Imaging Mass Spectrometry: What’s Now? What’s Next?

Acquiring spatially-resolved measurements of molecules in substrates using mass spectrometers is not a new endeavor – the first such experiments were performed in the 1960s via secondary ion mass spectrometry (SIMS) ionization sources (1). Those early instruments were termed “ion microscopes” and consisted of ion optic collection systems that maintained the spatial positions of ions desorbed from the sample surface through the mass analyzer and to a detector. Modern-day experiments are now typically performed in “scanning microprobe” modes of analysis, where a raster sampling of a tissue surface enables the collection of mass spectra from discrete x, y positions. In both microscope and microprobe modes, the goal is the same: to produce maps of intensities for compounds of interest across the sample surface. But the sophistication of the instrumentation and applications of imaging mass spectrometry, also termed mass spectrometry imaging (MSI), has advanced tremendously over the last 60 years, and has greatly accelerated over the last 20 years. So, what is the current status of the field? And where are we headed over the next 20 years? Herein, I highlight current research directions and trends in the field of imaging MS. These include new developments in technology, including the rise of spatial-omics approaches, multimodal analyses, high spatial resolution techniques, and isomer imaging, as well as new and exciting applications to molecular pathology. I’ve highlighted research from my own lab (since it is what I know best!) as well as exciting recent reports from others. This account is not intended to be exhaustive – there are too many stellar researchers and reports to name individually; for more information, I direct you to several excellent, recent reviews (2, 3, 4, 5). My intention here is to offer a personal perspective on the most impactful future developments in the world of imaging mass spectrometry.

What’s now?
 

Metabolomics and lipidomics strategies once relied on separation (for example, through solvent extraction, selective derivatization, and/or chromatography) prior to introduction to the mass spectrometer to effectively sample the breadth and depth of the cellular metabolome. This limited the scope of early imaging MS analyses of these compounds, which required direct (without prior separation) sampling from tissue surfaces. Investigators had to focus on mapping a few compounds of interest at a time. Today, high resolving power mass analyzers, rapid gas-phase separation techniques (such as ion mobility), and multiplexed tandem mass spectrometry (MS/MS or MSⁿ) approaches with improved peak capacities enable the mapping of hundreds to thousands of discrete compounds in a single experiment. MS has thus entered an age where “spatial-omics” measurements can be made – that is to say the multiplexed detection of entire classes of biomolecules with spatial context. For example, so-called “soft” ionization techniques, such as matrix-assisted laser desorption/ionization (MALDI) and desorption electrospray ionization (DESI), enable detection of metabolite and lipid analytes directly from tissue surfaces. Investigators now routinely detect these compounds in situ within the spatial context of the tissue, which has greatly aided molecular analyses of biology and pathology. 

Spatial proteomics measurements are also on the rise. Some protein imaging analyses are performed directly using MALDI and liquid surface extraction techniques, while others are performed indirectly (or “offline”), following solvent-based microextractions. Though spatial transcriptomics measurements are more frequently made with fluorescence microscopes, the use of MS to study oligonucleotides is currently experiencing a resurgence that may inspire mapping of genetic information using imaging MS.

The past decade has seen a tremendous rise in the availability of these spatial-omics technologies. Investigators are now integrating data from multiple orthogonal techniques that provide complementary chemical information. Data from spatial-omics workflows have been combined with microscopy (for example, optical, fluorescence, and particle-based), spectroscopy (for example, infrared and Raman), as well as electrochemical imaging. Multiple technologies are often used to compensate for the deficiencies of the complement modalities. For example, magnetic resonance imaging (MRI) is low in molecular specificity, but allows for in vivo measurements while the subject is still alive, unlike most imaging MS workflows. Multi-modal imaging workflows are also providing more holistic views of tissue biochemistry, as well as aiding in validating molecular observations. We have used imaging performed by MALDI, laser ablation inductively coupled plasma (LA-ICP), and bioluminescence to image immune response proteins, nutrient metals, and bacterial expression, respectively, in a mouse model of systemic Staphylococcus aureus infection (6). And co-registration of these images allowed us to confirm co-localization of metal-binding proteins detected by MALDI with nutrient metals detected by LA-ICP and areas of bacterial niche within abscesses. This multimodal imaging platform identified regions of metal starvation within soft tissue abscesses observed during infection and helped to advance our understanding of inflammatory response and host–pathogen interactions. 

Ambitious programs are underway from talented teams of scientists to create even larger multimodal spatial maps of human tissues that can serve as reference atlases for the scientific community.

A major challenge associated with integrating data from multiple imaging sources is the significant disparities in spatial resolution obtained from each of the modalities. Imaging MS is typically limited to 10–100 µm spatial resolutions, while other modalities can vary by orders of magnitude; for example, fluorescence and SEM imaging approaches can easily reach sub-1 µm and sub-1 nm resolutions, respectively. A variety of “single-cell” iterations of omics workflows have emerged; though a few microprobe single-cell approaches have been described, the majority of single-cell workflows are fluorescence-based or perform MS analysis following cell dispersions or solvent-based microextractions. The limited spatial resolution of imaging MS then generally limits the structural level to which chemical information can be assigned. Improvements in MALDI laser optics have enabled sampling beam diameters down to approximately 1 µm in diameter, but the cost and expertise required to build and maintain these customized platforms can be significant, making high spatial resolution imaging experiments unfeasible for the broader scientific community. Even these specialized setups can run into significant limitations, such as poor limits of detection due to less material ablated during the MALDI process. But creative approaches to combat this issue exist – secondary ionization (for example, MALDI-2) and the use of antibody-conjugated amplification detection strategies (for example, imaging mass cytometry and MALDI-immunohistochemistry). We and others are seeking to address these challenges by physically magnifying the tissue substrate to improve the effective spatial resolution of the imaging experiment. These polymer-based protocols are built from the framework of expansion microscopy (ExM) and can provide for imaging MS spatial resolution enhancements of 20-fold, providing exciting opportunities for single-cell and subcellular measurements. We and others have also explored the use of computational image fusion approaches, which enable the predictive upsampling of imaging MS data by building cross-modality mathematical relationships with high spatial resolution microscopy images (7, 8, 9).

Despite the high molecular specificity afforded by the mass spectrometer, severe deficiencies still remain in the differentiation and identification of small molecules, where a multitude of isobaric and isomeric compounds exist. The failure to adequately separate and identify these compounds results in composite images and limits accurate understanding of metabolism and cellular biochemistry. Conventional MS/MS performed using collision induced dissociation (CID) has become an important resource in proteomics, lipidomics, and metabolomics workflows due to the high level of specificity afforded by fragmenting compounds of interest and then analyzing the product masses. However, CID alone is not successful at resolving these compounds in all instances, necessitating alternative approaches. A number of groups have explored on-tissue chemical derivatization prior to or during ionization. This derivatization process changes the type of lipid ion that is ultimately sampled into the mass spectrometer and subjected to CID, resulting in commentary fragmentation pathways. For example, classical Paternò-Büchi (PB) photochemical derivatization has been used to specifically form adducts at lipid carbon–carbon double bonds (C═C) under ultraviolet (UV) irradiation (10, 11, 12). Low-energy CID then results in diagnostic product ions specific to double bond isomers, allowing for the identification and discrete imaging of each isomer. Separation and identification has also been performed in the gas-phase following ionization using ion mobility coupled to mass spectrometry (IM-MS), alternative ion dissociation approaches (for example, ultraviolet photodissociation and electron induced dissociation), ion/molecule reactions (for example, ozone-induced dissociation), and ion/ion reactions. In this area, we have used gas-phase charge inversion ion/ion reactions to transform protonated phosphatidylcholine (PC) monocations into more structurally-informative demethylated anions to map sn-positional lipid isomers in MALDI imaging MS (13). It is likely that each phospholipid has as many as 10–20 individual isomers! Resolving these isomers may reveal important insight into canonical and non-canonical patterns of metabolism and could serve as important biomarkers and potential targets for therapeutic intervention.

Overall, this cohort of recent technologies highlight the unique ability of imaging MS – and multi-modal, spatial-omics approaches in general – to serve as both hypothesis testing and hypothesis generating modalities of research. Investigators have made astounding inroads on understanding a wide variety of diseases using these new tools for molecular pathology. Pharmaceutical companies and clinical chemists are frequently using imaging MS to better understand drug pharmacokinetic-pharmacodynamic (PK-PD) relationships. Imaging MS is also being used to understand neurodegenerative diseases, such as Parkinson’s disease and Alzheimer’s disease. We and others have used lipid and metabolite imaging to better understand metabolic dysfunction in diabetes and cancer (14). Exciting progress has been made in the use of imaging MS to study a wide variety of infectious diseases. For example, we have recently mapped the metabolic cross talk between microbiota and Clostridioides difficile during systemic infection and demonstrated a metabolic remodeling in the mouse gut during Enterococcus and C. difficile coinfection (15,16). These findings are important for our understanding of the conditions impacting the outcome of C. difficile infection, the risk for recurrence, and the factors impacting successful treatment efforts. Such applications provide roadmaps for the discovery of innovative translational strategies to improve human health.

What’s next?
 

Spatial-omics approaches, multimodal analyses, high spatial resolution techniques, isomer imaging, and biological and clinical applications will all continue to grow and evolve. The increased complexity of “big data” produced via imaging MS and multi-modal technologies presents new and important challenges to data analysis and data integration. Pipelines that can import and visualize data from multiple imaging modalities are emerging. Software and databases that can intelligently mine data from different classes of biomolecules (for example, genes, proteins, and metabolites) and map discrete biochemical pathways will be invaluable for maximizing the potential impact of multimodal datasets. It is highly likely that artificial intelligence (AI) will be a major player in enabling these analyses. Of course, increased expression of a gene does not always correlate with downstream concomitant increase in high expression of a metabolite – biology is rarely so simple! Complex and overlapping transport, synthesis, degradation, and modification pathways make disease and drug pathway analyses challenging to unravel. Still, imaging MS holds significant promise for contributing to systems biology approaches to understanding human health and disease, as is evident by the steady increase in clinical applications of imaging MS over the past 15 years. Though translational acceptance of new analytical technologies is often slow, multiple dedicated labs are undertaking the important work of perfecting robust workflows, sampling technology, and data analysis. For example, several groups are perfecting in vivo sampling approaches to enable clinical applications. Liquid micro-junction surface sampling (LMJ-SSP), DESI, the iKnife, and the MasSpec Pen have been used to classify tumors and have even been used in real-time surgeries for monitoring margins during tumor resection. Extraordinary success has been enabled for breast, colorectal, thyroid, and lymph node tissues by these dedicated research teams.

The continued success and adoption of imaging MS approaches lies in ensuring rigorous approaches to these measurements and techniques as they expand beyond the subfields of mass spectrometry and analytical chemistry. As the scopes and complexities of the studies increase, so too do the chances of error. As a fellow technology developer, I stress to my own research group the importance of a fundamental and thorough comprehension of the underlying technology (in our case, the mass spectrometer). This depth of understanding allows us to creatively explore the limits of the technology, position ourselves for serendipitous discoveries, and be wary of spurious results. Generating reproducible and reliable data from controlled sample sources using verifiable statistical tools and software is imperative. The emergence of big data repositories and universal file formats will aid in this dissemination and validation, but it remains the responsibility of individual investigators to simultaneously serve as quality control checkpoints for existing techniques and to expand the frontiers of imaging mass spectrometry development and applications.