While genome-wide association studies tie variants to disease risk, they don’t explain how those variants influence gene expression or pathology. In the March 21 Nature Genetics, scientists led by Vilas Menon and Philip De Jager at New York’s Columbia University reported that many genetic variants affect gene expression not only in certain types of brain cells but in specific subtypes of those cells. They found 8,099 genes so regulated among 424 people in the ROS/MAP longitudinal cohorts. Some of the variants overlapped with AD and Parkinson’s disease loci, hinting that they might modulate pathology. A variant upregulated APOE expression only in microglia. Carriers had worse cerebral amyloid angiopathy than noncarriers but not any more plaques or tangles.

  • Gene variants influence expression by brain cell type and subtype.
  • One that drives TMEM106B expression reduced numbers of excitatory neurons.
  • An SNP near APOE upregulates its expression in microglia.
  • It correlated with worse CAA but not plaques or tangles in AD.

“The work opens new avenues of investigation,” Oleg Butovsky of Brigham and Women’s Hospital in Boston told Alzforum. Guojun Bu at Hong Kong University of Science and Technology, China, agreed. “The new information generated from these association studies provides new clues for functional neurobiologists to validate using animal, cellular, or other model systems,” he wrote.

De Jager, Menon, and colleagues previously ran single-nucleus RNA sequencing on cortical tissue from 192 adults. The samples were from two brain banks in the U.K., and one in the Netherlands. They identified quantitative trait loci (eQTLs) containing single-nucleotide polymorphisms (SNPs) that tinkered with the expression of 7,600 nearby genes. About half of these so-called eGenes were specifically expressed in one of eight brain cell types (Bryois et al., 2022). What about cell subtypes?

To find out, co-first authors Masashi Fujita, Zongmei Gao, and Lu Zeng analyzed whole-genome sequencing and snRNA-Seq data on postmortem dorsolateral prefrontal cortex tissue from 424 ROS/MAP participants. One-third of participants had been cognitively normal, 26 percent were mildly impaired, and 40 percent had dementia at death. Two-thirds were women, and 63 percent of participants were diagnosed with AD based on pathology.

Cells from the prefrontal cortex fell into seven types and 64 subtypes. At the cell type level, the scientists identified 10,000 eGenes, i.e., genes whose expression was changed by eQTLs. For about half of these, expression changes were limited to one cell type. Looking at cell subtypes, they found 8,099 eGenes, including 2,000 not identified at the cell-type level, suggesting that scientists might miss functional effects if they study change in bulk cell types. “Sequencing a larger number of nuclei per participant may be more important than increasing [participant] sample size,” the authors concluded.

Cells and eGenes. Excitatory neurons (Exc) have the most eGenes, i.e., those influenced by genetic variants. Inhibitory neurons (Inh) were next, followed by astrocytes (Ast), oligodendrocytes (Oli), oligodendroglial progenitor cells (OPCs), microglia (Mic), and, finally, endothelial cells (End). [Courtesy of Fujita et al., Nature Genetics, 2024.]

The authors did not report any pattern or functional characteristics among these genes, but they studied some AD-relevant candidates, including APOE. Though all but endothelial cells express the gene, one SNP, rs2288911, upregulated it specifically in microglia. A methylation and an acetylation site, indicative of an active promoter and enhancer, respectively, lay near the SNP, suggesting that may modulate APOE expression. These results also indicate that while genes can be expressed in multiple cell types, or subtypes, some genetic variants may control them in just one subset of cells.

GWAS had flagged rs2288911 as associated with increased AD risk (Mar 2019 news; Apr 2022 news). Intriguingly, participants with or without this variant had similar amyloid plaque and neurofibrillary tangle loads. However, carriers had worse cerebral amyloid angiopathy (CAA). This was independent of APOE4, an allele that increases both forms of amyloid deposition. “[This] raises the possibility that microglial ApoE [regardless of genotype] plays a key role in CAA development,” Bu noted.

The authors think this SNP might contribute to AD by increasing the risk of microhemorrhages. In this vein, they propose rs2288911 might help stratify people who are at risk of amyloid-related imaging abnormalities in anti-amyloid immunotherapy trials (Aug 2023 conference news; Jan 2024 news; Mar 2024 conference news).

Other SNPs were more selective, influencing expression in just a single subtype of cell. The major allele of the intronic SNP, rs5011436, in the TMEM106B  associated with AD in a massive genome-wide association study (Apr 2022 news). The SNP also controlled not just an eGene, but a whole subtype of excitatory neurons marked by expression of the transcription factor Cux2, indicating they likely reside in the upper layer of the cerebrum (Yang et al., 2020). The major allele at rs5011436 decreased numbers of these excitatory neurons while increasing expression of TMEM106B. Too much of this lysosomal protein leads to bloated, dysfunctional lysosomes and can cause frontotemporal dementia (Aug 2012 news). This fine control over neurons and lysosomes suggests how the SNP might impart AD risk.

Risk eGenes? GWAS SNPs for AD (left) and PD (right) co-localized with specific eGenes. Their expression (red scale bar) varied by cell type (color coded bars at top, labels at bottom), with microglia (purple) having the most affected genes in AD and neurons (green) in PD. [Courtesy of Fujita et al., Nature Genetics, 2024.]

Might eGenes coincide with other known AD risk genes? The scientists compared eGenes with a previously published GWAS of AD and dementia (Sep 2021 news). The 20 AD loci they tested mapped to 21 eGenes expressed in various cell types (image above), including, Bin1, Picalm, and CDAP2 in microglia. These cells had the most eGenes, once again implicating them in AD.

A repeat of the analysis using Parkinson’s disease GWAS data mapped 11 PD loci to 13 eGenes (image above; Nalls et al., 2019). Many of these were expressed in excitatory neurons, hinting that these cells are particularly susceptible to damage in PD.

All told, the findings suggest that scientists might need to expand the hunt for functional effects of risk variants to cell subtypes.—Chelsea Weidman Burke


  1. In this comprehensive work by the de Jager team, the authors addressed the effects of genetic variants on cell type- and cell subtype-specific gene expression in human brains. Using a large number of human brains and a wealth of genetic and new single-nucleus RNA-Seq datasets, they discovered a large number of “eGenes” at the cell type and cell subtype levels. Among them, a new gene variant is associated with the expression of the APOE gene but only in the microglia, not in astrocytes. Further, this event is linked to the greater amount of cerebral amyloid angiopathy (CAA), raising the possibility that microglial ApoE plays a key role in CAA development.

    ApoE is known to seed parenchymal amyloids, with ApoE4 exhibiting a greater propensity than ApoE3. The experimental evidence obtained from animal modelling is consistent with human pathological outcomes whereby carriers of the APOE4 allele typically have earlier and more abundant amyloid. APOE is also linked to the abundance of CAA in both animal studies and humans, with APOE4 associated with more CAA than APOE3. How microglia are linked to CAA is not clear but studies from multiple labs including ours have shown that microglial ApoE plays an important role in limiting amyloid plaque development where ApoE4 microglia are less functional. That the new variant affecting microglial ApoE does not associate with parenchymal amyloids could be an indication that it is the APOE genotype and astrocytic ApoE rather than the relatively small changes in microglial ApoE that impact parenchyma amyloids. Nonetheless, the new association discovered here between microglial ApoE level and CAA warrants a test of a new hypothesis—that either microglial ApoE and/or the function of microglia impacted by ApoE level can affect CAA. Microglia could be responsible for seeding CAA as they do for parenchymal amyloids. They could also respond to CAA in a way analogous to, or different from, how they respond to amyloids. Alternatively, microglial ApoE diffused into the vasculature could impact CAA formation by interacting with Aβ to affect its catabolism, and/or it may interact with ApoE receptors to impact the function of vascular cells.

    Overall, I applaud the tremendous work the authors put forward in this study using a large amount of data, combining genetic, epigenetic, and gene expression approaches in human brains. The new information generated from these association studies provide new clues for the functional neurobiologists who will need to validate the findings using animal, cellular, or other model systems.

Make a Comment

To make a comment you must login or register.


News Citations

  1. Paper Alerts: Massive GWAS Studies Published
  2. Paper Alert: Massive GWAS Meta-Analysis Published
  3. Is ARIA an Inflammatory Reaction to Vascular Amyloid?
  4. Brain of Woman Who Died on Leqembi Shows Worst-Case Scenario
  5. FTD Risk Factor Confirmed, Alters Progranulin Pathways
  6. From a Million Samples, GWAS Squeezes Out Seven New Alzheimer's Spots

Mutations Citations

  1. APOE C130R (ApoE4)

Paper Citations

  1. . Cell-type-specific cis-eQTLs in eight human brain cell types identify novel risk genes for psychiatric and neurological disorders. Nat Neurosci. 2022 Aug;25(8):1104-1112. PubMed.
  2. . Cux2 expression regulated by Lhx2 in the upper layer neurons of the developing cortex. Biochem Biophys Res Commun. 2020 Jan 22;521(4):874-879. Epub 2019 Nov 8 PubMed.
  3. . Identification of novel risk loci, causal insights, and heritable risk for Parkinson's disease: a meta-analysis of genome-wide association studies. Lancet Neurol. 2019 Dec;18(12):1091-1102. PubMed.

Other Citations

  1. )

Further Reading

No Available Further Reading

Primary Papers

  1. . Cell subtype-specific effects of genetic variation in the Alzheimer's disease brain. Nat Genet. 2024 Mar 21; PubMed.