The mind-boggling variety of cell subtypes in the brain is strongly tied to local environments. Two recent studies offer a glimpse of this heterogeneity by wielding multiplex spatial proteomics at single-cell resolution within fixed postmortem samples from human brain tissue. Scientists led by Bahareh Ajami at Oregon Health & Science University in Portland and Oliver Braubach of Akoya Biosciences in Marlborough, Massachusetts, adapted a method called CODEX, which makes use of antibodies affixed with DNA barcodes, to unveil the cytoarchitecture of the frontal cortex, and to zoom in on cell-cell interactions there. The authors pinpointed distinct microglial subtypes encircling Aβ plaques. The second study, led by Sean Bendall of Stanford University in Palo Alto, California, used antibodies tagged with metal isotopes that can be detected via mass spectrometry to survey proteins across entire coronal sections of the human hippocampus. They found microglia tangoing with tau tangles, and a potentially resilient population of neurons packed with mitofusin-2, which stabilizes mitochondria. Besides hinting at novel cellular mechanisms underlying neurodegenerative disease, these proof-of-concept studies grace the field with remarkable images of cells going about their business within the grand metropolis of the brain.

  • Spatial proteomics detects dozens of proteins at single-cell resolution in fixed brain samples.
  • Microglia subtypes surround Aβ plaques, tau tangles, and sickly neurons.
  • Neurons expressing MFN2, a mitochondrial protein, may be resilient to tau pathology.

“The level of detail achieved with these techniques is stunning,” said Betty Tijms of Amsterdam University Medical Center, who conducts proteomics studies in biofluids. Spatial proteomics adds critical information missing from conventional proteomics or neuropathological studies, said Tijms. “It can tell us which cells are producing the proteins detected in biofluids, and how they are interacting with brain pathology.”

“There is great need for a better mechanistic understanding of RNA expression, metabolites, and proteins at the single-cell level, without sacrificing tissue architecture or cell-cell interactions,” said Costantino Iadecola of Weill Cornell Medical College in New York. “These two papers address that kind of technology gap.”

Ajami posted a preprint on Research Square on May 2, while Bendall’s study was published last November in Acta Neuropathologica Communications.

Conventional immunohistochemistry measures just a few proteins at a time. Proteomics can measure thousands at once in tissue extracts, but say nothing about where in the brain the proteins come from or which cells make them (May 2020 news). Spatial transcriptomics can help, and has just started to integrate proteins—think Aβ plaques and tau tangles—into the equation (Feb 2023 news). Yet transcripts do not always translate into proteins.

Enter spatial proteomics—a collection of new techniques that quantify many proteins at once, with single-cell resolution, and all within intact, fixed human brain tissue. For their part, Ajami’s group improved on CODEX, aka Co-detection by indexing, which can label up to 100 proteins in a sample (Black et al., 2021). Co-first authors Paula Sanchez-Molina of OHSU and Aditya Pratapa of Akoya pretreated the brain slices with a trip to the spa, of sorts, bathing them in hydrogen peroxide under LED lights. This suppresses notorious autofluorescence that typically drowns out fluorescent antibody signals in paraffin-embedded tissue. These refreshed samples were then labeled with a panel of 32 oligonucleotide-barcoded antibodies to identify proteins chosen for their cell specificity, activation state, anatomical location, or toxicity. The samples came from the frontal cortices of four cognitively healthy people and four with AD.

Cell-Cell Contacts. GFAP astrocytes engaged with collagen IV blood vessels, while touching NeuN neurons (left) or Iba1 microglia (middle) at the same time. Microglia were also spotted encircling neurons (right). [Courtesy of Sanchez-Molina et al., 2023.]

The improved technique, dubbed CODEX-CNS, illuminated the anatomical features of the frontal cortex at single-cell resolution. The scientists could clearly visualize gray and white matter, the various cortical layers, blood vessels, the blood-brain barrier and its associated cells, and the layers of the meninges, where they spotted T cells on patrol.

CODEX-CNS also zoomed in on individual interactions among different cells in the brain. Sanchez-Molina and colleagues spotted GFAP-positive astrocytes interacting with neurons and microglia, while at the same time contacting blood vessels with their end feet. The scientists also spied Iba1-positive microglia in the act of consuming sick or dying neurons (see image above).

Plaque Proteome. CODEX identifies an Aβ plaque (white) closely surrounded by astrocytes (green) and microglia (red), with oligodendrocytes (turquoise) and neurons (yellow) farther afield. Orange tubes of collagen IV mark blood vessels. [Courtesy of Sanchez-Molina et al., 2023.]

Sanchez-Molina and colleagues also investigated the cellular neighborhood around Aβ plaques, where they spotted astrocytes clustering with what appeared to be microglia. However, some of the latter expressed CD163, a marker of perivascular macrophages. While some of these CD163-positive cells had sprouted branches typical of ramified microglia, and expressed the homeostatic microglial marker TMEM119, others lacked TMEM119 and had rounded up. Ajami thinks that perivascular macrophages may have infiltrated the parenchyma to surround plaques, or that microglia transition into a macrophage-like state in the plaque environment. Iadecola recently reported that perivascular macrophages contribute to ARIA, the inflammation that occurs when amyloid gets trapped in blood vessels (Apr 2023 news).

To examine microglial states in more detail, Ajami’s group developed an algorithm called Otsu’s Thresholding-based Segmentation and Merge (OTSM). Pronounced “awesome,” it traces the many tentacles of microglial cells, distinguishing them from those projected by their neighbors. Trying to match cell bodies to other seemingly disconnected cell parts seen in the same plane is a major challenge for immunohistochemistry. From nearly 50,000 microglia across the eight brain samples, OTSM identified 10 subpopulations based on protein expression , including some that aligned with clusters of microglia reported in single-cell transcriptomics studies (May 2019 news; Nguyen et al., 2020; Dec 2020 news).

Around plaques, the researchers found a dearth of microglia expressing CD68, a marker of phagocytosis. These cells were more abundant farther afield, suggesting that the plaque environment may hobble phagocytic activity. Homeostatic microglia—designated by high expression of TMEM119—were also sparse around plaques. The most abundant microglia in the vicinity expressed both CD163 and HLA-DR (image below). These cells were scantly detected farther away, and were not found lingering around diffuse plaques in control brains.

Hard Core Microglia. Microglia gathering around a dense-core plaque expressed CD163 and HLA-DR (top). The cells did not express these markers around diffuse plaques in a control brain (bottom). [Courtesy of Sanchez-Molina et al., 2023.]

“The technology presented in this paper not only improves our understanding of different microglial types, but also their interactions with the environment showing that there may be more detailed characteristics of their morphology that might be relevant for Alzheimer’s disease,” commented Joana Pereira of the Karolinska Institute in Stockholm. “This will certainly have an important impact in the field that is now increasingly recognizing the role of neuroinflammation played by microglial cells in neurodegenerative and other diseases,” she wrote (comment below).

Tight Beams
Bendall’s group bypassed fluorescence all together. Co-first authors Kausalia Vijayaragavan, Bryan Cannon, and colleagues used multiplex ion beam imaging by time of flight (MIBI-TOF) mass spectrometry, which images antigens targeted by antibodies labeled with elemental isotopes (Liu et al., 2022). The authors validated a panel of 36 isotope-labeled antibodies specific for a range of brain cell types and neuropathological proteins, including Aβ aggregates and tau tangles. Using MIBI, they could detect all of them at once, at nanometer resolution, across an entire coronal section of the human hippocampus.

Hippocampal Mass Spec. MIBI-TOF markers delineate different subregions of the hippocampus (top), including the molecular layer (Lmol), CA1, and hilus. Aβ42 aggregates (pink) appear in the hilus and molecular layer, while tau tangles (blue) inundate the CA1 region (bottom). [Courtesy of Vijayaragavan et al., Acta Neuropathologica Communications.]

The study surveyed myriad hippocampal features of three people who died in their 90s, including one who was cognitively normal, one who was cognitively impaired but did not have dementia, and one with AD. In one analysis, the researchers examined the proximity of different cell types to Aβ plaques or to tau pathology, including large neurofibrillary tangles and smaller neuropil threads. They detected both types of pathology in all three samples, but the burden was highest in the person with AD. Microglia hovered around plaques and tangles. Tangle-embedded microglia expressed more ApoE, CD33, and Iba1 relative to microglia farther away. Notably, while these apparently reactive microglia associated with tangles in all samples, they were most prevalent around tangles within the AD brain (image at right).

Comparing tangle-laden and tangle-free zones, the scientists identified some neurons that, despite residing in the middle of a tauopathy-ravaged “neighborhood,” somehow managed to escape having tangles themselves. These potentially resilient neurons were the ones expressing high levels of mitofusin 2, a mitochondrial protein that prevents the organelles from splitting apart (image below). Bendall hypothesized that MFN2 could rise in neurons in response to tauopathy-induced stress, potentially bolstering neuronal defenses against tau entanglement, at least for a time. Recent work has suggested tau and mitochondria are intimately involved (Jan 2022 newsApr 2023 news).

To Nicholas Seyfried of Emory University in Atlanta, who uses proteomics to study thousands of proteins within brain tissue and biofluids, the high-resolution spatial component offered by both techniques is a welcome advance. “They show us, beautifully and elegantly, how proteins are painted across the human brain,” he said. He thinks that for these complementary proteomics techniques, next crucial steps will involve scalability and ease of use. If researchers can adapt these protocols to study samples in brain banks, then the techniques could help answer important questions about disease mechanisms and biomarker development, he said.

Last Stand Against Tangles? In a hippocampal section from a person with AD (left), MIBI identifies a region rife with tau tangles (purple). From this area, individual neurons (numbered, middle panel) expressed paired-helical filaments of tau (blue) and MFN2 (yellow). Neurons with high MFN2 expression were likelier to be tangle-free (compare 3 and 4, right panel). [Courtesy of Vijayaragavan et al., Acta Neuropathologica Communications.]

Along those lines, CODEX is highly accessible, requiring light microscopy. It can also be used on archived brain samples, unlike some spatial transcriptomics methods that require specialized sample preparation. Antibodies are the biggest expense, Ajami said.

Iadecola agreed that scalability, and application to existing brain samples, will be crucial for these techniques to come to fruition. For example, using samples from a deeply phenotyped cohort such as ROS-MAP, spatial proteomics could examine cellular responses along the spectrum of Braak stages, he said. That data could prove useful for biomarker development. Tijms agreed, noting that the techniques could help scientists interpret findings from biofluids. “It’s really a great new way of looking at what’s going on in the brain,” she said.

CODEX-CNS and MIBI-TOF join a growing number of spatial omics techniques being applied to the brain, noted Adam Bachstetter of the University of Kentucky in Lexington. He has developed a spatial proteomics technique called QUIVER, which detects multiple proteins in a sample by serially staining with different antibodies (Shahidehpour et al., 2023). “The field is advancing rapidly, and the potential to increase our understanding of devastating neurodegenerative diseases by examining neuropathological changes using these spatial proteomic tools is truly exciting,” he wrote.—Jessica Shugart


  1. This study by Ajami and colleagues shows that a modification of CO-Detection by indEXing (CODEX) called CODEX-CNS identified several microglial phenotypes that varied depending on their surrounding microenvironment (gray and white matter) both in healthy and in AD brain tissue.

    This impressive technology was additionally able to detect different cells, such as astrocytes and macrophages, each with specific phenotypes around Aβ plaques.

    It was well known that microglia adopt different morphologies and express certain markers that are often expressed in distal segments of their processes but not within the cytoplasm adjacent to the nucleus. These distal segment expression patterns vary in disease and it is therefore especially important to take them into account.

    So, the technology presented in this paper not only improves our understanding of different microglial types but also their interactions with the environment, suggesting that there may be more detailed characteristics of their morphology that might be relevant for Alzheimer’s disease. This will certainly have an important impact in the field, which is increasingly recognizing the role played by microglial cells in neuroinflammation in neurodegenerative and other diseases.

  2. The array of tools available for spatial proteomics is continually expanding, yet autofluorescence remains a significant impediment to fluorescence-based techniques in the human brain. These two recently published offer innovative solutions to this issue.

    Ajami et al. introduces a method they have dubbed CODEX-CNS. Building on the well-established CODEX method, they've incorporated a simple chemical and light autofluorescence reduction step borrowed from Du et al. (2019). This approach is complemented by digital background subtraction. For robust antigens such as GFAP, Aβ, and MAP2, they managed to produce exceptionally clear images using an epifluorescence microscope. Leveraging a suite of open-source tools, including QuPath and their newly developed segmentation method, OTSM (pronounced "awesome"), they were able to identify unique clusters of plaque-associated CD163 positive+ IBA1+ cells.

    In contrast, Vijayaragavan et al. employed metal-conjugated antibodies and multiplex ion beam imaging. They analyzed 36 antibodies on their tissue samples, including pathology and cell type markers. Despite their cellular resolution not matching that of the CODEX-CNS method, they identified unique populations of neurons expressing the mitochondrial protein MFN2.

    These two publications add to the growing body of literature on spatial proteomics, including our method, QUIVER (Shahidehpour et al., 2023). It's an exciting time to study human neuropathology. As the field progresses, I anticipate a move beyond well-known markers, such as IBA1, GFAP, and MAP2. Instead, I foresee these technologies used to study proteins that can provide deeper insights into the functional state of cells and their interactions with neuropathology. This pursuit will necessitate integrating and expanding digital pathology tools, such as HALO, OTSM, histoCAT, and CytoMAP, among others, to quantify staining patterns.

    The field is advancing rapidly, and the potential to increase our understanding of devastating neurodegenerative diseases by examining neuropathological changes using these spatial proteomic tools is truly exciting.


    . Qualifying antibodies for image-based immune profiling and multiplexed tissue imaging. Nat Protoc. 2019 Oct;14(10):2900-2930. Epub 2019 Sep 18 PubMed.

    . The localization of molecularly distinct microglia populations to Alzheimer's disease pathologies using QUIVER. Acta Neuropathol Commun. 2023 Mar 18;11(1):45. PubMed.

  3. Both studies illustrate commercial, multiplexed immunohistochemistry techniques and their application toward postmortem AD brain tissue. While only a few cases are studied, the kind of high-dimensional data generated with those approaches is intriguing. Similar efforts have been illustrated for spatial transcriptomics (Wood et al., 2022), as well as for combined spatial transcriptomics and proteomics (Calafate et al., 2023) in other biomedical applications (He et al., 2022).

    The codex technique is interesting because it provides cost-efficient cyclic immunohistochemistry analysis. The imaging mass cytometry (IMC) approach is much faster and represents truly correlative imaging but requires quite delicate equipment. As for all antibody-based techniques, extensive validation is required and the systems are limited to targets that can be delineated with antibodies as well as by antibody availability. It will be very interesting to see the application of those systems toward more functional studies in, for example, AD model systems, and their expansion beyond of the shelf panels toward more tailored targets, such as different tau epitopes in concert with the corresponding neural cell architecture.


    . Plaque contact and unimpaired Trem2 is required for the microglial response to amyloid pathology. Cell Rep. 2022 Nov 22;41(8):111686. PubMed.

    . Early alterations in the MCH system link aberrant neuronal activity and sleep disturbances in a mouse model of Alzheimer's disease. Nat Neurosci. 2023 Jun;26(6):1021-1031. Epub 2023 May 15 PubMed.

    . Integrative in situ mapping of single-cell transcriptional states and tissue histopathology in a mouse model of Alzheimer's disease. Nat Neurosci. 2023 Mar;26(3):430-446. Epub 2023 Feb 2 PubMed.

    . High-plex imaging of RNA and proteins at subcellular resolution in fixed tissue by spatial molecular imaging. Nat Biotechnol. 2022 Dec;40(12):1794-1806. Epub 2022 Oct 6 PubMed.

  4. The study by Sanchez-Molina and collaborators offers a notable multiplexing approach that is, in principle, capable of visualizing up to 100 different proteins simultaneously in brain tissue, using a novel methodology called CODEX-CNS, a variation of the currently existing CODEX technology (Black et al., 2021). The authors designed a panel of 32 oligonucleotide-conjugated antibodies that were incubated simultaneously on fixed, paraffin-embedded postmortem human brain tissue from Alzheimer’s disease patients and age-matched, otherwise healthy individuals. The primary antibodies were thereafter detected using complementary oligonucleotide-conjugated fluorophores, incubated on the tissue in cycles of three fluorophores. After image acquisition in each cycle, the fluorophores were removed by isothermal wash. Tissue autofluorescence due to lipofuscin and fixative, among others, was dramatically reduced by incubating the tissue with H2O2 while simultaneously exposing it to broad-spectrum LED light prior to antibody staining, allowing for the acquisition of images with dramatically increased signal-to-noise ratios. The method is convenient not only for its multiplexing capabilities but also for the high image quality capable of rendering.

    Simultaneously using specific markers such as Iba-1, NeuN, Olig2, GFAP, and Collagen IV, the authors were able to identify and differentiate brain-cell populations and tissue features such as blood vessels. Abundant claudin-5 was detected in blood vessels as well as astrocytic end feet using GFAP. The combined use of these and additional markers could constitute a novel avenue of research in blood-barrier dysfunction during AD. The authors also describe the presence of CD163-positive, TMEM119-negative cells co-localizing with Aβ plaques, which could indicate monocyte/macrophage infiltration. Given the multiplexing capacity of CODEX-CNS, it should be possible to additionally stain these tissues with markers of phagocytic activity (CD68 staining was included in the study) and lysosomal function such as v-ATPase, Clc-7 and LAMPs, which combined would provide insight into the degradative capacity of resident microglia versus presumably infiltrating myeloid cells. This is important, since microglia do not seem particularly good at degrading Aβ in cell culture, whereas macrophages are (Boissonneault et al., 2009Majumdar et al., 2011). 

    The authors showed images of ApoE and GFAP staining near Aβ plaques. In these areas, ApoE staining seems most intense around plaques with reduced astrogliosis, which is an interesting observation. Astrocytes have been suggested to phagocytose Aβ in mouse brain, and ApoE-deficient astrocytes fail to respond or internalize Aβ deposits when compared with wild-type (Gomez-Arboledas et al., 2018; Koistinaho et al., 2004). 

    Using a threshold-based segmentation approach, the authors dissected nearly 50,000 microglia cells, which could be sorted into a number of different subpopulations based on expression of activation markers. One of the subpopulations, with high CD68 expression, seemed to predominantly exist away from Aβ deposits, suggesting that proximity to Aβ reduced phagocytic capacity. This could well be the case. Microglia near Aβ deposits may trim diffuse structures by extracellular digestion in a process called digestive exophagy, rather than by phagocytosis (April 2023 conference news). Although controversial, microglia could also internalize some Aβ material by phagocytosis, rapidly reach their peak degradative capacity, then, no longer being able to internalize additional material, downregulate phagocytosis markers.

    Another subpopulation of interest expressed high levels of CD163, which spatially correlated with Aβ deposits. This specific subpopulation might be associated with cerebral amyloid angiopathy, and might also even represent infiltrating myeloid cells—some of the plaques associated with these cells presented the typical morphology of vascular Aβ.

    It would be fantastic to see this methodology eventually applied to the study of lysosomal markers. For instance, the study of v-ATPase and Clc-7 expression in microglia near to and far from plaques would provide insight into the effect of fibrillar, extracellularly deposited Aβ on microglial degradative capacity. This is important, because lysosomal acidification is reduced during AD (Majumdar et al., 2011Lee et al., 2022). 


    . CODEX multiplexed tissue imaging with DNA-conjugated antibodies. Nat Protoc. 2021 Aug;16(8):3802-3835. Epub 2021 Jul 2 PubMed.

    . Powerful beneficial effects of macrophage colony-stimulating factor on beta-amyloid deposition and cognitive impairment in Alzheimer's disease. Brain. 2009 Apr;132(Pt 4):1078-92. Epub 2009 Jan 17 PubMed.

    . Phagocytic clearance of presynaptic dystrophies by reactive astrocytes in Alzheimer's disease. Glia. 2018 Mar;66(3):637-653. Epub 2017 Nov 27 PubMed.

    . Apolipoprotein E promotes astrocyte colocalization and degradation of deposited amyloid-beta peptides. Nat Med. 2004 Jul;10(7):719-26. PubMed.

    . Faulty autolysosome acidification in Alzheimer's disease mouse models induces autophagic build-up of Aβ in neurons, yielding senile plaques. Nat Neurosci. 2022 Jun;25(6):688-701. Epub 2022 Jun 2 PubMed.

    . Degradation of Alzheimer's amyloid fibrils by microglia requires delivery of ClC-7 to lysosomes. Mol Biol Cell. 2011 May 15;22(10):1664-76. PubMed.

Make a Comment

To make a comment you must login or register.


News Citations

  1. Massive Proteomics Studies Peg Glial Metabolism, Myelination, to AD
  2. Higher-Resolution Spatial Transcriptomics Maps Mayhem Near Plaques
  3. Macrophages Blamed for Vascular Trouble in ApoE4 Carriers
  4. When It Comes to Alzheimer’s Disease, Do Human Microglia Even Give a DAM?
  5. Most Detailed Look Yet at Activation States of Human Microglia
  6. Survey of Tau Partners Highlights Synaptic, Mitochondrial Roles
  7. By Unleashing Microglial cGAS, Tau STINGs Neurons

Paper Citations

  1. . CODEX multiplexed tissue imaging with DNA-conjugated antibodies. Nat Protoc. 2021 Aug;16(8):3802-3835. Epub 2021 Jul 2 PubMed.
  2. . APOE and TREM2 regulate amyloid-responsive microglia in Alzheimer's disease. Acta Neuropathol. 2020 Oct;140(4):477-493. Epub 2020 Aug 25 PubMed.
  3. . Multiplexed Ion Beam Imaging: Insights into Pathobiology. Annu Rev Pathol. 2022 Jan 24;17:403-423. Epub 2021 Nov 9 PubMed.
  4. . The localization of molecularly distinct microglia populations to Alzheimer's disease pathologies using QUIVER. Acta Neuropathol Commun. 2023 Mar 18;11(1):45. PubMed.

Further Reading

No Available Further Reading

Primary Papers

  1. . Single-cell spatial proteomic imaging for human neuropathology. Acta Neuropathol Commun. 2022 Nov 4;10(1):158. PubMed.
  2. . Single-cell spatial proteomic analysis by multiplexed imaging enables identification of microglial heterogeneity in Alzheimer’s disease human brain. Research Square