At this year’s Alzheimer’s Association International Conference, scientists reported the first substantial single-nucleus RNA-Seq analysis of brain tissue from people who had had familial Alzheimer’s disease. They found subclusters among all main cell types that were unique to autosomal-dominant forms of AD. Similarly, they found subclusters enriched in people who carried known risk variants for late-onset AD. The ADAD and LOAD subclusters were not the same, suggesting perhaps that the brain mounts different cellular responses to different forms of AD.

  • snRNA-Seq identifies unique subclusters of cells in ADAD.
  • LOAD subclusters are driven by genetic variants—and are different.
  • PU(.1)rification yields massive microglial nuclei haul—with some surprises.

Also at AAIC, researchers debuted one of the largest subclustering analyses of microglia to date, based on approximately 90,000 nuclei. Previously identified disease-associated transcriptome signatures did not fall neatly into any one cluster. This implies it may be difficult to therapeutically target specific subclusters of microglia.

ADAD vs. LOAD
The ADAD snRNA-Seq data came from the lab of Bruno Benitez and Oscar Harari at Washington University, St. Louis. First author Logan Brase and colleagues isolated almost 300,000 nuclei from samples of parietal lobes taken postmortem from 67 volunteers registered at the Knight ADRC and DIAN brain banks. Sixteen of these donors had had ADAD, carrying mutations in either the amyloid precursor protein gene or presenilin 1. Other samples came from nine healthy controls, eight people with other diseases, three with presymptomatic AD, and 31 who had had sporadic AD. The latter included people who carried one of four different genetic variants of TREM2, or the rs1582763 protective single nucleotide polymorphism that lies near the MS4A locus.

The Usual Blotches. From almost 300,000 nuclei, snRNA-Seq analysis identifies the main cell types in the brain. [Image courtesy Logan Brase, WashU.]

Brase analyzed gene expression patterns to identify subclusters of brain cells. He found nine distinct types of microglia among 17,089 nuclei, five subclusters among 38,815 astrocytes, four types of excitatory neurons among 38,176, and three subtypes from 15,414 inhibitory neurons. However, looking specifically at nuclei from people who had had ADAD, the subclustering pattern was different.

Among their microglia, subcluster 4 stood out as an ADAD-specific cadre. It predominated in ADAD but was teeny in late-onset AD samples, and all but absent from healthy controls, people with presymptomatic AD, and from brains with other diseases (see image below). Brase found that 5xFAD mouse brain samples contained a microglial subcluster analogous to the human subcluster 4, hinting that the APP/PS mutations might drive the expression signature—5xFAD mice contain three APP and two PS1 mutations.

What does microglial subcluster 4 do? It upregulates genes involved in lipid biology/atherosclerosis, autophagy, and necroptosis, Brase told Alzforum, though he does not know yet what it does in ADAD. Curiously, this cluster bore little resemblance to the DAM, MGnD, and HAM microglial transcriptomes previously linked to AD (Jun 2017 news; Sep 2017 newsMay 2019 news). Instead, those latter AD-related genes overlapped with transcripts upregulated in Brase’s microglial subcluster 1. This subcluster was not enriched in ADAD, but was well-represented in all brain samples analyzed.

Not only microglia distinguished themselves in ADAD. Three oligodendrocyte precursor cell subclusters, two astroglial subclusters, and one each of the oligodendrocyte and neuronal subclusters were enriched in samples from FAD mutation carriers. The transcriptome of one of the astrocyte subclusters resembled disease-associated astrocytes that were recently identified in transgenic mouse models of AD and in human brains (Habib et al., 2020). 

Enriched in ADAD. Among nine microglial subclusters found in ADAD (left), cells in cluster 4 were the most plentiful (orange scale, see asterisk). This cluster was barely detectable (right) in samples from controls (CO), sporadic AD (sAD), presymptomatic AD (Pres), or people with other diseases (OTH). (Asterisk denotes significant differences.) [Image courtesy Logan Brase, WashU.]

What about sporadic AD? The scientists found no subclusters enriched in this group as a whole. However, when they looked at nuclei from people who carried genetic risk variants, differences did pop out. Brains from people with either the R47H, R62H, or H157Y mutations that reduce TREM2 activation had distinct microglial and oligodendrocyte profiles. Specifically, they contained more microglial subcluster 2 and oligodendrocyte subcluster 5 cells. The researchers saw the same pattern among TREM2 carriers in the Religious Orders Study and Memory and Aging Project (ROSMAP) data set.

For the most part, what biological effect these cellular shifts might have in AD remains to be seen. The oligodendrocyte subcluster 5 cells increased expression of TFEB, a regulator of the autophagy/lysosomal pathway that can become compromised in many neurodegenerative disorders. Other WashU scientists have found that these TREM2 variants tone down gene expression responses within microglial clusters in samples from AD prefrontal cortices (Jan 2020 news). 

In contrast to the R47H/R62H/H157Y cases, one person who carried the R136W TREM2 mutation had subcluster densities for microglia, astrocytes, oligodendrocyte precursors and oligodendrocytes that more closely matched the profile of ADAD. “It’s hard to draw any definitive conclusions about this rare mutation because we only had a single subject,” Benitez told Alzforum. “We don’t know what that mutation does, but it would be really interesting to see if it might have a gain of function.” This person was diagnosed with early onset AD, i.e., before age 60, so there could be some links between ADAD to explore, said Benitez.

Next, Brase and colleagues studied the rs1582763 SNP at the MS4A locus. An adenine (A) for a guanine (G) nucleotide swap protects against AD. Here again, carriers shared a unique microglial profile. The GG homozygotes had no microglia subcluster 3. Only a few microglia sorted into this subcluster in the AG heterozygotes, while the cluster bloomed eightfold in the AA homozygotes. The same pattern repeated in the ROSMAP sample, albeit weakly.

Subcluster 3 microglia exhibited a transcription profile consistent with activation, upregulating genes such as NPC1 and TGFβ receptors. Curiously, subcluster 3 comprised nearly half of all microglia in people who had been diagnosed with presymptomatic AD, making these cells fivefold more abundant than in controls. It had all but vanished in LOAD and was absent in ADAD, hinting that perhaps this subcluster reflects an attempt by microglia to combat pathology early in the disease process.

Finally, Brase leveraged this subcluster dataset to test if expression of other LOAD variants might be important to specific cell types. While geneticists have identified at least 75 genetic variants that associate with AD in genome-wide analysis, they do not know what the functional locus is for most of them, or what cell types express the disease-associated variant (Feb 2021 news). 

Brase focused on expression of several dozen of the most notable GWAS hits, including Bin1, SORL1, PICALM, etc., in cell subclusters. “The idea is that if a gene is differentially expressed in subcluster 1 vs. subcluster 2, then that would highlight the importance of that gene in that cell type,” he said. In this analysis, some expression patterns turned out as expected. For example, differential expression of APOE was high among astrocytes and microglia, cells known to produce ApoE.

But there were surprises, too. PLGC2, thought to be predominantly microglial, was differentially expressed across the board, and strongly in neurons and oligodendrocytes. SORL1, typically associated with neurons, had a similar pattern. “The findings suggest that many GWAS genes may have a much broader role in disease than in the cells we typically associate them with,” said Brase.

It's Not All About Subclusters
Cell subclusters are defined by their shared preponderance of commonly expressed genes, but that doesn’t mean the expression pattern within a given subcluster is fixed. Katherine Prater, who works in the lab of Suman Jayadev at the University of Washington, Seattle, emphasized this when she presented one of the largest subclustering analyses of microglia to date.

Prater and colleagues got around a central problem facing scientists who study microglia using snRNA-Seq, namely that these cells represent only 3 to 5 percent of the nuclei typically recovered from the human brain. Hence, some of the early RNA-Seq analysis of AD microglia came from as few as several hundred or thousand cells. To grow that paltry catch, Prater and colleagues developed a fluorescent activated sorting method that homes in on nuclei carrying the essential microglial transcription factor PU.1.

Their system worked. Prater greatly boosted the number of microglia isolated from brain. In samples from four people, she was able to identify about 1,000 microglial nuclei using traditional methods and 23,300 after PU.1 sorting.

Prater used this enrichment method to isolate microglial nuclei from flash-frozen prefrontal cortices obtained postmortem from 22 people, average age 86. Twelve had had AD as determined by neuropathology, 10 were controls. Prater isolated about 4,000 microglia nuclei from each sample for a whopping 90,000 in toto. Researchers in Denmark recently achieved a similar enrichment, not by capturing microglial nuclei, but by throwing away those that come from neurons or oligodendrocytes (Gerrits et al., 2021). 

Prater identified 10 microglial subclusters, each with distinct expression profiles that suggest specific functions, such as homeostasis, cell cycling, stress responses, immune responses, and cell death. Notably, some subclusters that had been identified from previous small hauls of microglia did not emerge here (May 2019 newsJun 2017 news). Instead, their expression signatures spread across several of Prater’s subclusters. DAMs fit this scenario. “We probably don’t want to be fixated on DAMs,” Prater told Alzforum. “There is more to the changes seen in AD than just ‘DAMs appear,’ or ‘DAMS are dysregulated.’ Other subtypes are changing that we can’t just ignore.”

Indeed, Prater found that even within a given subcluster, differences emerge in AD. Cells in the homeostatic cluster, for example, were more likely to express antigen-processing and antigen-presenting genes, or immune response or endocytosis genes if the cells came from a donor with AD. Ditto for cells in the metabolism and endocytosis microglia subcluster.

“We see in AD an enrichment of inflammatory pathways and genes across many, but not all subtypes of microglia,” Prater said. “We need a lot more data to find out if we can target specific subtypes, or if it’s even possible to influence ‘bad’ microglia without affecting other microglia.”—Tom Fagan

Comments

No Available Comments

Make a Comment

To make a comment you must login or register.

References

Research Models Citations

  1. 5xFAD (B6SJL)

News Citations

  1. Hot DAM: Specific Microglia Engulf Plaques
  2. ApoE and Trem2 Flip a Microglial Switch in Neurodegenerative Disease
  3. When It Comes to Alzheimer’s Disease, Do Human Microglia Even Give a DAM?
  4. Human and Mouse Microglia React Differently to Amyloid
  5. Massive GWAS Meta-Analysis Digs Up Trove of Alzheimer’s Genes

Mutations Citations

  1. TREM2 R47H
  2. TREM2 R62H
  3. TREM2 H157Y
  4. TREM2 R136W

Paper Citations

  1. . Disease-associated astrocytes in Alzheimer's disease and aging. Nat Neurosci. 2020 Jun;23(6):701-706. Epub 2020 Apr 27 PubMed.
  2. . Distinct amyloid-β and tau-associated microglia profiles in Alzheimer's disease. Acta Neuropathol. 2021 May;141(5):681-696. Epub 2021 Feb 20 PubMed.

Further Reading