Spatial Proteomics and Multiplexed Tissue Imaging: A Methods Guide

Spatial proteomics maps proteins directly within the architecture of intact tissue sections, revealing not just which proteins are expressed but where they reside within and between cells in the environment where they actually function. While transcriptomics captures the molecular messages a cell is producing, spatial proteomics maps the physical machinery: the receptors, scaffolds, and effectors. Multiplexed tissue imaging technologies now permit the simultaneous detection of dozens to more than 100 proteins in a single tissue section. This technology has made spatial proteomics one of the most rapidly expanding disciplines in modern cell biology.

Why map proteins in space

Spatial proteomics preserves the tissue architecture that dissociation-based protein profiling discards, enabling measurement of protein identity, abundance, and location simultaneously within the same tissue section. Mass spectrometry-based proteomics extracts proteins from tissue lysates, dissolving cellular architecture in the process. Conventional immunofluorescence (IF) and immunohistochemistry (IHC) preserve tissue structure but are limited to detecting two to four targets per staining run, which is insufficient for resolving the phenotypic diversity of complex, heterogeneous tissues.

The shortfall is consequential for multiple areas of biology. Whether an immune cell can physically infiltrate a tumor depends on its position within tissue architecture, not just its transcriptional profile. Whether two populations with similar single-cell expression signatures interact within a functional niche depends on whether they are co-localized within the same tissue domain. Spatial proteomics addresses questions such as these by coupling high-plex protein detection with location information that traditional methods discard.

Nature Methods named spatial proteomics as its Method of the Year for 2024, citing its critical role in revealing the organization of complex tissues and its foundational contributions to atlas-scale projects in health and disease. That recognition reflects how multiplexed protein imaging has moved from a specialized tool to a critical component of tissue biology.

For researchers aiming to understand how spatial proteomics relates to multiomics approaches, a complete guide to spatial biology is available. For those approaching the field from a transcriptomics background, a spatial transcriptomics methods guide maps the RNA-detection side of the field in depth.

Fluorescence-based multiplexed tissue imaging

Fluorescence-based multiplexed imaging methods share a common principle: antibodies labeled with fluorescent reporters or DNA barcodes are applied to tissue sections, and their signals are read out across multiple imaging rounds using optical microscopes. The critical innovation enabling high-plex detection was the separation of the antibody-binding step from the readout step. When all antibodies are applied to the tissue simultaneously in a single staining reaction, epitopes are preserved and tissue quality is maintained across rounds. However, sequential detection cycles allow more targets than any fixed number of fluorescence channels would otherwise accommodate.

Co-detection by indexing (CODEX), commercialized as PhenoCycler by Akoya Biosciences, conjugates antibodies to unique DNA oligonucleotides rather than directly to fluorophores. A CODEX tissue imaging protocol published in Nature Protocols described how complementary fluorescent reporter probes are cyclically hybridized, enabling detection of up to 60 markers from a single staining reaction. Because all antibodies are bound before imaging begins, the approach avoids the cumulative epitope damage that accrues across successive staining rounds in other formats. This high-plex fluorescence tissue imaging is underpinned by a robust workflow.

Cyclic immunofluorescence (CyCIF), developed by the Sorger laboratory at Harvard Medical School, achieves high-plex detection through a different mechanism. Conventional fluorophore-conjugated antibodies are applied in successive rounds, with chemical inactivation of fluorophores between each imaging cycle rather than probe removal and replacement. Implementation of CyCIF in imaging of human tissues and tumors has demonstrated the assembly of up to 60-plex images from iterative four-channel acquisitions. The technique was performed on standard slide scanners and widefield microscopes, without requiring dedicated spatial biology instrumentation. That compatibility with existing laboratory equipment distinguishes CyCIF from platform-specific systems and has driven adoption in settings where capital equipment budgets are a constraint. Adequate workflows must be implemented, and registration steps taken, to integrate CyCIF approaches into standard laboratories successfully.

Both CODEX and CyCIF are compatible with formalin-fixed, paraffin-embedded (FFPE) tissue sections, making them well-suited for retrospective studies using archival clinical material. Both techniques also require computational image registration across imaging cycles to maintain precise spatial alignment of cell positions as successive rounds of signal are acquired. In practice, the number of markers that can be reliably recovered from a given sample is determined by tissue quality and the preservation of antigenicity across rounds, not solely by the number of antibodies in the panel.

Mass-based multiplexed tissue imaging

Mass-based spatial proteomics methods detect proteins using stable metal isotopes rather than fluorophores, eliminating spectral overlap and tissue autofluorescence in a single step. Where fluorescence-based methods must contend with overlapping emission spectra and background signal from collagen, lipofuscin, and red blood cells, metal-isotope detection distinguishes channels by mass-to-charge ratio, a physical property with no equivalent background in biological tissue. Because heavy metal isotopes do not occur at detectable concentrations in biological tissue, the background signal is essentially absent.

Imaging mass cytometry (IMC), developed by Bernd Bodenmiller's group and commercialized by Standard BioTools on the Hyperion imaging system, ablates tissue row by row using a pulsed ultraviolet laser. Metal-isotope-conjugated antibodies bound at each ablated pixel are vaporized along with the surrounding cellular material. The resulting aerosol is swept into a time-of-flight (TOF) mass spectrometer for element-specific detection.

Standard IMC operates at a spatial resolution of approximately 1 micrometer. A 2025 study in Nature Methods demonstrated subcellular resolution through oversampling and deconvolution, enabling visualization of nuclear foci and mitochondrial networks. IMC supports detection and visualization of up to 40 different protein markers per acquisition, a ceiling set by the number of non-overlapping metal isotopes available rather than by optical constraints. Researchers should consider the full workflow and panel design when conducting metal-tagged tissue imaging.

Multiplexed ion beam imaging (MIBI), developed by Michael Angelo's laboratory and commercialized by Ionpath on the MIBIscope, uses a primary ion beam to sputter material from labeled tissue sections rather than a laser. Secondary ions ejected from the surface are detected by TOF mass spectrometry. The MIBI-TOF platform described in Science Advances achieves spatial resolution down to 260 nanometers (nm), with sensitivity approaching single-molecule detection. That resolution, along with dynamic range, distinguishes MIBI from IMC, particularly for applications where subcellular protein localization is a defined experimental requirement rather than a secondary consideration. However, there are instrumental and practical trade-offs to consider with MIBI imaging.

Both IMC and MIBI require antibodies conjugated to metal isotopes rather than fluorophores, and both are compatible with FFPE tissue. The absence of autofluorescence background eliminates a major source of noise, giving mass-based detection a practical advantage in high-autofluorescence organs such as lung and liver where fluorescence-based methods require careful spectral unmixing or tissue-specific optimization.

Antibody panels and multiplexed imaging experimental design

The quality of a multiplexed imaging experiment is largely determined before the first tissue section reaches the platform. Panel design and validation, which encompasses selecting protein targets, identifying validated antibody clones, conjugating antibodies to the appropriate detection labels, and assigning targets to channels, sets the ceiling for what an experiment can reveal. A poorly validated panel cannot be rescued by platform sensitivity; an antibody that performs inconsistently in the target tissue type will produce an unreliable signal regardless of instrument specifications.

Clone selection is the first critical decision. Antibodies validated for conventional IF or IHC frequently perform differently under multiplexed conditions, particularly when fixation chemistry, tissue handling, or the conjugation chemistry required by a specific platform alters antigen accessibility. Specificity checks using positive and negative control tissues, titration experiments to identify the minimum effective antibody concentration, and signal-to-noise assessment in the actual target tissue type are standard validation steps regardless of whether the detection platform is fluorescence-based or mass-based.

Channel assignment introduces a second layer of platform-specific complexity. In fluorescence-based methods, antibodies must be assigned to fluorophore channels in a way that minimizes spectral bleed-through, and high-abundance targets should occupy channels with lower background to avoid saturation. In IMC, the metal isotope assigned to each antibody must be selected to avoid mass overlap with other isotopes in the panel. In MIBI, dynamic range provides additional flexibility for accommodating both abundant and rare targets in the same panel without signal compression.

A structured approach to antibody panel design applies across both fluorescence and mass-based platforms:

• Define biological questions first and select markers that directly address each question before filling the remaining detection channels.
• Validate every antibody clone in the specific tissue type and fixation chemistry to be used in the study, not only on reference cell lines or control tissues from a different organ.
• Assign high-abundance targets to lower-sensitivity channels and low-abundance targets to high-sensitivity channels to prevent signal compression or saturation.
• Run a single-stain titration for each antibody individually before constructing the full panel.
• Include positive and negative control tissues in every acquisition batch to track panel performance and detect batch-to-batch variability.

Selecting a multiplexed imaging method

Choosing a spatial proteomics platform is a matching exercise rather than a ranking problem: no single multiplexed imaging system performs optimally across all tissue types, biological questions, and laboratory environments. The relevant criteria to consider when comparing multiplexed imaging methods span plex, spatial resolution, tissue type compatibility, acquisition throughput, reagent access, and the computational infrastructure available for data analysis (Table 1).

Table 1: Major multiplexed tissue imaging platforms used in spatial proteomics and their principal characteristics.

Platform	Detection principle	Typical plex	Nominal resolution	Tissue compatibility
PhenoCycler (CODEX)	Fluorescence, DNA-barcoded cycling	Up to 60 markers or more	Near-subcellular (light-limited)	FFPE and fresh-frozen
CyCIF (t-CyCIF)	Fluorescence, iterative stain-image-bleach	Up to 60 markers	Near-subcellular (light-limited)	FFPE and fresh-frozen
IMC (Hyperion)	Metal-tag, laser ablation, mass spectrometry	40 or more markers	~1 µm (subcellular with high-res protocols)	FFPE and fresh-frozen
MIBI (MIBI-TOF)	Metal-tag, ion beam, TOF mass spectrometry	Up to 40 markers	Down to 260 nm	FFPE and fresh-frozen

Plex is often the first metric researchers consider, but it is rarely the binding constraint: many platforms support 40 or more markers per acquisition, a number that already exceeds what most biological questions require for comprehensive cell-type characterization.

Resolution and throughput are more frequently decisive. Mass-based platforms offer lower effective resolution in standard operating configurations compared with FFPE-optimized fluorescence systems, but that trade-off is largely irrelevant for studies focused on cell-level phenotyping rather than subcellular protein localization. Acquisition speed differs substantially: fluorescence cycling methods can cover large tissue areas in a single imaging session, while IMC's row-by-row laser ablation process is slower and is best suited to smaller regions of interest rather than whole-tissue scanning. A multiplex protein imaging review published in Nature Reviews Cancer covers platform trade-offs in detail, and tissue autofluorescence is a commonly cited selection factor. High-autofluorescence organs such as the lung and liver can substantially degrade signal-to-noise in fluorescence-based methods, giving mass-based detection a practical advantage in those contexts.

Analyzing spatial proteomics data

Spatial proteomics data analysis begins where image acquisition ends, and the computational steps between raw image and biological interpretation are as critical as the experimental design itself. Raw output from any of these platforms, whether gigapixel-scale fluorescence images or pixel-level mass spectrometry maps, requires substantial processing before yielding cell-level measurements. Cell segmentation, the step of defining where one cell ends and the next begins in a densely packed tissue section, is among the most consequential of these processing decisions, and errors introduced here propagate into every downstream analysis.

Segmentation approaches for multiplexed tissue imaging typically rely on nuclear markers such as DAPI (4′,6-diamidino-2-phenylindole) or nuclear proteins to define cell centers. This is followed by boundary expansion to approximate the cytoplasmic extent of each cell. Deep-learning segmentation models have largely replaced classical threshold-based approaches, and their performance scales with the quality and tissue-specificity of training data. Once cells are segmented, each cell inherits a vector of protein expression values from the pixels it contains. Cell typing then uses these expression vectors, typically through clustering or marker-threshold-based gating, to assign cell identity labels such as cytotoxic T cells, macrophages, or tumor epithelial cells.

Spatial statistics interrogate the arrangement of cell types, asking whether specific combinations are found in proximity more often than expected by chance and whether that proximity correlates with clinical outcomes or biological state. A spatial omics methods review published in Nature Methods surveyed computational approaches to spatial proteomics data, highlighting cell segmentation, normalization, and cell-type assignment as analytical steps where methodological choices propagate into downstream biological conclusions. Open-source toolkits, including Squidpy, Seurat, and Giotto—developed at the Technical University of Munich, the Satija Lab, and the Dries–Yuan lab, respectively—have become standard frameworks for this analysis work, with toolkit choice often driven by whether a lab's broader computational infrastructure runs in Python or R.

Data volumes add a practical layer of complexity: a single multiplexed imaging session across multiple tissue sections can produce terabyte-scale outputs when image files are combined with cell-by-marker expression matrices and spatial coordinate tables. Pipelines must be designed with data management in mind from the outset, including decisions about file formats, storage architecture, and version-controlled analysis workflows that can be reproduced by collaborators over time.

Spatial proteomics as a framework for tissue biology

Spatial proteomics has become a standard component of modern tissue biology, with applications now extending across oncology, immunology, neuroscience, and translational medicine. Cancer research adopted the approach rapidly because the tumor microenvironment is intrinsically spatial: whether immune cells can reach a tumor, how stromal cells signal to malignant cells, and which immune subsets are excluded from a tumor core are questions that dissociated single-cell methods cannot answer. Multiplexed imaging studies have demonstrated that the spatial organization of immune cells within the tumor microenvironment predicts response to immunotherapy in ways that bulk or single-cell sequencing alone cannot, a finding with direct implications for biomarker development across tumor types. A growing body of work on spatial omics in oncology continues to extend this evidence base and to connect spatial protein mapping to therapeutic response prediction.

Spatial proteomics methods, platforms, and data analysis workflows are each evolving rapidly, and the platform that is the right fit for a project today may not remain optimal as the field develops. Researchers building a first multiplexed imaging project are well served by starting with the specific biological question and working backward to the platform whose resolution, plex, tissue compatibility, and throughput actually match the experimental design, rather than defaulting to the highest-specification instrument available.

Additional context on where spatial protein mapping sits within the broader landscape of spatial biology in disease research is available. The methods, trade-offs, and experimental considerations covered across this collection are intended to give researchers the grounding needed to design experiments that are matched to their biology, their tissue types, and their infrastructure.

This content includes text that has been created with the assistance of generative AI and has undergone editorial review before publishing. Technology Networks' AI policy can be found here.