The full complexity of the blood-brain barrier, with its different cell types acting together, has been difficult to model in vitro. In the June 8 Nature Medicine, researchers led by Li-Huei Tsai at the Massachusetts Institute of Technology, Cambridge, debuted a model that combined endothelial cells, astrocytes, and pericyte-like cells in a three-dimensional gel. The endothelial cells formed capillaries that had high electrical resistance and held back the escape of solutes, suggesting the presence of tightly sealed junctions like those in the brain-blood barrier.

  • Blood-brain barrier model combines human astrocytes, pericytes, and endothelial cells.
  • APOE4 in pericytes, but not the other cells, promoted cerebral amyloid angiopathy.
  • Blocking NFAT signaling protected pericytes.

Other researchers called the model an important advance. “This is a very innovative, versatile blood-brain barrier model, with human relevance,” said Guojun Bu at the Mayo Clinic in Jacksonville, Florida. Costantino Iadecola of Weill Cornell Medical College in New York noted, “This is a step forward in terms of the ability to reproduce what we see in vivo.”

Using this model, Tsai and colleagues found that APOE4, a risk factor for cerebral amyloid angiopathy, worked its mischief through pericytes, but not the other cell types. When pericytes in the model expressed APOE4, they ramped up a transcriptional pathway that boosted ApoE secretion and led to amyloid deposition on capillary walls. Blocking this pathway prevented deposition. “Emerging evidence indicates an important role for pericytes in the physiology and pathology of blood vessels in the brain,” Tsai told Alzforum.

Capillary in a Dish. Two weeks after combining brain endothelial cells, mural cells, and astrocytes in hydrogel, mural cells (middle, green) line blood vessels (red), while astrocyte end-feet (right, green) touch vessels (red). Nuclei are blue. [Courtesy of Blanchard et al., Nature Medicine.]

A paper in the June 15 Nature Cell Biology highlights this. Researchers led by Asifa Akhtar at the Max Planck Institute of Immunobiology and Epigenetics in Freiburg, Germany, generated mice with a defective chromatin-regulating complex. Their neuronal metabolism went haywire, and this in turn inflamed pericytes, leading to a disrupted vasculature. The finding may help explain why people who carry a mutation in the same complex are at elevated risk for neurodegeneration.

Researchers have established numerous in vitro models that capture various aspects of the blood-brain barrier. Cheryl Wellington at the University of British Columbia in Vancouver, Canada, used human cells to build artificial blood vessels on a tube-like support. The tubes transported fluid and mimicked CAA when Aβ was added (Oct 2017 news). Don Ingber and Kit Parker at Harvard study interactions between brain cells and blood vessels on polymer chips (Aug 2018 news). Doo Yeon Kim and Rudolph Tanzi at Harvard Medical School juxtaposed an endothelial cell layer with a three-dimensional neuronal culture to model the BBB (Oct 2014 news). In-Hyun Park and colleagues at Yale School of Medicine in New Haven, Connecticut, grew cerebral organoids with a functioning vasculature (Oct 2019 news). 

Few of these models focus on the interactions of astrocytes, pericytes, and endothelial cells, however. To generate these cells, Tsai and colleagues separately differentiated human induced pluripotent stem cells into either endothelial cells, astrocytes, or mural cells. Mural cells come in different varieties, including the smooth muscle cells that hug larger vessels and the pericytes that line capillaries. Researchers disagree on where the dividing line is between them (Jun 2015 news; Feb 2017 news; May 2017 news). Because no specific markers have been identified for mural cell types, the authors characterized them by RNA-Seq, finding that they had an expression profile very similar to in vivo pericytes, and distinct from that of smooth muscle cells.

First author Joel Blanchard combined all three cell types in a Matrigel matrix and allowed them to grow for two weeks in the presence of platelet-derived growth factor BB and vascular endothelial growth factor A. At the end of that time, the endothelial cells had assembled into capillary-like structures surrounded by basement membrane. Pericytes had migrated to perch on these tubes, as they do in vivo, while astrocytes extended end-feet to touch the vessels (see image above). The capillaries expressed tight junction proteins, retained solutes, and exhibited a high electrical resistance, suggesting the presence of a blood-brain barrier.

APOE4 Acts Through Pericytes. When only pericytes carry APOE4 (right), blood vessels (red) accumulate amyloid (green); but when pericytes carry APOE3 and endothelial cells and astrocytes carry APOE4, no amyloid forms (left). Nuclei are blue. [Courtesy of Blanchard et al., Nature Medicine.]

The authors then used this system to examine how APOE4 boosts the risk for cerebral amyloid angiopathy (Premkumar et al., 1996; Rannikmäe et al., 2014). They generated endothelial cells, pericytes, and astrocytes from people who were either APOE3/3 or APOE3/4, then used different combinations to model the BBB. To promote CAA, the authors exposed their models to conditioned media from neuronal cultures carrying an APP duplication.

When the pericyte-like cells were APOE3/3, little amyloid deposited, no matter what genotype the astrocytes and endothelial cells were. When the pericyte-like cells expressed APOE4, however, extensive CAA occurred (see image above).

What did APOE4 do to pericytes? By RNA-Seq, the authors found about 2,300 genes elevated and almost 2,000 repressed in APOE3/4 pericyte-like cells compared with APOE3/3. They looked for possible master regulators, and noticed upregulation of nuclear factor of activated T cells (NFAT). The phosphatase calcineurin activates NFAT, allowing it to move into the nucleus and initiate transcription. Calcineurin and NFAT are elevated in Alzheimer’s brain, and NFAT interacts with the APOE promoter (Maloney et al., 2007; Reese and Taglialatela, 2011). APOE3/4 pericytes, but not the other cell types, had elevated calcineurin, nuclear NFATc1, and APOE expression. The data hinted at a positive feedback loop, with APOE4 in pericyte-like cells stimulating NFAT signaling to boost its own expression. It is unclear why APOE4 could have such different effects in distinct cell types.

Elevated APOE expression was the culprit behind CAA, the authors found. Overexpressing ApoE3 in the blood-brain barrier model mimicked the APOE4 genotype, causing more CAA, while knocking out the gene prevented it. Blocking NFAT signaling was equally effective. Treating the BBB model with calcineurin inhibitors for two weeks lowered ApoE, normalized the expression profile of APOE3/4 pericytes to that of APOE3/3 cells, and prevented CAA after exposure to Aβ. The same strategy worked in vivo. The authors crossed APOE4 knock-in mice with 5xFAD mice, and treated the offspring with calcineurin inhibitors for three weeks. Treated mice expressed less ApoE in the brain and had less CAA compared with untreated littermates. Likely, ApoE seeds amyloid aggregation in blood vessels as it does in brain, Tsai believes.

Does the mechanism hold in people? In postmortem brain samples from APOE3 and APOE4 carriers, the authors found higher levels of NFAT and ApoE in pericytes in the latter. However, commenters questioned how well findings from this model will translate. Iadecola noted that most CAA in the brain occurs in larger vessels, not capillaries. Bu noted that APOE4 carriers produce less ApoE in parenchyma than do noncarriers, the opposite of what was seen in the pericytes in this model. Because ApoE is secreted, the parenchymal protein might reach blood vessels, complicating the picture. “Is the vasculature exposed to a lower or higher amount of ApoE in ApoE4 carriers? We don’t know,” Bu said.

Nonetheless, researchers found the data exciting. “The interaction between neural-derived Aβ and pericyte-derived ApoE is incredibly novel … it suggests that neural-pericyte interactions may drive the pathology observed in APOE4 carriers,” Richard Daneman at the University of California, San Diego, wrote to Alzforum (full comment below). Steven Greenberg at Massachusetts General Hospital, Boston, wondered if the potent effect of ApoE4 on vascular amyloid drives the increased ARIA seen in APOE4 carriers on anti-amyloid immunotherapy (comment below).

Researchers were curious about therapeutic implications. Calcineurin inhibitors such as cyclosporin A and FK506 are approved by the Food and Drug Administration for use as immunosuppressants, typically to prevent rejection of transplanted organs. Intriguingly, people taking these drugs have a lower risk of dementia (Taglialatela et al., 2015). Christopher Norris of the University of Kentucky, Lexington, noted that calcineurin inhibitors exert beneficial effects on brain cells, calming reactive astrocytes and promoting synaptic plasticity in neurons. “Calcineurin is a major mediator of cellular dysfunction in numerous neurodegenerative diseases,” he said. Norris collaborates with Elizabeth Head at UC Irvine to test calcineurin inhibitors in a dog model of amyloidosis. In preliminary data, they see benefits to neuronal connectivity at drug levels low enough not to suppress the immune system, suggesting the therapy could be worth testing in people.

Tsai’s findings train a spotlight on pericytes, a cell type that has been obscure in Alzheimer’s disease until recently. ApoE4 has previously been implicated in boosting the pro-inflammatory cytokine cyclophilin A in pericytes, causing them to produce proteinases that damage the blood-brain barrier and make it leak (May 2012 news). Recently, researchers led by Axel Montagne and Berislav Zlokovic at the University of Southern California in Los Angeles tied this pericyte-mediated damage to cognitive decline (May 2020 news). 

Montagne noted that cyclosporin A squelches expression of pro-inflammatory cyclophilin A, the culprit behind this damage. “The vascular contribution to dementia is increasingly recognized, and having a drug that can protect blood-brain barrier function as well as reduce amyloidosis is very exciting for the field,” he wrote to Alzforum (full comment below).

The findings dovetail with the report by Akhtar and colleagues. First author Bilal Sheikh generated transgenic mice that lacked one of the three members of the nonspecific lethal (NSL) chromatin complex, but only in neurons. Conditional knockouts of Mof, Kansl2, and Kansl3 all had the same phenotype, developing cerebrovascular problems and brain hemorrhaging. The authors traced the cause to a breakdown of neuronal metabolism, which led to the buildup of long-chain fatty acids. These fatty acids were secreted, triggering the TLR4 receptor in pericytes. In response to TLR4 signaling, pericytes became inflamed, changed shape, and detached from blood vessels. The vessels became leaky or collapsed.

The same thing may happen in people who carry a Kansl1 mutation, the authors suggested. This mutation causes Koolen-de Vries syndrome, marked by enlarged ventricles and a poorly developed corpus callosum. In the postmortem brain of a fetus who died with the mutation, the authors found evidence of brain hemorrhaging, consistent with what they saw in mouse models. Kansl1 lies close to MAPT, the tau gene, and has been linked to progressive supranuclear palsy (Nov 2018 conference news). 

Tsai noted that in toto, the data suggest pericytes may be sensitive to the distress of other cell types such as neurons, and respond by ramping up inflammation. “An anti-inflammatory, immunosuppressant [treatment] approach could be beneficial,” Tsai suggested.—Madolyn Bowman Rogers

Comments

  1. The Tsai lab provides some remarkable insight into the generation of cerebral amyloid angiopathy, which may identify potential therapeutics for Alzheimer's disease. A large percentage of people with AD develop amyloid plaques along their blood vessels. This cerebral amyloid angiopathy can lead to severe pathology and may play a critical role in cognitive decline. There has long been speculation about why amyloid builds up around the vessels, with two main models: 1) vascular amyloid builds up due to overproduction by vascular cells including endothelial cells, pericytes, and vascular smooth muscle cells, 2) vascular amyloid buildup is due to a dysfunction of vascular clearance pathways including efflux across the blood-brain barrier and by the glymphatic system.

    The Tsai lab examined vascular amyloid deposition with a three-dimensional model of the blood-brain barrier using endothelial cells, mural cells (pericytes and vascular smooth muscle cells), and astrocytes generated from human iPSC cells, and came up with a really novel finding. They were able to show that vascular amyloid builds up through unique neural-pericyte interactions. They found significant vascular amyloid buildup only when exogenous amyloid-β was given to the BBB models, and this buildup was much greater when the vascular cells were derived from APOE4/4 iPSCs than from APOE3/3 iPSCs. By mixing and matching the genotypes of the different cells, they were able to identify that the critical vascular cells were the mural cells, as amyloid deposition occurred when these cells were APOE4/4 regardless of the genotype of the other cells, and that this genotype led to overproduction of APOE4 by the mural cells. Thus, vascular amyloid deposition occurred through the interaction of neural derived Aβ with pericyte-derived APOE.

    This is very interesting, as it is currently not known how neurons and pericytes may interact in the healthy brain. The finding of an interaction between neural-derived Aβ and pericyte-derived APOE is incredibly novel and it will be interesting to understand whether these interactions take place in the normal healthy brain (for instance to regulate blood flow) or whether this interaction is specific to Alzheimer's pathology. It is also not clear how excess mural cell ApoE leads to vascular amyloid deposition, whether this drives precipitation of the peptides or alters vascular amyloid clearance. This also is very interesting in the context of the recent finding that APOE4 carriers have pericyte damage and early blood-brain barrier breakdown that correlate with cognitive decline, because it suggests that neural-pericyte interactions may drive the pathology observed in the APOE4 carriers. The Tsai lab also shows that inhibition of calcineurin-NFAT signaling with approved drugs was able to limit APOE production by APOE4/4 pericytes, and thus limit the vascular buildup of amyloid. This is a truly remarkable finding as it suggests a novel potential treatment for AD.

  2. Joel Blanchard and Li-Huei Tsai have developed an interesting in vitro BBB model made of human induced pluripotent stem cells (iPSCs) differentiated into endothelial cells, pericytes, and astrocytes, that they named iBBB. First, they found that iPSC-derived pericytes from isogenic APOE4 carrier cells significantly increased their expression of ApoE as compared with iPSC-derived pericytes from isogenic APOE3 carrier cells. Then, they uncovered that elevated ApoE levels trigger more Aβ to accumulate through calcineurin–nuclear factor of activated T cells (NFAT) signaling. Interestingly, the authors used FDA-approved drugs targeting this pathway, including cyclosporine A (CsA), and were able to reduce the levels of ApoE proteins and the buildup of Aβ after administering the drugs in their in vitro iBBB model and also in vivo in mice. Even more interesting, there are known publications following those individuals who received organ transplants under medication with this drug and these people turned out to have a reduced risk of developing dementia.

    Another interesting point is that CsA is a known inhibitor of Cyclophilin A (CypA). Long story short, we have recently shown that the pericytic pro-inflammatory CypA-matrix metalloproteinase 9 (MMP9) pathway is activated in individuals carrying the APOE4 gene (not APOE3) which leads to basement membrane disruption and tight junction loss, ultimately increasing the BBB permeability (Montagne et al., 2020). 

    The vascular contribution to dementia is increasingly recognized. Having a drug that can protect BBB function as well as reduce amyloidosis is very exciting for the field.

    References:

    . APOE4 leads to blood-brain barrier dysfunction predicting cognitive decline. Nature. 2020 May;581(7806):71-76. Epub 2020 Apr 29 PubMed.

  3. This is a very elegant study, both technically in establishing a model of blood-brain barrier from induced brain endothelial, astrocyte, and pericyte-like cells, and then applying it to study the effects of ApoE4 on amyloid deposition and calcineurin-NFAT-dependent upregulation of APOE expression.

    ApoE4 has been known to associate with increased vascular and plaque Aβ accumulation for more than 25 years, but somehow the precise mechanisms haven’t been fully established. The effect appears to be even stronger for vascular than for plaque amyloid, possibly the reason for increased ARIA among APOE4 carriers in anti-amyloid immunotherapy trials. E4-induced increases in APOE expression seems like a plausible—and potentially treatable—mechanism.

  4. The authors developed methods to generate a BBB-like structure by putting together brain endothelial cells (BEC), mural cells (iMC), and astrocytes differentiated from human iPSCs. Because each of the BBB components can be derived individually, the authors were able to investigate the contribution of each cell type to cerebral amyloid angiopathy. Importantly, they discovered that the major cell type in which ApoE4 contributes to CAA is mural cells. They showed that Aβ accumulation became prominent in the BBB-like structure when mural cells were derived from iPSCs with the APOE4 genotype.

    By performing gene expression analysis, the authors found that NFAT signaling was highly upregulated in APOE4 mural cells. The data leads to the important hypothesis that inhibition of NFAT signaling, via inhibitors such as Cyclosporine A (CsA) or FK506, can rescue the ApoE4-mediated accumulation of Aβ. Indeed, these inhibitors rescued the phenotypes.

    It is very interesting that in previous clinical studies, patients chronically treated with CsA or FK506 showed significantly less incidence of dementia. Optimization of CsA or FK506, or related compounds that inhibit NFAT pathway but have a limited side effect, could be a potential drug for CAA.

  5. Tsai and colleagues present several converging lines of evidence to bring new insight to the role of ApoE and calcineurin–nuclear factor of activated T cells (NFAT) signaling in cerebral amyloid angiopathy (CAA) and Alzheimer’s disease (AD). Evidence to support a pathogenic mechanism for ApoE4 is drawn from multiple sources including in vitro, in vivo, and ex vivo blood-brain barrier (BBB) models and supported by human postmortem single-nucleus transcriptomic data. The authors developed a novel, 3D, induced pluripotent stem cell (iPSC)-derived BBB model (iBBB) to demonstrate greater amyloid accumulation in the APOE4 iBBB. Furthermore, they demonstrated dysregulation of calcineurin–NFAT signaling and APOE in pericyte-like mural cells leading to APOE4-associated CAA pathology.

    The novel iBBB is elegantly formed by spontaneous reorganization of iPSC-derived cell types into a model that recapitulates both anatomical and physiological properties of the human BBB. There is clear evidence of mural cells (MC) migrating to positions proximal to the brain endothelial cells (BEC) forming vessel-like structures. Additionally, aquaporin-4 positive astrocytes surround the vessels and extend GFAP+ projections into the perivascular space. Definition of pericyte markers has improved with the availability of single-cell transcriptomic datasets (Vanlandewijck et al. 2018), and the authors add to this body of literature showing their iMCs to be highly similar to hippocampal pericytes, identifying pericyte markers useful for future investigations.

    Populating the iBBB with iPSC-derived cell types allowed the authors to leverage genetically engineered isogenic pairs of APOE3/3 and APOE4/4 lines and additional APOE3/E4 donors to investigate the effects of APOE genotype on amyloid deposition. APOE4 is a prominent risk factor for capillary CAA (Thal et al., 2002), which is characterized by deposition of amyloid in the BBB basement membrane. The authors demonstrate that iBBB cultures remodel the extracellular matrix, acquiring proteins found in the BBB basement membrane—the site of amyloid deposition in capillary CAA. It would be interesting to investigate if the amyloid deposition observed in the APOE4 iBBB, much of which appears non-vascular or within the BEC and MCs, localizes to the basement membrane generated by the iBBB.

    Vessels in other self-assembly BBB models (Campisi et al., 2018) have perfusable lumens enabling studies of amyloid clearance. Does Tsai’s iBBB model have the ability to study clearance? We acknowledge that not all BBB models are able to answer multiple biological questions and strongly support the use of complementary models to supplement functional analyses as the authors have done with their transwell model to explore questions of barrier permeability and efflux transporters.

    Tsai and colleagues convincingly demonstrate that APOE4 pericytes produce more ApoE than do APOE3 pericytes in their iBBB and in mouse and human brain sections. From conditioned media experiments, they conclude that ApoE secreted from APOE4 pericytes is sufficient to result in increased amyloid accumulation in the iBBB. Bu and colleagues (Yamazaki et al., 2020) also recently reported BBB dysfunction resulting from APOE4 pericytes, where they observed impaired basement membrane formation. However, noncontact co-culture—a similar paradigm to conditioned media—was not sufficient to induce changes in endothelial expression of basement-membrane components in Bu’s study, whereas conditioned media was sufficient to increase amyloid accumulation in the iBBB. This may suggest that APOE4 pericytes cause BBB dysfunction by multiple mechanisms, both via secreted factors and via direct contact with endothelial cells. Notably, Bu and colleagues did not see the same trends in ApoE protein levels that Tsai and colleagues have reported with APOE transcripts, a relative decrease in astrocyte ApoE in APOE4, and a relative increase in pericyte ApoE in APOE4.

    The lipidation status of the ApoE secreted by APOE4 pericytes in this iBBB model is of interest given the 2018 findings of Holtzman and colleagues that an antibody specifically targeting nonlipidated, aggregated ApoE could inhibit amyloid accumulation in ApoE4-knockin, APP/PS1 mice (Liao et al., 2018). Whether the neutralized ApoE in the Holtzman study came primarily from pericytes and whether the pericyte ApoE in the study by Tsai is nonlipidated and aggregated are therefore interesting questions for future investigation.

    With regards to novel therapeutic targets arising from this paper, this group demonstrated that inhibiting the protein phosphatase calcineurin in the NFAT signaling pathway was sufficient to reduce the pathogenic APOE expression, and Aβ deposition. These results are enticing because calcineurin inhibitors are already commercially available, and the authors cite a paper in the discussion that reported decreased incidence of AD/dementia in patients treated with calcineurin inhibitors (Taglialatela et al., 2015), but there are considerable risks of long-term treatment with an immunosuppressant.

    A more recent paper found that long-term use of calcineurin inhibitors in patients who received liver transplants is associated with impaired cognitive function, and white-matter hyperintensities (Pflugrad et al., 2018), and there is emerging evidence of other potentially neurotoxic effects. Especially given the current climate with COVID-19, it is unlikely that an immunosuppressant would be widely accepted as a preventative treatment strategy. Calcineurin inhibitors may not be the way forward, but if we can find targets in this pathway more specific to the AD pathology, this may be a viable preventative treatment avenue.

    Overall, the authors bring together observations from multiple different BBB models and complementary transcriptomic analyses for a thorough investigation of pathogenic mechanism for APOE4 in pericytes and present new targets for intervention.

    References:

    . 3D self-organized microvascular model of the human blood-brain barrier with endothelial cells, pericytes and astrocytes. Biomaterials. 2018 Oct;180:117-129. Epub 2018 Jul 12 PubMed.

    . Targeting of nonlipidated, aggregated apoE with antibodies inhibits amyloid accumulation. J Clin Invest. 2018 May 1;128(5):2144-2155. Epub 2018 Mar 30 PubMed.

    . Longterm calcineurin inhibitor therapy and brain function in patients after liver transplantation. Liver Transpl. 2018 Jan;24(1):56-66. PubMed.

    . Reduced Incidence of Dementia in Solid Organ Transplant Patients Treated with Calcineurin Inhibitors. J Alzheimers Dis. 2015;47(2):329-33. PubMed.

    . Two types of sporadic cerebral amyloid angiopathy. J Neuropathol Exp Neurol. 2002 Mar;61(3):282-93. PubMed.

    . Author Correction: A molecular atlas of cell types and zonation in the brain vasculature. Nature. 2018 Aug;560(7716):E3. PubMed.

    . ApoE (Apolipoprotein E) in Brain Pericytes Regulates Endothelial Function in an Isoform-Dependent Manner by Modulating Basement Membrane Components. Arterioscler Thromb Vasc Biol. 2020 Jan;40(1):128-144. Epub 2019 Oct 31 PubMed.

Make a Comment

To make a comment you must login or register.

References

News Citations

  1. Artificial Human Blood Vessels: A Model for Cerebral Amyloid Angiopathy?
  2. 'Organ on a Chip' Models the Ins and Outs of the Blood-Brain Barrier
  3. Alzheimer’s in a Dish? Aβ Stokes Tau Pathology in Third Dimension
  4. It Bleeds! New Mini-Brains Sport Functioning Blood Vessels
  5. Smooth Muscle Cells, Not Pericytes, Control Brain Blood Flow
  6. Pericytes Don’t Go With the Flow—They Change It
  7. Finally, a Dye to Visualize Pericyte Function
  8. ApoE4 Makes Blood Vessels Leak, Could Kick Off Brain Damage
  9. Even Without Amyloid, ApoE4 Weakens Blood-Brain Barrier, Cognition
  10. International Symposium Puts PSP/CBD on the Map

Research Models Citations

  1. 5xFAD (C57BL6)

Paper Citations

  1. . Apolipoprotein E-epsilon4 alleles in cerebral amyloid angiopathy and cerebrovascular pathology associated with Alzheimer's disease. Am J Pathol. 1996 Jun;148(6):2083-95. PubMed.
  2. . APOE associations with severe CAA-associated vasculopathic changes: collaborative meta-analysis. J Neurol Neurosurg Psychiatry. 2014 Mar;85(3):300-5. Epub 2013 Oct 25 PubMed.
  3. . Important differences between human and mouse APOE gene promoters: limitation of mouse APOE model in studying Alzheimer's disease. J Neurochem. 2007 Nov;103(3):1237-57. PubMed.
  4. . A role for calcineurin in Alzheimer's disease. Curr Neuropharmacol. 2011 Dec;9(4):685-92. PubMed.
  5. . Reduced Incidence of Dementia in Solid Organ Transplant Patients Treated with Calcineurin Inhibitors. J Alzheimers Dis. 2015;47(2):329-33. PubMed.

Further Reading

Primary Papers

  1. . Reconstruction of the human blood-brain barrier in vitro reveals a pathogenic mechanism of APOE4 in pericytes. Nat Med. 2020 Jun;26(6):952-963. Epub 2020 Jun 8 PubMed. Correction.
  2. . Neural metabolic imbalance induced by MOF dysfunction triggers pericyte activation and breakdown of vasculature. Nat Cell Biol. 2020 Jun 15; PubMed.