Immunity: dual developmental features of disease-associated microglia and disease-inflammation-associated macrophages in aging and neurodegeneration

Guide:

The intracerebral macrophage population includes parenchymal microglia, border-associated macrophages, and recruited monocyte-derived cells
.
Together, these cells control brain development and homeostasis, and are also implicated in
the pathogenesis of aging and neurodegeneration.
However, the phenotype, localization, and function of each cell population in different contexts have not been resolved
.
In a recent research paper published in the journal Immunity entitled "Dual ontogeny of disease-associated microglia and disease inflammatory macrophages in aging and neurodegeneration," the authors' team established a mouse brain myeloid scRNA-seq integration to systematically describe the brain macrophage population
。 It was found that the disease-associated microglia (DAM) populations previously identified in mouse Alzheimer's disease models actually included two distinct cell lineages that were genetically and functionally distinct on the part: i.
e.
, embryonic-derived trigger receptor myeloid cell 2 (TREM2)-dependent DAM expressing neuroprotective signals, and disease inflammatory macrophages (DIMs)
expressing monocyte-derived TREM2 that accumulate in the brain during aging.
And these two different cell populations also seem to be conserved in the human brain
.
Here, the authors establish an analytical model of the heterogeneity of brain myeloid cells in development, homeostasis, and disease, and identify cell targets for the treatment of neurodegeneration
.

1.
M-Verse: Generate a universal ensemble to identify murine brain macrophage subsets across multiple datasets

Given that different single-cell RNA sequencing (scRNA-seq) studies have different approaches and conclusions on macrophage subsets in mouse brains, the authors' team first aimed to integrate the six datasets collected in other studies and the two internal datasets generated using Seurat (Silvin IS, Silvin C1).

。 To ensure the resulting integration—the authors' team named it "M-Verse" (Figure 1A) and selected a published dataset to cover a wide range of conditions within the central nervous system: one focused on comparing microglia and neurons, interneurons, oligodendrocytes, and astrocytes in juvenile and adult mice (Zeisel); the other tracked the transition from embryos to senescent microglia at E14.
5, P4, P5, P30, P100, and P540 (Hammond); There is also a microglia in neurodegeneration, which contains the first description of DAM (Keren-Shaul).

The dataset from Van Hove also adds a spatial dimension to the analysis as it includes cells from different brain regions: choroid plexus, dura, subdural regions (SDMs), and brain parenchyma (whole brain).

The authors' team's dataset includes microglia isolated from embryonic (E12.
5, E16.
5, and E18.
5), early postnatal (P7, P14), and adult (P60) mouse brains, as well as CD45+ macrophages
from adult mice (P60).
The authors' team expects to put these disparate datasets together to build a powerful tool to resolve the heterogeneity
of mouse brain macrophages.

MVerse (https://macroverse.
gustaveroussy.
fr/2021_M-VERSE) allowed the author team to extract expression data
for 14,794 common genes detected across all datasets 。 By comparing these datasets, the authors' team generated an unbiased global map of myeloid cells in the brains of developing and adult mice and identified five regions in the map (Figure 1A): one corresponding to non-myeloid cells (including a cluster corresponding to neurons and one cluster corresponding to interneurons, oligodendrocytes, and astrocytes); One corresponds to T cells and NK cells; a cell rich in expressing the characteristics of developmental microglia; a cell rich in characteristics of mature microglia; and a fifth myeloid cell
containing non-microglia.

When the authors' team projected data from each study into M-Verse separately (Figure 1B), they found that the Van Hove dataset was rich in myeloid cells, possibly due to scRNA-seq effects
in the meningeal and choroid plexus regions.
The Silvin C1 and Hammond datasets are the only two datasets containing embryonic and early postnatal microglia, and both produce a well-defined cluster
of developmental microglial characteristics 。 The Zeisel dataset contains a high proportion of non-myeloid cells (differentiated neurons [Tubba2, Sox11], interneurons [Slc32a1, Pnoc], oligodendrocytes [Hapln2, Opol1], and astrocytes [Slc1a2, Slc1a3], clearly visible near the neuronal region of M-Verse [Stmn1, Tub2b] (Figure 1B)).

Nonetheless, the expression profiles of microglia in the Zeisel dataset overlapped completely with those in other datasets, indicating that the integration was successful
.
Notably, the Zeisel dataset also contains some microglia that cluster in regions of developmental characteristics, possibly due to their use of juvenile or juvenile mice (from 21 days to 1 month).

。 Similarly, the authors' team observed a small number of cells in the characteristic region of developing microglia in 60-100-day-old mice in the Hammond, Silvin C1, and Silvin IS datasets (Figures 1A, 1B), and increased cells in that region in aged mice with neurodegenerative diseases
compared to older WT mice (Keren-Shaul and Van Hove) (Figure 1B).
This suggests that in these neurodegenerative disease models, some microglia may acquire development-like programs
.

2.
Refine M-Verse using directed index sorting data from mouse brain macrophages

The scRNA-seq method provides valuable data that can facilitate more precise identification
of cell subpopulations by in-depth analysis of a small number of cells to complement the results of extensive analyses.
Therefore, the authors' team next analyzed the Silvin IS dataset, which combines protein expression information with deep gene expression measurement information
, using Seurat V2 (shown in red in Figure 1A).
This dataset contains information from 213 adult murine brain CD45+Ly6C-F4/80+CD11b+ macrophages and is generated
using the Smart-Seq2 method.
The method is able to measure a similar number of genes as 10X Genomics, with a higher number of
fragment reads per cell.
This enables binding with IS to isolate individual cells
with known characteristics such as defined size, particles, and selected expression markers.
To definitively identify border-associated macrophages (BAMs) in traditional gated microglia populations, the authors' team used cells from 2-month-old Lyz2GFP mice containing GFP-labeled lysozyme-expressing BAMs (as opposed to microglia).

The 213 sequenced cells were then clustered using Seurat V2 and Phenograph methods and three clusters were determined based on their scRNA-seq expression profiles: one large cluster 1 and two small clusters 2 and 3
.

To further characterize these three clusters, the authors' team obtained the expression
of each set of "index" surface molecules by flow cytometry.
Secondary clusters 2 and 3 expressed higher CD45 and Lyz2GFP, while major cluster 1 expressed moderate amounts of CD45, F4/80, CD11b and Ly6C and a small amount of Lyz2GFP (a microglia-like phenotype).

To better understand the composition of clusters 2 and 3, the authors reanalyzed Lyz2GFP+ cells (BAMs)
independently of other cells using previous parameters.
Based on gene expression profiles, cluster 2 is resolved into two subclusters and annotated as 2a and 2b (Figure 1C, left).

Then, based on the flow cytometry data, it can be found that cluster 2a is more like microglia (cluster 1), while cluster 2b is clearly separated from cluster 1, especially in which it expresses a large number of MHCII proteins
.
To further investigate the characteristics of the cluster, the authors' team used differentially expressed gene (DEG) analysis, which showed that cluster 1 corresponds to microglia (high expression of Tmem119).

Although Cluster 2 is characterized by high expression of the BAM marker Mrc1 (Cd206), Cluster 2b exhibits higher H2-Ab1, H2-Aa, and Cd74 expression in RNA and protein amounts than Cluster 2a (Figure 1D).

Finally, cluster 3 represents a unique cell population characterized by the expression of Cd11a, Cd44, Cd300e, Plac8, Nr4a1, Ear2, and Fcgr4 (Figure 1D).

Among them, there is a correlation between the RNA and CD11a proteins of Cd11a in cluster 3 cells (Figure 1D).

The authors' team then used the new depth data from IS scRNA-seq (Silvin IS) to generate reliable features
for each cell population in M-Verse.
This allowed the authors to confirm that CD11a+ (Cluster 3) and CD206+ macrophages (Cluster 2a and 2b) and microglia (Cluster 1) were significantly aggregated in M-Verse (Figure 1C).

These features are then projected onto M-Verse to extend and verify their heterogeneity on other datasets
.
Probing CD11a+, CD206+MHCII- and CD206+MHCII+ macrophage features enabled the authors' team to highlight the number of cells containing similar gene patterns in specific myeloid cell regions (Figure 1E).

By focusing on these regions (Figure 1F), the authors' team identified regions corresponding to BAMs common to all datasets (Figure 1G).

This approach also enables better identification of cells in other published studies: when the authors used the Van Hove choroid plexus BAM (CPepiBAM) marker to study the regions of its dataset containing choroid plexus cells, they found that the cells expressing this trait were actually located primarily within the mature and developing microglial region defined by M-Verse (Figure 1H).

Therefore, the cells that were thought to be BAMs in the original study are likely to be microglia: the published results of the expression of the transcription factor Sall1 by the choroid plexus 'BAM' further support this idea
.
Because it was previously found that the transcription factor is only expressed
in microglia in brain macrophages in the central nervous system of mice.

In summary, integrating the scRNA-seq dataset of six brain macrophages generated using different techniques enabled the authors' team to perform a global cross-comparison
of developing and adult mouse brains.
A consensus set of mRNA marker transcripts was determined by further application of IS single-cell analysis to improve the resolution of M-Verse, enabling reliable and accurate identification of mouse microglia and brain macrophage populations
across multiple life stages.

3.
M-Verse directional analysis solves the problem of the origin of BAMs

Based on the analysis of M-Verse and signature differentially expressed genes (DEGs)-encoded proteins, the authors' team designed a rigorous flow cytometry gating strategy to identify brain macrophage populations
as defined above.
These marker genes and proteins include Mrc1 (encoding CD206), Itgal (encoding CD11a), and H2-Ab1 (encoding MHCII) (Figure 1D).

The authors first evaluated cell suspensions of the entire young adult mouse brain (brain, pimater, arachnoid and dura), the meninge-removed brain (brain and pimater), and the meninges alone (arachnoid and dura
).
As expected, CD11a+ and CD206+MHCII+ macrophages are enriched in the meningeal fraction, while the brain-only fraction is enriched with CD206+ MHCII- cells (Figure 1I).

Therefore, this gating strategy based on comprehensive analysis to identify marker DEGs enabled the authors' team to accurately identify two subsets of microglia and differentially MHCII-expressing CD206+BAMs, as well as macrophages subsets
expressing integrin CD11a.

After establishing these baselines, the authors tracked the contribution of blood monocytes to the cell population using the recently described Ms4a3 fate mapping model: less than 1% of microglia were labeled, while more than 87% of CD11a+ macrophages were labeled, suggesting that the
latter were predominantly monocyte-derived.
The monocyte contribution rate of CD206+ MHCII+ macrophages (20%) was higher than that of CD206+ MHCII-macrophages (<5%) (Figure 1J).

However, the authors' team also observed that the percentages of CD206+ MHCII+ monocyte-derived cells (26%) and microglioid monocyte-derived cells (2%-3%) were slightly higher in mice older than 6 months (Figure 1K).

Together, these results suggest that meningeal CD206+MHCII+ macrophages have a higher monocytic composition and are therefore replaced to a greater
extent than previously thought, compared to perivascular macrophages.

Fig.
1 Overall consensus on the heterogeneity of macrophages in M-verse rat brain

4.
Phenotypic heterogeneity within DAM populations

DAM is a group of CNS macrophages thought to exert a protective role in mouse models of Alzheimer's disease (AD) and is also present in the brains of
human AD patients.
Another study showed that DAMs also accumulate in the brains of aged mice, but whether they have a protective effect is still controversial
.
Initially scRNA-seq analysis was used to determine that DAM was highly expressed in Cd11c, Csf1, and Cd9 and low in Tmem119, P2ry12, and Cx3cr1, and depended on trigger receptors
expressed on myeloid cell 2 (TREM2).
Currently, the proprioceptive development of DAMs remains elusive, and recent studies have come to conflicting conclusions that they may be related to
BAMs or monocytes.

Therefore, the authors' team sought to investigate the properties
of DAM in the extended brain macrophage population throughout development, adulthood, aging, and neurodegeneration through M-Verse.
The authors projected Keren-Shaul's DAM gene expression signatures onto a Keren-Shaul dataset with M-Verse common coordinates and found that cells with high DAM gene signatures were located in two distinct regions: some in the developmental microglial region (DAM) and others in the mature microglial region (Figure 2A).

Because these cells are highly expressed in inflammatory genes, the authors define them as disease-inflammatory macrophages (DIMs).

To assess whether these two subpopulations expressing DAM traits are equally dependent on TREM2, the authors' team compared their relative abundances from the Keren-Shaul dataset in the brains of elderly C57BL/6 WT, 5XFAD (AD model), TREM2-/- and 5XFAD-TREM2-/- mice
.
The results showed that the proportion of DAM cells in the cerebral cortex of AD-TREM2-/- mice was about half that of AD mice, while the proportion of DIMs in all TREM2-/- mice increased significantly (Figure 2B).

At the RNA level, both populations highly express Trem2 (Figure 2C), but there are differences in the expression of other markers: DAM has relatively low expression of P2ry12 and Cx3cr1 compared to DIMs, while Itgax (Cd11c) has higher expression (Figure 2C).

Therefore, M-Verse, combined with the comparison of cell molecules in two regions, can not only identify the DAM marker gene expression status shared by the two cell populations, but also uniquely describe the molecular functional differences in gene expression that coincides with it
.
These data strongly suggest that DAMs, rather than DIMs, represent the real DAM
described by Keren-Shaul.

5.
True mature DAM shows gene expression characteristics that overlap with YAM

The authors then went on to investigate whether DAM cells could be detected in other datasets used to generate M-Verse and found that they were also present in the Van Hove dataset containing cells from APP-PS1 AD mouse models (Figure 2D).

In both datasets, cells expressing DAM characteristics were contained within the developing microglial region, suggesting that mature true DAM shares the same gene expression pattern
as typical embryonic microglia.
Therefore, the authors took the DEG analysis a step further, comparing DAMs and DIMs in the two datasets in an attempt to define a set of genes that are more expressed in DAM and represent their truly specific characteristics (Figure 2E).

Results Eight DAM-specific and highly expressed genes were found in two datasets: Dkk2, Fabp5, Gpnmb, Igf1, Itgax, Mamdc2, Spp1 and Gm1673
.
The authors' team then projected the average expression of these characteristic genes onto two datasets within M-Verse (Keren-Shaul and Van Hove overlay) to visualize their expression specificity (Figure 2F) and assess their proportion in the different mice used in these studies (Figure 2G).

。 Cells from the Silvin C1 and Ham-mond datasets (WT mice) appear in the DAM region (Figure 2H), but in this case, also correspond to cells from embryonic (E12.
5, E14.
5, and E18.
5) and early postnatal (P4, P5, and P7) time points (Figure 2I).

In conclusion, the authors' analysis showed that adult DAM shares common gene expression characteristics with developmental (fetal and early postnatal) microglia in non-AD mice, including the expression of integrin Itgax (CD11c), which has been shown to be expressed
on a subset of microglia called P7 CD11c+.
To understand the possible relationship between these two separately described cell types, the authors' team projected P7 CD11c+ transcriptome features onto M-Verse and observed significant overlap with DAM features from the Keren-Shaul dataset (Figures 2J and 2A).

。 And five of the seven genes in the DAM conserved profile (Dkk2, Gpnmb, Igf1, Itgax, and Spp1) were also expressed in P7 CD11c+ cells from the Hagemeyer and Wlodarczyk datasets (Figure 2K), as well as in the Hammond and Sil-vin C1 datasets at embryonic time points (Figures 2H and 2J).

Thus, the DAM conservation features seen in neurodegenerative diseases are very similar
to the group of cells present in healthy wild-type mouse embryos and newborn brains.
Based on these observations, the authors' team recommends calling CD11c+P7 microglia youth-associated microglia (YAM).

Next, the authors' team wanted to better understand the similarities and differences
between these two subsets of microglia with similar gene expression characteristics but in very different environments.
So they performed a DEGs comparison path analysis
between YAM from the Silvin C1 and Hammond datasets and DAM from the Keren-Shaul and Van Hove datasets.
Both YAM and DAM express a large number of genes associated with phagocytosis, autophagy, and mitochondrial metabolism, but YAM additionally expresses genes associated with fatty acids, proline, and glutamate, while DAM alone expresses genes associated with anti-inflammatory response (Figure 2L).

Thus, mature DAM appears to undergo fetal-like reprogramming, transcribing fusion into embryonic and neonatal YAM
.
However, although YAM and DAM share a common gene expression pattern, the former occurs naturally in embryogenesis and early postnatal life, representing a population of microglia associated with brain development; DAM, on the contrary, appears against the background of neurodegeneration and may be involved in restoring brain homeostasis
.

Figure 2 The DAM feature includes two distinct cell populations

6.
DIMs appear during aging and develop in neurodegenerative diseases

The authors' team then went on to explore the nature of the DIM population and first detected the presence of DAM during neurodegeneration in mice (Figure 2A).

When examining other scRNA-seq datasets, they also detected DIMs in healthy aged mice in the Hammond, Keren-Shaul, and Van Hove datasets (Figure 3A).

To understand the dynamics of how these cells appear during the mouse life cycle, the authors' team quantified their relative abundance at different time points in the Hammond and Silvin C1 dataset: DIMs appeared after birth and increased dramatically in mice older than one year (Figure 3B).

In the neurodegenerative state studied in the Van Hove and Keren-Shaul datasets, DIMs were more abundant in disease mouse models compared to age-matched WT mice (Figure 3C).

Surprisingly, the proportion of DIMs in AD-TREM2-/- mice in the Keren-Shaul dataset increased dramatically compared to DAM, showing their TREM2 independence (Figure 2B, 3C).

。 By comparing gene expression data from the Hammond, Keren-Shaul, and Van Hove datasets, the authors' team then established a DIM conserved marker feature that contains genes associated with pro-inflammatory response and immune activation, including Il1a, Il1b, Tnf, Nfkbia, Cd49f, Cd54, and Cd83 (Figure 3D, 3E).

This feature is enriched in the DIM region compared to other regions of M-Verse
.
Therefore, given the high expression of inflammation-related genes and their increased proportion in the context of neurodegenerative diseases, the authors' team named this cell population as inflammatory macrophages (DIMs)
that express TREM2 (Figure 2C) but do not depend on TREM2 (Figures 2B, 3C).
。 By analyzing pathways specific to DAMs and DAMs, the authors observed that DIMs expressed a number of genes involved in inflammation-related pathways, including Tnf, Il6, and Il1, as well as toll-like receptor signaling and production of nitric oxide (NO) and reactive oxygen species (ROS) (Figure 3F), consistent
with potential pro-inflammatory states 。 However, the authors' team also observed several immunosuppressive pathways, including IL-10 signaling, glucocorticoid receptor signaling, and peroxisome proliferator-activated receptor (PPAR) signaling pathways (Figure 3F), primarily due to the fact that the expression of genes such as Ccr5, Il10ra, Tnf, Hsp90aa1, Cited2, Icam1, Ccl2, Stat1, FCGR1a, and Adbr2 may reflect a feedback loop that integrates the inflammatory environment
。 However, the authors did not observe significant expression
of immunosuppressive cytokines and chemokine-coding genes.

Figure 3 Increased proportion of DIMs in aging and AD mouse models

7.
Brain inflammation induces the accumulation of CD83+DIMs derived from monocytes

In addition to the inflammatory environment characterized by neurodegeneration in mice, the authors' team also wanted to understand whether DAM/DIMs-like cells
develop when CNS inflammation occurs in other situations.
In humans, exposure to skull ionizing radiation causes progressive cognitive dysfunction and neuroinflammation, which is associated
with the release of pro-inflammatory cytokines such as TNF-a, IL-1a, and IL-6 in vivo.
Radiation exposure and homologous bone marrow (BM) metastasis after depletion of microglia in mice lead to the penetration of monocytes into the brain and the production of microglia-like cells
.
To link these observations, the authors' team studied the presence of monocyte-derived macrophages and microglia-like cells in the brain after ionizing radiation to see if they might be related to
DIMs or DAMs.
BM chimeras were generated by irradiating 1-month-old C57BL/6 WT mice and reconstructing their BMs with cells from Ms4a3CreRosaTomato mice, and then their brains
were analyzed after 5 or 10 months.
After transplantation, the number of Ms4a3-TdT+ monocytes exceeded 96%, but in the steady state, even 10 months after reconstruction, less than 25% of brain macrophages (according to CD45, CD11b, F4/80 expression and low-expression Ly6C) expressed Ms4a3CreRosaTomato
.
Conversely, monocyte-derived macrophages increased significantly after irradiation, reaching more than 20% at
subsequent time points.
These Ms4a3-TdT+ macrophages express more F4/80 and CD83 and less CD206
than non-monocyte-derived macrophages.
The authors then confirmed the presence of Ms4a3-TdT+ macrophages in the brain
through microscopic observation.
This suggests that Ms4a3-TdT+ cells also express IBA1+ and are distributed in different brain regions
.

Of particular excitement to the authors' team is the high expression of CD83 on Ms4a3-TdT+ macrophages in the hindbrain, as CD83 is also one of
the genes specifically highly expressed in DIMs by M-Verse.
Therefore, the authors' team classified and compared
CD83+Ms4a3-TdT+ macrophages, CD83+Ms4a3-TdT-microglia, and CD83-Ms4a3-TdT-microglia by transmission electron microscopy (TEM) and H&E staining.
It was found that CD83+Ms4a3-TdT+ macrophages were larger and contained more intracellular vesicles than microglia, which may correspond to lipid droplets
visible in TEM.
The authors also performed batch RNA sequencing on these populations to compare their gene expression profiles and understand their relationship to DAMs and DIMs (Figure 4A).

。 Classification according to CD45, F4/80, CD11b, Ly6C, CD83, and Ms4a3-TdT (gating strategy shown in Figure 4B): approximately 30% of brain macrophages correspond to CD83-Ms4a3-TdT-microglia, 30% to CD83+Ms4a3-TdT-microglia, and 40% to CD83+Ms4a3-TdT+ macrophages (Figure 4C).

。 The authors' team first used gene set enrichment analysis (GSEA) based on the broader DIM gene signature established by M-Verse to investigate whether any of these three populations correspond to RNA-level DIMs.

GSEA shows significant enrichment of DIM features in CD83+Ms4a3-TdT+ macrophage populations (Figure 4D).

Extending this to protein content analysis, CD83+Ms4a3-TdT+ macrophages expressed similar amounts of CD11c and CD11b compared to CD83-Ms4a3-TdT-microglia, but had slightly lower amounts of P2RY12 and higher MHCII levels at their surface (Figure 4E).

。 There were significant differences in the expression of 68 genes between CD83+Ms4a3-TdT+ macrophages and CD83+Ms4a3-TdT-microglia, including Axl, Irf7, and Clec12a, as well as genes involved in neuroinflammatory signaling pathways that were uniquely upregulated
in CD83+Ms4a3-TdT+DIMs 。 These results suggest that irradiated hindbrain monocyte-derived macrophages are characterized by high expression of CD83 and F4/80 on their surface and expression of genes
corresponding to the DIM population defined in M-Verse.

Finally, the authors' team used the Ms4a3CreRosaTomato model to monitor the number of DIMs in the brains of mice 2 months old and older than 6 months of age and to confirm their accumulation
during aging.
Using a similar gating strategy (Figure 4F), the proportion of DIM in young adult mouse (2 months old) macrophages (Figure 4F and 4G) was significantly increased
compared to adult mice (> 6 months of age).
In contrast, the proportions of CD83-Ms4a3-TdT- and CD83+Ms4a3-TdT-microglia did not change
during this period.
These DIMs are localized in the hippocampus, thalamus, and cortex (Figure 4H).

In summary, DIMs are monocyte-derived macrophages that accumulate
in the brain with age.

8.
Monocyte-derived DIMs and YS-derived DAM accumulate in the brain of 5XFAD mice

The authors' team then used flow cytometry techniques to evaluate the presence of DIMs and DAMs and their ontogeny
in the brains of 12-month-old C57BL/6-Ms4a3CreRosaTomato, 5XFAD-Ms4a3CreRosaTomato, C57BL/6-TREM2-/- and 5XFAD-TREM2-/- mice 。 Applying the previous gating strategy (Figure 4B) and adding antibodies against CX3CR1, CD11c, CD206, and CD11a, the authors observed an increase in the percentage of DIMs (CD83 and F4/80 high expression, dark red dots) and DAM (CD11c high expression and CX3CR1 low expression, dark blue dots) in AD mouse brain macrophages (Figure 4I and 4J).

。 As described in the Keren-Shaul dataset, the proportion of DAM was significantly reduced in TREM2-/- and AD-TREM2-/- mice, and the authors found a significant increase in the proportion of DIM in both models (Figures 4I and 4J).

Most DIMs express large amounts of the Tomato reporter, confirming their monocytic origin; DAMs, on the other hand, confirmed their embryonic origin (Figures 4K and 4L).

Phenotypically, the expression of CD45, F4/80, CD83, and Ly6C of DIMs was increased in the AD model compared to microglia, while the expression of CD45, CD83, CD206, CD11c and CX3CR1 decreased in DAM (Figure 4M).

。 These observations suggest that DAM and DIMs are two different cell lineages on proprioception, distinguished by the expression of F4/80, CD83, CX3CR1, CD11c, and CD206, and exhibit different dependencies on TREM2, but both accumulate
during neurodegeneration.

Figure 4 DIMs are derived from monocytes, while DAMs are derived from embryos

9.
DIMs and amyloid β protein polymers are co-localized in the brain of AD mice

The authors' team next used the 5XFAD-Ms4a3CreRosaTomato model to study the
distribution of DIMs in the AD brain.
Whole brain sections were studied by immunofluorescence microscopy to observe that DIMs are not uniformly distributed, but are distributed in clusters in the hippocampus, meninges, almond basolateral nucleus, and intracortex (Figure 5A and 5B).

The authors also confirmed their phenotype in vivo: Ms4a3-TdT+DIMs expressed similar amounts of Iba1 compared to Ms4a3-TdT-microglia (Figures 5C and 5D), while the amounts of F4/80 increased significantly (Figures 5C and 5E).

The authors' team then examined the DIMs clusters in more detail using spatial density estimation and found significant overlap between the DIMs clusters and some brain regions with high amyloid β (Aβ) aggregation densities (Figure 5F and 5G).

At the single-cell level, 3D image reconstruction shows DIMs in close contact with Aβ aggregates of various sizes, sometimes almost as large as the DIMs themselves (Figures 5H and 5I).

However, there appears to be population-level limitations to the degree of colocalization of DIM with Aβ aggregates: local DIM density increases until a certain regional aggregate density is reached (plaque density from 0.
0010 to 0.
0013).

The number of DIMs above this density decreases sharply while Ms4a3-TdT-microglia continue to aggregate (Figure 5J).

These results suggest that DIMs accumulate in specific brain regions (hippocampus and basolateral amygdala), inconsistent
with DAM showing aggregation areas in the cerebral cortex.
Thus, DIMs can be found near Aβ plaques, but only at a certain aggregation density, above which microglia are more likely to come into contact with
Aβ plaques.

Figure 5 DIM aggregation in specific brain regions in 12-month-old 5XFAD mice

10.
The equivalents of DIM, DAM and YAM are present in the human fetus and AD brain

To understand how these observations relate to the human brain, the authors' team first reanalyzed the adult scRNA-seq brain dataset of Thrupp et al.
using Seurat V3 and identified 18 cell clusters, including BAMs (MRC1, F13A1), CD14+ and CD16+ monocytes (FCGR1A, FCGR3A, HMGB1), microglia (CX3CR1, P2RY12) and contaminating astrocytes (SNAP25) and oligodendrocytes (PLP1).

When the authors looked at the microglial cluster in more detail, nine subclusters emerged (Figure 6A) and similar DEGs were found in mouse DIMs or DAM signatures: clusters 11Up, 11Down, and 9 differentially expressed some DAM characteristic genes, while clusters 12, 14, and 15 differentially expressed some DIM characteristic genes (Figure 6B).

。 However, only cluster 11Up cells had significantly lower expression of CX3CR1 and P2RY12, while expression of GPNMB, SPP1, APOE, TREM2, CD63, and ITGAX increased, which corresponded to murine DAM expression profiles (Figure 6B).

In addition, CD83, EGR1, IER2, FOS, JUN, and AXL were all expressed higher in 12 clusters than in other clusters, corresponding to murine DIM expression profiles (Figure 6B).

。 Overall, cluster 11Up highly expressed 8 of the 20 genes (B2M, CD63, MAMDC2, CCL3, GPNMB, SPP1, TYROBP, and TREM2) with extensive DAM characterization in mice, while cluster 12 expressed 15 of the 40 genes with a wide range of mouse DIM characteristics (CCL4, CD14, CD83, CSF2RA, EIF1, FOS, IER2, JUN, JUNB, IL1B, TNF, PLAUR, SAT1, and BTG2) (Figure 6C).

The authors' team then used the total DEGs of clusters 11Up and 12 to identify the pathways for enrichment of these cells and again found a large amount of overlap: 8 of the first 15 pathways in cluster 11up were also enriched in mouse DAM (pathways highlighted in blue Figure 6D), and 7 of the top 19 paths in cluster 12 were shared with human DIMs (paths highlighted in red in Figure 6D).

Therefore, the authors suggest that human cluster 11up may be comparable to rat DAMs, while cluster 12 may correspond to
mouse DIMs.

To identify human YAM, DAM-like cell populations present in normally developing brains, the authors' team integrated the Thrupp dataset with the scRNA-seq dataset (13- to 18-week fetus) from Kracht et al.
and clustered
the Kracht dataset using parameters similar to the Thrupp dataset 。 By applying the human DIMs (cluster 12) and DAM (cluster 11Up) parameters generated above, the authors were able to not only identify these clusters in the integrated Thrupp dataset (Figure 6E), but also assess their similarity in the integrated Kracht dataset (Figure 6F).

This revealed clear clusters of DAM-like cells in the fetal brain from the Kracht dataset, thus representing the human equivalent of murine YAM
.
Human YAM characteristics can be defined by looking at the transcriptional data of the cells in this cluster (Figure 6F).

The authors' team also found a similar pattern in the relative abundance of DIMs, DAMs, and YAM in embryos and older adults: a higher proportion of DIMs and a lower proportion of DAM in the adult Thrupp dataset (Figure 6G); In the fetal Kracht dataset, the proportion of YAM is high and the proportion of DIMs is low (Figure 6H).

Human YAM also expresses a large number of genes associated with metabolic pathways similar to murine YAM, including pathways
related to proteolysis, mitochondrial tissue, glutathione redox, and oxidative phosphorylation (OXPHOS).
These cells further showed higher lipid alteration and degradation pathway enrichment compared to murine cells
.
Among the ubiquitous cellular processes and signaling pathways of mouse and human YAM, the authors identified neurodegenerative diseases and brain development pathways, further suggesting that DAM procedures and functions are indeed programs
shared with YAM during embryonic development.
In conclusion, the scRNA-seq profiles of DIMs, DAM, and YAM identified in mice can be extended to humans, and the programming of these three different populations in brain development, aging, and neurodegenerative processes appears to be conserved between the two species
.

11.
Accumulation of CD83+TNF-α+DIMs in the pia mater in AD patients

The authors next investigated whether DIMs exhibited the same distribution pattern and Aβ plaque colocalization
as in human mouse models of AD.
By immunohistochemistry of P2RY12, CD83, Aβ, and TNF-α brain slices from AD patients and non-AD patients, it was found that there were P2RY12+CD83+ cells in the pia mater (Figures 6I and 6J) in AD patients and not
in the parenchyma.
When looking at the localization of Aβ and DIMs, the authors found that Aβ clustered around pia mater vessels (Figure 6J) and within the parenchymal cortex; Also compared to non-AD patients, patients with AD have significantly higher density of P2RY12+CD83+TNF-α+ cells in the pia mater (Figure 6J and 6K).

And P2RY12 signal was higher in pia mater samples from AD patients compared to non-AD controls (Figure 6K).

In addition, P2RY12+CD83+ cells expressing the DIMs characteristic-associated molecule AXL in the pia mater of AD have significantly higher cell densities compared to non-AD brains (Figures 6L and 6M).

Thus, cells expressing DIM markers, including CD83 and TNF-α, accumulate around the pia mater in AD patients, and they may be associated with
neuroinflammation common in these regions during AD.

Figure 6 Increased number of P2RY12+CD83+TNF+ cells in cortical pia mater in patients with AD

Brief summary:

By integrating 6 scRNA-seq brain datasets, the authors' team generated a myeloid atlas called M-Verse that spans embryonic development to senescence and neurodevelopment to depict the heterogeneity
of macrophage populations.
M-Verse reveals two distinct macrophage populations that express published disease-associated microglial (DAM) characteristics: embryo-derived TREM2-dependent DAMs and monocyte-derived TREM2-independent inflammatory macrophages (DIMs).

Therefore, TREM2 as a potential therapeutic target for AD needs to be handled with great caution, as any changes could affect neuroprotective DAM and potentially neurodegenerative DIM populations in currently unpredictable ways
.

： ，
。   ， ，