Four years ago, researchers made a splash with the first in vitro model of Alzheimer’s pathology. In that advance, dubbed “Alzheimer’s in a dish,” neurons and astrocytes grown in a three-dimensional gel matrix produced amyloid plaques and hyperphosphorylated tau, proving that amyloid somehow begets tau. Now, Doo Yeon Kim and Rudy Tanzi of Massachusetts General Hospital in Charlestown and Hansang Cho of the University of North Carolina, Charlotte, have taken the model to a new, lethal dimension. Using a sophisticated, multi-chambered microfluidic chip, they mimicked the microglia activation, invasion, and cell death seen in AD brain. Their “AD on a chip” suggests that neuron- and astrocyte-derived chemokines lure microglia to sites of amyloid accumulation, where the microglia then set off a neuroinflammatory pandemonium that leads to severed axons and dead cells. Blocking migration or downstream inflammatory mediators spared neurons, the authors showed. The work appeared June 27 in Nature Neuroscience.

  • AD on a chip: Tri-culture of neurons and glia breeds neuroinflammation.
  • Chemokines, cytokines spur microglia activation, infiltration, astrogliosis, and neuronal death.
  • Microfluidic system can help unravel pathways of neurodegeneration.

“This new and improved three-dimensional culture holds great promise for understanding the contribution of innate immunity to AD evolution and for preclinically evaluating experimental therapeutics that target it,” wrote Terrence Town, University of Southern California, in an email to Alzforum (see full comment below).

The work “lays the foundation for some important future studies,” wrote Christopher Henstridge and Tara Spires-Jones in a commentary accompanying the paper. “Many AD risk factors are in genes highly expressed in microglia, such as ApoE, Trem2 and Clusterin, and the system could be used to discover how these genes affect microglial influence on AD progression,” they wrote.

Clean Cut. In the three-dimensional tri-culture, microglia (red) appear to sever the axon of a neuron (green). [Courtesy of Park et al., Nature Neuroscience 2018.]

In their previous three-dimensional system, Doo and Tanzi demonstrated unequivocally that amyloid deposition came first, and led to formation of tau deposits resembling neurofibrillary tangles (Oct 2014 news). If the scientists blocked amyloid production, they never saw tau pathology emerge. However, the cells happily co-existed with plaques and tangles, giving no hint of neurodegeneration. Clearly, the model lacked a critical piece. Could it be inflammation?

Channeling Microglia.

In a three-dimensional culture, APP-expressing neurons and astrocytes in the central chamber of a microfluidic chip give off chemokines that attract microglia to migrate in from an outer chamber. [Courtesy of Henstridge and Spire-Jones, Nature Neuroscience, 2018.]

To add that dimension, the researchers used a microfluidic, chip-based system. While a postdoc at MGH, Cho, now an assistant professor in Charlotte, had developed a model for microglia activation and chemotaxis using a circular chip with an inner central chamber connected to a surrounding outer space by narrow channels, like spokes on a wheel. Previously, Cho’s group showed they could load the central chamber with Aβ or other chemoattractants, which would diffuse out through the spokes. That activated microglia waiting in the outer chamber, and induced them to migrate into the central space (Cho et al., 2013). 

In the new study, first author and Cho postdoc Joseph Park used the three-dimensional neuron and astrocyte culture as the central lure. He seeded human neuron progenitors overexpressing human APP harboring the Swedish and London familial AD (FAD) mutations into a three-dimensional gel matrix, which he loaded into the middle chamber. There, over three weeks, the cells differentiated into a mix of neurons and astrocytes. After nine weeks, ELISA assays revealed the cells produced excess soluble Aβ40, 42, and phospho-tau. The cells also accumulated aggregated Aβ42. Using PHF-1 and AT8 p-tau antibodies, the investigators found the protein building up in cell bodies and neurites, and the cultures showed neuronal hyperactivity.

Park found that by week nine, the neurons and astrocytes produced immune mediators, including threefold more of the monocyte chemotactic factor CCL2 (aka MCP-1) and eight times more TNFα compared with similar cultures of cells without human APP. The cultures also made IFNγ, a microglia-activating cytokine absent from two-dimensional cultures of the same cells.

Space Invaders.

Microglia (red) infiltrate a three-dimensional culture model of plaque and tangle pathology in neurons and astrocytes. Top panel shows extensive encroachment after six days. Lower panels show infiltration through narrow channels connecting the inner and outer chambers. [Courtesy of Park et al., Nature Neuroscience 2018.]

To ask if CCL2/MCP-1 made by the three-dimensional cultures could recruit microglia, the investigators added human microglia to the outer chamber, and watched. Within 48 hours, the microglia lost their resting morphology, became elongated and enlarged, upregulated the activation marker CD11b, and began a mass migration though the channels into the central chamber.

A neutralizing antibody to CCL2/MCP-1 partially blocked the migration, indicating that it was one, but not the only, substance responsible for attracting microglia. Other factors, such as Aβ itself and ATP, could also play a role, the authors suggested.

Once the microglia reached their destination, chaos ensued. The cultures ramped up expression of chemotactic factors and inflammatory cytokines, including MCP-1, IL-6, and Il-8. Astrogliosis set in. The microglia attacked neurons outright, damaging axons and causing neurite retraction. At the same time, they boosted levels of neurotoxic TNFα and nitric oxide (NO) by 1.5- and ninefold, respectively. Six days after the microglia invaded, the cultures had lost a fifth of their neurons and astrocytes. This was not seen in otherwise identical tri-cultures in which the neurons did not express APP, or in less-mature cultures where Aβ levels were lower.

Could damping down microglia prevent this destruction? The answer was yes. When the researchers added neutralizing antibodies to IFNγ, or antagonists of toll-like receptor 4 (TLR4) to the tri-culture, they saw a reduction in TNFα and NO, and less cell death. TLR4 knockdown in microglia alone partially protected the tri-cultures. Neither treatment affected microglia infiltration, suggesting that the TNFα and NO released by microglia delivered the deadly blow.

The results begin to unravel the sequence of complex interactions that underlie neuroinflammation and cell death, Tanzi told Alzforum. “This shows us the first release of cytokines and chemokines is starting from neurons and astrocytes. This excites and recruits the microglia, but only when they get to the party does all hell break loose. They start producing tons of cytokines, and only then do you see reactive astrocytes and widespread neuroinflammation and neurodegeneration,” he said.

Importantly, this all starts with amyloid and tau. “If you have wild-type neurons, the microglia hang out and they don’t care about coming in. But if the neurons are making Aβ and tangles, they come swimming in fast,” Tanzi said.

Major Retraction: Microglia (red) swarm a neuron (green), which promptly withdraws its neurite. [Courtesy of Park et al., Nature Neuroscience 2018.]​

At the moment, the results raise more questions than answers. “We see now there’s a complicated dance going on, a back-and-forth between neurons, astrocytes, and microglia,” Tanzi said. “Now we can start deconvoluting those relationships, in a system where we can ask questions one at time or several at a time, and genetically or pharmacologically manipulate each of the cells. And we can do those experiments in five or six weeks, rather than waiting a year or more for a mouse.”

Death and Doom.

With time, invading microglia (red cells) thin out the ranks of neurons and astrocytes (green cells) in the three-dimensional tri-culture system. [Courtesy of Park et al., Nature Neuroscience 2018.]

Knowing the sequence of events offers new possibilities for intervening early to choke off pathology. For example, Tanzi said his group is exploring small molecules that inhibit MCP-1 release from astrocytes. “That may be a nice early step after neurons have made Aβ and tangles, and prior to microglia infiltration. If you can prevent that chemokine response, you can nip neuroinflammation in the bud,” he said.

The investigators replicated their results using a different source of neurons and astrocytes. They started with neural stem cells newly generated from human IPSCs overexpressing the same APP mutants. When they differentiated the cells in the same culture system, the astrocytes and neurons were able to entice microglia to migrate, and produced cell loss. This result opens the door to using a wider variety of cells, including patient samples, in the system.

An important next step will be to incorporate other types of microglia, Malu Tansey at Emory University in Atlanta wrote to Alzforum (full comment below). In the current configuration, the scientists use a convenience line of immortalized human microglia. Recently, several groups have developed protocols to efficiently generate microglia from human iPSCs (Mar 2018 news; Muffat et al., 2016; Abud et al., 2017; Haenseler et al., 2017). Tanzi said his group is switching over to primary IPSC-derived microglia, which recently became commercially available.

For drug discovery, the new system opens up the possibility of screening for drugs that hit not only plaques and tangles, but also neuroinflammation at the level of microglia and astrocyte activation. The current microfluidic chip harbors 25 individual chambers on a single plate. Cho told Alzforum his group is developing a second-generation version with 120 devices per plate. The technology is generating interest from pharmaceutical companies, he said.—Pat McCaffrey



  1. This study is an impressive technological attempt to interrogate the neuro-immune axis within a three-dimensional tri-culture disease model system. Their elegant mechanistic experiments implicate TLR4 activation and microglial-derived IFN-γ and TNF as important factors contributing to neuronal loss in this three-dimensional model culture of Alzheimer’s disease. This study is a significant advance because it is one of the first neurodegenerative disease models to use human-derived cells in three-dimensional culture that included both neural ectoderm and myeloid lineages, and to conclusively demonstrate that myeloid cells can initiate the inflammatory cascade and neuronal loss. As systems like these improve, they'll provide powerful avenues for interrogating pathogenic mechanisms underlying human neurodegenerative diseases and eventually for screening therapeutic candidates.

    However, as the authors aptly point out, overexpression of APP mutations that cause familial AD in immortalized human neural progenitor cells to trigger activation of immortalized microglia is a significant caveat/limitation of this study. Therefore, extrapolation of the conclusions in this paper to three-dimensional models involving primary cells needs to be explored and confirmed empirically.

    Specifically, primary microglia are highly heterogeneous and represent a vastly different cell population than SV40-immortalized microglia. Furthermore, to consider these cells activated requires a suite of data including functional properties as well as characterization with multiple surface protein and transcription factor markers (not just CD68 immunocytochemistry). 

    Within the spatiotemporal context of this study, the large amount (>20 percent) of neuronal death reported is surprising and raises the interesting possibility that the infiltrating microglia formed and/or replaced a cellular niche within the three-dimensional tri-culture rather than distributing homogeneously throughout the culture, although this was not investigated. Moreover, while the News and Views article points out that “the microglia are plated days or weeks after the neurons and astrocytes, which mimics the waves of microglial infiltration into the developing brain and allows investigation of neuroinflammation in a more mature culture of functional neurons supported by astrocytes,” it is now well accepted that microglia progenitors migrate from the yolk sac to the brain prior (~E8.5) to brain circuit formation and are intimately involved in shaping and sculpting neuronal development.

    The use of iPSCs in some of the experiments is a good attempt to extend the findings beyond immortalized cell lines, although these cells also overexpress mutated APP. One significant caveat with human iPSCs is their functional transcriptional age. Specifically, the authors mention that their three-dimensional tri-culture model exhibits late-stage AD markers. Given that iPSC-derived neurons are reset to model fetal brain transcriptomes, as opposed to the aged transcriptome of the donor, adult protein isoforms, including some forms of tau, are not expressed or expressed at very low levels. Therefore, caution must be used when interpreting the significance of late-stage AD markers in this model with APP overexpression. Aging the cells or using direct differentiation approaches may provide a better model system in future iterations.

    In short, the three-dimensional tri-culture presented in this study is a significant advance toward efforts in the neuroscience community to more accurately model the human brain and AD in a dish.  Future studies may integrate three-dimensional co-cultures of neurons, microglia, and astrocytes that are all derived from the same iPSC lines generated from individuals with various mutations that cause or increase the risk of AD.

    In the future the refinement and extension of this type of in vitro three-dimensional model system will allow scientists to examine the role of other physiological components (i.e., vascular system, peripheral immune system) in AD and other neurodegenerative diseases. This will be important given the growing body of evidence that peripheral immune cells modify or contribute to both inflammation and neurodegeneration in many of these diseases.

  2. The authors have devised an ingenious upgrade to their three-dimensional human brain culture system. Not only do they now capture critical neuroinflammatory/neuroimmune AD components, but they are also modeling brain migration of microglia. This new and improved three-dimensional culture holds great promise for understanding the contribution of innate immunity to AD evolution and for preclinically evaluating experimental therapeutics that target it.

  3. This paper represents a significant advancement in the generation of human cell models of Alzheimer's disease. Building upon previous work demonstrating that three-dimensional neuronal cultures overexpressing APP with mutations linked to familial AD (fAD) develop Aβ and tau pathologies more rapidly than two-dimensional cultures (Choi et al., 2014), Park et al. extend their three-dimensional model to include the introduction of microglial cells into the co-culture via the elegant use of microfluidic chambers, which allow the activation and migration of microglia to be investigated. This is one of the first studies to investigate the AD inflammatory response in a human cell culture system, and is a shift away from studying single cell types in isolation and toward more complex co-culture models. This system permits the investigation of non-cell-autonomous disease mechanisms, the importance of which is highlighted through the high representation of microglial genes as risk factors for AD.

    The work in this paper uses immortalized neuronal precursor cells overexpressing APP with fAD mutations, iPSC engineered to overexpress mutant APP, and an immortalized microglial cell line. 

    It will be interesting to see if the phenotypes described here are recapitulated using iPSC-derived neurons/microglia with genotypes relevant to both familial and sporadic AD. These have the important advantage of endogenous levels of gene expression, avoiding the overexpression of genes with multiple mutations in order to drive a cellular phenotype. Of particular note, several papers have shown that three-dimensional cerebral organoids derived from fAD iPSC develop Aβ deposits and increased tau phosphorylation, supporting the premise that patient-derived iPSC capture disease relevant phenotypes in vitro (Raja et al., 2016, and Lin et al., 2018).

    Up until recently, a barrier in the development of such a model has been a lack of protocols for the differentiation of iPSC into microglia, however this is no longer the case and iPSC-microglia can now be robustly generated (Abud et al., 2017; Haenseler et al., 2017).   

    Taken together with the availability of multiple transcriptomic data sets from microglia in aging and disease states (e.g., Olah et al., 2018; Keren-Shaul et al., 2017), the scene is set for the development of iPSC tri-cultures and deep interrogation of how closely the in vitro models capture signatures of aging and disease.


    . A three-dimensional human neural cell culture model of Alzheimer's disease. Nature. 2014 Nov 13;515(7526):274-8. Epub 2014 Oct 12 PubMed.

    . Self-Organizing 3D Human Neural Tissue Derived from Induced Pluripotent Stem Cells Recapitulate Alzheimer's Disease Phenotypes. PLoS One. 2016;11(9):e0161969. Epub 2016 Sep 13 PubMed.

    . APOE4 Causes Widespread Molecular and Cellular Alterations Associated with Alzheimer's Disease Phenotypes in Human iPSC-Derived Brain Cell Types. Neuron. 2018 Jun 27;98(6):1141-1154.e7. Epub 2018 May 31 PubMed.

    . iPSC-Derived Human Microglia-like Cells to Study Neurological Diseases. Neuron. 2017 Apr 19;94(2):278-293.e9. PubMed.

    . A Highly Efficient Human Pluripotent Stem Cell Microglia Model Displays a Neuronal-Co-culture-Specific Expression Profile and Inflammatory Response. Stem Cell Reports. 2017 Jun 6;8(6):1727-1742. PubMed.

    . A transcriptomic atlas of aged human microglia. Nat Commun. 2018 Feb 7;9(1):539. PubMed.

    . A Unique Microglia Type Associated with Restricting Development of Alzheimer's Disease. Cell. 2017 Jun 15;169(7):1276-1290.e17. Epub 2017 Jun 8 PubMed.

  4. Caution is warranted here. We must not jump to the conclusion, based on this data, that the "inflammatory" or microglial response is the cause of destruction. It accompanies destruction, to be sure, but there are reasons to wonder whether the microglial reaction is a response to the injury, rather than the cause.

Make a Comment

To make a comment you must login or register.


News Citations

  1. Alzheimer’s in a Dish? Aβ Stokes Tau Pathology in Third Dimension
  2. Derived Human Microglia Manage Without Functional TREM2

Paper Citations

  1. . Microfluidic chemotaxis platform for differentiating the roles of soluble and bound amyloid-β on microglial accumulation. Sci Rep. 2013;3:1823. PubMed.
  2. . Efficient derivation of microglia-like cells from human pluripotent stem cells. Nat Med. 2016 Nov;22(11):1358-1367. Epub 2016 Sep 26 PubMed.
  3. . iPSC-Derived Human Microglia-like Cells to Study Neurological Diseases. Neuron. 2017 Apr 19;94(2):278-293.e9. PubMed.
  4. . A Highly Efficient Human Pluripotent Stem Cell Microglia Model Displays a Neuronal-Co-culture-Specific Expression Profile and Inflammatory Response. Stem Cell Reports. 2017 Jun 6;8(6):1727-1742. PubMed.

Other Citations

  1. Swedish

External Citations

  1. commercially available

Further Reading

No Available Further Reading

Primary Papers

  1. . A 3D human triculture system modeling neurodegeneration and neuroinflammation in Alzheimer's disease. Nat Neurosci. 2018 Jul;21(7):941-951. Epub 2018 Jun 27 PubMed.
  2. . Modeling Alzheimer's disease brains in vitro. Nat Neurosci. 2018 Jul;21(7):899-900. PubMed.