Mini Review: Spatial Transcriptomics to Decode the Central Nervous System

Abstract

Spatial transcriptomics (ST) is revolutionizing our understanding of the central nervous system (CNS) by providing spatially resolved gene expression data. This mini review explores the impact of ST on CNS research, particularly in neurodegenerative diseases like Alzheimer’s, Parkinson’s, multiple sclerosis, and amyotrophic lateral sclerosis. We describe two foundational ST methods: sequencing-based and imaging-based. Key studies are reviewed highlighting the power of ST data sets to map transcriptomes to disease-specific histomorphology, elucidate molecular mechanisms of regional and cellular vulnerability, integrate single-cell data with tissue mapping, and reveal receptor-ligand interactions. Despite current challenges like data interpretation and resolution limits, ST holds promise for identifying novel drug targets, evaluating their therapeutic potential, and bridging gaps between animal models and human studies to advance development of CNS-targeting compounds.

The complex biology of the central nervous system (CNS) has been extensively studied to deepen our understanding of its intricate processes, encompassing its developmental, homeostatic, disease progression stages, and toxicology.^11,12,23,28 Conventional pathology methods like hematoxylin and eosin (H&E) staining, immunohistochemistry (IHC), and in situ hybridization (ISH) offer valuable insights into neurodegeneration, neurotoxicity, and related alterations in key protein or RNA markers. While these classical techniques provide single-cell resolution, their limited multiplexing capacity (~3-5 target proteins/transcripts) restricts their ability to fully unravel the complexity of the CNS.

Classical transcriptomics techniques, such as bulk RNA sequencing, analyze brain structures as, e.g., the different cortical layers as single samples, providing average gene expression profiles from diverse cell types. This cost-effective approach enables the identification of differentially expressed genes (DEGs) across conditions, including specific disease states and finely micro-dissected brain regions. However, interpreting these results is challenging due to the heterogeneous mix of cell types and brain structures, where dominant signals (highly expressed genes of most abundant cell types) can mask changes in specific cell populations.

Single-cell and single-nucleus RNA sequencing (sc/snRNA-seq) address this limitation by capturing cell-specific gene expression profiles, allowing for more precise analysis. Nonetheless, scRNA-seq lacks information on the spatial origin of cells, which is crucial for understanding cellular interactions and phenotypic characteristics in their tissue context.

To overcome the limitations of previous methods, the concept of spatial transcriptomics (ST) emerged in 2016, ²⁶ combining conventional histology methods, such as H&E or fluorescent IHC with the high-throughput capabilities of advanced sequencing technologies. ⁷ There are various ST methods that provide RNA information along with spatial coordinates within tissues. These technologies are broadly classified into two categories: sequencing-based and imaging-based methods (Figure 1). Sample preparation follows standard histopathology practices, with most listed technologies offering solutions for both fresh frozen (FF) and formalin-fixed paraffin-embedded (FFPE) samples. While commercial protocols are optimized for human and mouse, many technologies are species-agnostic. Vendor-specific variations may require alignment with pathologist workflows. Ensuring RNA preservation is crucial for experimental outcomes, which should be verified in all cases. Tissue samples should be collected as soon as possible after autopsy or sacrifice. During processing, under-fixation and over-fixation should be avoided. For FF tissue, low-temperature control is essential, as repeated handling and sectioning may compromise RNA quality. In addition, when preparing slides for ST experiments, all equipment must be kept sterile and RNase-free. A more detailed comparison of these technologies is available in the review article: “Points to Consider from the ESTP Pathology 2.0 Working Group: Overview on Spatial Omics Technologies Supporting Drug Discovery and Development.”¹⁰

Figure 1.

Basic principles of sequencing-based (sST) and imaging-based (iST) ST technologies, after standard histopathology sample preparation protocols. For sST, tissue samples are stained using H&E or immunofluorescence (IF) protocols, RNA transcripts are released and captured with spatially barcoded oligonucleotides, and the final product oligonucleotides containing gene and spatial barcode sequences are sequenced. For iST, sample preparation includes pretreatment, probe panel application, and signal amplification. Fluorophore-conjugated oligonucleotides hybridize to visualize the probe panel, followed by imaging and signal recording over several rounds. The signal sequence at each coordinate is decoded, and the single-cell or subcellular resolution readout can be combined with additional staining techniques. Created in BioRender. Pesti, B. (2025) https://BioRender.com/m00q059.

The most commonly used commercially available sequencing-based ST (sST) platforms include 10x Genomics Visium, Stereo-Seq, Nanostring GeoMx, and Curio Seeker. ²⁵ These technologies can generate readouts for thousands of genes (18,000-20,000), potentially covering the entire transcriptome and enabling comprehensive gene expression analysis with multiplexing capabilities. These platforms capture transcripts using pre-defined spatial coordinate layouts and do not attempt to identify individual cells or their nuclei. However, novel high-definition sST technologies with 0.5 to 10 µm resolution are approaching single-cell resolution. Despite this advancement, the transcriptomic information obtained from each capture location can still originate from multiple different cells. In addition, the high costs associated with novel sST assays can significantly limit experimental scope, restricting the number of samples and conditions that can be explored.

The most widely used imaging-based ST (iST) platforms include 10x Genomics Xenium, Merscope, and NanoString CosMx. Unlike most sST methods, imaging-based methods offer single-cell or even subcellular resolution, though achieving this resolution can be limited by factors, such as cell segmentation challenges (requires advanced image analysis algorithms or tissue-specific membrane staining) and optical crowding (e.g., highly expressed gene signals overlap with other transcripts). These platforms typically target only 500 to 5000 genes (of ~22,000), and increasing the number of targeted transcripts can further exacerbate optical crowding issues.^4,23 In addition, these technologies require significant investment in high-cost instruments, limiting broader accessibility. While both types of platforms are expensive, iST often requires additional optical systems and specialized equipment, which may increase costs relative to sST approaches.

By incorporating spatial dimensions into transcriptomic data sets, researchers can align these readouts with classical staining techniques, such as H&E or IHC, ISH, or even Matrix-Assisted Laser Desorption/Ionization (MALDI) imaging. Staining on the same or consecutive tissue sections is pivotal to get a multimodal readout from the very same cell or cell-niche. This alignment enables the annotation of ST data points based on imaging characteristics, such as anatomical regions, disease hallmarks, or the presence of a therapeutic compound. These annotations facilitate the grouping of data points and the identification of DEGs that define specific phenotypic areas, such as the cerebral cortex, histopathologic features like gliosis, or treatment-affected regions. However, sST technologies often face challenges in fully separating signals from distinct cells within these annotated areas due to the limited spatial resolution. To address this, several deconvolution methods have been developed that integrate sST with sc/snRNA-seq data to map the abundance of specific cell types within the tissue. ⁶ Figure 2 illustrates how cell type composition per spot is estimated from mixed cell populations. An additional strength of spatially resolved technologies is their ability to facilitate detailed studies of cell communication events, owing to the spatial component of the transcriptomics information. This enables the examination of various receptor-ligand interactions across different anatomical regions, as well as changes in specific receptor-ligand interactions within the same anatomical region under different conditions, such as disease versus control, or early-stage versus late-stage disease samples. ⁵ Figure 2 illustrates the application of above described ST analysis concepts in the CNS, with a particular focus on neurodegenerative diseases.

Figure 2.

Spatial transcriptomics workflow applied to CNS samples with neurodegenerative diseases, highlighting four key analyses: 1. Spot annotation by phenotypic hallmarks as disease associated features and/or brain regions. 2. Differential gene expression by annotation. 3. Integration of sc/snRNA-seq: deconvolve cellular composition or map cell types to tissue regions. 4. Cell-cell interaction enrichment through receptor-ligand co-localization. Created in BioRender. Pesti, B. (2025) https://BioRender.com/m00q059.

ST-enabled analysis has shown great potential for generating spatial cell atlases of the healthy brain and examining cell types in brain subregions, primarily in mice, and also in humans and non-human primates.^2,20,29 Moffitt et al. ²² used iST and snRNA-seq to study the preoptic regions of the mouse hypothalamus, identifying around 70 distinct neuronal populations—many previously unknown—and revealing significant heterogeneity in their abundance within cell populations once thought to be functionally uniform. Similarly, Chen et al. used high-resolution sST and snRNA-seq to map cell-type organization in the macaque cortex. Comparative analysis of transcriptomic data from the cortices of humans, macaques, and mice has identified cell types unique to primates that are predominantly found in layer 4, with their marker genes exhibiting region-specific expression patterns. ² These studies highlight the potential of ST CNS atlases to uncover novel CNS biology and provide valuable reference data for assessing region-specific target expression across preclinical animal species and humans.

Beyond healthy brain mapping, ST is increasingly used to investigate CNS tissues in mouse models and patients with neurodegenerative diseases, including Alzheimer’s disease (AD), Parkinson’s disease (PD), multiple sclerosis (MS), and amyotrophic lateral sclerosis (ALS). The unique ability of ST to align neuropathological features with molecular signatures provides critical insights, as detailed in the following sections.

ST Uniquely Maps the Transcriptome to Histomorphology, Enabling the Study of Disease-Specific Patterns Across Brain Regions and Translating Findings From Mouse Models to Human Diseases

This is exemplified by investigations of the plaque environment in AD using patient samples and/or AD mouse models.^4,16,21 sST and iST combined with plaque-detecting IHC in App^NL-G-F mice (mimics plaque accumulation) of different ages revealed early changes in a gene co-expression network featuring myelin and oligodendrocyte genes. ⁴ At more advanced stages, a multicellular gene network characterized by the complement system, oxidative stress, lysosomal activity, and inflammation becomes more pronounced. ⁴ sST was also combined with Aβ and pTau IHC to examine tangle-associated and plaque-associated gene signatures. ¹⁶ In addition, sST was integrated with fluorescent amyloid imaging to study gene expression related to amyloid-beta plaques and fibrils, revealing shared genes between 5xFAD mice (mimics plaque accumulation) and humans. ²¹ This study also included patient samples from various AD stages and Down syndrome, to investigate the DEGs while grouping the data by cortical layers and chromosomes, thus genes with conserved spatial patterns were identified (e.g., QKI upregulation in upper cortical layers). It also highlighted sex-specific differences, demonstrating ST’s potential to characterize disease forms, progression, and sex-specific patterns. ²¹

Similarly, sST provided novel insights into ALS progression by comprehensively assessing spinal cord samples from ALS SOD1-G93A (superoxide dismutase 1-glycine 93 to alanine mutation) mouse models (mimics motor neuron degeneration) and SOD1-WT (superoxide dismutase 1-wild type) mice at presymptomatic, onset, symptomatic, and end-stage phases, as well as human patients. ¹⁹ The study identified dysregulated signaling pathways during the presymptomatic phase, differential spatiotemporal expression patterns of astrocytes and microglia (e.g., pro-inflammation, phagocytosis), and a common dysregulation of sphingolipid signaling in both mouse models and human patients. ¹⁹

Moreover, analysis in MS research underscores the value of sST technologies for characterizing disease progression and its potential link to serum biomarkers. ¹ sST revealed spatiotemporal differences in neuroaxonal damage and neurofilament light chain (NfL) release in the experimental autoimmune encephalomyelitis (EAE) rodent model, comparing normal-appearing white matter (NAWM) and white matter lesions (WMLs) at early and late disease stages. ¹ The analysis showed a shifting pattern from WML to NAWM (centrifugal) of immune responses regulation and an opposite directional change from NAWM to WML (centripetal) of gliogenesis and humoral adaptive immune regulation. Early EAE was marked by widespread cytokine and chemokine responses, while late stages showed significant glial accumulation in WML. ¹ In addition, sST studies on MS patients and other CNS diseases revealed common glial dysregulation across conditions, with a unique pro-inflammatory microglial signature in MS (HIF1A+/SPP1+), highlighting the plasticity of microglial states even in tissue without acute demyelination. ¹⁷ Overall, these findings demonstrate ST’s potential to investigate disease mechanisms and morphological changes in apparently normal tissue, as evaluated by H&E and luxol fast blue stainings.

ST Assesses Molecular Mechanisms Underlying Vulnerability in Specific Brain Regions and Cell Types

In an AD study using patient tissue from the middle temporal gyrus and sST, unique gene signatures and pathways were identified that may contribute to early cellular and regional predisposition. ³ sST combined with pS129 α-synuclein (pSyn) IHC revealed that specific excitatory neurons are particularly vulnerable to Lewy pathology in both PD mouse models and patient samples, with conserved gene expression changes observed in Lewy body (pSyn+)-bearing neurons. ⁸ In line with this, high-resolution sST of the mouse brain identified markers of healthy and aging dopaminergic neurons and other cell types, revealing Parkinson’s-associated pathways and potential therapeutic targets. ¹⁵ Similarly, high-resolution sST was used to map ALS-associated genes in the primary motor cortex of a patient and controls, revealing layer 5 as the source of vulnerability, where pyramidal neurons are susceptible to cell death and have early TDP-43 pathology in ALS. ⁹ Characterizing these vulnerable regions and cell types may reveal pathways amenable to therapeutic interventions.

ST Supports Target Discovery by Enabling the Mapping of Single-Cell Data to Tissue

sST has mapped CNS-homing immune cells from the blood of MS patients to postmortem brain tissue, linking these cells to white matter demyelination and highlighting them as potential therapeutic targets. ¹³ Integration of scRNA-seq data with the sST analysis can be valuable even with high-resolution sST technologies, complementing the ST data with cell type-specific gene signatures. Deconvolution techniques have proven useful in various contexts, such as resolving the cellular composition of heterogeneous plaque niches in AD, ¹⁸ analyzing the vulnerable layer 5 of the motor cortex in ALS, ⁹ and distinguishing pSyn+/pSyn– cerebral cortex segments in PD. ⁸ Integrating sST with sc/snRNA-seq data sets, whether generated within the same study from the same patients or mice obtained from publicly available studies involving different samples, has demonstrated the complementary nature of these technologies.

ST Supports the Identification of Novel Targets by Enabling the Study of Receptor-Ligand Interactions

A study on brain tissue from progressive MS patients used sST data to computationally predict receptor-ligand interactions and link them to gene set modules representing biological processes. This analysis identified receptors like GPR37L1, TYRO3, SIRPA, and FGFR3 as potential therapeutic targets, strongly correlated with modules, such as synapse assembly, neurotransmitter secretion, myelination, and myeloid cell differentiation. ¹⁴ In addition, iST and sST were used to study microglia-astrocyte crosstalk in the plaque niche in the hippocampus, revealing disrupted astrocytic signaling in response to increasing microglia density. ¹⁸

Besides supporting the analysis of human or animal tissue, ST can be used with complex CNS ²⁷ in vitro organoid models, ²⁸ and to associate cell types and signatures with morphological features. ²⁴

In conclusion, ST technologies are transforming CNS research by integrating spatial and gene expression data, revealing unique transcriptional patterns across brain regions. This advancement enhances our understanding of CNS disorders, including AD, ALS, MS, and PD, and improves the characterization of critical mouse models used to study these conditions. Although translating findings from mice to humans remains challenging, ST, combined with organoid systems, offers a promising bridge to overcome this gap and advance drug development. ST technologies inherently focus on the intermediate stage of the central dogma of molecular biology (DNA to RNA to protein), which limits their scope. Therefore, we emphasize the importance of integrating multimodal technologies in future research to provide a more comprehensive understanding. This includes high-plex (~30-40) protein marker detection with oligo-conjugated antibodies for cell type detection, or mass spectrometry imaging integration for more exploratory purposes. While ST has great potential to advance CNS research, barriers, such as high costs, the need for sophisticated instruments (e.g., the Xenium platform), bioinformatics expertise, and limited automation hinder broader accessibility. As ST technologies continue to evolve, they hold robust potential for delivering precise insights into disease mechanisms and progression, as well as for investigating compound-associated neurotoxicity, translating from animal data to humans, and ultimately paving the way for more effective clinical interventions.