique advantages and limitations in suitable chemical species, spatial resolution, matrix use, or ambient/vacuum operation. Selecting a suitable ionization source should systematically consider the physiochemical properties of the drug, the specific physical size of the sample, and the physiologic structure of interest. Generally, MALDI, DESI, and SIMS, as three major ionization sources in MSI, are top choices for imaging biotherapeutics, low molecular weight organic synthesis drugs, and metal-based drugs, respectively. DESI has a typical lateral resolution of 50-100 μm, whereas MALDI has a typical resolution of 20-50 μm. Therefore, DESI and MALDI can be used for conventional analyses of whole-body animals, organs, or tissues at the mm scale. Recent reports showed that DESI resolution had substantial improvement and robust performance by modifying the co-axial spraying capillary setup or introducing a ring-electrode for focusing highly-dispersed spraying [8, 9]. Introducing a well-designed optical pathway and a focusing system could also sharpen the impact laser point to a few microns or even a sub-micron scale, thus enabling laser-based MSI (e.g., MALDI and laser ablation-inductively coupled plasma-mass spectrometry [LA-ICP-MS]) collect the cellular-resolved images [10, 11]. SIMS has the highest spatial sub-micron resolution (approximately 300 nm) compared to the above-mentioned sources. This advantage makes SIM the first choice for observing cellular or sub-cellular biological process for in vitro drug research [12, 13]. Considering that this review emphasizes pharmaceutical applications, more details in ionization principles, instrumentation, and underlying mechanisms will not be exhaustively elaborated, which can be referenced to the corresponding literature according to the reader’s special interests [14–16].
5. MSI SENSITIVITY
Sensitivity is another challenging issue facing MSI practical use in pharmaceutical analysis. For technical reasons, the MSI advantage in free-of-sample pretreatment is built on the in situ ionization of drug molecules from the biological sample. Thus, the complex composition of a biomatrix will inevitably compete with the drug molecule and suppress ionization efficiency, which is known as the matrix effect . From a pharmacology perspective, an effective drug candidate with a higher potency means a lower dose to reach the minimum therapeutic level, which may also cause false-negative detection by MSI. From the pharmacokinetic perspective, first-pass elimination lowers the effective concentration of a drug administered through the non-intravenous route, particularly the oral pathway. Given these reasons, drug distribution results acquired using the MSI technique need to be carefully validated via conventional liquid chromatography-tandem mass spectrometry (LC-MS/MS)-based drug measurement as the complementary reference to avoid false-negative results, especially when the single dose is less than the mg/kg level.
Enormous effort has been expended in instrumentation and the MSI methodology to overcome limited sensitivity. Photon-based post-ionization (UVPD) or secondary ionization (MALDI-2) techniques were developed to fully ionize non-polar, neutral molecules [11, 18, 19]. For ambient ionization, the DESI prototype configuration has undergone innovative improvement (air-flow-assisted DESI [AFA-DESI]). A local vacuum chamber is coupled with the MS inlet and connected with the transport tube from the sampling end in this set-up. The gas pressure gradient focuses and directs random airflow to pick up additional molecules splashed from the tissue into the MS system for detection [20, 21].
New functional matrices have been developed for MALDI to avoid severe interference at the lower mass range and facilitate laser desorption ionization of chemical species . Various nanomaterials are fabricated as the sample loading/imprinting substrate or energy absorption/transfer medium to enhance the target detection . On-tissue chemical derivatization gains increasing attention for those species which have low abundance, less stability, weak polarity, or poor ionization efficiency, such as steroids, thiols aldehydes, and alkenes [24, 25]. Derivatization reagent can be introduced through matrix spraying, vaporization, or photon irradiation in a closed chamber [26–28]. Solvent immersion can be a cleanup process to remove those unwanted species and enhance the target analyte’s sensitivity [29, 30]. Alternatively, the hydrogel has also been proven to be a biocompatible and tissue-friendly medium to conduct in situ desalting, digestion, cleavage, and derivatization with no severe tissue disruption .
6. MSI DATA ACQUISITION
The peak capacity of an MSI method relies heavily on the performance of a mass analyzer. Currently, to fulfill the task of collecting increasing data volume in spatial omics research, a high-resolution mass spectrometer (HR-MS) has become the mainstream device with an orbitrap, quadrupole time-of-flight (QTOF), and Fourier-transform ion cyclotron resonance (FTICR) as three representative types. A full scan is the default mode to acquire as much abundant phenotype information as possible to present untargeted metabolic or proteomic profiles. HR-MS also provides the exact m/z value for a more convincing annotation of the ion identity. A more authentic spatial information specific to a relatively “pure” ion can be achieved by filtering the interfering ions that m/z are extremely close to the target .
The triple quadrupole and linear ion trap are low-resolution tandem mass analyzers that are mainly used for quantitative and qualitative analyses of drug concentration and metabolism in an LC-MS/MS system. The triple quadrupole and linear ion trap can also be used to conduct MSI studies in special situations. Selected reaction monitoring (SRM) has become the alternative choice for targeting drugs and the active/toxic metabolic products if the drug exists in a low concentration and is too weak to be detected by the full scan or the target ion has severe isobaric ion interference [32, 33]. An MS(n) scan performed by a linear ion trap is mainly used for identifying those region-specific marker ions.
Additionally, ion mobility mass spectrometry (IMS) has gained popularity among MSI users in recent years because of unique technical features in isomer separation . These isomers sharing the identical m/z can be differentiated to achieve a separate spatial distribution. Interfering components will also be filtered out of the expected traveling time window to increase the target ion intensity. Coupling in the front end of an MS system creates a new dimension for molecular profiling and broadens the scope of observable species by MSI .
7. MSI DATA ANALYSIS
An MSI experiment generates gigabytes of big data that contains thousands of ion information for follow-up data processing. The collected MSI data relies heavily on the software or a self-written code to be translated into spatially-resolved molecular information. There is more than one choice of available MSI software developed for reader reference and practice ( Table 2 ). Basic MSI data pre-processing includes mass peak picking, alignment, and normalization. The basic function of MSI software includes ion image construction, region of interest (ROI) delineation, and average mass spectrum generation. Commercial and open-access software is a user-friendly choice to researcher convenience .
|Software||Availability||Ionization dependence||Data format||Featured functions apart from basic visualization**||Ref.|
|M2aia||Open-source||NS*||imzML||Prep; MIO; Seg; manual alignment feature extraction; dimension reduction|||
|MSiReader||Open source||NS*||imzML||QMSI; MIO; ROI, Mass Measurement Accuracy Heatmap and Histograms; Multiple sample loading||[148, 149]|
|BioMap||Open-source||MALDI||Analyze 7.5, TIF, DICOM, PNP||MIO; 3D; ROI|||
|omniSpect||Open-source||DESI, MALDI||Analyze 7.5, mzXML, cdf||MVA|||
|Mirion||Open-source||NS*||imzML||Automatic processing; Automatic image generation; MIO|||
|msIQuant||Open-source||MALDI||QMSI; BDP; MIO|||
|Cardinal||Open-source||DESI, MALDI||imzML||Seg; Stata|||
|Pew2||Open source||LA-ICP||Agilent Batch;|
|MIO; ROI; QMSI|||
|DataCube Explorer||Open source||NS*||imzML||Prep; 3D; ROI|||
|OpenMSI||Open source||NS*||xml||Web-based API for remote access and processing,|
URL-based data analysis sharing; BDP
|MITICS||Open-source||MALDI||xml||MSI; AVG; ROI|||
|rMSIannotation||Open source||NS*||imzML||Peak annotation; isotopic peak removal; adduct ion recognition|||
|SpectralAnalysis||Open source||DESI, MALDI||imzML||Prep; MVA|||
|LipostarMSI||Open source||NS*||imzML||Prep; ROI; Stata; MVA; MIO; automated identification|||
|MassExplorer||Open source||DESI||csv||Prep; Stata; ML; FS|||
|Massimager||Commercial||NS*||cdf, mzXML||Prep; ROI; AVG; MIO; FS; MVA; I/E||[160, 161]|
|FlexImaging||Commercial||MALDI||imzML||Prep; ROI; AVG; MVA|||
|SCiLS Lab||Commercial||MALDI||imzML||Prep; ROI; AVG, Seg; MIO; 3D; I/E; Extensive API|||
|TissueView||Commercial||MALDI||imzML||AVG; ROI; MIO||—|
|HDImaging||Commercial||DESI, MALDI||imzML||MIO; Streamline processing||—|
*NS: not specified.
**3D:3D imaging; QMSI: quantitative MSI; AVG: average mass spectrum generation; BDP: big data processing; ROI: region of interest selection; Pre: spectral pre-processing steps including peak picking, baseline smoothing, data transform; MIO: Multimodular images overlay; Seg: spatial segmentation; Stat: statistical description and feature extraction; ML: machine learning; I/E: import and export; FS: feature selection; API: application programming interface.
Apart from the graphic user interface (GUI)-based software, self-programing is another choice for developers who are engaged in the new MSI function and computation method . The new MSI data computation method can be promptly translated into real practice when implemented in environments, such as RStudio, Python, MATLAB, and Delphi. Multivariate analytic methods, machine learning, and deep learning models are introduced into the MSI data processing tasks, such as automatic spatial segmentation [38, 39], discriminative ion picking and latent feature extraction [40–42], region-specific matrix effect normalization , 3D image construction [44, 45], high-quality image recovery by over- or sparse-sampling [46, 47], co-localization (co-registration) among multi-modal images [48, 49], and spatial multi-omics imaging data integration [50, 51].
8. QUANTITATIVE MSI
MSI presents the relative abundance of a drug according to the normalized ion intensity across the tissue section. Unfortunately, tissue structural and compositional heterogeneity causes severe variance in the ionization efficiency of the drug. This region-specific matrix effect may distort the linear response of a drug ion intensity with the local concentration [43, 52, 53], thus biasing the comparison of drug distribution across different regions. The quantitative MSI method (QMSI) has been well-developed to overcome this issue and obtain an accurate spatially-resolved content difference of the drug. The tissue-specific ion suppression can be well-compensated by deuterated internal standards, chemometric calibration , or external evaluation of the tissue signal extinction coefficient . A dilution series of drug and internal standards can be incorporated into a blank tissue to serve as the simulative dose one by several strategies, including the on dry tissue method (the most frequently used) , under wet tissue method , in-tissue homogenate , and in biomimetic 3D tissue . Another alternative strategy is to introduce laser-capture microdissection (LCM) to harvest numerous tiny tissue regions of interest for follow-up LC-MS/MS measurement [60, 61]. The drug content is reported as the molar amount per area (i.e., pmol/mm2) or the weight percentage (i.e., μg/g). Although each strategy cannot guarantee completely identical ionization behavior with a native drug, the LC-MS/MS cross-validation result reveals a system error that is usually within an acceptable range (±15%) for drug quantitation in the drug metabolism and pharmacokinetics (DMPK) study.
9. SPATIAL CO-LOCALIZATION
Precise localization of drug molecules in a physiologic structure is a prerequisite to learn about the distribution in targeted and non-targeted regions. The drug ion image cannot provide an accurate map of intact tissue outlines and structural microregions. A guiding map of the region can be accessed from three strategies: (1) region-specific endogenous metabolite as the marker, such as heme, which can delineate the blood vessels [62, 63]; (2) MSI data-driven spatial segmentation, which can be performed with the aid of a machine learning or deep learning method to split all biological image pixels into several groups of micro-regions according to the molecular profile patterns; frequently used methods include principal component analysis (PCA), partial least square discriminant analysis (PLS-DA), t-stochastic neighbor embedding (t-SNE), KMeans clustering, uniform manifold approximation and projection (UMAP), and convolutional neural network ; (3) a complementary image that can be acquired from an optical microscopy image of a biological sample hematoxylin and eosin stain (the most frequently used) [65, 66], bright-field polarimetry [67, 68], immunohistochemistry (IHC) , molecular imaging, such as fluorescence microscopy , and Fourier transfer infrared microscopy , nuclear medicinal imaging technique, such as magnetic resonance imaging (MRI) , and positron emission tomography/computer tomography (PET/CT) .
10. DRUG METABOLISM AND PHARMACOKINETICS
In preclinical drug research and development (R&D), the assessment of the drug candidate’s distribution is a central concern for a better understanding of its efficacy and possible toxicity at the early stage of in vivo investigation. DMPK testing became the first in vivo study to decide whether a drug candidate deserves further investigation or suspension. Although drug concentrations in plasma/serum are indicative of DMPK properties, it has been argued whether the plasma/serum concentrations alone can accurately reflect the localized drug concentration in tissue. Drug enrichment located in the target region may indicate the potential efficacy, whereas the unexpected accumulation in non-target tissue probably causes adverse effects or even toxicity. For conventional drug distribution measurement by LC-MS, bloodletting is a necessary step to diminish the influence of vascularization and the amount of blood infused. The unique advantage of MSI in spatially-resolving ability cannot only differentiate the drug and metabolites, but also the blood vessel and non-vessel regions, making MSI well-suited for the demands of DMPK studies, although MSI still depends on how well an MSI system can achieve spatial resolution. Currently, an MSI marker, such as heme (m/z 616), has been introduced to localize the blood vessel distribution or differentiate the blood vessel distribution from non-blood vessel regions within an organ . Other imaging methods, such as immunohistochemical vessel staining, can also help to co-locate blood vessels and rule out their influence on blood vessel distribution .
Currently, MSI has been extensively applied to visualize the spatial distribution of a drug in animal models ( Table 3 ). The distribution of a drug and its metabolites within intricate sub-organ structures or micro-compartments can be well-discerned. The MSI-observable object has a huge expansion that varies from mm to sub-microns in physical size, including in vivo samples, such as whole-bodies of rats or zebrafish, organs, tumors, tissues, and ex vivo samples, such as organoids and cultured cells ( Figure 2 ). The Dome and Marko-Varga groups collaboratively characterized the distribution and metabolism of sunitinib in solid tumors, as well as kidney and liver by MALDI-MSI. Drug content was quantified, and the three metabolites were found to have close distribution patterns with the administrated parent drug ( Figure 3a-b ). Tumor growth and intra-tumor VEGF receptor-2 expression were significantly inhibited by sunitinib-treatment. There were intra-tumor areas where the signal intensity of sunitinib correlated with expression of VEGF receptor-2 . Prideaux et al. used MALDI-MSI to investigate the targeting efficiency of the anti-tuberculosis (TB) drug, moxifloxacin, in TB-infected rabbit lungs. The drug was shown to accumulate granulomatous lesions at levels higher than the surrounding lung tissue from 1.5 h post-administration, which was quantitatively validated by LC-MS/MS ( Figure 3c ) .
The Zeper and He groups presented the use of air flow-assisted desorption electrospray ionization (AFA-DESI) in accurately quantifying the distribution of paclitaxel (PTX) and its prodrug (PTX-R) in whole-body sections of xenografted mice based on a previously developed virtual calibration quantitative MSI method (VC-QMSI) [43, 77]. Based on the machine learning-based spatial segmentation and signal calibration, the drug distribution in each organ and tumor micro-region was accurately discerned and quantified. AFA-DESI-MSI indicated that PTX is widely distributed in multiple organs throughout the dosed body in the PTX group, both in the form of through direct injection and liposome. In contrast, the PTX prodrug was enriched in the tumor tissue and catabolized to release more active PTX for anti-tumor action, particularly in the poorly differentiated intratumor and necrotic areas ( Figure 4a ). The relative targeting efficiency of PTX-R was increased approximately 50-fold compared to the plain PTX injection . The Römpp group developed a new MALDI-MSI method with excellent performance in mass (240k @ m/z 200 full width at half maximum [FWHM]) and spatial resolution (10 μm pixel size). Thereafter, four first-line anti-tuberculosis drugs (pyrazinamide, rifampicin, ethambutol, and isoniazid) and two second-line drugs (moxifloxacin and clofazimine) in murine lungs were successfully visualized in the expected action site of necrotizing granulomas. For the first time, clofazimine was imaged and revealed to be accumulated in lipid deposits around airways ( Figure 4b ) .
Apart from preclinical pharmaceutical studies, MSI has also been successfully translated into clinical drug trials and therapeutic drug monitoring for a better prognostic evaluation. Eberlin et al. used DESI-MSI to visualize the transverse distribution of several sodium modulators on the harvested human skin. The transdermal permeation process can be spatially resolved from the drug amount across the epidermis, dermis, and hypodermis ( Figure 4c ). Then, the skin structure-specific lipid compositions were characterized and further used to evaluate the compound protection and healing effects on the skin impairment after long-term ultraviolet exposure . The Lobinski group combined LA-ICP-MSI and MALDI-MSI to evaluate the penetration of cisplatin and oxaliplatin in tumors resected from patients with colorectal or ovarian peritoneal carcinomatosis. The behavioral differences between oxaliplatin and cisplatin were confirmed by LA-ICP MSI. Oxaliplatin was enriched at the tumor periphery, whereas cisplatin penetrated deeper into the tumor core region . Menetrey et al. used MALDI-QMSI to quantitatively evaluate the penetration and concentration of an apoptosis protein antagonist, xevinapant, in tumors resected from squamous cell carcinoma of head and neck (SCCHN) patients in a preoperative window-of-opportunity (WoO). The overall tumor-to-plasma ratio of 25 was determined, indicating a tumor penetration for xevinapant in the treatment of SCCHN patients ( Figure 4d ) .
11. IN VIVO DRUG EFFICACY AND MOLECULAR MECHANISM
Elucidation of the drug candidate action of mechanism is fundamental to evaluate drug efficacy. MSI can simultaneously acquire different types of endogenous metabolites apart from drug molecules. These spatially-resolved phenotypic changes contain functional molecular information associated with disease progression and drug action. Pharmacologists can evaluate drug efficacy by analyzing downstream metabolites that are regulated by the target enzyme or transporter. An untargeted spatial metabolomics study performed using MSI can also help to explore potential drug targets, screen the functional metabolite markers, and investigate the possible molecular mechanism.
The Agar group conducted a complementary analysis on the glioblastoma patient-derived xenograft (PDX) model. The erlotinib ion image was co-registered with the H&E staining, MRI, and immunohistochemistry (IHC) results to correlate the targeting distribution of the EGFR inhibitor, erlotinib, and resulting tumor phenotype changes, including phosphoproteomics, for assessing protein signaling response and mRNA sequencing for evaluating the transcriptional response. This comprehensive multi-omics information provides insight into pharmacokinetics/pharmacodynamics (PK/PD) correlation .
A hyperspectral imaging method was proposed to simultaneously monitor an anti-insomnia drug candidate, NHBA, and endogenous metabolic profile in whole-body rat sections by AFA-DESI-MSI . The PCA-based hyperspectral imaging spatially revealed metabolic perturbation in response to the NHBA stimulation. Six functional metabolites were discovered to be significantly changed after drug intervention, including neurotransmitters (γ-aminobutyric acid, neurotransmitter precursors choline, and glycerophosphocholine), an endogenous sleep factor, adenosine, and creatine, which are closely associated with insomnia or other neurologic disorders ( Figure 5a ). These findings provide insight into the mechanism underlying NHBA action. It has also been shown that the power of MSI-based spatial metabolomics in studying the molecular mechanism of drug action, especially for drug candidates with multiple or undefined targets in the preclinical study stage. AFA-DESI-MSI combined with the metabolic network mapping method was developed to spatially-resolve the metabolic alterations from the micro-region of the rat brain. Key functional metabolites involved in the complex regulatory network in the central nervous system have been elucidated ( Figure 5b ) after intervention by scopolamine which is a drug candidate for treating Alzheimer’s disease .
12. IN VITRO ACTIVITY EVALUATION AND HIGH THROUGHPUT SCREENING
In the early stage of drug discovery, hundreds of compounds were needed to go through in vitro high-throughput screening (HTS) to select small molecular hits with binding activity to the biological target before conducting the hit-to-lead (H2L) structural optimization and follow-up in vitro and in vivo DMPK studies. A qualified drug candidate should perform an ideal IC50 towards the target protein. This step requires a robust, and reliable in vitro evaluation method with a convenient signal readout system to reflect the compound’s activity with varied chemical structures and test concentrations. In addition, the in vitro tested object should also be highly consistent among different batches to guarantee the comparability in compounds and reproducibility of the whole screening process. The traditional HTS system relies on the single signal readout system. In recent years, the use of MSI has also been put into more practice in the high-content HTS process because of label-free and multiplex detection.
The 2D media-cultured single cells are frequently used testing objects for in vitro activity evaluation and drug candidate screening. These cells contain stably- or transiently-expressed target protein (normally an enzyme, receptor, or transporter) through transfection. To characterize the cell uptake process, the cutting-edge MSI technique based on a laser or ion beam needs to be used to obtain the sub-cellular resolved drug ion image . The Winograd group realized subcellular imaging of antibiotics in single E. coli. bacteria using time-of-flight secondary ion mass spectrometry (ToF-SIMS) with C60 cation as the ion beam. The spatial resolution reached 300 nm. The dose-response of ampicillin and tetracycline in E. coli cell aggregates was confirmed . The Hang group developed a micro-lensed fiber (MLF) to focus the laser desorption post-ionization (LDPI) size to the 300 nm level . Furthermore, the 3D distribution of two anti-tumor drugs, proflavine and methylthioninium chloride, within a single cell were realized at a voxel resolution of 500 × 500 × 500 nm3 using micro-lensed fiber laser desorption post-ionization-mass spectrometry imaging [MLF-LDPI-MSI] ( Figure 6 ) .
Recently, the 3D culturing-based organoids and tumor spheroids became a promising topic in the biomedical engineering field and gains the increasing interest of researchers. Unlike the 2D cultured cells, cells in the 3D organoid, and tumor spheroids have more similar topological structures as the native organ or tumor in the body because of the gradient nutrient supply and mass exchange with the external environment . Paper becomes an emerging 3D culture medium because the microporous fiber structure can simulate the extracellular matrix (ECM) environment where endogenous cells live and grow. The potential of MALDI-MSI with paper-based 3D cell cultures has been demonstrated as a rapid, effective, and large-scale drug screening platform .
The Hummon group evaluated the penetration of doxorubicin-encased liposomes into 3D cell spheroids over the course of 72 h using MALDI-MSI combined with fluorescence microscopy. The free and liposomal doxorubicin, as well as its three active metabolites throughout the spheroids after 12 h of treatment, were identified . Cho et. al constructed the self-assembling multicellular spheroids that have reproducible blood-brain barrier (BBB) features and functions for rapid screening of brain-penetrating drugs. The developed BBB spheroids were then successfully applied to the ex vivo transport of BKM120 (a BBB-penetrant drug) and dabrafenib (a non-penetrant drug) by MALDI-MSI . Superhydrophobic–hydrophilic droplet microarrays (DMAs) were fabricated for the high-throughput cell assay platform for the purpose of MALDI-MSI-based high speed. The size of a spot in DMA was reduced to an internal diameter (ID) of 0.5 mm, allowing for as few as 10 A549 lung cancer cells confined in a 40-nL droplet for high-content drug profiling. Finally, a fatty acid synthase (FASN) inhibitor, GSK2194069, was successfully screened out as the drug candidate ( Figure 7 ) .
13. SAFETY AND TOXICITY EVALUATION
Safety and toxicity tests are important items for assessing possible physiologic damage caused by a drug candidate in the preclinical trial. Many promising drug candidates may have to be suspended for further development because of the severe side effects or even toxicity. The side effects can be induced by the intensive or persistent dosing of the drug or the generation of a toxic metabolite. Special attention should be paid when the local accumulation of a drug in a non-target organ is observed, or its toxic metabolite is monitored. MSI cannot only provide the in situ evidence of the parent drug and its toxic metabolite accumulation, but also correlate the regional molecular profile with the histopathologic information, facilitating the molecular elucidation behind the side effect or toxicity on the impacted tissue region.
There is an increasing number of MSI application cases investigating organ accuracy and chronic toxicity, such as hepatotoxicity, nephrotoxicity , pulmonary toxicity , ocular toxicity, and neurotoxicity . The Cai group integrated MALDI-MSI and MS-based lipidomics to study bisphenol S nephrotoxicity in mice after a 56-day long-term exposure. Glomerular necrosis in the renal cortex, cloudy swelling in the renal medulla, and interstitial collapsing in the renal pelvis were diagnosed. Dysfunctional lipids were those lipids which serve signaling, such as ceramide, sphingosine, and sphingomyelin, in the renal cortex, implicating the vulnerability of renal cortex lipids to the bisphenol S exposure than those in the renal medulla and pelvis . Castellino et al. led a team in GlaxoSmithKline to investigate the metabolism and toxicity of an HIV non-nucleoside reverse transcriptase inhibitor, fosdevirine (FDV), in the central nervous system (CNS). The MALDI-MSI was applied to map the spatial pattern of FDV and its metabolites in the brains of rabbits, minipigs, and monkeys. A cysteine conjugate FDV metabolite was discovered to be the predominant component and persistent in the CNS for an extended period after the last dose in patients with seizure and minipigs. This metabolite was highly enriched in the white matter of rabbit and minipig brains, whereas the predominant component in monkey CNS was FDV, which was specifically located in the gray matter ( Figure 8a ) .
Organ toxicity after long-term or high-dose exposure to a drug can be tested on vertebrate models, such as rats, mice, minipigs, rabbits, and monkeys. In addition to these model animals, zebrafish, and its larvae or embryos are another important vertebrate model in genetics and toxicology ( Figure 8b ) [96, 97]. Shao et al. examined the distribution and metabolism of diazepam (DZP) and its chlorination disinfection byproduct, 2-methylamino-5-chlorobenzophenone (MACB), in zebrafish by DESI-MSI. MACB was shown to accumulate in the spinal cord of the female group. Thereafter, the enriched MACB also transversed through the BBB and induce microglial phagocytosis of neurons .
14. MSI FOR MEDICINAL PLANT AND NATURAL PRODUCT
Terrestrial and marine plants, along with their symbiotic microorganisms on earth, are the largest natural library containing abundant chemical species to be explored. The structural diversity of natural products provides a valuable molecular template for medicinal chemist optimization to satisfy the drug-like properties as a qualified lead compound . MSI has shown its potential in localizing those previously known natural products within the active medicinal part of plants. Additionally, the MSI can also be used for differentiating the multiple steps for active natural product biosynthesis or monitoring the spatiotemporal chemical changes of active natural products during steaming or boiling process.
Compared to the extensively studied biological tissues, the MSI application in plant chemistry is still relatively lagging. In the author’s personal view, the reason behind this phenomenon arises from several technical aspects. First, the natural product in a plant is also known as a secondary metabolite. They have a relatively low content compared to the universally existing primary metabolites, such as amino acids, lipids, carbohydrates, and nucleotides. The more potent a natural product, the less content existing within a plant/microorganism. Second, a plant cell has a unique cellulose wall providing hydrophobic mechanic protection. It is not as easy as a biological tissue for sectioning and adhesively mounted on the substrate for MSI analysis. Third, the availability of plant metabolomics databases is still short in hand for a confident assignment of detectable species from plant imaging. In most situations, natural products derived from a certain type of parent structure are isomers with the only positional difference in substitute groups. The identity of an ion is still hard to accurately assign without the aid of LC-MS/MS or nuclear magnetic resonance (NMR) as complementary. All these reasons pose a methodologic challenge in sensitive detection and accurate structural identification by MSI.
Nevertheless, technical improvement efforts have been made to address these putative issues, including the development of nanomaterials for plant imprinting and on-tissue chemical derivatization of natural products with a specific functional group. A new composite substrate, a hydrophobic polydopamine-modified TiO2 nanotube coated with plasmonic gold nanoparticle, was constructed as a dual-polarity surface-assisted laser desportion/ionization (SALDI) substrate. The intricate structural details of plant samples can be preserved and transferred for the MSI study. The flavonoids, coumarins, and anti-tumor medicine, vinblastine, were successfully visualized from the imprinted petal of J. integerrima  ( Figure 9a ).
Apart from the method development for plant imaging, there were also mounting reports of MSI in evaluating the in vivo isolated natural product properties and a localization study . The MSI application for the in vivo structure-activity relationship (SAR) of natural products has been reported recently. The Guo group quantitatively evaluated the distribution of 7 monoterpene indole alkaloids in 13 rat brain microcompartments by DESI-MSI. These alkaloids were isolated from Uncaria species (Rubiaceae), a traditional Chinese medicine (TCM) for treating CNS diseases. The distribution trend of these Uncaria alkaloids and the R/S-configuration epimer influence were summarized to provide valuable SAR information . Sun et al. used MALDI-MSI to successfully map distributions of notoginsenoside-R1, notoginsenoside-Fc, ginsenoside-Re in the rhizome, and root of Panax notoginseng ( Figure 9b ), a highly valuable herbal medicine used in Asian for the treatment of cardiovascular and hematologic diseases. Furthermore, the spatiotemporal changes during the 2 h steaming process. Series of metabolites were also discovered to significantly change by streaming ( Figure 9c-d ) . Kusari and Spiteller collaboratively reported the use of MALDI-MSI in probing several maytansinoids (maytansine, maytanprine, maytanbutine, maytanvaline, and normaytancyprine) distribution in Putterlickia pyracantha plants. These active maytansinoids show remarkable antibiotic activities along with high cytotoxicity. Many of the maytansinoids have been put into the front line of chemotherapy or clinical trials for treating breast cancer .
The spatially-resolved biosynthesis pathway for the active natural product is another important field for plant chemistry. Qiu and Chen studied the spatial distributions of phenolic acids, flavonoids, and tanshinones across the underground and aerial parts of Salvia miltiorrhiza. The flavonoids and phenolic acids were chosen for in-depth visualization of phenylpropanoid biosynthesis pathways in S. miltiorrhiza. DESI-MSI highlighted key reactions of flavonoid biosynthesis in flowers and identified the core precursor for phenolic acid biosynthesis in Salvia species .
15. PROSPECTIVE OUTLOOK
MSI has been shown to be a powerful tool in pharmaceutical research at the preclinical and clinical phases. Nonetheless, MSI still has room for further improvement in technique and methodology. The technical advances in MSI and its application in pharmaceutical research are summarized in Figure 10 for an overview to what extent MSI has been developed in the two decades of common efforts by the whole community, and which specific research topics MSI is still poorly investigated and unsolved in future. The author would like to share some personal viewpoints on the direction in which the MSI can be further developed as a better multiplex molecular imaging technique.
Sensitivity: A more sensitive in situ soft ionization method needs to be further developed to acquire molecular profile, even from the sub-micron-sized impact region without severe loss in sensitivity. Efforts should be continuously made in the following: the novel sample-mounting nanomaterials; well-designed secondary-ionization; intricate ion focusing lens system; sample-friendly pre-processing; and chemical derivatization. This is always a critical research field that may drive the entire MSI research community forward. The territory of applicable pharmaceutical research will also be expanded due to the technical advancement.
Spatial resolution: This is another critical MSI metric to determine how precise an ion image can be acquired to learn more detailed spatial information. In most situations, this is a paradox issue conjugating with the sensitivity because a smaller sampling size usually means a very limited molecular amount and the corresponding signal response given a constant ionization efficiency. Currently, spatial resolution improvement can be achieved by a multi-modular image fusion strategy. Molecular information in a smaller region can be predicted by co-registering pixels from a rough MS image with other cell-resolved digital images acquired from H&E, IF, and MRI, [70, 75, 106, 107]. This high-resolution image is artificially predicted other than being really observed. Once soft ionization gains another evolutionary breakthrough in sensitivity, a more universe, cost-effective MSI method for sub-micron-resolved drug/metabolite imaging will not be the bottleneck for the single-cell, spatial phenotypic-omics.
Coverage of detectable species: MSI is expected to be a universally-applicable molecular imaging method to all types of functional proteins, glycans, metabolites, and drugs. Thus, the complex networks consisting of all molecules for signaling regulation, metabolism, and interaction can be spatiotemporally explored. From the of proteomic imaging, especially for those protein complexes or membrane proteins, although there have been native MSI methods proposed for direct detection of those functional membrane proteins or complexes [108, 109], the mass tag-based ionization and imaging provide a feasible and universal strategy to indirectly report the locations for these functional macromolecules [110–114]. Thus far, compared to the small molecular drugs, the, such as monoclonal antibody drug (mAb), antibody-drug conjugate (ADC), and RNA drugs, still lack reports in the MSI field.
Diversity of testable objects: The biospecimens for MSI study have widely covered the whole-body animal sections from rats, mice, minipigs, zebrafish, monkeys, tumor spheroids, and xenografted tumors from model animals or patients, fragile organs like lungs, and eye lenses, cultured single cells, and 3D organoids. An way to make MSI compatible with the formalin-fixed paraffin embedding tissue sections is limited, and more widely used biospecimens in clinical pathology, particularly for ambient ionization-based drug/metabolomics imaging studies.
Data collection mode: An industrialized robot system for automatic management of batch tissue samples, streamlined MSI data acquisition, and quality control is expected to establish a more standardized procedure. It is surprising to notice that effort has been made in this industrial design direction . Currently, the microprobe-based spot-by-spot or line-by-line scan mode remains the predominant way for MSI data collection, which usually takes hours or even a whole day to complete for a larger size sample or when the ultrahigh-resolution image is required. A rapid, microscopic snapshot-mode MSI data collection will definitely accelerate the MSI working efficiency . This requires a more sophisticated MS instrumentation design in the micro-arrayed ion detection system.
Isobaric ion differentiation and reliable identification: MSI has the difficulty in recognizing isobaric ions due to the in situ detection principle. Ions with identical formula composition cannot be differentiated. The invention of ion mobility mass spectrometry (IMS) provides a feasible direction to overcome this limitation because it actually plays a role in gas-phase ion “chromatography” separation. The separating ability is still under further improvement. Ion identification is another issue posed to the MSI method. A relatively confident result for an ion identification could be assured by chemometrics based on multiple matching criteria, including exact mass tolerance, isotope abundance, adduct ion types, and featured fragments [117–119]. Apart from the post-acquisition dry method, the online wet experiments also provide complementary chemical structure to help isobaric ion differentiation, such as hydrogen-deuterium exchange (HDX) during the ionization process in the liquid phase or and post-ionization traveling process in the gas phase [120, 121].
Functional Imaging: Mass spectral profile acquired by an MSI experiment is not simply the list of separated irrelevant peaks; there are biological connections among these peaks. These bio-informative connections could provide insight into the drug-phenotype interaction, and spatially resolved molecular composition-environment-biological function. It is believed that MSI can also become a functional imaging tool to visualize biological events, such as proliferation, metastasis, invasion, autophagy, ferroptosis [122, 123], and biomolecular activity, such as enzyme catalysis [124–127], and biological tissue microenvironments, such as acidity , by analyzing and organizing the acquired big data.
Artificial intelligence: Deep learning has already been introduced to learn the molecular profile pattern and guide precise spatial recognition . It is expected to introduce more cutting-edge data science techniques and computation methods to facilitate the integration of spatial multi-omics information, such as spatial genomics and transcriptomics.
Open-access to spatial omics database or platform: An open-access MSI data platform should be proposed and encouraged for MSI researchers’ free exploration and communication . It is also expected for an open-access spatial omics database containing molecular phenotype atlas from key organs and cancer tissue. This, of course, needs the common efforts and consensus of the entire MSI and spatial omics community. It is critical not only for the MSI community, but also for multimodality imaging analysis between MS images and other molecular imaging techniques for the entire spatial omics community.
|Application||Sample||Drugs||Metabolic product||Quantitation||Ionization||Mass analyzer||Complementary methods||Ref.|
|First-pass effect||CD57 mice and SD rat whole-body||Clozapine, terfenadine, fexofenadine, ketoconozole||Y||N||MALDI||QqTOF||LC-MS/MS|||
|Blood-brain barrier||Mouse brain with PVL311 xenograft tumor||Erlotinib, and two drug candidates (BKM120, RAF265)||N||N||MALDI||TOF||Fluorescence|
|Blood-tumor barrier||Mouse brain with metastases of the lung cancer||Epertinib, lapatinib||Y||Y||MALDI||LTQ||Bioluminescent image|
|Dermal permeation||Human/rat skin||Saxitoxin, neosaxitoxin, batrachotoxin, aconitine||N||N||DESI||Orbitrap||H&E Staining|||
|Targeting||Xenografted tumor||Paclitaxel derivatives||Y||Y||AFA-DESI||Orbitrap||H&E Staining|||
|Targeting||Patient-Derived Xenografts (PDX)||Paclitaxel||N||Y||MALDI||TOF||H&E Staining|
|Sustained release||Implanted hydrogel||Celecoxib nanoparticles||N||N||DESI||orbitrap||Immunofluorescence|||
|Penetration||Necrotic Granulomas in lung tissue||Clofazimine, moxifloxacin, bedaquiline, pyrazinamide, rifampicin||N||N||AP-MALDI||orbitrap||H&E Staining|||
|Penetration||Triple negative breast cancer xenograft tumor on rat||Veliparib and carboplatin||N||Y||MALDI, LA-ICP||TOF|
|MRI, LC-MS, H&E Staining|||
|Absorption||Rat intestine||Polyphenol (theaflavin-3′-O-gallate; epicatechin-3-O-gallate)||Y||N||MALDI||TOF||LC-MS|||
|CNS distribution||Monkey, Rabbit brain||Fosdevirine (GSK2248761)||Y||N||MALDI||TOF||LC-MS/MS|||
|Drug distribution||Neoplastic and normal tissues||Pioglitazone||N||Y||AP-MALDI||orbitrap||HPLC-UV|||
|Drug distribution||Colon cancer murine model; transversal whole-body tissue||Irinotecan and its active metabolite SN-38||Y||N||MALDI||TOF, FT-ICR||Immunohistochemical staining|||
|Drug distribution||Cervical tissue model||Emtricitabine||N||Y||IR-MALDESI||orbitrap||LC-MS/MS|||
|Drug distribution||Rat liver, and kidney||Atorvastatin||Y||N||MALDI||orbitrap||LC-MS|||
|Drug distribution||Neonate mice whole-body||Clozapine, and three unknown compounds||N||N||DESI||orbitrap||H&E staining|
|Drug distribution||Rat whole-body|
Mouse bearing neuroglioma
|Deoxytylophorinidine (CAT)||Y||Y||AFA-DESI||QTRAP||H&E staining|
|Drug distribution||Human tumor||Cisplatin, oxaliplatin||N||N||LA-ICP|
|Pharmacokinetics||Mouse spleen||Cucumarioside A2-2||N||Y||MALDI||TOF||Radiospectroscopy|||
|Correlation of drug distribution with vascularization||Xenografted tumor||Afatinib, erlotinib, sorafenib||N||N||MALDI||TOF||Immunohistochemical staining|||
|Cassette-Dosing High throughput PK||Rat kidney||Haloperidol, bufuralol, midazolam, clozapine, terfenadine, erlotinib, olanzapine, moxifloxacin||N||N||MALDI, DESI, LESA||QqTOF, QTRAP, FTICR||Optical microscope|||
|Drug-drug interaction||Mice brain||Propranolol, loperamide||N||Y||DESI|
|SAR||Rat brain||Uncaria indole alkaloids||N||Y||DESI||QTOF||H&E staining|||
|Cyclosporine||Y||Y||DESI / MALDI||Orbitrap|||
|Metabolism and hepatotoxicity||C57BL/6J mice liver||Acetaminophen||Y||N||DESI||QTOF||LC-MS/MS|||
|PK/PD||Squamous Cell Carcinoma of the Head and Neck from patients||Xevinapant||N||N||MALDI||FT-ICR||LC-MS/MS|
|Safety and Toxicology||Zebrafish||Diazepam|
|Safety and Toxicology||Rat kidney||Bisphenol S||N||N||MALDI||TOF||LC-MS/MS|||
|Safety and Toxicology||Wistar Hannover Galas rat kidney||Two tested compounds and their common metabolite bisulphonamide||Y||N||MALDI||QTOF||NMR|
|Mechanism||Rat whole-body||NHBA||N||N||AFA-DESI||TOF; QqQ||LC-MS|
|Mechanism||Rat brain||Scopolamine||N||N||AFA-DESI||Orbitrap||Nissl Staining|||
|Cellular uptake||In vitro cultured HeLa cells||Carboplatin||N||N||LA-ICP||Quadrupole||Microscopy|||
|Pharmaco-metabolomics||Rat brain||Scopolamine||N||N||AFA-DESI||Orbitrap||H&E staining|||
|Pharmaco-metabolomics||Mouse lung, liver, kidney||Pirfenidone||Y||Y||MALDI||FTICR||—|||
|In vitro evaluation||HCT 116 colon carcinoma multicellular spheroids||Irinotecan||N||Y||MALDI||FTICR||Nano-LC-MS|||
|In vitro high-throughput screening||Live A549 lung cancer cells||Fatty acid synthase inhibitor GSK2194069||N||N||MALDI||TOF||Microscopy|||