15 Jun 2018

Despite years of research on ApoE4, questions remain about exactly how this gene variant raises risk for Alzheimer’s disease. To boot, much of the data to date has come from mouse models and may not reflect human brain biology. Two new papers seek to rectify this by describing what the ApoE4 allele does to human neurons and glia derived from induced pluripotent stem cells. The two studies take different approaches but arrive at similar findings. They confirm some previous research on ApoE4, while also holding a few surprises.

In iPSC-derived human neurons, ApoE4 raises Aβ production and tau phosphorylation.
In human astrocytes and microglia, ApoE4 slows uptake and clearance of Aβ.
ApoE4 exerts its toxic effects through multiple cell types.

In the May Nature Medicine, researchers led by Yadong Huang at the Gladstone Institute of Neurological Disease, San Francisco, reported that ApoE4 in human neurons boosted production of Aβ40 and Aβ42. It does not do that in mouse neurons. Independent of its effect on Aβ, ApoE4 triggered phosphorylation and mislocalization of tau. In mixed neuronal cultures, GABAergic neurons seemed particularly sensitive to this tau toxicity, perishing at high rates. The other study, led by Li-Huei Tsai and colleagues at Massachusetts Institute of Technology, Cambridge, took a broader approach, comparing gene-expression profiles of ApoE4 and ApoE3 neurons, astrocytes, and microglia. Published online May 31 in Neuron, these scientists likewise report a rise in Aβ42 production in ApoE4 neurons. In astrocytes and microglia, the ApoE4 allele slowed Aβ uptake and clearance. The data suggest that ApoE4 promotes AD pathology through distinct effects on different cell types, Tsai told Alzforum.

Both groups stressed that iPSC-derived cultures can answer mechanistic questions about AD and are suitable for screening potential therapies. “Human neurons may work better than mouse neurons to determine compound safety and efficacy,” Huang noted.

Others said the cell-type-specific data make an important contribution. “[The findings] suggest that despite low levels of ApoE expression by neurons, ApoE4 can still result in a neuronal phenotype … What is not clear is which phenotypes are relevant in the in vivo setting, where there are many other cell types present including glia, which produce much more ApoE than neurons,” David Holtzman at Washington University in St. Louis wrote to Alzforum (full comment below).

ApoE4 Triggers Tau Pathology.

ApoE4 neuronal cultures (left) accumulate more pathological phosphorylated tau (yellow) than do ApoE3 cultures (right). Normal tau is red. *[Courtesy of Wang et al., Nature Medicine.]*

Earlier mouse and human imaging studies have suggested that the ApoE4 allele slows Aβ clearance, allowing amyloid to accumulate faster (Apr 2009 news; Jul 2010 conference news). It remained unclear, however, if these effects were due to loss of ApoE’s normal functions, or to a gain of toxic function. Studies conflict on this point, with some finding better brain function in mice from lowering ApoE, others from raising it (Dec 2011 news; Feb 2012 news; Feb 2012 conference news).

Huang and colleagues wanted to know what ApoE does in human cells. First author Chengzhong Wang generated iPSC lines from three people who were homozygous for ApoE4, and three who were homozygous for ApoE3. The scientists differentiated each line into mixed excitatory and inhibitory neuronal cultures and looked for differences between E3 and E4 cells.

The first surprise was that ApoE4 cultures secreted twice as much Aβ40 and Aβ42 as did E3 cultures. A direct effect on Aβ production does not happen in mouse models, Huang said. Although some papers report higher Aβ levels in ApoE4 knock-ins, this may be due to slower clearance of the peptide, he added (Pankiewicz et al., 2014; Boehm-Cagan and Michaelson, 2014; Luz et al., 2016). An in vivo microdialysis study that directly measured Aβ dynamics in knock-in mice confirmed that the E4 allele only affected clearance, not production (Jun 2011 news).

Huang and colleagues detected more soluble amyloid precursor protein as well, suggesting that enhanced processing of APP was responsible for the uptick. They found no change in APP expression, contradicting earlier findings in human neurons (Jan 2017 news).

Other findings from this study reinforced previous research. E4 neurons retained more ApoE inside their cell bodies than did E3s, and broke it down more, in agreement with reports that the E4 isoform fragments more readily (Apr 2012 news). E4 neurons developed tau pathology even when Aβ processing was inhibited, confirming a direct effect of ApoE4 on tau as first reported in mouse models (Apr 2017 conference news; Sep 2017 news). Huang noted that tau pathology was more intense in these human neurons than in mice. Antibodies against paired helical filaments, tau’s most pathological form, rarely give much signal in mice, but lit up the human neurons strongly, he told Alzforum.

In the mixed neuronal cultures, GABAergic neurons appeared particularly vulnerable to the effects of tau. They developed more signs of tau toxicity, for example hyperphosphorylation and axonal degeneration, and subsequently perished at higher rates than did glutamatergic neurons in the same cultures. This, too, matches in vivo findings from mice, where degeneration of inhibitory interneurons has been blamed for learning and memory deficits (Feb 2012 conference news). It also agrees with long-standing findings on inhibitory neuron loss in human AD brain samples. However, the iPSC culture system now gives researchers tools to dissect the mechanisms behind this phenomenon, Huang believes. He is examining whether inhibitory neurons have less tau phosphatase than excitatory ones, allowing more p-tau to accumulate.

The iPSC system also answered the long-standing debate over loss-of-function versus gain-of-function mechanisms for ApoE4. When the researchers generated iPSCs from a person lacking ApoE, the resulting neuronal cultures resembled ApoE3 lines with regard to Aβ, tau, and GABAergic neuron viability. This benign knockout phenotype jibes with some human data (Aug 2014 news). The findings argue against a loss-of-function mechanism for ApoE4, hinting that E4 may be toxic directly, Huang said. Indeed, transfecting the knockout cells with ApoE4 triggered Aβ production, tau pathology, and GABAergic degeneration, whereas changing ApoE4 cells to ApoE3 using CRISPR restored their health. “Based on these data, we should lower ApoE4 to treat AD,” Huang said.

Likewise, the authors rescued ApoE4 neurons by treating them with a molecule that acts as a chaperone, refolding E4 into the E3 configuration. Huang noted that this “structural corrector” molecule is unsuitable for therapeutic development (Brodbeck et al., 2011; Chen et al., 2012). He is working on derivatives that are more potent and better enter the brain, and hopes to take them to the clinic.

Lars Ittner at the University of New South Wales in Sydney said he was particularly intrigued by the fact that the structural corrector lowered tau pathology in ApoE4 cells, whereas reducing the cells’ Aβ levels did not reduce tau pathology. “We certainly need to see if a small molecule corrector of ApoE4 will improve deficits in relevant in vivo models of Alzheimer’s disease,” Ittner wrote to Alzforum.

ApoE4 Promotes Amyloid.

ApoE4 cerebral organoids (top) accumulated many more Aβ deposits (green) than did ApoE3 organoids (bottom). Nuclei are blue, neurons red. *[Courtesy of Neuron, Lin et al.]*

For their part, Tsai and colleagues focused on gene-expression changes in ApoE4 cells. Previous studies found that ApoE4 protein can act both directly on DNA and indirectly via the transcription factor AP-1 to stimulate transcription (Theendakara et al., 2016; Huang et al., 2017). Joint first authors Yuan-Ta Lin and Jinsoo Seo generated iPSCs from a healthy human donor homozygous for ApoE3, then altered both alleles to E4 in some lines. They differentiated these isogenic stem cells into pure cultures of glutamatergic neurons, astrocytes, or microglia, and analyzed gene-expression profiles for each cell type.

In the ApoE4 excitatory neuron cultures, 445 genes differed from their levels in E3 cultures. ApoE4 turned down genes involved in cell proliferation, while dialing up genes involved in cell differentiation. How did this change the cells? E4 cultures secreted 20 percent more Aβ42 and accumulated more early endosomes, suggesting more APP was being processed. In addition, ApoE4 neurons formed 25 percent more synaptic connections than E3s did, with a corresponding bump in excitatory electrical discharges. “ApoE4 neurons seem to be hyperactive. That surprised us,” Tsai said.

Astrocyte and microglial cultures showed many more gene changes than did neurons. In astrocytes, ApoE4 upregulated 418 genes and downregulated 909. Overall, genes involved in lipid metabolism were up, while genes involved in lipid transport and tissue development were down. ApoE4 astrocytes produced much more cholesterol than E3s, while making about half as much ApoE. They cleared only half as much Aβ42 from culture media over two days as did E3s, and degraded it more slowly.

In microglia, ApoE4 boosted expression of 329 genes, while dampening 1,131 more. Upregulated genes mostly pertained to inflammation, while downregulated genes were involved in cell movement and development. Overall, the transcriptome data reflects a shift from a phagocytic phenotype to a more inflammatory one, Tsai said. Supporting this, ApoE4 microglia had fewer and shorter processes than E3s, and took up less than half the Aβ as E3s. These findings dovetail with previous data that suggested ApoE promoted microglial inflammation and curbed phagocytosis (Feb 2015 conference news; Feb 2015 conference news). “[The gene changes] are consistent with several recent results suggesting effects of ApoE on microglia under certain conditions can affect the innate immune response and drive neurodegeneration,” Holtzman noted.

To learn how these different ApoE4 cell types might influence each other, the authors placed iPSCs into a blob of Matrigel in a rotating chamber, which provided a steady flow of nutrients and oxygen. Under these conditions, iPSCs differentiate and self-assemble into mini brain-like structures known as cerebral organoids (Aug 2013 news; Raja et al., 2016). Initially, the organoids contained only neurons and did not accumulate Aβ. Six months later, astrocytes appeared, and at that point ApoE4 organoids accumulated twice as much Aβ and p-tau as did E3 organoids. This demonstrates that the ApoE4 allele alone can cause AD pathology, Tsai noted.

In ongoing work, she is analyzing the gene-expression data to find signaling pathways that are altered in ApoE4 cells and might suggest potential therapeutic targets. Edwin Weeber at the University of South Florida, Tampa, agreed this holds potential, but added, “Time will tell if the utility of human cell lines to better understand fundamental AD pathogenic processes will translate into desperately needed successful human clinical trials.”—Madolyn Bowman Rogers

Comments

- David Holtzman
  Washington University
- Posted: 15 Jun 2018
In Wang et al., the authors clearly demonstrate that there are several phenotypes seen in ApoE-4 versus ApoE3-expressing neurons derived from iPSC. They include increased tau phosphorylation, increased Aβ production, and a decrease in GABAergic neurons. The phenotypes appear clear because they can be reversed by gene editing. This is interesting as it suggests that, despite low levels of ApoE expression by neurons, ApoE4 can somehow still result in a neuronal phenotype, at least under these culture conditions. This provides a new model system to test agents that can reverse these phenotypes as well as study mechanism. What is not clear is which phenotypes are relevant in the in vivo setting when there are many other cell types present, including glia, which produce much more ApoE than neurons.

In Lin et al., the authors assess iPSC-induced neurons, astrocytes, and microglia derived from ApoE3/E3 and ApoE4/E4 individuals. Interestingly, all the cell types have gene-expression phenotypes indicating that ApoE can have effects in all cells in a cell-autonomous fashion. The gene changes are most extensive between genotypes in microglia. This is consistent with several recent results suggesting effects of ApoE on microglia, under certain conditions, can affect the innate immune response and drive neurodegeneration. As with the Lin et al. paper, what is not clear is which phenotypes are relevant in the in vivo setting when there are many other cell types present together. However, this does provide a new model system as well to ask specific questions about ApoE biology.

View all comments by David Holtzman

- Edwin J. Weeber
  University of South Florida
- Posted: 15 Jun 2018
The recent emphasis on Alzheimer’s disease drug development has produced therapeutics with promising results for various stages of pathology in AD mouse models, but unfortunately these strategies fail in human clinical trials. Using human iPSCs (hiPSC) from homozygous ApoE3 and ApoE4 individuals, Huang and colleagues were able to compare AD-related pathology between these two genetic cell lines and in gene-edited isogenic and ApoE-deficient hiPSC lines. Human ApoE4-expressing cells showed many results confirming previous mouse model data in regard to increased tau phosphorylation, greater soluble APP detected, and enhanced Aβ production.

With these cell lines, the question of whether ApoE4 represents a loss of normal ApoE function or a gain of toxic function could be proposed. Reintroduction of ApoE4 into ApoE-deficient hiPSCs recapitulated the ApoE4 hiPSC phenotype, suggesting the latter of the two possibilities. These results were supported by the group’s introduction of a small molecule “structural corrector” that can change the ApoE4 structure in that of ApoE3. Its use ameliorates the pathology established in the ApoE4 hiPSCs, indicating that the secession of the toxic effects from ApoE4 can rescue the major established pathological phenotypes.

What could the mechanism of toxic action be that is specific to ApoE4? Lin et al. may provide insight. Similar hiPSCs from homozygous individuals are used in this study, as well, and also reveal amelioration of pathology in both neurons and glia when a CRISPR/Cas9 approach is used to convert ApoE4 expression to that of ApoE3. Lin et al. observed a reduction of both ApoE4 protein and mRNA in their ApoE4-astrocyte cell line, suggesting a negative transcriptional regulation of ApoE4 on itself. Further transcriptional analysis of ApoE4-expressing neurons, glia, and astrocytes showed significant changes in numerous genes, in particular those genes associated with lipid metabolism and protein transport.

Clearance of Aβ has historically been the focus of ApoE isoform function, but recent reports have minimized the potential for significant in vivo interactions of ApoE and amyloid. Regardless, ApoE4 remains as the greatest genetic contributor for sporadic AD risk. These studies nicely demonstrate the usefulness of hiPSCs and provide insight into the possible consequences of altered gene expression dependent on presence of ApoE4.

Questions of ApoE2 actions on gene expression, peripheral gene targets for AD intervention and small molecule therapeutic design in light of these results have yet to be addressed. Moreover, time will tell if the utility of human cell lines to better understand fundamental AD pathogenic processes will translate into desperately needed success in human clinical trials.

View all comments by Edwin J. Weeber

- Lars M. Ittner
  University of New South Wales
- Posted: 20 Jun 2018
These are two interesting studies shedding new light on APOE4 and its role on Aβ production/clearance by different CNS cell types and neuronal tau pathology.

I was particularly intrigued by the finding in Shi et al. that not reduction of Aβ but the small molecule-mediated change to APOE4 reduced tau phosphorylation in the cells. This provides further evidence for an intimate connection between APOE4 and tau in Alzheimer’s disease, as recently nicely shown by the Holtzman lab (Shi et al., 2017).

We certainly need to see next if a small-molecule corrector of APOE4 will improve deficits in relevant in vivo models of Alzheimer’s disease.

Eds: This comment has been edited to correct the citation.

References:

Shi Y, Yamada K, Liddelow SA, Smith ST, Zhao L, Luo W, Tsai RM, Spina S, Grinberg LT, Rojas JC, Gallardo G, Wang K, Roh J, Robinson G, Finn MB, Jiang H, Sullivan PM, Baufeld C, Wood MW, Sutphen C, McCue L, Xiong C, Del-Aguila JL, Morris JC, Cruchaga C, Alzheimer’s Disease Neuroimaging Initiative, Fagan AM, Miller BL, Boxer AL, Seeley WW, Butovsky O, Barres BA, Paul SM, Holtzman DM. ApoE4 markedly exacerbates tau-mediated neurodegeneration in a mouse model of tauopathy. Nature. 2017 Sep 28;549(7673):523-527. Epub 2017 Sep 20 PubMed.

View all comments by Lars M. Ittner

- Florent Ginhoux
- Posted: 22 Jun 2018
These two complementary papers provide new insight on the mechanisms by which APOE4 could drive Alzheimer’s disease. Both studies used CRISPR/Cas9 and iPSCs to examine APOE4 effects on human brain cell types. In Lin et al., a striking observation is that APOE4 microglia-like cells exhibit impaired Aβ peptide uptake and a clear inflammatory transcriptomic profile. Such new insights were only made possible with the advent of new protocols to differentiate macrophages from iPSCs.

Beyond these exciting APOE4 variant-driven mechanistic insights, and especially the realization that each brain cell type behaves differently, these two studies used an experimental approach that can be extended to other molecules/variants involved in AD as well as other neurological diseases.

With the combined resolution of bulk transcriptomic analysis, as well as the advent of CRISPR-CAS9 gene editing and the development of advanced tissue organoids, it will become possible to understand the specific molecular factors involved in the pathophysiology of disease and how their dysregulation can lead to pathologies in humans as well as lead to novel therapeutic targets. These are exciting prospects for the future.

View all comments by Florent Ginhoux

- Gregory Petsko
  Harvard Medical School and Brigham&Women's Hospital
- Posted: 25 Jun 2018
See also Jessica Young's recent paper, which also shows tau effects in AD patient-derived neurons independent of APP (Young et al., 2018).

As for a small molecule corrector of ApoE4 deficits, yes, that is one approach, but I would argue that a more profitable line of attack based on human genetics is likely to be trying ApoE2 gene therapy.

References:

Young JE, Fong LK, Frankowski H, Petsko GA, Small SA, Goldstein LS. Stabilizing the Retromer Complex in a Human Stem Cell Model of Alzheimer's Disease Reduces TAU Phosphorylation Independently of Amyloid Precursor Protein. Stem Cell Reports. 2018 Mar 13;10(3):1046-1058. Epub 2018 Mar 1 PubMed.

View all comments by Gregory Petsko