Neuronal death is the indisputable endgame of Alzheimer’s disease, but what does the killing, and how? To try to solve the mystery, researchers have staged increasingly complex crime scenes in three-dimensional co-culture systems. Microglia were previously exposed as ruthless neuron manglers, and now, a study published August 24 in Nature Neuroscience finds that CD8+ T cells make a bad situation worse. Using a sophisticated system to track the movements of individual cells, scientists led by Doo Yeon Kim and Rudolph Tanzi of Massachusetts General Hospital in Charlestown report that T cells infiltrate an AD triculture model and, in cahoots with microglia, set off a neurotoxic inflammatory barrage. The chemokine CXCL10, dispatched by “panicked” astrocytes, seems partially responsible for summoning the T cells, and blocking this signal substantially spared neurons.

  • In a three-dimensional culture system, neurons expressing AD mutations activated glia.
  • CD8+ T cells infiltrated the culture.
  • When both CD8+ T cells and microglia were present, neurodegeneration ensued.
  • Reducing T cell recruitment by blocking a chemokine receptor spared neurons.

The findings add to evidence implicating T cells in the pathogenesis of neurodegenerative disease. They also imply new therapeutic targets to prevent these adaptive immune cells from inadvertently attacking neurons.

In AD and other neurodegenerative diseases, neuroinflammation has emerged as both a critical response to, and driver of, neuronal damage. But which immune cells are involved, and what turns them on? Microglia, the brain-resident immune cells, are being intensely investigated in this regard. For their part, the MGH scientists previously created a microfluidic, three-dimensional co-culture system, in which iPSC-derived neurons and astrocytes overexpressed APP and PS1 carrying familial AD mutations. Despite having to cope with Aβ and tau pathology, the neurons in these cultures managed to survive until the researchers added in microglia (Jul 2018 news). Once the microglia caught wind that the neurons were struggling with AD pathology, they infiltrated the cultures and destroyed the cells.

Might peripheral immune cells, such as monocytes, B cells, and T cells, worsen this neurotoxicity? This is the question co-first authors Mehdi Jorfi and Joseph Park and colleagues went after. An engineer by training, Jorfi designed the peripheral immune chip. PiChip features a central “brain” compartment, surrounded by four “peripheral” microfluidic chips, each connecting to the central chamber via 100 microchannels (image below). At 10 x 10 microns, each is small enough to allow the scientists to clock the movement of individual cells from the peripheral to the central zone.

AD Model. A three-dimensional co-culture of neurons, astrocytes, and microglia grows in the central cylindrical compartment. Four surrounding microfluidic chambers house peripheral cells (left). One hundred microchannels (right) connect each peripheral chamber to the center. [Courtesy of Jorfi et al., Nature Neuroscience, 2023.]

Before introducing peripheral cells, the researchers checked how brain cells residing in the central chamber fared in AD versus control cultures, as they had done previously. APP/PS1 neurons churned out massive amounts of Aβ peptides, and phospho-tau accumulated in their cell bodies and neurites. When added directly to the AD cultures, induced human microglia (iMGL) rounded up and increased Iba1 expression, as has been seen in AD (for example, Feb 2019 news). The microglia also incited expression of pro-inflammatory genes within the cultures, including complement C3, IFN-γ receptor, and CSF1R. CCL2, an astrocyte-secreted chemokine previously pegged as a microglial recruiter, was upregulated in AD co-cultures regardless of whether microglia were there. Essentially, microglia seemed to exacerbate existing cell stress and inflammatory pathways in AD neurons and astrocytes.

What would happen if peripheral immune cells entered the scene? The researchers isolated peripheral blood mononuclear cells (PBMCs) from healthy donors, added them to the peripheral microfluidic chips, and tracked their progress across the microchannels into the central chamber. Monocytes, B cells, and T cells readily squeezed through the microchannels. While T cells and monocytes infiltrated AD cultures more than they did control cultures, the opposite was true of B cells, which migrated more rapidly into control cultures. Microglia in the AD 3D culture did not affect T cell migration but they slowed monocytes and B cells.

Microchannel Migration. The central “brain” chamber (left) contains AD neurons and astrocytes (yellow) and induced microglia (blue). Over 17 hours, CD8+ T cells (pink) migrate from peripheral chambers through microchannels (white dots, middle) and into the central chamber (right), where they mingle with neurons, astrocytes, and microglia. [Courtesy of Jorfi et al., Nature Neuroscience, 2023.]

Because T cells preferentially migrated into AD cultures, the scientists focused on how they affect neurons. In a nutshell, they found that cytotoxic, CD8+ T cells incited a neuronal massacre within 72 hours of their arrival. While either microglia or CD8+ T cells alone exacerbated neuronal death, the mayhem was most severe with both cell types there.

The scientists next turned to single-cell RNA sequencing to tease out the contributions of microglia and T cells to the inflammatory cascade. A few of the highlights? Microglia with a disease-associated (DAM) signature were substantially more abundant in AD relative to control cultures. However, upon the infiltration of T cells, microglia ditched the DAM signature in favor of a more inflammatory one. They ramped up expression of chemokines and cytokines, particularly those involved in IFN-γ signaling. Reciprocally, the presence of microglia riled infiltrating CD8+ T cells, boosting their own expression of IFN-γ, TNFα, and granzyme B, a cytotoxic protease that triggers apoptosis in target cells.

Ultimately, Jorfi and colleagues zeroed in on the chemokine CXCL10 as a lure for T cells. Churned out by astrocytes, CXCL1 binds the receptor CXCR3. While 77 percent of CD8+ T cells expressed this receptor, only 39 percent of CD4+ T cells did. Yet blocking CXCR3 with a monoclonal antibody only stemmed recruitment of CD8+ T cells by about a third, suggesting other signals contribute to T cell recruitment. Still, this antibody reduced neuronal loss more than fivefold, and lessened neuronal damage 36-fold, as gauged by preservation of surface area taken up by GFP-expressing neurons.

Death Squads? Based on findings from the PiChip system, scientists propose that in the AD brain, neurons churning out Aβ incite astrocytes to summon peripheral T cells via CXCL10. Microglia also become activated. Once T cells arrive, the two together set off a pro-inflammatory cascade that kills neurons. [Courtesy of Jorfi et al., Nature Neuroscience, 2023.]

To Tanzi’s mind, the neurodegenerative cascade is an immune response gone wrong. Astrocytes and microglia respond to the stressed, Aβ-producing neurons as they would to an infection—by transforming into a reactive state and calling in T cells from the periphery to vanquish the threat. The havoc that ensues ends up severely harming neurons, rather than protecting them. Within the brain, cells that make up the blood-brain barrier would be intimately involved in this process, and the researchers are already building the next iteration of PiChip, which will include a brain vasculature component.

Tanzi views the CXCL10-CXCR3 axis as a potential therapeutic target, and plans to put PiChip and similar systems to use to find more targets and to screen drugs.

The study casts more suspicion on T cells in neurodegenerative disease. Scientists in David Holtzman’s lab at Washington University in St. Louis blamed T cells for neuronal death in a mouse model of tauopathy (Mar 2023 news). Tony Wyss-Coray at Stanford University and David Gate, now at Northwestern University in Chicago, detected CD8+ T cells patrolling the brains and CSF of people with AD and PD (Jan 2020 news). Gate subsequently spotted CXCL16-expressing monocytes encircling Aβ plaques along with T cells expressing the cognate receptor, CXCR6, in brain samples from people with AD (Dec 2022 news). The findings, along with Jorfi’s, support the idea of a collaboration between myeloid cells in the brain and T cells, Gate said.

Gate called Jorfi’s study an impressive advancement in the field, noting that the PiChip system allows for mechanistic investigations not possible in human samples. He also noted a caveat: The T cells came from different donors than did the iPSCs that were used to derive the brain cells, meaning their human leukocyte antigens (HLA) were mismatched. This could lead to a substantial allogeneic response in the T cells, Gate said. Jorfi acknowledged this, but noted that T cells did not kill neurons in the control cultures, suggesting that the HLA mismatch alone did not set off a significant cytotoxic response.—Jessica Shugart



  1. In this study, the authors developed a three-dimensional human neuroimmune axis model, which includes human stem-cell-derived neurons, astrocytes, and microglia, together with human peripheral immune cells, to model AD pathology and progression. They found that 1) the co-existence of CD8+ T cells and microglia resulted in a synergism that exacerbated neuronal and glial damage; 2) in the presence of infiltrating T cells, microglia were significantly activated in IFN-associated pathways and antigen presentation; 3) CXCL10 and its T cell receptor, CXCR3, played key roles in mediating selective T cell infiltration into the three-dimensional model, and treatment with MAB160, an anti-CXCR3-neutralizing antibody, significantly prevented cellular damage. This comprehensive study illustrates that T cells, and their interactions with glia, significantly contribute to AD-related neurodegeneration.

    It is now well recognized that both innate and adaptive immune repones are present in neurodegenerative diseases. In AD, adaptive immunity is an important component in both Aβ and tau pathogenesis (Chen and Holtzman, 2022). Several studies have found an increase of T cells in the CSF, leptomeninges, and hippocampus in AD patient postmortem tissue and in both Aβ and tau mouse models (Gate et al. 2020; Laurent et al., 2017; Merlini et al., 2018). Our study found T cells are markedly increased in the brain of a mouse model of tauopathy as well as in the human AD brain in regions with tauopathy. Immune depletion of T cells significantly ameliorates brain atrophy, neuronal loss, and behavioral impairment (Chen et al., 2023). 

    Mouse models and human samples pose limitations for monitoring the interaction between peripheral immune cells and brain-resident cells in a high-throughput way. This work provides a useful tool to capture brain-immune interactions in vitro and pave the way for mechanistic studies and drug screening for therapeutics development. For example, it will be extremely interesting to study which specific T cell subpopulations, including but not limited to, effector CD4+ T cells, Treg, cytotoxic CD8+ T cells, exhausted CD8+ T cells, and which immune modulator or checkpoint blockade molecules are essential in AD progression.  

    This work, together with previously published research in the AD field, raises intriguing questions. Would blocking interactions between microglia and T cells serve as a therapeutic target for preventing neurodegeneration in AD? IFNγ receptors were known to be expressed in both neurons and microglia in the brain (Filiano et al., 2016). We found that in the brains of P301S; APOE4 mice, IFNγ transcripts were enriched in T cells, especially CD8+ T cells. Inhibition of IFNγ significantly ameliorated brain atrophy (Chen et al., 2023). It is interesting to investigate if blocking IFNγ receptors in specific cells, including microglia, would modulate neuroinflammation or would ameliorate neurodegeneration in the setting of amyloid or tau pathology. 

    This research provides insightful context for immunotherapy and highlights new therapeutic strategies for AD. The authors found key roles of CXCL10 and its receptor CXCR3 in regulating T cell infiltration and neuronal damage. We have found the removal and modulation of T cells rescued tau-linked brain atrophy, highlighting an important role of T cells in neurodegeneration. Several reports also indicate that PDCD1 immune checkpoint blockade decreases cognitive impairment in mouse AD models (Baruch et al., 2016; Kumagai et al. 2020). Future studies could extend immunotherapies for AD by targeting immune signaling, even with engineered immune cells.


    . Emerging roles of innate and adaptive immunity in Alzheimer's disease. Immunity. 2022 Dec 13;55(12):2236-2254. Epub 2022 Nov 8 PubMed.

    . Clonally expanded CD8 T cells patrol the cerebrospinal fluid in Alzheimer's disease. Nature. 2020 Jan;577(7790):399-404. Epub 2020 Jan 8 PubMed.

    . Hippocampal T cell infiltration promotes neuroinflammation and cognitive decline in a mouse model of tauopathy. Brain. 2017 Jan;140(1):184-200. Epub 2016 Nov 5 PubMed.

    . Extravascular CD3+ T Cells in Brains of Alzheimer Disease Patients Correlate with Tau but Not with Amyloid Pathology: An Immunohistochemical Study. Neurodegener Dis. 2018;18(1):49-56. Epub 2018 Feb 7 PubMed.

    . Microglia-mediated T cell infiltration drives neurodegeneration in tauopathy. Nature. 2023 Mar;615(7953):668-677. Epub 2023 Mar 8 PubMed.

    . Unexpected role of interferon-γ in regulating neuronal connectivity and social behaviour. Nature. 2016 Jul 21;535(7612):425-9. Epub 2016 Jul 13 PubMed.

    . PD-1 immune checkpoint blockade reduces pathology and improves memory in mouse models of Alzheimer's disease. Nat Med. 2016 Feb;22(2):135-7. Epub 2016 Jan 18 PubMed.

    . The PD-1 expression balance between effector and regulatory T cells predicts the clinical efficacy of PD-1 blockade therapies. Nat Immunol. 2020 Nov;21(11):1346-1358. Epub 2020 Aug 31 PubMed.

Make a Comment

To make a comment you must login or register.


News Citations

  1. Invading Microglia Unleash Neurodegeneration in 3D AD Culture
  2. In Pathology Cascade, Microglia Rev Up After Plaques but Before Tangles
  3. Neurodegeneration—It’s Not the Tangles, It’s the T Cells
  4. Attack of the Clones? Memory CD8+ T Cells Stalk the AD, PD Brain
  5. In AD, CSF Immune Cells Hint at Mounting Mayhem in the Brain

Further Reading

No Available Further Reading

Primary Papers

  1. . Infiltrating CD8+ T cells exacerbate Alzheimer's disease pathology in a 3D human neuroimmune axis model. Nat Neurosci. 2023 Sep;26(9):1489-1504. Epub 2023 Aug 24 PubMed.