Background
Alzheimer’s disease (AD) is accompanied by a robust inflammatory response [
1]. However, until recently, it has been unclear whether myeloid cells (including brain-resident microglia and possibly infiltrating monocytes) actively contribute to AD pathogenesis and progression. Recent Genome Wide Association Studies have linked single nucleotide polymorphisms (SNPs) in inflammation-related genes to increased AD risk [
2], including a SNP encoding the R47H variant in Triggering Receptor Expressed on Myeloid cells 2 (
TREM2). The
TREM2 R47H variant not only constitutes one of the strongest single allele genetic risk factors for AD [
3,
4], but also confers elevated risk for Parkinson’s disease, amyotrophic lateral sclerosis, and frontotemporal dementia [
5]. Furthermore, homozygous
TREM2 variants cause Nasu-Hakola disease, which is characterized by extensive white matter loss and frontotemporal-like dementia [
6]. These genetic studies definitively demonstrate that myeloid cell perturbations can contribute to neurodegenerative disease. However, it remains unclear how the
TREM2 R47H variant alters myeloid cell function to enhance disease risk.
In the brain, TREM2 is expressed exclusively by myeloid cells [
7,
8] and has been implicated in a diverse range of myeloid cell functions [
5]. A number of studies have investigated the role of TREM2 in AD pathogenesis using
Trem2 deficient mice. Myeloid cells accumulate around amyloid plaques in the AD brain, but the abundance of these plaque-associated myeloid cells is substantially diminished in AD mice lacking
Trem2, consistent with its known roles in myeloid cell survival and proliferation. Yuan et al. postulate that the loss of plaque-associated myeloid cells promotes plaque expansion and damage to surrounding neurites in
Trem2 deficient mice [
9]. In support of this hypothesis,
Trem2 deficient AD mice exhibit enhanced amyloid pathology at late stages in disease [
10,
11] accompanied by increased plaque-associated neuritic dystrophy [
9,
11,
12]. However, at early stages of disease progression,
Trem2 deficiency reduces amyloid burden [
10,
13].
While these studies have elucidated some important aspects of TREM2 function in the context of AD, how these studies relate to disease-associated TREM2 variants has only recently begun to be investigated. In vitro studies have demonstrated that the TREM2 R47H variant reduces affinity for TREM2 ligand binding [
9,
11,
14‐
18], and alters glycosylation [
19,
20], leading to speculation that the TREM2 R47H variant may result in a loss of TREM2 function. The function of the R47H variant was recently assessed for the first time in vivo. Song et al. expressed the human
TREM2 R47H variant using a bacterial artificial chromosome (BAC) transgenic and found that the R47H variant could not rescue aspects of TREM2 function in AD mice lacking endogenous
Trem2 expression [
21]. This study is in agreement with the in vitro data suggesting the
TREM2 R47H variant results in a loss of TREM2 function. However, because of the approach used in this study, it is unclear whether the loss of function phenotypes observed could be attributed to impairments in association of human TREM2 with mouse signaling pathways. In addition, these mice expressed eight copies of the
TREM2 gene and, because TREM2 overexpression has previously been associated with changes in microglial function and pathology [
22], it is difficult to determine which phenotypes observed in this study were due to the TREM2 R47H variant or overexpression of the TREM2 protein. In the current study, we use a complementary approach that maintains endogenous regulation of
Trem2 expression. We address the critical question of how the R47H
Trem2 variant alters TREM2 function in vivo
, including AD-associated myeloid cell responses, using AD mouse models in which CRISPR/Cas9 was used to knock the R47H variant into the endogenous mouse
Trem2 gene. Using this model, we demonstrate that the
Trem2 R47H variant dramatically reduces TREM2 expression, compromising myeloid cell responses to AD-like amyloid pathology. Furthermore, we are the first to demonstrate that these myeloid cell changes with the R47H
Trem2 variant alter plaque structure to enhance neuritic dystrophy.
Discussion
In order to investigate how the
Trem2 R47H variant affects TREM2 function and AD pathology, we developed a CRISPR/Cas9 knock-in of the R47H variant into the mouse
Trem2 gene. Because this approach maintains endogenous regulation of TREM2 expression, we were able to determine that expression of one copy of the R47H variant reduces
Trem2 expression in a wild-type background and further impairs upregulation of
Trem2 expression in an AD mouse model. This finding differs from a previous study that found no changes in TREM2 expression in postmortem tissue from human AD patients heterozygous for the
TREM2 R47H variant [
33]. While many factors could contribute to this discrepancy,
Trem2 levels are known to change throughout disease progression [
13], and our study evaluates
Trem2 changes at a relatively early stage in pathology in the APPPS1–21 model, while the postmortem samples are from humans at a late stage in disease. It will be interesting to assess in future studies whether
Trem2 levels are differentially affected by the R47H variant throughout disease progression.
Importantly, this finding also merits consideration when interpreting studies of TREM2 R47H function in vitro, which have all used systems where
Trem2 R47H expression is maintained at WT levels, and the recent evaluation of
Trem2 R47H variant function using BAC transgenics where
Trem2 was overexpressed. It is possible that the observed loss-of-function phenotypes may arise, at least in part, through reduced expression of TREM2. Furthermore, by knocking the R47H variant into the mouse
Trem2 gene, we maintain the appropriate interaction of mouse TREM2 with its endogenous ligands and signaling molecules. However, despite a high degree of homology between human
TREM2 and mouse
Trem2 genes, it is possible that the R47H variant affects human TREM2 differently than it affects mouse TREM2 structure and function. This caveat of our approach is addressed by complementary work using a BAC to express human TREM2 R47H in
Trem2-deficient AD model [
21]. Notably, our CRISPR/Cas9 knock-in approach and their BAC transgenic yield comparable results in myeloid cell accumulation around plaques, Together, these findings suggest that the
Trem2 R47H variant confers phenotypes consistent with loss of TREM2 function in a mouse model of AD-like amyloid deposition.
AD mice expressing the
Trem2 R47H variant exhibit reduced plaque-associated myeloid cells. We find that this is, in part, due to reduced proliferation. In addition, we demonstrate a selective reduction in plaque-associated cells expressing high levels of CD45 in mice expressing the
Trem2 R47H variant. It remains unclear whether the reduction in myeloid cell number represents impaired recruitment or survival of peripherally derived macrophages in the AD brain or diminished phenotypic conversion of resident microglia to adopt expression of this marker. Other possible mechanisms may also contribute to the reduction of myeloid cells around plaques in mice expressing the
Trem2 R47H variant, including deficits in myeloid cell migration [
34] and survival [
11].
The alterations in myeloid cell accumulation are also reflected by changes in inflammation-related gene expression. While changes in hippocampal gene expression are largely similar between APPPS1–21;Trem2
+/R47H
mice and APPPS1–21;Trem2
+/−
mice, in the cortex, increases in mRNA levels of Arg1, Fizz1, Ym1 and IL-6 are specific to mice expressing the Trem2 R47H variant. This demonstrates that there are some functional measures in which the R47H variant does not completely phenocopy loss of one copy of Trem2. These differences in cortical gene expression between APPPS1–21;Trem2
+/R47H
and APPPS1–21;Trem2
+/−
mice are not reflected in differences in the other myeloid cell phenotypes or features of pathology assessed in this manuscript. Additional experiments will be required to fully address whether these region-specific alterations in gene expression relate to other meaningful differences in myeloid cell function and pathology.
Our data show that the Trem2 R47H variant does not alter 6E10 positive plaque burden, but does reduce compact, thioflavin S positive plaques, suggesting that the changes in myeloid cell function mediated by the Trem2 R47H variant result in altered plaque structure. Yuan et al. suggested that this could be due to impaired accumulation of myeloid cells around plaques, which may normally limit plaque growth. However, it has also been shown that TREM2 influences the phagocytic activity of myeloid cells, which could also contribute to changes in plaque structure.
It has been previously postulated that myeloid cells form a barrier around plaques, protecting surrounding neurites from damaging Aβ species [
35], leading to the prediction that impaired association of myeloid cells with plaques would increase neuritic dystrophy. Indeed, studies have previously shown enhanced neuritic dystrophy with reduced myeloid cell plaque coverage in AD mice deficient for
Trem2, and in AD patients carrying the
TREM2 R47H variant [
9,
12]. Consistent with these findings, we observed an increase in dystrophic neurites, relative to plaque size, in mice expressing the
Trem2 R47H variant. However, it has also been shown that larger plaques typically have less microglial coverage and more neuritic dystrophy. Thus, we expected that reduced myeloid cell accumulation around plaques with changes in
Trem2 genotype would preferentially increase neuritic dystrophy around small plaques, and have less impact on larger plaques, since these plaques already exhibit little myeloid cell coverage. In contrast, however, we find that dystrophic neurite area correlated just as strongly with plaque size in both APPPS1–21;
Trem2+/R47H and APPPS1–21;
Trem2+/− mice. Furthermore, there was a trend toward an increase in the slope between dystrophic neurite area and plaque size in APPPS1–21;
Trem2+/R47H and APPPS1–21;
Trem2+/− mice relative to controls, suggesting that larger plaques may be even more strongly affected by the loss of TREM2 function, and consequently reduced accumulation of plaque-associated myeloid cells. Together, these data are suggestive of additional roles for TREM2 in modulating neuritic dystrophy other than limiting access of plaque species to surrounding neurites. These findings suggest that TREM2 may be involved in other mechanisms of dystrophic neurite formation, or perhaps more likely, given its demonstrated role of phagocytosis in vitro, in the clearance of these dystrophic neurites [
5]. It will be important to determine whether the enhanced neuritic dystrophy also correlates with neurodegeneration and cognitive deficits.
A central question arising from this work is how the changes observed in our study relate to the approximate three-fold elevation in AD risk in heterozygous carriers of the TREM2 R47H variant. Our data demonstrate that the Trem2 R47H variant impairs TREM2 function, in part by reducing TREM2 expression. This results in a reduced myeloid cell response to AD pathology, and increased neuritic dystrophy. Our results highlight the important functional roles of myeloid cells in AD pathogenesis and progression, and suggest that enhancing TREM2 signaling may be beneficial in the context of sporadic AD. In addition, because the TREM2 R47H variant confers risk for other neurodegenerative diseases, this study also provides a basis for understanding important myeloid cells functions and provides potential avenues for therapeutic targets in other disease contexts. Collectively, understanding the mechanism by which the Trem2 R47H variant affects myeloid cell function and pathology across multiple disease models promises to decipher common mechanisms by which myeloid cells modulate neurodegenerative disease pathology.