Advances in microsurgery and assisted reproduction for management of male infertility.

Текст сохранён. Часть иллюстраций пока доступна только в первоисточнике; загрузка будет проверена повторно.

Introduction

Alzheimer’s disease (AD) is a progressive age-dependent dementia disorder that is associated with the accumulation of extracellular β-amyloid (Aβ) plaques and intracellular neurofibrillary tangles (NFTs) comprising hyperphosphorylated tau [ 1 , 2 ]. The accumulation of Aβ and tau in the human brain follow distinctive spatial and temporal characteristics during aging and AD onset, where Aβ typically accumulates within the precuneus and cortical regions within the brain at least a decade or more prior to the manifestation of clinical symptoms [ 3 ]. Tau accumulation, however, correlates more closely with cognitive impairment during AD onset [ 4 , 5 ]. Interestingly, tau pathology can appear in brain regions such as the entorhinal cortex and temporal lobe in normal aging in the absence of cognitive effects, and is a primary age-related tauopathy (PART) that is largely independent of memory impairment and Aβ pathology [ 6 ]. Cumulative pathological evidence indicates that cognitive decline correlates with the dispersion of tau pathology from the entorhinal cortex into neocortical regions [ 7 , 8 ]. This suggests that the spatial distribution and temporal sequence of pathological progression during AD onset embody distinct features that could indicate stage, severity and cellular states in AD. In line with this, Braak staging which has been used to define the clinical progression of AD pathology is based on the dispersion of neurofibrillary tau tangles from the transentorhinal region to limbic regions and the hippocampus at mid-stages, and spread to neocortical regions at late stages [ 9 ]. Because significant advanced-stage AD pathology is observed in a proportion of cognitively normal individuals [ 10 , 11 ], this has spawned a growing concept in the field based on the idea that some individuals may be resistant to Aβ and/or tau toxicity [ 12 ]. Although specific mechanisms underlying how individuals with normal cognition can overcome AD-associated proteotoxicity are still unclear, concepts such as “cognitive reserve” describe enhanced brain function and efficiency which may enable individuals to sustain cognitive function under pathological conditions [ 13 ]. Other concepts include “brain reserve” or “brain resilience” where enhanced anatomical features related to synaptic function or circuitry can confer resistance to degeneration [ 14 ]. Thus, in addition to pathology-related features that may define clinical decline in AD, unique characteristics may be seen in cognitively normal individuals with AD pathology potentially due to resilience mechanisms.

One central question in AD is how progressive accumulation of Aβ and tau pathology affects cellular behavior and function in AD brain. Early bulk RNA-seq studies have defined changes in transcriptomic profiles in AD and healthy brain from varying regions and progressive pathological states [ 15 ]. Single-cell RNA-seq (scRNA-seq; used here to collectively refer to both single-cell and single-nucleus RNA sequencing unless otherwise specified) studies have also characterized transcriptomic changes in various cell types in varying brain regions [ 16 – 22 ]. While bulk RNA-seq [ 23 ] and scRNA-seq methods [ 24 , 25 ] have seen considerable technological and methodological development and optimization in AD/neurodegeneration research during the last few years [ 26 ], these methods rely on coupling pre-characterization of pathological features with transcriptomic analysis. With the development and use of spatial transcriptomic technologies to characterize changes in healthy and AD brain tissue, cellular expression profiles can be established in direct relation to pathological neurodegenerative landmarks in brain. Here we discuss the implementation, development and challenges related to spatial transcriptomic analysis in AD.

An overview of spatial transcriptomics in the current era

Spatial transcriptomics is used to profile mRNA expression in a spatial context, either at small-bulk or single-cell resolution. Use of hybridization methods to visualize DNA-RNA hybrids in situ with radioactive RNA dates back to the late 1960 s [ 27 , 28 ]. Thanks to the commercialization of various spatial transcriptomics platforms, a surge in implementation of current-era spatial transcriptomics is seen ~ 2020 [ 29 ]. Based on how spatial information is preserved, current-era spatial transcriptomic methods fall into five categories including: (i) methods involving tissue microdissection or region of interest (ROI) selection, (ii) methods that preserve relative local positions followed by de novo construction of global spatial positioning, (iii) methods using spatial barcoded-arrays with next-generation sequencing (NGS), (iv) in situ hybridization (ISH) and (v) in situ sequencing (ISS) methods [ 29 ] (Table 1 ). Many reviews provide in-depth summaries describing the development of spatial transcriptomic technologies [ 29 – 35 ] as well as analysis [ 33 , 35 – 38 ]. Here, we focus on the latter three categories of current-era spatial transcriptomic methods (Fig. 1 , Table 1 ), which can profile a large number of spots or cells with either high gene throughput or high spatial resolution, and discuss their implementation in characterizing AD brain.

Current-era array-based and imaging-based spatial transcriptomic technologies

Spatial barcoded-arrays with NGS, or array-based spatial transcriptomic approaches, capture mRNA in spatially-divided spots, beads or bins using spatially barcoded arrays, and are well-suited for transcriptome-wide profiling (Fig. 1 , Table 1 ). Each spot typically captures multiple cells, and therefore does not produce profiles at single-cell resolution. Currently, 10 × Genomics Visium, developed based on an early array-based spatial transcriptomics method [ 39 ], is the most widely used commercial array-based platform which offers whole-transcriptomic analysis of spots of 55 μm in diameter, with each spot typically containing 1–10 cells. In 2024, 10 × Genomics released Visium HD [ 40 ] which offers increased resolution (2 μm × 2 μm barcoded squares). In addition to Visium, other representative barcode-based or array-based spatial transcriptomic approaches include Slide-seq and the upgraded version, Slide-seqV2 [ 41 , 42 ], “deterministic barcoding in tissue for spatial omics sequencing” (DBiT-seq) [ 43 ] and “SpaTial EnhancedResolution Omics-sequencing” (Stereo-seq) [ 44 ]. Slide-seq uses bead barcodes to capture whole-transcriptome with a spatial resolution of around 10 μm, and its detection efficiency was further improved in Slide-seqV2 with sensitivity approaching droplet-based single-cell RNA-seq methods. Instead of using a pre-patterned array, DBiT-seq generates spatial barcodes directly on tissue via microfluidic channels, and can profile the whole-transcriptome with a spatial resolution from 50 μm up to 10 μm per pixel. In Stereo-seq applications, the slide is patterned with DNA nanoballs (DNBs), each carrying a unique spatial barcode. Individual DNBs are around 220 nm in diameter, and spot center-to-center distance is around 500 to 715 nm, offering a cellular/subcellular resolution of around 500 nm. Due to the low gene detection rate in each DNB, Stereo-seq data is typically grouped into bins (e.g. Bin 50 aggregates every 50 × 50 DNBs yielding a bin size of 25 μm × 25 μm). In general, current array-based spatial transcriptomic methods can capture large tissue areas and enable whole-transcriptome profiling but typically measure transcripts from multiple cells per spatial spot. Although several high-resolution array-based platforms (e.g., Visium HD and Stereo-seq) have improved spatial resolution toward single-cell or subcellular levels, these platforms are also limited with respect to detection sensitivity (Table 2 ).

ISH and ISS are imaging-based spatial transcriptomic approaches that detect transcripts in situ and can achieve subcellular resolution (Fig. 1 , Table 1 ). ISH methods detect target sequences by hybridization of complementary fluorescent probes, and ISS directly reads transcript sequences in situ (Fig. 1 ). Both ISH and ISS approaches use high-resolution imaging to spatially locate transcripts, together with additional staining to identify cell or nucleus boundaries. Single-cell gene expression profiles are reconstructed typically by assigning transcripts to cells based on their boundaries. Popular commercial imaging-based spatial transcriptomic platforms include CosMx from NanoString, Multiplexed error-robust FISH (MERFISH) [ 45 ] on the MERSCOPE platform from Vizgen, and Xenium from 10 × Genomics. CosMx, MERFISH and Xenium are ISH-based methods. CosMx can achieve subcellular resolution with a pixel-edge as small as 120 nm, and offers pre-designed gene panels of 1,000–6,000 genes, or even whole transcriptome panels as demonstrated in prototype. MERFISH on the MERSCOPE platform can achieve a pixel-edge length of 100 nm and offers gene panels of up to 1000 genes. Xenium also achieves subcellular resolution with a pixel-edge ~ 200 nm wide, and supports gene panels of up to 5,000 genes. Other representative imaging-based spatial transcriptomic approaches include ISH-based seqFISH [ 46 ] and seqFISH + [ 47 ], and ISS-based STARmap [ 48 ] platforms. In addition to custom probe design, some platforms provide pre-designed gene panels for specific applications, such as mouse neuroscience gene panel offered by CosMx and MERSCOPE. Although imaging-based spatial transcriptomic approaches can characterize cell-level gene expression profiles through the detection of transcripts at subcellular resolution, analysis using these methods is often limited to panels comprising several hundreds to thousands of genes. Capacity to capture profiles approaching the whole-transcriptome has been demonstrated for certain imaging-based spatial transcriptomic approaches [ 47 , 49 ], however, their popularity is limited due to both technical and practical concerns including increased error rates due to optical crowding, increased probe-design complexity and costs. Additional challenges include prolonged experimental run time due to limited imaging throughput, and increased computational complexity of image preprocessing (Table 2 ) [ 50 , 51 ]. Further limitations and considerations with array-based and imaging-based spatial transcriptomic technologies are discussed below (“ Limitations, challenges and future directions with spatial transcriptomics ”).

Use and considerations for spatial transcriptomics in AD research

Spatial transcriptomics has experienced a recent boost in prevalence in AD research in both human AD brain and mouse AD models (Table 3 ), and can be used in a number of analytical applications (Fig. 2 ). Emergence of AD-related research using current-era spatial transcriptomic technologies from ~ 2020 continues to grow, thanks to the accessibility of commercialized spatial transcriptomic platforms. Although the total number of studies in AD using spatial transcriptomics is still limited, the field is seeing a shift from multicellular to subcellular-resolution spatial transcriptomic methods. This reflects expanding accessibility to various spatial transcriptomic platforms combined with the necessity to investigate AD pathobiology at the single-cell level.

Although substantial efforts are driving spatial transcriptomic technologies toward both subcellular resolution and whole-transcriptomic profiling, current well-established spatial transcriptomic platforms still see a tradeoff between resolution and multiplexity. In general, spatial transcriptomic platforms achieving subcellular resolution feature limited multiplexity of up to several thousand genes, which limits the depth of downstream analysis. Conversely, whole-transcriptomic spatial transcriptomic platforms generally lack subcellular resolution, and do not support analysis at single-cell resolution (Table 2 ). To address these limitations, scRNA-seq is often used concurrently to impute whole-transcriptomic profiles of spatial transcriptomic datasets at cellular resolution; alternately, spatial organization of cells profiled by scRNA-seq can be computationally inferred by mapping to spatial transcriptomic profiles based on similarities in gene expression. In this context, scRNA-seq and spatial transcriptomic datasets generated from adjacent tissue sections are advantageous as they often share similar cell populations, tissue architecture, pathology deposition and cellular states. Thus, scRNA-seq and spatial transcriptomic data were frequently combined to additively enable in-depth whole transcriptome analysis at single cell resolution with spatial context [ 52 – 57 ].

Given that brain tissue preservation typically involves formalin fixation and paraffin embedding (FFPE), compatibility of spatial transcriptomic methods with FFPE tissues is critically important (Table 1 ). Unlike mouse brain tissues which are typically fresh-frozen with relatively high RNA quality, human brain samples are commonly processed as FFPE blocks and require specialized library preparation procedures that accommodate RNA crosslinking and fragmentation resulting from formalin fixation, paraffin embedding and long-term storage. Therefore, compatibility of spatial transcriptomic procedures with FFPE samples is essential. The Visium platform employs a probe-based chemistry pipeline that uses specifically-designed probes to capture degraded RNA fragments; however, detection is limited to protein-coding genes. Similarly, MERFISH incorporates additional steps for FFPE samples including deparaffinization, decrosslinking, and other steps to preserve fragmented RNAs and improve probe hybridization to RNA fragments.

Compatibility of histological staining on the same tissue section is another important consideration for spatial transcriptomic experiments, particularly in AD studies as it enables direct analysis of transcriptional changes within pathological microenvironment (Table 1 ). Immunofluorescence (IF) is widely used for pathology staining due to its high specificity and resolution. And it is compatible with most current-era spatial transcriptomic platforms. For ROI-based platforms such as the GeoMx Digital Spatial Profiler (DSP), IF staining is integrated into its workflow and performed before transcript profiling to guide ROI selection. For array-based platforms including Visium, DBiT-seq and Stereo-seq, IF staining is performed before transcript acquisition. For imaging-based platforms including CosMx and Xenium, IF staining is typically performed after transcriptomic profiling, whereas MERFISH allows IF staining either before or after transcriptomic profiling (Table 1 ).

In addition to resolution, multiplexity, FFPE compatibility and pathology-staining compatibility, spatial transcriptomic platform selection also involves other considerations including tissue coverage, scalability, throughput, and cost, as well as software to support raw data processing and preliminary data quality assessment [ 58 , 59 ].

Processing and analyzing spatial transcriptomics datasets in AD

Spatial transcriptomics and scRNA-seq share many analytical computation modules, and routine pipelines for both data types have been extensively reviewed [ 33 , 35 – 38 , 60 – 63 ]. This section briefly describes standard spatial transcriptomic data processing procedures with a particular focus on adaptations required for brain tissues in AD research.

Cell segmentation in spatial transcriptomic datasets

Analysis of gene expression profiles at the cellular level in spatial transcriptomic datasets requires preparatory steps to define and annotate cell boundaries through cell segmentation. In this review, cell boundary is used as a generic term referring to either cell boundary or nuclear boundary. Based on cell boundaries, spatially-defined transcripts are then assigned to cells, and aggregated to generate gene expression profiles for each cell. Cell segmentation is typically based on morphological images [ 64 – 68 ], transcript distribution [ 69 , 70 ], or both [ 70 – 73 ].

Cell segmentation in brain is particularly challenging due to high cell density, heterogeneous cell morphology, and substantial background signals from neuropil and AD pathologies. The brain comprises densely-packed cells, resulting in overlapping cell bodies which can hinder accurate cell segmentation. Computational methods can be used to identify inaccuracies due to overlapping cells based on gene expression profiles [ 74 ] or imaging features [ 75 ]. Given the substantial morphological diversity of brain cell types, use of a single model to perform cell segmentation across all cell types may be challenging; use of specialized segmentation models that account for morphological differences across brain cell types may improve segmentation accuracy. Additionally, excessive background signals from AD pathology, as well as neuropils comprising a dense network of axons, dendrites, synapses and glial processes interspersed between cell bodies can also complicate cell segmentation. These challenges may be further exacerbated in regions with high pathological burden, particularly in late-stage AD brain tissue; for example, DAPI which primarily stains nuclei, was found to also stain Aβ plaques [ 76 ]. To address issues related to background noise, staining strategies can be optimized to reduce false positive signals and improve segmentation accuracy [ 77 , 78 ]. For example, a combinatorial staining strategy (e.g., DAPI + polyT) can improve cell segmentation accuracy compared to single staining alone [ 77 ]. In addition, existing machine learning models can be further refined using a “human-in-the-loop” approach where user-annotated or adjusted cell boundaries can be incorporated to improve segmentation performance [ 67 , 68 ].

Data processing in spatial transcriptomics

As spatial transcriptomic and scRNA-seq data both comprise cell/spot × gene matrices, they share many core data processing steps, including general data quality control, normalization, clustering and cell type annotation. Toolkits or libraries for spatial transcriptomic data analysis include Seurat [ 79 ] and Scanpy [ 80 ], which were originally designed primarily for scRNA-seq analysis and later extended to support spatial transcriptomic datasets. Other analysis toolkits specifically for spatial transcriptomic data analysis are also available [ 81 – 83 ]. This section discusses data processing procedures unique to spatial transcriptomics and used in studies related to AD.

Similar to doublet artifacts in scRNA-seq datasets where two cells are sequenced as a single cell, inaccurate segmentation of neighboring or overlapping cells can compromise downstream spatial transcriptomics analyses. These artifacts can be inferred using gene expression-based or image-based approaches. In addition, data pertaining to cell size acquired during cell segmentation in spatial transcriptomics can be used to remove abnormally small or irregularly large and potentially overlapping cells due to inaccurate segmentation. Filtering such artifacts improves the accuracy of downstream analyses.

To normalize transcript levels in spatial transcriptomic data, common strategies include global scaling using the total number of transcripts per cell, use of cell-specific scaling factors estimated by pooling similar cells together to handle heterogeneous dropouts [ 84 ], applying variance-stabilizing transformations [ 85 ], and using cell area or volume as scaling factors [ 86 ]. Using cell area or volume was found to be more suitable for spatial transcriptomic datasets with skewed or limited gene panels since overall transcript abundance cannot be accurately estimated from a small number of genes [ 87 ].

Clustering and annotation of spatial transcriptomic datasets can employ different approaches depending on both data resolution and specific endpoints required. Clustering based on gene expression profiles is often used for identifying cell types and subtypes by examining top differentially expressed genes (DEG) within each cluster or through marker gene expression. Joint clustering of gene expression profiles and spatial coordinates can be used to identify spatial tissue domains or anatomical structures [ 88 – 91 ]. Annotation of cells at single-cell resolution in spatial transcriptomics can be inferred by knowledge-guided approaches using cell marker genes, or by mapping annotated reference data to spatial transcriptomics data to infer cell types which typically use all available genes instead of only depending on marker genes [ 92 , 93 ]. For multicellular spatial transcriptomic datasets, cell type composition in spots/bins/beads can be estimated using reference-based approaches [ 94 , 95 ], as well as reference-free methods [ 96 ]. scRNA-seq datasets from large atlas databases, such as mouse brain datasets from Allen brain cell atlas [ 86 ] and human brain cell datasets from the BRAIN Initiative Cell Census Network (BICCN) [ 97 ], provide valuable references to annotate brain cell types in spatial transcriptomics.

Anatomical annotation of brain regions or subregions

Spatial transcriptomic profiles in brain can be annotated into anatomical regions or subregions for region-specific analysis and comparisons across different regions [ 52 , 76 , 98 – 109 ] (Table 3 ). To this end, brain slices are often aligned to a common coordinate framework (CCF) of an anatomy-defined high-resolution 3D brain map, such as the Allen Mouse Brain CCF [ 110 ]. CCF alignment typically uses histological images such as Hematoxylin and Eosin (H&E)-stained images, anatomical landmarks, and cytoarchitectural features such as cell morphology and density. In addition, anatomical region-specific molecular features (e.g., gene expression and protein markers) can be used to supplement CCF mapping or guide annotation of an anatomical region. For example, cortical layers can be annotated using layer-specific marker genes, together with H&E and IF staining without using a defined CCF [ 100 ]. Both CCF-based and region-specific marker gene-based approaches depend heavily on prior knowledge pertaining to anatomical brain characteristics and region-specific features. De novo approaches have also been used to segment anatomical regions using only transcriptomic expression profiles and their spatial coordinates. For example, joint clustering of transcriptomic profiles and spatial coordinates using BayesSpace [ 88 ] can effectively annotate cortical layers and white matter (WM) with high consistency [ 52 ]. Anatomical annotation of brain regions in spatial transcriptomics enables characterization of region-specific molecular, cellular, and pathological heterogeneity in AD brain.

CCI inference with spatial context

Various databases comprising curated ligand/receptor binding pairs such as CellPhoneDB [ 111 ], CellChatDB [ 112 ], LRIDB [ 113 ] and CITEdb [ 114 ] have been established. Cell–cell interaction (CCI) analysis is widely performed on scRNA-seq data, and potential CCIs are inferred solely based on enrichment of ligand or receptor gene expression [ 112 , 115 – 117 ]. The spatial context provided in spatial transcriptomic data can be used as additional information to infer CCIs [ 115 , 116 , 118 – 122 ] supported by spatial co-localization or potential direct physical contact which is essential for classical ligand-receptor signaling.

Pathology staining, segmentation and quantification

Recent advances in spatial transcriptomics have enabled discoveries directly linking spatial transcriptomic profiles with pathological features such as Aβ plaques and tau pathology [ 52 , 76 , 98 , 100 – 102 , 104 – 106 , 123 – 127 ]. AD pathology is usually stained and imaged concurrently on tissues used for spatial transcriptomic analysis, or stained and imaged on adjacent brain slices. Many spatial transcriptomic platforms such as Visium, DBiT-seq, Stereo-seq, CosMx, MERFISH and Xenium support pathology staining on the same tissue sections used for spatial transcriptomic data acquisition.

While the majority of published AD studies using spatial transcriptomics so far have characterized gene expression profiles in proximity to Aβ [ 52 , 76 , 98 , 102 , 104 – 106 , 123 , 125 ]; relatively fewer studies consider Aβ and tau pathology in combination [ 100 , 101 ]. Various antibodies have been used to characterize Aβ plaque deposition in spatial transcriptomic applications such as 6E10 [ 98 , 104 , 105 , 123 ], MOAB-2 [ 124 ], 4G8 [ 125 , 128 ], or Aβ 6F/3D (M08762) monoclonal antibodies [ 106 ]. Aβ plaques have also been stained using dye-based stains such as the X-34 Congo red derivative [ 101 , 102 ] or Amylo-Glo [ 52 ] in spatial transcriptomic studies. So far, most spatial transcriptomic studies characterizing tau pathology use the AT8 phosphorylated tau (ptau) antibody as an indicator of pathological tau phosphorylation [ 100 , 101 ]. The AT8 antibody detects pS202/T205 [ 129 ], and pS208 [ 130 ] tau phosphoepitopes which are commonly seen in tau paired-helical filaments. Strong AT8 ptau signals are observed in AD brain, with relatively less reactivity in healthy adult brain by western blotting [ 131 – 133 ]. Phosphoproteomic analysis indicates that phosphopeptides containing the AT8 pS202/T205 phosphoepitope can be detected in a majority of AD cases, as well as in some (~ 20–40%) control cases, potentially indicating that AT8 ptau may be phosphorylated in prodromal phases of onset [ 134 ]. This suggests that variability in AT8 ptau detection may arise between individuals and cohorts, as well as detection methods.

While identification and quantification of Aβ and tau pathology in stained images previously involved manual annotation methods [ 100 , 104 ], computational methods are now widely used. Custom pipelines using image analysis software are widely implemented in spatial transcriptomics [ 52 , 98 , 101 , 105 , 125 ]. For example, histogram-based threshold methods have been implemented in ImageJ to generate binary annotations for Aβ plaques [ 98 ]. Automated custom imaging analysis protocols to segment Aβ have also been applied using the NIS-Elements software package from Nikon [ 52 ], and computational pipelines have been used to quantify ptau intensity using Fiji [ 101 ]. In addition to the integrated use of image analysis software, computational methods utilizing image analysis algorithms [ 106 ] and machine learning methods [ 124 ] can also be implemented.

Aβ and tau pathology significantly differ with respect to spatial distribution, biological characteristics and morphological features. Thus, differing methods are required for identifying, annotating and quantifying Aβ and tau pathology. Given that Aβ plaques form compact extracellular structures, most methods segment individual Aβ plaques [ 52 , 76 , 100 , 101 , 104 – 106 , 124 , 125 ]. Alternatively, some studies annotate Aβ plaques using an overall intensity score within a small region, such as cell body area [ 98 , 102 ]. In contrast to Aβ, tau pathology is largely intracellular and manifests in neurons as small and irregular structures. Therefore, methods to annotate and quantify tau pathology is primarily based on staining signal intensity [ 100 , 101 , 106 ]. Spatial annotation and quantification of AD pathology can then be integrated with anatomically mapped spatial transcriptomic profiles to characterize pathology-related regional features, including regional pathology burden, transcriptomic heterogeneity across regions with pathology, local pathological microenvironment, and associated cell–cell crosstalk (Fig. 2 ).

Combining and integrating spatial transcriptomic and scRNA-seq datasets

Whole-transcriptome, matched scRNA-seq profiles are often used in parallel with spatial transcriptomics [ 52 , 53 , 104 , 106 , 125 ], thereby complementing and providing cross-validation between platforms in brain studies related to AD. scRNA-seq can serve differing roles in spatial transcriptomics depending on the resolution and multiplexity of the spatial transcriptomics dataset.

For high-resolution single-cell spatial transcriptomics datasets probed with a limited gene panel, scRNA-seq data can be used as a reference to design a probe panel which can effectively distinguish different cell populations [ 86 , 135 ]. In addition to guided selection of cell markers, scRNA-seq can be used to aid and optimize gene selection in probing specific biological pathways or processes [ 76 ]. In these instances, spatial transcriptomics is primarily used to investigate or validate specific hypotheses derived from scRNA-seq data within a spatially resolved context. scRNA-seq data is also widely used as a reference to annotate cell types in spatial transcriptomics [ 106 ]. In addition, scRNA-seq profiles can be mapped to spatial transcriptomics datasets to impute whole-transcriptomic gene expression for each cell, or alternatively, spatial organization of cells profiled by scRNA-seq can be inferred using spatial transcriptomic data [ 86 , 135 ].

For low-resolution multi-cellular spatial transcriptomic datasets, scRNA-seq profiles can be used to infer cell type composition for “cell type-aware” analyses involving cell abundance and gene expression changes [ 94 , 95 ]. Multicellular spatial transcriptomics can also be used to infer spatial organization of cells in scRNA-seq datasets [ 52 ]. For example, snRNA-seq profiles can be mapped to multicellular spatial transcriptomic datasets by predicting spatial coordinates of each snRNA-seq cell with CellTrek [ 136 ], thereby consolidating whole-transcriptomic spatial profiles at single-cell resolution [ 52 ].

Biological insights from spatial transcriptomics in AD brain

Spatial transcriptomics studies have enabled the characterization of healthy and diseased cellular states within intact tissue, and have provided insight into cellular gene expression profiles in relation to brain region, pathology, and neighboring cells. Here, we discuss recent work in AD using spatial transcriptomics (Table 3 ) focusing on current-era high-resolution or high-throughput platforms, and highlight their biological insights in AD brain. We apologize for the excellent studies in AD or AD-related dementia utilizing spatial transcriptomic methods outside the scope of this review.

Regional vulnerability and resilience

A defining feature of AD is the selective vulnerability of specific brain regions and cell populations to aging-related stressors. Neurofibrillary tau pathology follows a highly stereotyped progression pattern defined by Braak staging, initiating in transentorhinal and entorhinal regions before spreading into the hippocampus and neocortex [ 9 ], whereas Aβ deposition originates in the precuneus and other cortical regions [ 3 , 137 ]. Notably, substantial AD pathology can also be observed in cognitively normal individuals, raising important questions about mechanisms of resilience and resistance to neurodegeneration. However, molecular features that render specific brain regions and cell populations vulnerable or resistant to degeneration remain unclear. Given that spatial transcriptomics preserves anatomical organization while resolving regional and cell type-specific transcriptional states in relation to pathology, use of spatial transcriptomics is particularly well suited to addressing these questions.

Spatial transcriptomic studies have revealed substantial regional heterogeneity in both pathological burden and molecular responses across AD brain regions. In human hippocampus, increased deposition of Aβ plaques was observed in the CA1 region, coinciding with a reduction in CA1 neurons, as well as enrichment of microglia and astrocytes in proximity to Aβ plaques [ 106 , 137 ]. Further Stereo-seq analysis coupled with snRNA-seq analysis revealed increased expression of genes related to synaptic pruning in the fimbria of AD, as well as increase in genes related to energy generation in the CA region [ 106 ]. Similarly in mouse hippocampus, LPL (lipoprotein lipase), a microglia disease-associated marker [ 138 ] that facilitates lipid and Aβ uptake and degradation [ 139 ], featured enhanced up-regulation in the CA1-3 subregions in 3xAD (3xTg AD, APP K670N/M671L, MAPT P301L, PS1 M146V) or 3xPB (3xTg AD/Polβ+/-) AD mouse models [ 99 ]. In frontal cortex and posterior cingulate cortex of the human brain, spatial transcriptomic analysis revealed downregulation of genes modules associated with neurotransmission, neurodevelopment and amyloid-β formation exclusively in white matter and cortical L3/L4 layers in early-stage AD [ 52 ].

Region- and cell type-specific depletion has also been described in AD brain, implicating selective vulnerability in neurons with disease progression and pathological burden. In human brain, neuronal subpopulations in upper cortical layers within the middle temporal gyrus (MTG) and Brodmann area 9 (A9) were found to be depleted with progressive onset of AD pathology [ 53 ]. A vulnerable neuronal Sst + subpopulation, localized primarily to supergranular layer II and III, featured downregulation of specific kinases including tyrosine kinase and calcium 2+ /calmodulin-dependent kinase families, and E3 ubiquitin ligases in early stages of pathological onset compared to unaffected Sst + subpopulations [ 53 ]. In human temporal cortex, Parvalbumin ( PVALB ) neurons were depleted in gray matter in late Braak stage [ 54 ]. Other studies have noted subregion-specific vulnerable cell types in human hippocampus [ 106 ]. In particular, excitatory neurons were depleted in CA1, CA2 and DG, but not in CA3/4 subregions in AD hippocampus, suggesting resistance of CA3/4 excitatory neurons to degeneration during AD onset [ 106 ].

In contrast, selective resilience and molecular signatures associated with resilience features have also been characterized using spatial transcriptomics. In human brain, a resilient excitatory neuron population co-expressing RORB , CUX2 , and EYA4 was identified in layer 4 of primary visual cortex (BA17) [ 57 ]. Although Layer 4 is associated with amyloid plaque deposition, it is also associated with reduced tau pathology, and is considered to comprise neurons resistant to degeneration [ 57 ]. Layer 4 neurons feature upregulated expression of KCNIP4 , a voltage gated potassium channel–interacting protein associated with reduced excitotoxic hyperexcitability. Functional validation in an AD mouse model demonstrated that KCNIP4 upregulation suppresses activity-dependent genes (Arc, c-Fos), consistent with decreased neuronal excitability [ 57 ]. This suggests that high KCNIP4 expression attenuates neuronal firing and activity-dependent signaling, and potentially acts as a functional barrier limiting both intrinsic vulnerability and activity-dependent trans-synaptic propagation of tau pathology across connected circuits.

Together, these studies suggest that selective vulnerability and resilience in AD likely reflects interactions between intrinsic transcriptional programs, regional microenvironment, pathological burden, and circuit-level connectivity. Spatial transcriptomics has enabled AD research to progress beyond descriptive anatomical staging by identifying molecular programs associated with both vulnerability and resilience across various cell types and anatomically distinct brain regions.

Spatially-defined cellular responses to AD pathology

While regional and cell type-specific vulnerability and resilience provides an important framework for understanding selective degeneration in AD, gene expression changes in direct proximity to Aβ and tau pathology may represent more immediate cellular responses to pathological stress. Spatial transcriptomics has enabled direct characterization of transcriptional changes within pathological microenvironments, revealing localized cellular responses that are difficult to resolve using dissociative approaches.

Early studies used spatial transcriptomics [ 39 ] to characterize cell-specific changes associated with Aβ in humanized APP NL−G−F mouse models and identified 57 plaque-induced genes (PIGs) associated with the complement system, oxidative stress, lysosomes and inflammation [ 98 ]. Interestingly, further analysis of PIGs using ISS data [ 140 ] revealed that PIGs were largely derived from microglia, and to a smaller extent, astrocytes and other cell types [ 98 ]. These studies provided early evidence that Aβ plaques are associated with highly localized multicellular transcriptional responses.

Subsequent studies [ 101 , 105 , 123 , 125 ] have made significant inroads characterizing changes in various cell populations in relation to Aβ pathology across brain regions. The role of glia in AD pathogenesis, microglia and astrocytes in particular, has become a subject of interest in AD pathogenesis [ 141 , 142 ]. Microglia have an active role in sensing and clearing Aβ [ 143 ], and form a barrier around Aβ plaques, thereby limiting plaque diffusion [ 144 ]. Distinctive expression profiles comprising disease associated microglia (DAM) [ 138 ], or neurodegenerative microglia (MGnD) [ 145 ] signatures have been characterized in AD mouse models by snRNA-seq or bulk RNA-seq analysis in sorted microglia, which contrast from expression profiles in homeostatic microglia. Perhaps not surprisingly, spatial profiling methods have noted enrichment of microglia nearby plaques across brain regions in both human and mouse brains [ 76 , 101 , 105 , 106 ]. For example, microglia were enriched around plaques in CA1 of human hippocampus [ 106 ]. In TauPS2APP mouse brain cortex and hippocampus, spatial profiling using STARmap PLUS identified enrichment of DAM microglia in proximity to plaques [ 101 ]. And the homeostatic gene P2ry12 was found to be downregulated in Aβ-proximal microglia, reinforcing the notion that microglia are responsive to spatial distance to amyloid pathology [ 76 ]. Similar to microglia, astrocytes take up and clear Aβ [ 146 , 147 ], and are found in proximity to Aβ plaques in human [ 148 ] and AD mouse brain [ 149 ]. As disease-associated astrocytes (DAA) signatures have also been characterized in AD and other pathological mouse models [ 150 ], spatial profiling using STARmap PLUS analysis also revealed enrichment of DAA in TauPS2APP mouse cortex and hippocampus [ 101 ]. Interestingly, astrocyte proximity to Aβ plaques may vary between brain subregions; results from Stereo-seq analysis in human AD hippocampus revealed astrocyte enrichment nearby plaques in stratum lucidum, radiatum, and moleculare (SLRM) of the human hippocampus, whereas Aβ-associated astrocyte abundance was reduced in CA4 subregions [ 106 ]. In contrast to DAM and DAA, neurons overall were found to be depleted in proximity to Aβ plaques using STARmap PLUS in TauPS2APP AD mouse brain [ 101 ]. However, variations exist among neuron subtypes across regions. For example, an inhibitory neuron subtype, marked by Sst and Chodl genes, were found to have higher proportion nearby Aβ plaques in human cortical regions [ 151 ].

Beyond characterizing individual cell-type responses, spatial transcriptomics has enabled characterization of coordinated multicellular responses to Aβ plaques across cell types. For example, cells surrounding Aβ plaques featured DAM microglia within the inner core surrounding plaques, with DAA-like cells and oligodendrocyte precursors/oligodendrocytes localized to outer regions extending from Aβ plaques in TauPS2APP mouse brain [ 101 ]. DEG analysis in microglia, astrocytes, oligodendrocytes and neurons in proximity to plaques in TauPS2APP mouse brain revealed enrichment of genes involved in inflammation, gliogenesis, myelination and axon ensheathment [ 101 ].

Several studies reveal shared and distinct plaque-associated responses across varying genetic murine models of AD. In comparing humanized APOE3 or APOE4 knock-in mouse models crossed with 5xFAD animals, APOE4 /5xFAD animals featured exacerbated upregulation of DAM/MGnD gene signatures by scRNA-seq analysis; subcluster analysis of Visium spots revealed APOE4 /5xFAD enriched signatures featuring upregulation of reactive microglia signatures and genes related to microglia activation, synaptic pruning, neuronal apoptosis in APOE4 /5xFAD animals compared to APOE3 /5xFAD [ 102 ]. These studies highlight spatial sensitivity of microglia to pathological features such as Aβ plaques in AD pathology, which can be modulated by factors that enhance AD-onset and pathogenesis such as APOE4 . Characterization of 5xFAD mice carrying the AD-associated Trem2 R47H variant using MERFISH revealed region-specific cellular changes in proximity to Aβ, where DAM microglia in 5xFAD brain were higher in cortical layers V-VI compared to cortical layers II-III; 5xFAD/ Trem2 R47H also featured a slight decrease in DAM proportion compared to 5xFAD alone [ 76 ]. In human brain, triplication of APP on Chromosome 21 in DS individuals is thought to be causal to AD-like symptoms and early onset Aβ pathology [ 152 ]. Comparison of human AD and DS cortex by snRNA-seq and Visium spatial transcriptomic analysis revealed layer-specific changes in DEG profiles, and genes such as CLU and VIM were associated with regions enriched with Aβ plaques and fibrils [ 52 ]. Similarly, Clu and Vim were also found to be associated with amyloid in 5xFAD animals; other Aβ-associated genes such as NEFH, NEFM, ALDOC and MAFB are also found to show similar changes in 5xFAD and human AD/DS brain [ 52 ].

Transcriptional states in relation to tau pathology have also been defined using spatial transcriptomics; in contrast to Aβ and amyloid-related studies, fewer spatial transcriptomic studies have characterized profiles associated with tau. In TauPS2APP mouse hippocampus, ptau pathology is observed mainly in CA1 excitatory neurons and correlates with the local enrichment of oligodendrocytes [ 101 ]. In PS19 mouse brain, GeoMx DSP data analysis revealed region-specific transcriptomic changes in hippocampal CA3 subregion with up-regulation of genes involved in metabolic pathways before overt tau pathology [ 153 ]. DAM and DAA signatures emerge in parallel with the development of ptau pathology across multiple brain regions [ 153 ]. In human AD brain, analysis of the MTG region using the Visium platform identified gene signature profiles in regions enriched with AT8 ptau [ 100 ]. Visium spots containing AT8 ptau tangles and neuropil threads appeared to be enriched in upregulated genes/pathways associated with synaptic transmission, as well as downregulation of genes associated with apoptosis [ 100 ]. Further RNAscope analysis of human control and AD brain within the MTG region revealed elevation of CSRP1 and GLUL in layer V astrocytes in proximity to neurofibrillary tangles (NFTs) and AT8 + ptau neuropil threads, and upregulation in microglia C1QB and SPP1 , as well as SNCG neurons in layer II/III in proximity to both Aβ plaques and NFT/AT8 ptau pathology [ 100 ]. Recent characterization of NFT-bearing (AT8⁺) CA1 hippocampal neurons in late stage (Braak V–VI) AD using spatial transcriptomics [ 154 ] reported coordinated upregulation of synaptic genes ( VAMP2 , SYT1 ), molecular chaperones ( YWHAG , YWHAH , YWHAZ ), and genes associated with apoptosis, endoplasmic reticulum stress, and stress granule formation ( IFITM2 , CCDC88B , PTMA , G3BP2 , and ATF4 ). These findings are consistent with prior observations from NFT-bearing (AT8⁺) neocortical neurons in BA9 late-stage AD (Braak VI) [ 18 ]. Notably, a similar transcriptional profile is observed in PART, another 3R/4R secondary tauopathy characterized by minimal Aβ deposits, suggesting that tau-driven transcriptional changes in hippocampal CA1 neurons may share transcriptional signatures with regions associated with low Aβ plaque deposition [ 154 ]. In contrast, spatial transcriptomics analysis of NFT‑rich (AT8⁺) areas in primary motor cortex from progressive supranuclear palsy (PSP, 4R primary tauopathy) brain, revealed a distinct transcriptional landscape marked by downregulation of genes involved in insulin signaling, phagocytosis, cellular stress, neutrophil degranulation, neurotransmitter release, oxidative phosphorylation, and NF‑κB signaling [ 155 ].

Emerging evidence suggests that cellular responses may be shaped not only by proximity to Aβ or tau individually, but in combination. Unique upregulated and downregulated genes and pathways have been characterized in human AD brain by spatial spots annotated for Aβ or tau pathology and Aβ/tau in combination using the Visium platform [ 100 ]. Regions with only tau pathology feature up-regulation of genes and pathways associated with synaptic transmission, and downregulation of genes and pathways associated with “negative regulation of apoptosis”. Interestingly, further co-expression analysis revealed that certain gene modules changed gradually as the distance from either Aβ or tau pathology increases, but similar gradual changes were not observed when considering both Aβ and tau in combination. These results suggest that combined analysis of Aβ and tau pathologies within the same tissue sections is important for distinguishing pathology-specific responses from combinatorial effects and to understand how overlapping pathological landscapes drive cellular dysfunction during AD progression.

Spatially-resolved cell–cell interactions

In addition to cell-intrinsic responses to pathology, spatial transcriptomics has also allowed us to define how AD pathology reshapes communication or interaction profiles between neighboring cells. For example, knockdown of the dual-specificity phosphatase, Inpp5d (SHIP1) with nucleotide variants previously implicated in AD risk in PSAPP murine AD microglia increased plaque-associated microglia, and aggravated Aβ plaque burden [ 123 ]. Using the NICHES method [ 118 ] which can infer CCIs with spatial information, alterations were found in receptor-ligand pairs including genes involved in phagocytosis and lysosome activity (including C1q-Lrp1 , C3-Lrp1, Timp1-Cd63, Pros1-Axl, and Ly86-Cd180) in cell clusters in close proximity to plaques and altered by Inpp5d depletion [ 123 ]. In human AD dorsal-lateral prefrontal cortex, downregulation of ligand-receptor pairs related to synaptic pathways was observed in regions enriched with glia [ 125 ]. Moreover, upregulation of ligand-receptor pairs related to immune response (CLU-TREM2, PLXNA1-TREM2, APOE4-LDLR and GFAP-APLNR) as well as the extracellular matrix and vesicle signaling was observed in regions enriched with Aβ [ 125 ]. Interestingly, transcriptomic changes indicative of synaptic degeneration were enriched in low Aβ regions, as well as dramatic changes in ligand-receptor pairs in low Aβ regions, suggesting that neurodegenerative effects are associated with areas with fewer Aβ plaques, and regions enriched with Aβ may be associated with late degeneration and cell death [ 125 ]. Targeted CCIs between reactive microglial and DAA-like astrocyte subclusters in Stereo-seq datasets from human AD hippocampus using StereoSiTE [ 119 ] revealed enrichment of ligand-receptor pairs involved in neuroinflammatory response such as SPP1-ITGB1 and MIF-CD44 in fimbria [ 106 ]. In APP NL−G−F mouse brain, CCIs inferred using the CellPhoneDB method [ 117 ] on CosMx datasets revealed stronger receptor-ligand PIGs in proximity to plaques [ 105 ]. In particular, homotypic microglia-microglia DAM gene interactions such as Csf1-Csf1r and Cd44-Tyrobp (DAP12), as well as DAA astrocyte-microglia DAM interactions such as Apoe-Trem2, or Clu-Apoe were found to be enhanced in proximity to plaques [ 105 ]. In another study using APP NL−G−F mouse brain, weighted gene co-expression network analysis [ 156 ] of multicellular spatial transcriptomic profiles also implicated that PIGs including secreted microglia DAM components such as C1q and Apoe , as well as DAM receptors such as Trem2 and Tyrobp correlate with astrocyte activation profiles ( Gfap, C4, Clu, Prdx6, Cst3, Serpina3 ) in the proximity of plaques [ 98 ].

Inferred CCI profiles can be stratified into regions or subregions. For example, comparison of cell–cell communication networks inferred by CellChat [ 112 ] in human DS and control brain revealed changes in CCIs in upper/lower cortical and white matter regions [ 52 ]. Interestingly, NECTIN signaling, a component required for synaptic maintenance [ 157 – 159 ] and implicated in genetic AD risk [ 160 , 161 ] was found to be downregulated in DS and late-stage AD [ 52 ]. In transgenic R955-hTau mouse brain, BDNF neurotrophic signaling from DG in hippocampal regions was decreased compared to WT while neurotrophic signaling from cortical layer L2/3 to L4-6 was increased, suggesting a shift in neurotrophic signaling largely in the DG region in WT hippocampus to L2/3 cortex in R955-hTau animals [ 55 ].

Overall, use of spatial transcriptomics in AD research has uncovered important features in cell–cell communication networks that are impacted by pathology and vary across brain regions. However, CCI analysis is limited by the resolution and multiplexity of current spatial transcriptomic technologies. While imaging-based spatial transcriptomic technologies provide subcellular resolution for cell type-specific analysis, the limited gene panels constrain the number of CCIs that can be detected. Without full transcriptomic coverage, the interpretation of incomplete CCI networks can be potentially biased. Comprehensive characterization of CCI networks is in principle feasible using array-based spatial transcriptomic technologies with whole-transcriptomic data, but their multi-cellular resolution complicates interpretation due to limited cell-type specificity and reduced spatial precision (Table 2 ). A comprehensive and reliable characterization of CCI networks relies on the advancement of the spatial transcriptomics technologies towards simultaneous subcellular resolution and whole-transcriptomics coverage. Moreover, existing CCI inferences primarily rely on mRNA co-expression and physical proximity, lacking direct validation of the physical contact or functional interaction. Fundamentally, mRNA expression does not directly characterize post-transcriptional regulation, protein translation, secretion or receptor availability. In addition, spatial transcriptomics profiles frozen brain tissue, whereas CCIs are dynamic and temporally ordered. These limitations highlight the need to integrate spatial transcriptomics with complementary modalities, such as spatial proteomics and live imaging, to validate inferred CCIs.

Spatial–temporal dynamics and disease progression

Although most spatial transcriptomic studies in AD rely on static tissue snapshots, emerging work suggests that molecular responses evolve dynamically across disease stages rather than remaining static. A major challenge in AD research is reconstructing how cellular states transition as pathology spreads across anatomically connected brain regions over time. Spatial transcriptomics is particularly well positioned to study this question because it preserves information on anatomical location, local pathology burden, and neighboring cellular states simultaneously.

Emerging evidence suggests that early disease stages are marked by well‑defined, cell‑type–specific transcriptional programs, whereas later stages are characterized by convergence of gene expression across cell types [ 162 ], reflecting widespread network disruption and roles for glial/neuronal communication [ 100 , 163 ] with attenuated CCIs in AD normally involved in Aβ clearance [ 163 ]. In addition, accumulating evidence from spatial transcriptomic studies indicates that gene expression in Alzheimer’s disease is dynamically regulated, rather than uniformly up‑ or downregulated over disease progression. Instead, individual genes follow distinct, stage‑dependent trajectories characterized by transient phases of up‑ and downregulation. For example, the neuroprotective gene PSAP is upregulated in prefrontal cortex during early and intermediate stages of AD but progressively downregulated in advanced stages [ 163 ]. Such nonlinear expression patterns may help explain earlier, seemingly paradoxical observations of elevated synaptic gene expression—including SNAP25 , GABRA1 , GRIN2B , and SYT1 —in ptau–bearing (AT8⁺) neurons undergoing tau‑mediated neurodegeneration [ 18 ], despite expectations of synaptic gene loss in degenerating neurons. These findings suggest that gene expression profiling performed at different disease stages may capture transient compensatory responses aimed at counteracting tau‑induced dysfunction. Consistent with this idea, other synaptic genes such as SLC1A2 , GRM7 , and NRXN3 are upregulated at later stages of AD [ 163 ]. Why specific synaptic genes display distinct temporal sensitivities across disease stages remains a central open question in models of network vulnerability and pathological propagation. One plausible explanation is that expression of genes encoding synaptic receptors or ligands is dynamically shaped by changes in the availability of their cognate signaling partners, producing transcriptional responses that are inherently non–cell autonomous. Within vulnerable networks, tau-driven neurodegeneration may selectively disrupt afferent and efferent connections, altering synaptic input–output balance and depriving neurons of stabilizing circuit-level signals. Such connectivity loss could trigger compensatory transcriptional programs aimed at preserving synaptic function, excitability, or signaling capacity. In this framework, stage-dependent synaptic gene regulation reflects not only intrinsic neuronal stress responses but also evolving circuit context, which may in turn modulate susceptibility to further degeneration and constrain or facilitate the trans-synaptic propagation of pathology.

Given that spatial transcriptomics can characterize changes related to pathology in asymptomatic individuals, information from these studies may identify early prodromal events and cell‑specific signatures of vulnerability or resilience before they become obscured in advanced disease stages. Future efforts that integrate post‑mortem spatial transcriptomics with early‑stage human AD samples, longitudinal experimental models, and targeted perturbations will be critical for reconstructing disease trajectories and for defining molecular programs that drive, modify, or constrain neurodegeneration across AD stages.

Insights and mechanisms in AD, lessons from spatial transcriptomics

With the rapid rise of studies utilizing spatial transcriptomics in AD, an immense amount of data linking region-specific, pathology-associated and pathology-independent features is being accrued. Although these results can only offer a snapshot of region and/or pathology-associated dysfunction in human and mouse brain (Table 3 ), some underlying trends in AD have been established through spatial transcriptomics thus far.

First, glial responses to pathology are highly spatially restricted rather than uniformly distributed across affected tissue. Microglia and astrocytes consistently adopt pathology-associated states in close proximity to Aβ plaques, suggesting that local pathological microenvironments are major drivers of reactive state transition. Second, neuronal vulnerability is highly selective across brain regions and cell subtypes, which can be further linked to neuronal degeneration during AD progression. Spatial transcriptomics studies indicate that CA1 neurons are particularly vulnerable while CA3/CA4 neurons show attenuated degeneration [ 106 ], which are consistent with previous findings that hippocampal CA1, entorhinal cortex and subiculum subregions are particularly vulnerable during AD pathogenesis [ 164 ], and during early stages of degeneration, CA2, CA3 and hilus regions are relatively resistant to degeneration [ 165 ]. This implicates progressive heterogeneity in neuronal dysfunction during AD onset. Third, consistent disruption of local intercellular communication networks involved in synaptic maintenance, immune signaling, and Aβ clearance is observed in AD brain. Finally, genetic modifiers such as APOE , TREM2 , and INPP5D further shape pathology-associated cellular responses, highlighting how inherited risk factors may influence local disease microenvironments. Use of mouse models has also demonstrated alterations in expression through modulation of genes related to AD risk such as APOE4 in 5xFAD mouse models [ 102 ]; some models such as Inpp5d depletion in PSAPP animals show effects in enhancing microglia recruitment to plaques [ 123 ], while others such as Trem2 R47H in 5xFAD brain show only modest effect in plaque targeting [ 76 ].

Although definitive roles for specific expression of genes and their contribution to degenerative phenotypes will require further mechanistic evidence, spatial transcriptomics so far has demonstrated potential versatility in defining degenerative events in certain regions and cellular subtypes.

Limitations, challenges and future directions with spatial transcriptomics

Here, we discuss both experimental challenges in implementing spatial transcriptomics in AD brain tissues and computational challenges with spatial transcriptomic data analysis in AD.

Challenges related to AD brain tissue and tissue processing

Use of postmortem tissue in spatial transcriptomics can be challenging, especially where variables such as postmortem interval (PMI) can affect RNA integrity and downstream detection in both human [ 166 , 167 ] and mouse brain [ 168 ]. RNA Integrity Number (RIN) is a commonly used metric to assess RNA quality[ 169 , 170 ], with many spatial transcriptomic platforms requiring a RIN value greater than 7 for optimal performance. Studies have shown a negative correlation between RIN and PMI and have observed selective degradation of certain mRNA species, particularly neuronal synaptic mRNA as RIN declines [ 169 ]. Long-term storage of brain tissue can also impede RNA detection by affecting probe binding to target molecules [ 171 ]. Autofluorescence signals in aged brain tissue can also interfere with image acquisition; lysosomal accumulation of lipofuscin lipid, protein, sugar and ion aggregates with age is particularly problematic [ 172 ]. Photobleaching using light-emitting diode (LED)-based [ 173 ] or commercial Sudan Black-based reagents (TrueBlack Plus) [ 174 ] can be used to reduce lipofuscin-derived autofluorescence.

Spatial transcriptomics and challenges related to technology and analysis in AD

Considerations with current spatial transcriptomic platforms

Despite increasing use of spatial transcriptomics in AD, current spatial transcriptomic platforms face substantial limitations that warrant further improvement. A major limitation associated with commercialized platforms used in AD research is the trade-off between multiplexity and resolution. Array-based spatial transcriptomic technologies such as Visium offer whole-transcriptomic profiling without subcellular resolution and therefore lack cell type-specific analysis. In situ spatial transcriptomic methods such as MERFISH, Xenium and CosMx can achieve subcellular resolution but with limited multiplexity. Emerging efforts to develop methods to profile whole-transcriptome at subcellular resolution suffer from technical issues including low detection rates with increased resolution for array-based methods, and increased error rates with increased multiplexity for imaging-based methods. For example, Stereo-seq can detect transcripts at subcellular resolution (~ 0.5 μm) and can theoretically profile whole transcriptomic gene expression. However, Stereo-seq is limited to detection of ~ 300–400 genes per cell/nucleus in human brain samples [ 106 , 163 ], and fewer than 700 genes detected per cell in mouse brain [ 175 ]. To address this, spots detected by Stereo-seq can be aggregated into larger bins (e.g., 50 μm x 50 μm); while this can increase overall detection to 834 genes per bin in human hippocampus, subcellular resolution is lost by aggregating spot profiles [ 106 ]. Given the spatial resolution and transcript coverage of array-based and imaging-based methods, array-based methods are suited for discovery-based applications, while imaging-based methods may be more compatible with validation studies (Table 2 ).

Thus, there is a need to simultaneously improve both multiplexity and resolution to enable true whole-transcriptome and subcellular-resolution profiling. Additionally, FFPE compatibility for postmortem human brain samples and pathology staining compatibility are also critical in the analysis of AD brain tissue (Table 1 ). Current spatial transcriptomic experiments are expensive, and limited in throughput, scalability and tissue coverage. Substantial effort is needed to develop cost-effective spatial transcriptomic platforms that can generate high quality data for large cohorts.

Limitations related to computational analysis

Current computational pipelines used in spatial transcriptomics are presented with challenges with cell segmentation, cell type deconvolution and annotation, and in modeling spatial microenvironments. Both subcellular-resolution array-based and imaging-based spatial transcriptomic platforms rely on accurate cell segmentation to generate cell-level gene expression profiles. Cell segmentation in AD brain tissues is particularly difficult due to high cell density in the brain, irregular cell morphology and high background levels from neuropil and AD pathology. Ambiguous and unexpected gene expression profiles have been associated with inaccurate cell segmentation [ 76 ]. To improve cell segmentation accuracy, emerging efforts integrate morphological images, transcript distribution and gene expression profiles from reference scRNA-seq datasets to optimize cell segmentation pipelines [ 72 , 176 , 177 ].

Cell deconvolution and annotation are essential for cell type-specific analysis. For multicellular array-based spatial transcriptomic datasets, cell deconvolution is needed to infer cell type composition in each multicellular spot. Accurate deconvolution of cell populations requires high-quality reference scRNA-seq datasets with strong congruency with respect to cell types and signatures with spatial transcriptomic profiles. For subcellular spatial transcriptomic data, accurate annotation of cell populations is challenging due to the limited number of genes profiled. For both cell type deconvolution and annotation, scRNA-seq profiles in matched samples can improve accuracy by reducing errors due to biological heterogeneity across samples [ 178 , 179 ]. Additionally, batch correction may be needed to combine and integrate profiles from scRNA-seq and spatial transcriptomic platforms [ 180 , 181 ].

Modeling spatial microenvironments in AD brain involves the characterization of cellular response in the vicinity of AD pathology. Computational challenges arise in identification and segmentation of AD pathology, and associating pathology with gene expression. Various methods have been used to annotate pathology, including use of existing image processing software with custom modules, traditional image segmentation algorithms and implementing machine learning-based segmentation methods. Due to substantial variations in sample quality, pathological morphology, staining strategies, and imaging platforms, the field currently lacks a robust computational pipeline that can be used across a variety of datasets. Standardized procedures to establish relationships between pathological microenvironments and transcriptomic features require future development of toolkits and pipelines that streamline workflow from pathology segmentation to downstream analysis in spatial microenvironments.

Future directions

The full impact of spatial transcriptomics in Alzheimer’s disease is likely to emerge through its integration with other complementary data modalities. Spatial transcriptomics has provided a critical framework for linking gene expression with histological context, anatomical organization, and pathological hallmarks across brain regions. However, AD pathobiology is governed by regulatory mechanisms that extend beyond transcriptional states alone. Integration with spatial epigenomic [ 182 – 189 ] and proteomic data [ 190 , 191 ] offers an opportunity to capture regulatory architecture and protein-level signaling within pathological microenvironments in the AD brain. Spatial multi-omic integration may enable a more mechanistic interpretation of how pathological niches influence cellular states and intercellular interactions during AD progression. As cohort sizes continue to increase, integration of spatial transcriptomic profiles with genetic data is becoming increasingly feasible. Although genome-wide association studies have identified numerous AD risk loci, spatial transcriptomics can contextualize these variants within specific cell types, brain regions, and pathological microenvironments. Realizing this potential will require the development of computational frameworks capable of jointly modeling heterogeneous spatial data types [ 192 , 193 ], genetic profiles, pathology features, and regulatory hierarchies within complex brain tissue.

Although efforts have been made to standardize spatial transcriptomic data analysis [ 91 , 194 ], there is a need to standardize pathology‑associated computational pipelines for spatial transcriptomic analysis tailored for AD studies. Variability in tissue preservation, staining strategies, spatial transcriptomic platform-specific biases, cell segmentation accuracy, and pathology morphology currently limits cross‑study comparability. Community‑driven benchmarks, curated reference datasets, and shared annotation frameworks could improve reproducibility and facilitate quantitative comparison of pathology‑associated cellular states across cohorts and platforms. Looking forward, pipeline building and execution can be accelerated with use of artificial intelligence (AI) agent platforms or tools [ 195 , 196 ] powered with large language models (LLM). Recent advances in agentic bioinformatics systems can autonomously orchestrate multi-step analytical workflows, interface with specialized bioinformatics tools, and adaptively generate reproducible pipelines. The combination of curated standard pipelines by bioinformatics experts and adaptive applications by AI agentic systems can also enable rapid iteration and benchmarking of pathology-aware spatial transcriptomic pipelines, particularly as community standards and reference datasets continue to mature. Though AI agent platforms allow natural language specifications which greatly improve the ease of use, they inherit the primary limitations of LLM particularly hallucinations, therefore supervision and intervention from bioinformatics experts are necessary for current AI agent tools to avoid inappropriate analysis workflows resulting in misleading biological findings [ 197 , 198 ].

Although spatial transcriptomics captures static snapshots of brain tissue, AD manifests through spatially and temporally coordinated pathological progression across brain regions. Future efforts should therefore focus on modeling spatial–temporal disease trajectories by integrating spatial transcriptomic and pathological profiles to build computational frameworks that infer temporal ordering across regions, pathological burdens and disease stages. Existing efforts have demonstrated promising progress on modeling spatial and temporal gene expression patterns using spatial transcriptomics data. For example, trajectory inference methods such as SpaTrack [ 199 ] leverages optimal transport to integrate spatial and transcriptomic information across multiple samples over temporal intervals, enabling reconstruction of dynamic changes in cell state and tissue architecture over time. Specifically, modeling AD progression requires coupling of the pathological profiles into spatial–temporal models. Using these models, development of AD pathology can be inferred by cellular level changes including cell vulnerability, resilience and gene expression patterns across regions. Essentially, spatial–temporal modeling can reconstruct dynamic cellular state transitions and evolving pathological niches along the course of AD progression. Further integrating with longitudinal imaging and cerebrospinal fluid (CSF) data may provide insights into biomarkers for non-invasive clinical diagnosis.

Ultimately, the potential translational contribution of spatial transcriptomics in AD lies in its ability to anchor molecular alterations to defined anatomical regions and pathological contexts, providing a biological bridge between cellular mechanisms and clinically observable phenotypes. By resolving region- and pathology-specific cellular programs, spatial transcriptomics studies can inform the interpretation of in vivo imaging signals such as positron emission tomography (PET) imaging and magnetic resonance imaging (MRI), as well as CSF or plasma biomarkers, by linking them to their underlying cellular and molecular origins [ 106 , 127 ]. As spatial datasets expand across cohorts and disease stages, these context-resolved molecular profiles may support improved patient stratification, guide target prioritization towards pathology-relevant cellular niches, and inform the design of stage- or region-specific therapeutic strategies. While spatial transcriptomics is not itself a clinical tool, its translational value lies in providing a mechanistic framework that connects molecular pathology with biomarkers and disease progression in AD.

Collectively, these future directions highlight how spatial transcriptomics in AD moves beyond descriptive atlases toward an integrative framework for modeling disease-relevant molecular process in their native tissue context. Advances in spatial resolution, multiplexity, and pathology awareness, together with improved standardization, computational modeling and multi-omic integration, are expected to expand the analytical scope of spatial transcriptomic studies beyond static atlases. By enabling pathology-aware cellular profiling, spatial–temporal inference, and translational anchoring to imaging readouts and fluid biomarkers, spatial transcriptomics is playing an increasingly important role in connecting molecular pathology with disease progression and clinical phenotypes.

Conclusions

Over the past century, AD research has progressed from largely descriptive characterization of neuropathological features to increasingly detailed investigation of the molecular and cellular processes that underlie disease progression. Spatial transcriptomics represents an important step in this trajectory by enabling gene expression to be examined directly within intact tissue preserving anatomical organization and pathological context. As highlighted in this review, spatial transcriptomic studies have provided insight into region-specific vulnerability and resilience, cellular responses to Aβ and tau pathology, and spatially resolved cell–cell interactions and spatial–temporal trajectories that are difficult to capture using bulk RNA sequencing or dissociated single-cell approaches alone.

Despite these advances, current spatial transcriptomic technologies are limited by fundamental trade-offs between spatial resolution and multiplexity. Imaging-based platforms can achieve subcellular spatial resolution but are typically limited to a few hundred up to several thousand genes. In contrast, array-based approaches offer whole transcriptome coverage but often measure signals derived from multiple cells within each capture location. As a result, integration with scRNA-seq data has become a common strategy to accommodate limited spatial resolution or multiplexity, facilitate cell type annotation, and enable in-depth investigation of cell type-specific spatial patterns in AD brain. At the same time, computational challenges remain substantial, including accurate cell segmentation, reliable identification of pathological structures, and modeling gene expression in densely packed and structurally complex regions of the diseased brain.

Further progress in the field will likely depend on continued improvements in both experimental and analytical approaches. Technical advances that increase spatial resolution while maintaining whole transcriptome coverage, improve compatibility with FFPE tissue and histopathological staining will expand the scope of spatial transcriptomic studies in AD. In parallel, there is a need for standardized analytical frameworks that incorporate pathological features and support reproducible comparisons across platforms, cohorts, and disease stages. Integration of spatial transcriptomics with complementary modalities, including histopathological imaging data, spatial proteomics, spatial epigenomics, and genetic variation, may help connect molecular alterations to regulatory mechanisms and disease trajectories within defined brain microenvironments.

The potential translational value of spatial transcriptomics arises from its ability to anchor molecular and cellular changes to specific anatomical regions and pathological context. By providing a framework that connects regional cellular programs with clinically observable phenotypes, spatial transcriptomic studies may improve interpretation of neuroimaging and fluid biomarkers, support more refined patient stratification, and aid identification of cellular processes that represent potential therapeutic targets. As the field moves beyond descriptive mapping toward integrative and longitudinal analyses of disease progression, spatial transcriptomics is likely to play an increasingly important role in clarifying the cellular mechanism of AD and informing future therapeutic strategies.

Over a century has passed since Alois Alzheimer first described [ 200 – 203 ] unique symptoms and brain pathology from a patient, Auguste Deter, who had suffered from irregular memory deficits at the 37th meeting of South-West German psychiatrists in Tübingen, Germany in 1906 [ 203 ]. It is likely that Alzheimer had used a simple Zeiss light microscope mounted on a jug-handle stand, considered to be state-of-the-art at the time, which allowed him to publish a more complete work with illustrations of AD pathology in 1911 [ 201 ]. Although presentation of his work in 1906 solicited no questions or discussion and was initially deemed to be “unsuitable for publication” in the meeting proceedings [ 203 ], the meeting organizers later published a two-page summary describing Alzheimer’s clinical and pathological findings in 1907 [ 200 , 202 ] which would shape and define the course of AD research to this day.

As spatial transcriptomic technology continues to evolve, the AD field will surely reap the benefits of this growing technology. Although the cost of spatial transcriptomics remains high, perhaps we can be encouraged by Alzheimer who for years, worked without salary in Munich and at times, maintained his research from his personal funds [ 203 ]. Passion and dedication in the AD field is commonplace, and with Alois Alzheimer as a leading example, pathologists, engineers and computational biologists will surely find a way forward for AD and related dementias.

Abbreviations

Amyloid beta

Alzheimer’s disease

Artificial intelligence

Cornu Ammonis

Cell-cell interaction

Common coordinate framework

Cerebrospinal fluid

Disease-associated astrocyte

Disease-associated microglia

Differentially expressed gene

Dentate gyrus

DNA nanoballs

Down syndrome

Formalin-fixed paraffin-embedded

Hematoxylin and eosin

Immunofluorescence

In situ hybridization

In situ sequencing

Mild cognitive impairment

Multiplexed Error-Robust Fluorescence In Situ Hybridization

Neurodegenerative microglia

Magnetic resonance imaging

Middle temporal gyrus

Next-generation sequencing

Neurofibrillary tangle

Primary age-related tauopathy

Positron emission tomography

Plaque-induced gene

Postmortem interval

Progressive supranuclear palsy

Phosphorylated tau

Region of interest

Single-cell RNA sequencing/single-nucleus RNA sequencing

Stratum radiatum, lacunosum, and moleculare

Spatially Resolved Transcript Amplicon Readout Mapping

SpaTial Enhanced Resolution Omics-sequencing

White matter

Author contributions

C.H.S., K.Y.Y. and T.Y.H conceived the topic and general ideas for the article; and C.H.S. spearheaded most aspects of article organization. C.H.S., T.Y.H. and J.P.C. researched topics covered, and wrote the article. All authors contributed to editing the manuscript and provided input to the article content. Images were conceived and generated by C.H.S.

Funding

This work was supported by grants from National Institute of Health (R01 AG070391, R01 AG056130 and R01 AG085498 to T.Y.H.).

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

All contributing authors have approved this article for submission.

Competing interests

Timothy Y. Huang is an Editorial Board Member for Molecular Neurodegeneration Advances and was not involved in the editorial review or decisions regarding this manuscript.

Footnotes

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

References

1. Guo T, et al. Molecular and cellular mechanisms underlying the pathogenesis of Alzheimer’s disease. Mol Neurodegener. 2020;15(1):40.

2. Zheng Q, Wang X. Alzheimer’s disease: insights into pathology, molecular mechanisms, and therapy. Protein Cell. 2024;16(2):83–120.

3. Vermunt L, et al. Duration of preclinical, prodromal, and dementia stages of Alzheimer’s disease in relation to age, sex, and APOE genotype. Alzheimers Dement. 2019;15(7):888–98.

4. Giannakopoulos P, et al. Tangle and neuron numbers, but not amyloid load, predict cognitive status in Alzheimer’s disease. Neurology. 2003;60(9):1495–500.

5. Nelson PT, et al. Correlation of Alzheimer disease neuropathologic changes with cognitive status: a review of the literature. J Neuropathol Exp Neurol. 2012;71(5):362–81.

6. Crary JF, et al. Primary age-related tauopathy (PART): a common pathology associated with human aging. Acta Neuropathol. 2014;128(6):755–66.

7. Price JL, et al. Neuropathology of nondemented aging: presumptive evidence for preclinical Alzheimer disease. Neurobiol Aging. 2009;30(7):1026–36.

8. Price JL, Morris JC. Tangles and plaques in nondemented aging and “preclinical” Alzheimer’s disease. Ann Neurol. 1999;45(3):358–68.

9. Braak H, Braak E. Neuropathological stageing of Alzheimer-related changes. Acta Neuropathol. 1991;82(4):239–59.

10. Blessed G, Tomlinson BE, Roth M. The association between quantitative measures of dementia and of senile change in the cerebral grey matter of elderly subjects. Br J Psychiatry. 1968;114(512):797–811.

11. Katzman R, et al. Clinical, pathological, and neurochemical changes in dementia: a subgroup with preserved mental status and numerous neocortical plaques. Ann Neurol. 1988;23(2):138–44.

12. de Vries LE, et al. The concept of resilience to Alzheimer’s Disease: current definitions and cellular and molecular mechanisms. Mol Neurodegener. 2024;19(1):33.

13. Stern Y, et al. A framework for concepts of reserve and resilience in aging. Neurobiol Aging. 2023;124:100–3.

14. Pettigrew C, Soldan A. Defining cognitive reserve and implications for cognitive aging. Curr Neurol Neurosci Rep. 2019;19(1):1.

15. Heberle BA, et al. Systematic review and meta-analysis of bulk RNAseq studies in human Alzheimer's disease brain tissue. bioRxiv, 2024.

16. Mathys H, et al. Single-cell transcriptomic analysis of Alzheimer’s disease. Nature. 2019;570(7761):332–7.

17. Mathys H, et al. Single-cell atlas reveals correlates of high cognitive function, dementia, and resilience to Alzheimer’s disease pathology. Cell. 2023;186(20):4365-4385 e27.

18. Otero-Garcia M, et al. Molecular signatures underlying neurofibrillary tangle susceptibility in Alzheimer’s disease. Neuron. 2022;110(18):2929-2948 e8.

19. Morabito S, et al. Single-nucleus chromatin accessibility and transcriptomic characterization of Alzheimer’s disease. Nat Genet. 2021;53(8):1143–55.

20. Leng K, et al. Molecular characterization of selectively vulnerable neurons in Alzheimer’s disease. Nat Neurosci. 2021;24(2):276–87.

21. Lau SF, et al. Single-nucleus transcriptome analysis reveals dysregulation of angiogenic endothelial cells and neuroprotective glia in Alzheimer’s disease. Proc Natl Acad Sci U S A. 2020;117(41):25800–9.

22. Grubman A, et al. A single-cell atlas of entorhinal cortex from individuals with Alzheimer’s disease reveals cell-type-specific gene expression regulation. Nat Neurosci. 2019;22(12):2087–97.

23. Aguzzoli Heberle B, et al. Systematic review and meta-analysis of bulk RNAseq studies in human Alzheimer’s disease brain tissue. Alzheimers Dement. 2025;21(3):e70025.

24. Walter TJ, Suter RK, Ayad NG. An overview of human single-cell RNA sequencing studies in neurobiological disease. Neurobiol Dis. 2023;184:106201.

25. Jin W, et al. Comprehensive review on single-cell RNA sequencing: a new frontier in Alzheimer’s disease research. Ageing Res Rev. 2024;100:102454.

26. Piwecka M, Rajewsky N, Rybak-Wolf A. Single-cell and spatial transcriptomics: deciphering brain complexity in health and disease. Nat Rev Neurol. 2023;19(6):346–62.

27. Gall JG, Pardue ML. Formation and detection of RNA-DNA hybrid molecules in cytological preparations. Proc Natl Acad Sci U S A. 1969;63(2):378–83.

28. John HA, Birnstiel ML, Jones KW. RNA-DNA hybrids at the cytological level. Nature. 1969;223(5206):582–7.

29. Moses L, Pachter L. Museum of spatial transcriptomics. Nat Methods. 2022;19(5):534–46.

30. Asp M, Bergenstrahle J, Lundeberg J. Spatially resolved transcriptomes-next generation tools for tissue exploration. BioEssays. 2020;42(10):e1900221.

31. Liao J, et al. Uncovering an organ’s molecular architecture at single-cell resolution by spatially resolved transcriptomics. Trends Biotechnol. 2021;39(1):43–58.

32. Rao A, et al. Exploring tissue architecture using spatial transcriptomics. Nature. 2021;596(7871):211–20.

33. Williams CG, et al. An introduction to spatial transcriptomics for biomedical research. Genome Med. 2022;14(1):68.

34. Tian L, Chen F, Macosko EZ. The expanding vistas of spatial transcriptomics. Nat Biotechnol. 2023;41(6):773–82.

35. Jain S, Eadon MT. Spatial transcriptomics in health and disease. Nat Rev Nephrol. 2024;20(10):659–71.

36. Dries R, et al. Advances in spatial transcriptomic data analysis. Genome Res. 2021;31(10):1706–18.

37. Kleino I, et al. Computational solutions for spatial transcriptomics. Comput Struct Biotechnol J. 2022;20:4870–84.

38. Fang S, et al. Computational approaches and challenges in spatial transcriptomics. Genomics Proteomics Bioinformatics. 2023;21(1):24–47.

39. Stahl PL, et al. Visualization and analysis of gene expression in tissue sections by spatial transcriptomics. Science. 2016;353(6294):78–82.

40. Oliveira MF, et al. High-definition spatial transcriptomic profiling of immune cell populations in colorectal cancer. Nat Genet. 2025;57(6):1512–23.

41. Rodriques SG, et al. Slide-seq: a scalable technology for measuring genome-wide expression at high spatial resolution. Science. 2019;363(6434):1463–7.

42. Stickels RR, et al. Highly sensitive spatial transcriptomics at near-cellular resolution with Slide-seqV2. Nat Biotechnol. 2021;39(3):313–9.

43. Liu Y, et al. High-spatial-resolution multi-omics sequencing via deterministic barcoding in tissue. Cell. 2020;183(6):1665-1681 e18.

44. Chen A, et al. Spatiotemporal transcriptomic atlas of mouse organogenesis using DNA nanoball-patterned arrays. Cell. 2022;185(10):1777-1792 e21.

45. Chen KH, et al. RNA imaging. Spatially resolved, highly multiplexed RNA profiling in single cells. Science. 2015;348(6233):aaa6090.

46. Shah S, et al. In situ transcription profiling of single cells reveals spatial organization of cells in the Mouse Hippocampus. Neuron. 2016;92(2):342–57.

47. Eng CL, et al. Transcriptome-scale super-resolved imaging in tissues by RNA seqFISH. Nature. 2019;568(7751):235–9.

48. Wang X, et al. Three-dimensional intact-tissue sequencing of single-cell transcriptional states. Science. 2018. 10.1126/science.aat5691.

49. Cheng Y, et al. Sequencing-free whole-genome spatial transcriptomics at single-molecule resolution. Cell. 2025;188(24):6953-6970 e12.

50. Park HE, et al. Spatial transcriptomics: technical aspects of recent developments and their applications in neuroscience and cancer research. Adv Sci (Weinh). 2023;10(16):e2206939.

51. Valihrach L, et al. A practical guide to spatial transcriptomics. Mol Aspects Med. 2024;97:101276.

52. Miyoshi E, et al. Spatial and single-nucleus transcriptomic analysis of genetic and sporadic forms of Alzheimer’s disease. Nat Genet. 2024;56(12):2704–17.

53. Gabitto MI, et al. Integrated multimodal cell atlas of Alzheimer’s disease. Nat Neurosci. 2024;27(12):2366–83.

54. Gaur P, et al. Single-nucleus and spatial transcriptomic profiling of human temporal cortex and white matter reveals key associations with AD pathology. Nat Commun. 2025;16(1):10395.

55. Ji JX, et al. Intercellular signaling and synaptic deconstruction uncovered by single-cell and spatial transcriptomics in an AD tauopathy model. Commun Biol. 2025;8(1):1583.

56. van Olst L, et al. Microglial mechanisms drive amyloid-beta clearance in immunized patients with Alzheimer’s disease. Nat Med. 2025;31(5):1604–16.

57. Dharshini SAP, et al. Molecular signatures of resilience to Alzheimer’s disease in neocortical layer 4 neurons. Nat Commun. 2026. 10.1038/s41467-026-68920-4.

58. Lim HJ, et al. A practical guide for choosing an optimal spatial transcriptomics technology from seven major commercially available options. BMC Genomics. 2025;26(1):47.

59. Song W, et al. Spatial transcriptomics: integrating platforms and computational approaches for clinical insights. Biotechnol Adv. 2026;87:108791.

60. Luecken MD, Theis FJ. Current best practices in single-cell RNA-seq analysis: a tutorial. Mol Syst Biol. 2019;15(6):e8746.

61. Su M, et al. Data analysis guidelines for single-cell RNA-seq in biomedical studies and clinical applications. Mil Med Res. 2022;9(1):68.

62. Wang M, et al. Guidelines for bioinformatics of single-cell sequencing data analysis in Alzheimer’s disease: review, recommendation, implementation and application. Mol Neurodegener. 2022;17(1):17.

63. Heumos L, et al. Best practices for single-cell analysis across modalities. Nat Rev Genet. 2023;24(8):550–72.

64. Vincent LSP. Watersheds in digital spaces: an efficient algorithm based on immersion simulations. IEEE Transa Pattern Anal Mach Intell. 1991;13(6):583–98.

65. Stringer C, et al. Cellpose: a generalist algorithm for cellular segmentation. Nat Methods. 2021;18(1):100–6.

66. Bannon D, et al. DeepCell Kiosk: scaling deep learning-enabled cellular image analysis with Kubernetes. Nat Methods. 2021;18(1):43–5.

67. Greenwald NF, et al. Whole-cell segmentation of tissue images with human-level performance using large-scale data annotation and deep learning. Nat Biotechnol. 2022;40(4):555–65.

68. Pachitariu M, Stringer C. Cellpose 2.0: How to train your own model. Nat Methods. 2022;19(12):1634–41.

69. He Y, et al. ClusterMap for multi-scale clustering analysis of spatial gene expression. Nat Commun. 2021;12(1):5909.

70. Petukhov V, et al. Cell segmentation in imaging-based spatial transcriptomics. Nat Biotechnol. 2022;40(3):345–54.

71. Chen H, Li D, Bar-Joseph Z. SCS: cell segmentation for high-resolution spatial transcriptomics. Nat Methods. 2023;20(8):1237–43.

72. Fu X, et al. BIDCell: biologically-informed self-supervised learning for segmentation of subcellular spatial transcriptomics data. Nat Commun. 2024;15(1):509.

73. Jones DC, et al. Cell simulation as cell segmentation. Nat Methods. 2025;22(6):1331–42.

74. Xi NM, Li JJ. Benchmarking computational doublet-detection methods for single-cell RNA sequencing data. Cell Syst. 2021;12(2):176-194 e6.

75. Tiesmeyer S, et al. 2D, or not 2D? Investigating vertical signal integrity of tissue slices. bioRxiv. 2025;2025.01.13.632601.

76. Johnston KG, et al. Single-cell spatial transcriptomics reveals distinct patterns of dysregulation in non-neuronal and neuronal cells induced by the Trem2(R47H) Alzheimer’s risk gene mutation. Mol Psychiatry. 2025;30(2):461–77.

77. Wang Y, et al. A systematic evaluation of computational methods for cell segmentation. Brief Bioinform. 2024. 10.1093/bib/bbae407.

78. Oliveira VC, et al. Sudan Black B treatment reduces autofluorescence and improves resolution of in situ hybridization specific fluorescent signals of brain sections. Histol Histopathol. 2010;25(8):1017–24.

79. Hao Y, et al. Dictionary learning for integrative, multimodal and scalable single-cell analysis. Nat Biotechnol. 2024;42(2):293–304.

80. Wolf FA, Angerer P, Theis FJ. SCANPY: large-scale single-cell gene expression data analysis. Genome Biol. 2018;19(1):15.

81. Palla G, et al. Squidpy: a scalable framework for spatial omics analysis. Nat Methods. 2022;19(2):171–8.

82. Mah CK, et al. Bento: a toolkit for subcellular analysis of spatial transcriptomics data. Genome Biol. 2024;25(1):82.

83. Kumar A, et al. Intracellular spatial transcriptomic analysis toolkit (InSTAnT). Nat Commun. 2024;15(1):7794.

84. Lun AT, Bach K, Marioni JC. Pooling across cells to normalize single-cell RNA sequencing data with many zero counts. Genome Biol. 2016;17:75.

85. Hafemeister C, Satija R. Normalization and variance stabilization of single-cell RNA-seq data using regularized negative binomial regression. Genome Biol. 2019;20(1):296.

86. Yao Z, et al. A high-resolution transcriptomic and spatial atlas of cell types in the whole mouse brain. Nature. 2023;624(7991):317–32.

87. Atta L, et al. Gene count normalization in single-cell imaging-based spatially resolved transcriptomics. Genome Biol. 2024;25(1):153.

88. Zhao E, et al. Spatial transcriptomics at subspot resolution with BayesSpace. Nat Biotechnol. 2021;39(11):1375–84.

89. Long Y, et al. Spatially informed clustering, integration, and deconvolution of spatial transcriptomics with GraphST. Nat Commun. 2023;14(1):1155.

90. Singhal V, et al. BANKSY unifies cell typing and tissue domain segmentation for scalable spatial omics data analysis. Nat Genet. 2024;56(3):431–41.

91. Hu Y, et al. Benchmarking clustering, alignment, and integration methods for spatial transcriptomics. Genome Biol. 2024;25(1):212.

92. Aran D, et al. Reference-based analysis of lung single-cell sequencing reveals a transitional profibrotic macrophage. Nat Immunol. 2019;20(2):163–72.

93. Hao Y, et al. Integrated analysis of multimodal single-cell data. Cell. 2021;184(13):3573-3587 e29.

94. Biancalani T, et al. Deep learning and alignment of spatially resolved single-cell transcriptomes with Tangram. Nat Methods. 2021;18(11):1352–62.

95. Cable DM, et al. Robust decomposition of cell type mixtures in spatial transcriptomics. Nat Biotechnol. 2022;40(4):517–26.

96. Gaspard-Boulinc LC, et al. Cell-type deconvolution methods for spatial transcriptomics. Nat Rev Genet. 2025;26(12):828–46.

97. Siletti K, et al. Transcriptomic diversity of cell types across the adult human brain. Science. 2023;382(6667):eadd7046.

98. Chen WT, et al. Spatial transcriptomics and in situ sequencing to study Alzheimer’s disease. Cell. 2020;182(4):976-991 e19.

99. Navarro JF, et al. Spatial transcriptomics reveals genes associated with dysregulated mitochondrial functions and stress signaling in Alzheimer disease. iScience. 2020;23(10):101556.

100. Chen S, et al. Spatially resolved transcriptomics reveals genes associated with the vulnerability of middle temporal gyrus in Alzheimer’s disease. Acta Neuropathol Commun. 2022;10(1):188.

101. Zeng H, et al. Integrative in situ mapping of single-cell transcriptional states and tissue histopathology in a mouse model of Alzheimer’s disease. Nat Neurosci. 2023;26(3):430–46.

102. Lee S, et al. APOE modulates microglial immunometabolism in response to age, amyloid pathology, and inflammatory challenge. Cell Rep. 2023;42(3):112196.

103. Liu S, et al. Moxibustion improves hypothalamus Aqp4 polarization in APP/PS1 mice: evidence from spatial transcriptomics. Front Aging Neurosci. 2023;15:1069155.

104. Lu Y, et al. Multi-transcriptomics reveals brain cellular responses to peripheral infection in Alzheimer’s disease model mice. Cell Rep. 2023;42(7):112785.

105. Mallach A, et al. Microglia-astrocyte crosstalk in the amyloid plaque niche of an Alzheimer’s disease mouse model, as revealed by spatial transcriptomics. Cell Rep. 2024;43(6):114216.

106. Wang P, et al. Molecular pathways and diagnosis in spatially resolved Alzheimer’s hippocampal atlas. Neuron. 2025;113(13):2123-2140 e9.

107. Lee EJ, et al. Spatial transcriptomic brain imaging reveals the effects of immunomodulation therapy on specific regional brain cells in a mouse dementia model. BMC Genomics. 2024;25(1):516.

108. Gelber A, et al. Spatial transcriptomics reveals brain-wide circadian disruption in an Alzheimer's disease model. bioRxiv. 2026.

109. Karasik Y, et al. Super-resolved spatial transcriptomics reveals early changes in RNA localization in the 5xFAD hippocampus. bioRxiv. 2025.

110. Wang Q, et al. The Allen mouse brain common coordinate framework: a 3D reference atlas. Cell. 2020;181(4):936-953 e20.

111. Vento-Tormo R, et al. Single-cell reconstruction of the early maternal-fetal interface in humans. Nature. 2018;563(7731):347–53.

112. Jin S, et al. Inference and analysis of cell-cell communication using Cell Chat. Nat Commun. 2021;12(1):1088.

113. Yang W, et al. DeepCCI: a deep learning framework for identifying cell-cell interactions from single-cell RNA sequencing data. Bioinformatics. 2023. 10.1093/bioinformatics/btad596.

114. Shan N, et al. CITEdb: a manually curated database of cell-cell interactions in human. Bioinformatics. 2022;38(22):5144–8.

115. Wang X, Almet AA, Nie Q. The promising application of cell-cell interaction analysis in cancer from single-cell and spatial transcriptomics. Semin Cancer Biol. 2023;95:42–51.

116. Armingol E, Baghdassarian HM, Lewis NE. The diversification of methods for studying cell-cell interactions and communication. Nat Rev Genet. 2024;25(6):381–400.

117. Efremova M, et al. Cell PhoneDB: inferring cell-cell communication from combined expression of multi-subunit ligand-receptor complexes. Nat Protoc. 2020;15(4):1484–506.

118. Raredon MSB, et al. Comprehensive visualization of cell-cell interactions in single-cell and spatial transcriptomics with NICHES. Bioinformatics. 2023. 10.1093/bioinformatics/btac775.

119. Liu X, et al. StereoSiTE: a framework to spatially and quantitatively profile the cellular neighborhood organized iTME. Gigascience. 2024;13:giae078.

120. Wang S, et al. Gene interactions analysis of brain spatial transcriptome for Alzheimer’s disease. Genes Dis. 2024;11(6):101337.

121. Troule K, et al. Cell PhoneDB v5: inferring cell-cell communication from single-cell multiomics data. Nat Protoc. 2025;20(12):3412–40.

122. Jin S, Plikus MV, Nie Q. Cell chat for systematic analysis of cell-cell communication from single-cell transcriptomics. Nat Protoc. 2025;20(1):180–219.

123. Castranio EL, et al. Microglial INPP5D limits plaque formation and glial reactivity in the PSAPP mouse model of Alzheimer’s disease. Alzheimers Dement. 2023;19(6):2239–52.

124. Zhang G-W, et al. Integrative mapping of spatial transcriptomic and amyloid pathology in Alzheimer’s disease at single-cell resolution. bioRxiv. 2023;2023.05.07.539389.

125. Avey DR, et al. Uncovering plaque-glia niches in human Alzheimer’s disease brains using spatial transcriptomics. Mol Neurodegener. 2025;1(1):2.

126. Choi H, et al. Spatiotemporal characterization of glial cell activation in an Alzheimer’s disease model by spatially resolved transcriptomics. Exp Mol Med. 2023;55(12):2564–75.

127. Shen Y, et al. A spatial imaging-transcriptomics paradigm for deciphering the molecular basis of microscopic MRI in the normal brain and Alzheimer’s disease. Cell Rep. 2025;44(8):116073.

128. Hatami A, Monjazeb S, Glabe C. The anti-Amyloid-beta monoclonal antibody 4G8 recognizes a generic sequence-independent epitope associated with alpha-Synuclein and Islet Amyloid Polypeptide amyloid fibrils. J Alzheimers Dis. 2016;50(2):517–25.

129. Goedert M, Jakes R, Vanmechelen E. Monoclonal antibody AT8 recognises tau protein phosphorylated at both serine 202 and threonine 205. Neurosci Lett. 1995;189(3):167–9.

130. Malia TJ, et al. Epitope mapping and structural basis for the recognition of phosphorylated tau by the anti-tau antibody AT8. Proteins. 2016;84(4):427–34.

131. Mercken M, et al. Monoclonal antibodies with selective specificity for Alzheimer Tau are directed against phosphatase-sensitive epitopes. Acta Neuropathol. 1992;84(3):265–72.

132. Strang KH, et al. Generation and characterization of new monoclonal antibodies targeting the PHF1 and AT8 epitopes on human tau. Acta Neuropathol Commun. 2017;5(1):58.

133. Koss DJ, et al. Soluble pre-fibrillar tau and beta-amyloid species emerge in early human Alzheimer’s disease and track disease progression and cognitive decline. Acta Neuropathol. 2016;132(6):875–95.

134. Wesseling H, et al. Tau PTM profiles identify patient heterogeneity and stages of Alzheimer’s disease. Cell. 2020;183(6):1699-1713 e13.

135. Farah EN, et al. Spatially organized cellular communities form the developing human heart. Nature. 2024;627(8005):854–64.

136. Wei R, et al. Spatial charting of single-cell transcriptomes in tissues. Nat Biotechnol. 2022;40(8):1190–9.

137. Long JM, Holtzman DM. Alzheimer disease: An update on pathobiology and treatment strategies. Cell. 2019;179(2):312–39.

138. Keren-Shaul H, et al. A unique microglia type associated with restricting development of Alzheimer’s disease. Cell. 2017;169(7):1276-1290 e17.

139. Nishitsuji K, et al. Lipoprotein lipase is a novel amyloid beta (Abeta)-binding protein that promotes glycosaminoglycan-dependent cellular uptake of Abeta in astrocytes. J Biol Chem. 2011;286(8):6393–401.

140. Qian X, et al. Probabilistic cell typing enables fine mapping of closely related cell types in situ. Nat Methods. 2020;17(1):101–6.

141. Han J, et al. The roles of microglia and astrocytes in neuroinflammation of Alzheimer’s disease. Front Neurosci. 2025;19:1575453.

142. Singh D. Astrocytic and microglial cells as the modulators of neuroinflammation in Alzheimer’s disease. J Neuroinflammation. 2022;19(1):206.

143. Fruhwurth S, Zetterberg H, Paludan SR. Microglia and amyloid plaque formation in Alzheimer’s disease - evidence, possible mechanisms, and future challenges. J Neuroimmunol. 2024;390:578342.

144. Condello C, et al. Microglia constitute a barrier that prevents neurotoxic protofibrillar Abeta42 hotspots around plaques. Nat Commun. 2015;6:6176.

145. Krasemann S, et al. The TREM2-APOE pathway drives the transcriptional phenotype of dysfunctional microglia in neurodegenerative diseases. Immunity. 2017;47(3):566-581 e9.

146. Wyss-Coray T, et al. Adult mouse astrocytes degrade amyloid-beta in vitro and in situ. Nat Med. 2003;9(4):453–7.

147. Xiao Q, et al. Enhancing astrocytic lysosome biogenesis facilitates Abeta clearance and attenuates amyloid plaque pathogenesis. J Neurosci. 2014;34(29):9607–20.

148. Kato S, et al. Confocal observation of senile plaques in Alzheimer’s disease: senile plaque morphology and relationship between senile plaques and astrocytes. Pathol Int. 1998;48(5):332–40.

149. Wegiel J, et al. The role of microglial cells and astrocytes in fibrillar plaque evolution in transgenic APP(SW) mice. Neurobiol Aging. 2001;22(1):49–61.

150. Habib N, et al. Disease-associated astrocytes in Alzheimer’s disease and aging. Nat Neurosci. 2020;23(6):701–6.

151. Bayaraa O, et al. Multi-region spatial transcriptomics reveals region specific differences in response to amyloid beta (Abeta) plaque induced changes in Alzheimer’s disease (AD). Hum Genomics. 2025;20(1):2.

152. Wiseman FK, et al. A genetic cause of Alzheimer disease: mechanistic insights from Down syndrome. Nat Rev Neurosci. 2015;16(9):564–74.

153. Wang S, et al. Spatiotemporal transcriptomic profiling reveals upregulation of glycolysis pathway genes before overt tauopathy in the PS19 mouse model. Exp Mol Med. 2026;58(2):548–61.

154. Stein-O’Brien GL, et al. Transcriptional signatures of hippocampal tau pathology in primary age-related tauopathy and Alzheimer’s disease. Cell Rep. 2025;44(3):115422.

155. Martinez-Valbuena I, et al. 4R-tau seeding activity reveals molecular subtypes in Progressive Supranuclear Palsy. Nat Commun. 2025;17(1):1006.

156. Zhang B, Horvath S. A general framework for weighted gene co-expression network analysis. Stat Appl Genet Mol Biol. 2005;4:Article17.

157. Mizutani K, et al. Nectins and Nectin-like molecules in synapse formation and involvement in neurological diseases. Mol Cell Neurosci. 2021;115:103653.

158. Mizutani K, et al. Nectin-2 in general and in the brain. Mol Cell Biochem. 2022;477(1):167–80.

159. Tomorsky J, et al. Precise levels of Nectin-3 are required for proper synapse formation in postnatal visual cortex. Neural Dev. 2020;15(1):13.

160. Nazarian A, Yashin AI, Kulminski AM. Genome-wide analysis of genetic predisposition to Alzheimer’s disease and related sex disparities. Alzheimers Res Ther. 2019;11(1):5.

161. Sun Y, et al. A transcriptome-wide association study of Alzheimer’s disease using prediction models of relevant tissues identifies novel candidate susceptibility genes. Genome Med. 2021;13(1):141.

162. Chen KS, et al. Interneuron transcriptomics reveals pathologic markers of Alzheimer’s disease progression. Neurobiol Dis. 2025;213:107012.

163. Gong Y, et al. Stereo-seq of the prefrontal cortex in aging and Alzheimer’s disease. Nat Commun. 2025;16(1):482.

164. Bouwman MMA, et al. Hippocampal subfields: volume, neuropathological vulnerability and cognitive decline in Alzheimer’s and Parkinson’s disease. Alzheimers Res Ther. 2025;17(1):121.

165. Mrdjen D, et al. The basis of cellular and regional vulnerability in Alzheimer’s disease. Acta Neuropathol. 2019;138(5):729–49.

166. Birdsill AC, et al. Postmortem interval effect on RNA and gene expression in human brain tissue. Cell Tissue Bank. 2011;12(4):311–8.

167. White K, et al. Effect of postmortem interval and years in storage on RNA quality of tissue at a repository of the NIH NeuroBioBank. Biopreserv Biobank. 2018;16(2):148–57.

168. Olney KC, et al. Postmortem interval leads to loss of disease-specific signatures in brain tissue. eNeuro. 2025. 10.1523/ENEURO.0505-24.2025.

169. Tian J, et al. An analysis of RNA quality metrics in human brain tissue. J Neuropathol Exp Neurol. 2025;84(3):236–43.

170. Sonntag KC, et al. Limited predictability of postmortem human brain tissue quality by RNA integrity numbers. J Neurochem. 2016;138(1):53–9.

171. Howe S, et al. Long-term fixation impact on archived human nervous tissues for sequencing-based transcriptomics. Brain Commun. 2025;7(6):fcaf428.

172. Gray DA, Woulfe J. Lipofuscin and aging: a matter of toxic waste. Sci Aging Knowledge Environ. 2005;2005(5):re1.

173. Sapio MR, et al. Efficient removal of naturally-occurring lipofuscin autofluorescence in human nervous tissue using high-intensity white light. J Pain. 2025;30:105359.

174. Stillman JM, et al. Lipofuscin-like autofluorescence within microglia and its impact on studying microglial engulfment. Nat Commun. 2023;14(1):7060.

175. Han L, et al. Single-cell spatial transcriptomic atlas of the whole mouse brain. Neuron. 2025;113(13):2141-2160 e9.

176. Wu L, Beechem JM, Danaher P. Using transcripts to refine image based cell segmentation with FastReseg. Sci Rep. 2025;15(1):30508.

177. Wang Y, et al. GeneSegNet: a deep learning framework for cell segmentation by integrating gene expression and imaging. Genome Biol. 2023;24(1):235.

178. Zhou Z, et al. Spatial transcriptomics deconvolution at single-cell resolution using Redeconve. Nat Commun. 2023;14(1):7930.

179. Sun D, et al. STRIDE: accurately decomposing and integrating spatial transcriptomics using single-cell RNA sequencing. Nucleic Acids Res. 2022;50(7):e42.

180. Korsunsky I, et al. Fast, sensitive and accurate integration of single-cell data with Harmony. Nat Methods. 2019;16(12):1289–96.

181. Stuart T, et al. Comprehensive integration of single-cell data. Cell. 2019;177(7):1888-1902 e21.

182. Vandereyken K, et al. Methods and applications for single-cell and spatial multi-omics. Nat Rev Genet. 2023;24(8):494–515.

183. Russell AJC, et al. Slide-tags enables single-nucleus barcoding for multimodal spatial genomics. Nature. 2024;625(7993):101–9.

184. Zhang D, et al. Spatial epigenome-transcriptome co-profiling of mammalian tissues. Nature. 2023;616(7955):113–22.

185. Lee CN, et al. Spatial joint profiling of DNA methylome and transcriptome in tissues. Nature. 2025;646(8087):1261–71.

186. Takei Y, et al. Spatial multi-omics reveals cell-type-specific nuclear compartments. Nature. 2025;641(8064):1037–47.

187. Liu X, et al. Spatial multi-omics: deciphering technological landscape of integration of multi-omics and its applications. J Hematol Oncol. 2024;17(1):72.

188. Deng Y, et al. Spatial-CUT&Tag: Spatially resolved chromatin modification profiling at the cellular level. Science. 2022;375(6581):681–6.

189. Lu T, Ang CE, Zhuang X. Spatially resolved epigenomic profiling of single cells in complex tissues. Cell. 2023;186(10):2275–9.

190. Mrdjen D, et al. Spatial proteomics of Alzheimer’s disease-specific human microglial states. Nat Immunol. 2025;26(8):1397–410.

191. Walker JM, et al. Spatial proteomics of hippocampal subfield-specific pathology in Alzheimer’s disease and primary age-related tauopathy. Alzheimers Dement. 2024;20(2):783–97.

192. Long Y, et al. Deciphering spatial domains from spatial multi-omics with SpatialGlue. Nat Methods. 2024;21(9):1658–67.

193. Marconato L, et al. SpatialData: an open and universal data framework for spatial omics. Nat Methods. 2025;22(1):58–62.

194. Grases D, Porta-Pardo E. A practical guide to spatial transcriptomics: lessons from over 1000 samples. Trends Biotechnol. 2025.

195. Wang H, et al. SpatialAgent: An autonomous AI agent for spatial biology. bioRxiv, 2025;2025.04.03.646459.

196. Huang K, et al. Biomni: a general-purpose biomedical AI agent. bioRxiv. 2025. 10.1101/2025.05.30.656746.41669157

197. Alber S, et al. Cell Voyager: AI CompBio agent generates new insights by autonomously analyzing biological data. Nat Methods. 2026;23(4):749–59.

198. Bu D, et al. Empowering AI data scientists using a multi-agent LLM framework with self-evolving capabilities for autonomous, tool-aware biomedical data analyses. Nat Biomed Eng. 2026.

199. Shen X, et al. Inferring cell trajectories of spatial transcriptomics via optimal transport analysis. Cell Syst. 2025;16(2):101194.

200. Alzheimer A. Über eine eigenartige Erkrankung der Hirnrinde. Allegmine Zeitschrift für Psychiatrie und Psychisch-Gerichtliche Medizin. 1907;64:146–8.

201. Alzheimer A. Über eigenartige Krankheitsfälle des späteren Alters. Z Gesamte Neurol Psychiatr. 1911;4:56–385.

202. Alzheimer A, et al. An english translation of Alzheimer’s 1907 paper, “Uber eine eigenartige Erkankung der Hirnrinde.” Clin Anat. 1995;8(6):429–31.

203. Dahm R. Alzheimer’s discovery. Curr Biol. 2006;16(21):R906-10.

Associated Data

Data Availability Statement

No datasets were generated or analysed during the current study.