Though it carries nary a human transgene, a new mouse model developed many of the hallmark features of Alzheimer’s disease, according to a study published August 14 in Science Advances. Mice deficient in the WD domain of the autophagy protein Atg16L1 accumulated Aβ, phosphorylated tau in the brain, developed neuroinflammation and neurodegeneration, and had memory loss. In the mice, cells struggled to perform a newly discovered type of endocytosis, called LANDO, which helps microglia recycle cell-surface receptors, including TREM2. Researchers led by Douglas Green, St. Jude’s Children’s Hospital, Memphis, Tennessee, reported that an inflammasome inhibitor not only quelled neuroinflammation in these knockout mice, but lessened tau phosphorylation, spared neurons, and assuaged memory deficits.

  • Atg16L-WD knockout mice develop clumps of Aβ, neuroinflammation, and tau phosphorylation.
  • Neuronal loss, synaptic deficits, and learning/memory loss ensue.
  • Blocking inflammation corrects all but the Aβ buildup.

The findings underscore the nexus of endocytosis and neuroinflammation in AD. Even so, some commentators questioned if the knockouts truly model AD, which unfolds over decades. “Additional work is required to determine whether the authors have identified an exciting disease-relevant pathway or a genetic alteration that results in similar downstream phenotypes,” wrote Joseph Lewcock, Kate Monroe, and Lesley Kane from Denali Therapeutics, South San Francisco. Genetic studies have not, thus far, associated loss of the Atg16L1 WD domain with risk of AD.

Previously, Green reported a novel form of receptor uptake called LC3-Associated eNDOcytosis (LANDO). In this pathway, the Atg16L1 protein uses its WD domain to recruit LC3 proteins to endocytic vesicles. LC3s facilitate subsequent trafficking of the vesicles, and in the case of LANDO, the vesicles return to the cell surface for receptor recycling. Derailing LANDO worsened amyloid plaques, tau phosphorylation, neuroinflammation, and memory deficits in 5xFAD mice (Jun 2019 news). 

For the current study, first author Bradlee Heckmann and colleagues asked whether disabling this endocytosis pathway would lead to AD-like phenotypes in mice without AD transgenes. They acquired aged Atg16L1-WD-deficient mice and aged littermate controls from co-author Thomas Wileman of Norwich Medical School, U.K. Wileman’s group had been studying autophagy pathways in them and noticed what appeared to be neurological problems (Fletcher et al., 2018). 

With but a few old mice at their disposal, the researchers performed behavioral experiments and subsequent analyses of brain tissue in the same animals. For most experiments, the data represent four or five mice per group. Once the mice settled into their new environs in Memphis, the researchers put them through behavioral tests. Compared with controls, 2-year-old Atg16L1-WD knockouts had trouble distinguishing novel from familiar objects, and were less likely to move toward unexplored places, both signs of memory deficit.

Could AD-like neuropathology be to blame? The researchers found the hippocampi and cortices of the aged knockouts to be riddled with Aβ clusters as detected by immunofluorescence. Unlike true plaques, the clusters were more diffuse, and occurred inside and outside of neurons. Based on mean fluorescence intensity from Aβ-specific antibody binding in their hippocampi, the knockouts had about five times more Aβ than did controls. Compared with those from controls, brain lysates of Atg16L1-WD knockouts contained much more soluble Aβ40 and Aβ42, as seen on western blots. These were not quantitated. Knockouts and controls had similarly low levels of insoluble Aβ. This came as no surprise to the researchers, because mouse Aβ42 has less propensity to form β-sheets than the fibrillation-prone human version. Heckmann speculated that the Aβ deposits observed in the brain-tissue sections represent a loosely aggregated form that solubilizes easily.

Aβ Aggregates. Immunofluorescence detects deposits of Aβ peptides (red) in the hippocampi of Atg16L1-WD knockout mice (right) but not of littermate controls (left). Nuclei are stained blue. [Courtesy of Heckmann et al., Science Advances, 2020.]

The researchers also detected hyperphosphorylated tau in the hippocampi and cortices of the knockouts. Biochemical studies revealed multiple phosphorylations, including on serine residues 199, 202, 404, and 416.

In keeping with a lackluster LANDO, recycling of microglial receptors TREM2, CD36, and TLR4 were reduced by two- to fourfold in the knockouts. These receptors have all been implicated in uptake of Aβ. Primary microglia cultured from the knockouts readily consumed a single dose of Aβ peptides, but failed to internalize a second dose added 12 hours later. This suggested the receptors were unable to recycle back to the cell surface for a second helping. The microglia also rounded up, rather than extending their processes, and they bumped up expression of Iba1. The brains of the knockouts bathed in a pro-inflammatory stew of cytokines, including TNF-α, IL-1β, and IL-6, suggesting widespread neuroinflammation. The researchers did not look beyond effects on Aβ and tau, but Heckmann thinks deficiencies in LANDO could affect the aggregation of other proteins, such as α-synuclein. “Theoretically it is not a significant leap, because oligomeric α-synuclein can be recognized by TLR2 on microglia (Liu et al., 2007) and other receptors, including Fc receptors,” he wrote.

The recycling deficits events bode ill for brain cells. The knockouts had about 20 percent fewer hippocampal neurons than did controls, and markers of apoptosis were rampant among those that remained. LANDO-less mice also had defects in long-term potentiation, suggesting synapses in their remaining neurons were less plastic.

Might quelling the neuroinflammation help? The researchers treated 20-month-old knockouts and controls for eight weeks with MCC950, a brain-penetrant inhibitor of the inflammasome. While MCC950 did not affect Aβ deposition, it strongly reduced neuroinflammation, microglial activation, tau hyperphosphorylation, and even stemmed some of the neuronal loss. MCC950-treated knockouts had fewer apoptotic neurons than untreated counterparts, and they performed at near-wild-type levels on memory tests.

Overall, the findings suggested that disrupting the LANDO pathway triggered some sort of Aβ accumulation and neuroinflammation, which led to tau hyperphosphorylation and neurodegeneration. Countering neuroinflammation appeared to stop the neurodegenerative cascade. Heckmann speculated that a similar approach could work for people with AD, even if anti-inflammatory drugs were administered after the amyloid cascade was in full swing.

MCC950 has caused liver toxicity in some clinical trials. Other inflammasome inhibitors, such as Inzomelid and neflamapimod are in various stages of development for a range of diseases, including AD (Sep 2017 news; for review, see Zahid et al., 2019). In previous trials, non-steroidal anti-inflammatories failed to slow or stop progression of AD (Apr 2019 news). 

The researchers found hints from a small number of human tissue samples that deficiencies in the LANDO pathway could be at play in AD. Comparing postmortem brains from four healthy controls with brains from four people with AD, the researchers found reduced expression, at the RNA and protein levels, of Atg16L and other key players in LANDO, including Rubicon and Atg5. Master transcriptional regulators of autophagy, including TFEB and EIF5α, were also lower in AD brains. “Should these findings in human samples be substantiated through additional work in larger cohorts, this model could be of significant interest to the field,” wrote Lewcock and colleagues (full comment below). "It may better represent aspects of human disease and would not be as uniformly driven by Aβ levels or prone to the APP overexpression artifacts observed in other models.”

Takaomi Saido of RIKEN Brain Science Institute, Wako, Japan is not convinced that the Atg16L1-WD knockouts model AD. Saido pointed out that deletion of the domain is not something that happens in AD pathogenesis. He added that in people, neurodegeneration starts decades after Aβ deposition. “The phenotypes of the WD domain-deficient mice are indeed interesting, but they provide essentially no insight into establishing the cause-and-effect relationship or into identifying the pathological mechanisms in the AD etiology,” he wrote.—Jessica Shugart


  1. Commonly employed mouse models of Alzheimer’s disease (AD) are almost exclusively driven by transgenic expression of known genetic modifiers of the disease. In this study, Heckmann et al. present an exciting new model where a subtle change in the domain composition of the autophagy gene ATG16L (ATG16L∆WD) leads to spontaneous Aβ deposition, tau pathology, and neuronal degeneration in the absence of any AD-related pathological protein overexpression.

    To demonstrate that the findings in this model are consistent with human disease biology, the authors also show a decrease in ATG16L and other LC3-associated endocytosis (LANDO) proteins in the brains from a small cohort of AD patients. This encouraging finding will be critical to reproduce in larger, diverse cohorts of AD patients. Overall, although ATG16L∆WD is sufficient to endogenously lead to AD pathologies in mice, the relevance of this finding to AD patients more broadly remains an open question. Should these findings in human samples be substantiated through additional work in larger cohorts, this model could be of significant interest to the field. It may better represent aspects of human disease and would not be as uniformly driven by Aβ levels or prone to the APP overexpression artifacts observed in other models.

    The authors propose that neuronal death in this model is driven by a robust pro-inflammatory microglial response. Based on this hypothesis, they examine the efficacy of a brain-penetrant NLRP3 inhibitor, MCC950, which reduced pro-inflammatory cytokine IL-1β and p-tau levels, and rescued behavioral endpoints. Interestingly, MCC950 treatment did not alter amyloid levels contrary to previous data in a different AD model (Dempsey et al., 2017), highlighting the challenges associated with the utilization of a range of preclinical models to study a complex disease. The robust cytokine alterations observed in this model have also not yet consistently been seen in the CNS of AD patients, but it is possible patients who would benefit from NLRP3 inhibitors could be identified by more detailed analysis of biomarkers such as IL-1β. These studies would be extremely helpful in more conclusively determining the potential benefit of NLRP3 for treatment of AD.

    In order to investigate the mechanisms underlying the microglia activation, Heckmann et al. investigated the expression of AD-relevant cell-surface receptors such as Trem2 on primary microglia and found they were reduced. This led to decreased Aβ uptake using in vitro assays, but this mechanism was not investigated nor validated in vivo. Though intriguing, this mechanistic link between ATG16L∆WD, Aβ uptake, and inflammasome activation requires further study.

    Overall, the potential link between autophagy and AD-relevant phenotypes is an interesting one, especially as ATG16L function appears most relevant in microglia rather than neurons or other CNS cell types. However, as this is not a genetic variation that has been directly associated with increased risk of AD, additional work is required to determine whether the authors have identified an exciting disease-relevant pathway or a genetic alteration that results in similar downstream phenotypes.


    . Inhibiting the NLRP3 inflammasome with MCC950 promotes non-phlogistic clearance of amyloid-β and cognitive function in APP/PS1 mice. Brain Behav Immun. 2017 Mar;61:306-316. Epub 2016 Dec 18 PubMed.

  2. We have developed mouse models that accumulate Aβ without overexpressing APP or APP/PS1 (Saito et al., 2014). They are being used by more than 400 laboratories world-wide, so overexpression artifacts are not always a problem with AD models.

    Still, in our models, more than 30 percent coverage of the cerebral cortex by Aβ did not induce any neurodegeneration even after 18 months (Masuda et al., 2016). The primary question to ask is thus, how does Aβ deposition induce tau pathology and neurodegeneration over decades during preclinical AD? In other words, AD pathology is a biology of time, which is often underappreciated.


    . Single App knock-in mouse models of Alzheimer's disease. Nat Neurosci. 2014 May;17(5):661-3. Epub 2014 Apr 13 PubMed.

    . Cognitive deficits in single App knock-in mouse models. Neurobiol Learn Mem. 2016 Nov;135:73-82. Epub 2016 Jul 1 PubMed.

Make a Comment

To make a comment you must login or register.


Research Models Citations

  1. 5xFAD (C57BL6)

News Citations

  1. Gently Used: Can Recycled Microglia Receptors Prevent Plaque?
  2. New AD Target: Silencing the NLRP3 Inflammasome with Boron?
  3. Closing the Book on NSAIDs for Alzheimer’s Prevention

Therapeutics Citations

  1. Inzomelid
  2. Neflamapimod

Paper Citations

  1. . The WD40 domain of ATG16L1 is required for its non-canonical role in lipidation of LC3 at single membranes. EMBO J. 2018 Feb 15;37(4) Epub 2018 Jan 9 PubMed.
  2. . Identification of proteins involved in microglial endocytosis of alpha-synuclein. J Proteome Res. 2007 Sep;6(9):3614-27. PubMed.
  3. . Pharmacological Inhibitors of the NLRP3 Inflammasome. Front Immunol. 2019;10:2538. Epub 2019 Oct 25 PubMed.

Further Reading


  1. . LC3-Associated Phagocytosis and Inflammation. J Mol Biol. 2017 Nov 24;429(23):3561-3576. Epub 2017 Aug 25 PubMed.

Primary Papers

  1. . Noncanonical function of an autophagy protein prevents spontaneous Alzheimer's disease. Sci Adv. 2020 Aug;6(33):eabb9036. Epub 2020 Aug 14 PubMed.